JP7740878B2 - Distributed photobiomodulation therapy systems and methods - Google Patents

Description

CROSS-REFERENCE TO RELATED APPLICATIONS
This application claims priority to U.S. Provisional Application No. 62/653,846, entitled "Distributed Photobiomodulation Therapy System and Method," filed April 6, 2018. This application is related to the following applications: International Application No. PCT/US2015/015547, entitled "Sinusoidal Drive System and Method for Phototherapy," filed February 12, 2015; International Application No. PCT/US2016/058064, entitled "3D Bendable Printed Circuit Board with Redundant Interconnects," filed October 21, 2016; and U.S. Patent Application No. 16/377,192, entitled "Apparatus and Method for Distributed Photobiomodulation Therapy, Biofeedback, and Communication Protocol," filed April 6, 2019.

Each of the foregoing applications and patents is incorporated herein by reference in its entirety.

The present invention relates to biotechnology for medical and health applications, including photobiomodulation, phototherapy, and photobiomodulation therapy (PBT).

Related art discussion

Biophotonics is a biomedical field related to the electronic control of photons, or light, and their interaction with living cells and tissues. Applications of biophotonics include surgery, imaging, biometrics, disease detection, and photobiomodulation (PBM). Photobiomodulation therapy (PBT), also known as phototherapy, is the controlled application of photons. Photobiomodulation therapy (PBT), also known as phototherapy, involves the controlled application of photons (usually infrared, visible, or ultraviolet) to induce photobiomodulation for medical therapeutic purposes. Applications of PBT include the treatment of injury, illness, pain, and immune system ailments. More specifically, PBT involves exposing cells and tissues undergoing treatment to a stream of photons of specific wavelengths of light, either continuously or in repeated discrete pulses, to control the energy transfer and absorption behavior of living cells and tissues.

Figure 1 shows the elements of a PBT system capable of continuous or pulsed light operation. The PBT system includes an LED driver 1 that controls and drives an LED as a source of photons 3 emitted from an LED pad 2 that illuminates a patient's tissue 5. While the human brain is shown as tissue 5, any organ, tissue, or physiological system can be treated using PBT. Before, after, or during treatment, a physician or clinician 7 can adjust the treatment by controlling the settings of the LED driver 1 according to their observations.

While many potential mechanisms exist, it is generally accepted that the primary photobiological process 22 involved in photobiomodulation during PBT treatment using red and infrared light, as shown in Figure 2, occurs within mitochondria 21, organelles present in all eukaryotic cells 20, including both plants and animals, including birds, mammals, horses, and humans. Current understanding suggests that the photobiological process 22 involves photon 23 collisions with cytochrome c oxidase (CCO) molecules, which act as a charger to increase the cellular energy content by transforming adenosine monophosphate (AMP) into the higher-energy adenosine diphosphate (ADP) molecule, or by converting ADP molecules into the higher-energy adenosine triphosphate (ATP) molecule. The cytochrome c oxidase molecule 24 acts as a battery charger, storing energy in the cellular battery, increasing the stored energy in the AMP-to-ADP-to-ATP sequence 25, a process that can be considered "photosynthesis" in animals. The cytochrome cytochrome c oxidase molecule 24 can convert energy from glucose resulting from food digestion into fuel through an ATP charging sequence 25, or through a combination of digestion and photosynthesis. To fuel cellular metabolism, the ATP 26 molecule can release energy 29 through a discharge process 28 from ATP to ADP and AMP. Energy 29 is used to drive protein synthesis, including the formation of catalysts, enzymes, DNA polymerases, and other biomolecules.

Another aspect of the photobiological process 22 is that cytochrome c oxidase molecules 24 are scavengers of nitric oxide (NO) molecules 27, a key signaling molecule in neuronal communication and angiogenesis, the growth of new arteries and capillaries. Illumination of cytochrome c oxidase molecules 24 within cells treated during PBT releases NO molecules 27 near damaged or infected tissue. The released NO increases blood flow and oxygen delivery to the treated tissue, accelerating healing, tissue repair, and immune response.

To perform PBT and stimulate cytochrome c oxidase molecules (CCO) 24 to absorb energy from photons 23, intervening tissues between the light source and the light-absorbing tissue cannot block or absorb the light. As shown in Figure 3, the molecular absorption spectrum of electromagnetic radiation (EMR) in human tissue is plotted as a graph 40 of absorption coefficients versus wavelength of electromagnetic radiation λ (measured in nm). Figure 3 shows the relative absorption coefficients S of oxygenated hemoglobin (curve 44a), deoxygenated hemoglobin (curve 44b), cytochrome c (curves 41A and 41b), water (curve 42), and fat and lipids (curve 43) as a function of light wavelength. As shown, deoxygenated hemoglobin (curve 44b) and oxygenated hemoglobin, i.e., blood (curve 44a), strongly absorb light in the red portion of the visible spectrum (especially for wavelengths shorter than 650 nm). At longer wavelengths in the infrared portion of the spectrum, EMR above 950 nm is absorbed by water (H2O) as shown by curve 42, as shown by the transparent optical window 45. At wavelengths between 650 nm and 950 nm, human tissue is essentially transparent.

Aside from absorption by fat and lipids (curve 43), EMR containing photons 23 of wavelength λ within the transparent optical window 45 is directly absorbed by cytochrome c oxidase (curves 41a and 41b). Specifically, cytochrome c oxidase molecules 24 absorb the infrared portion of the spectrum, represented by curve 41b, without interference from water or blood. The secondary absorption tail for cytochrome c oxidase illuminated with light in the red portion of the visible spectrum (curve 41a) is blocked by the absorption characteristics of deoxygenated hemoglobin (curve 44b), limiting photobiological responses in deeper tissues but activating them in epithelial tissues and cells. Thus, Figure 3 shows that PBT for skin, internal organs, and tissues requires different treatments and light wavelengths for red and infrared light.

Current photonic delivery systems

To achieve maximum energy coupling to the tissue during PBT, it is important to devise a consistent delivery system to consistently and uniformly illuminate the tissue with photons. Early attempts used filtered lamps, but these lamps were very hot, uncomfortable for the patient, and could potentially cause burns to the patient and physician. Maintaining uniform illumination during prolonged treatments is extremely difficult. Lamps also have a short lifespan and, if constructed using dilute gases, can be expensive to replace periodically. Filters require lamps to operate at very high temperatures to achieve the photon flux necessary to achieve efficient treatment over a reasonable treatment period. Unfiltered lamps, like the sun, actually have a spectrum that is too broad, limiting the efficiency of the photons. Broad-spectrum light, especially in the ultraviolet portion of the electromagnetic spectrum, simultaneously stimulates both beneficial and undesirable chemical reactions. Because ultraviolet light damages DNA, prolonged exposure to UV light is also known to increase the risk of cancer. In the infrared spectrum, prolonged exposure to far-infrared electromagnetic radiation and heat can dry out the skin and destroy elastin and collagen, leading to premature aging.

Alternatively, lasers have been and continue to be used to perform PBT. This is commonly referred to by the term LLLT, an acronym for Low-Level Laser Therapy. Unlike lamps, lasers pose a risk of burning the patient not by heat but by exposing tissue to intense, concentrated light power, also known as ablation. To prevent this problem, the output of laser light is limited, and special care must be taken to avoid accidentally generating excessively high currents that would produce dangerous light levels. A second, more practical problem arises from the laser's small "spot size," or area of irradiation. Because lasers illuminate a small focal area, it is difficult to treat large organs, muscles, or tissues, and it is much easier to become overwhelmed.

Another problem with laser light is that "coherence," which prevents the laser beam from spreading, makes it more difficult to cover large areas during treatment. Research has shown that PBT with coherent light offers no inherent additional benefits. First, bacterial, plant, and animal life have evolved and naturally absorb scattered light rather than coherent light, since coherent light does not naturally occur from any known light source. Second, the first two layers of epithelial tissue already destroy optical coherence, so the coherent properties of an incident laser beam are quickly lost upon absorption by human or animal tissue. Laser manufacturers promote the premise that optical interference patterns in laser light, called "speckle," resulting from backscattering, enhance therapeutic effects, but no scientific evidence has been provided to support such marketing-motivated claims.

Furthermore, the optical spectrum of a laser is too narrow to fully excite all the beneficial chemical and molecular transitions necessary to achieve highly efficient PBT. Because a laser's spectrum is limited, typically within a ±1 nm range from the laser's central wavelength, it is difficult to adequately excite all the beneficial chemical reactions required for PBT. Covering a frequency range with a narrow-bandwidth light source is difficult. For example, referring again to Figure 3, the CCO absorption spectrum (curve 41b) associated with the chromophore (light-absorbing molecule) in the chemical reaction is clearly different from the reaction that produces the absorption tail (curve 41a). Assuming that both regions of the absorption spectrum are shown to be beneficial, it is difficult to cover this wide range with a light source limited to a 2 nm-wide wavelength spectrum.

Sunlight has a very broad wavelength spectrum, stimulating many competing chemical reactions photobiologically with many EMR wavelengths. In contrast, the wavelength spectrum of laser light is too narrow to stimulate enough chemical reactions to provide a full phototherapy effect. This subject is discussed in more detail in a related application entitled "Phototherapy System and Process Including a Dynamic LED Driver with Programmable Waveforms" by Williams et al. (U.S. Application No. 14/073,371), now U.S. Patent No. 9,877,361, issued January 23, 2018, which is incorporated herein by reference.

To deliver PBT by exciting the entire wavelength range within the transparent optical window 45, i.e., the entire width from approximately 650 nm to 950 nm, each light source requires a bandwidth of approximately 80 nm, even if four different wavelength light sources are used to extend the range. This is more than an order of magnitude wider than the bandwidth of a laser light source. This range is too wide for lasers to cover in a practical manner. Today, LEDs are commercially available to emit a wide range of light, from the deep infrared to the ultraviolet portion of the electromagnetic spectrum. With a bandwidth of ±30 nm to ±40 nm, it is much easier to cover the desired spectrum, as center frequencies lie in the red, long red, short near-infrared (NIR), and mid-NIR portions of the spectrum (e.g., 670 nm, 750 nm, 810 nm, and 880 nm).

Photobiomodulation therapy (PBT) can be clearly distinguished from photo-optical therapy. As shown in Figure 4A, PBT involves the direct stimulation of tissue 5 with photons 3 emitted from an LED pad 2. Tissue 5 is unrelated to the eye and may include organs related to the endocrine and immune systems, such as the kidney, liver, glands, and lymph nodes, or the musculoskeletal system, including muscles, tendons, ligaments, and even bones. PBT also directly treats and repairs neurons, including peripheral nerves, the spinal cord, and (as shown) the brain 5 and brainstem. PBT transcranial treatment penetrates the skull and demonstrates significant and rapid therapeutic effects in concussion recovery and repair of damage caused by mild traumatic brain injury (mTBI). In other words, PBT energy is absorbed by chromophores in cells not associated with the optic nerve. In contrast, photo-optical therapy is based on stimulating the retina with colored light or images to elicit cognitive or emotional responses or synchronize the body's circadian rhythm to the surroundings. In this case, the image 12 from the light source 12 stimulates the optic nerve in the eye 11, sending electrical signals, or nerve impulses, to the brain 5.

Some basic tests highlight the many significant differences between PBT and phototherapy. For one, phototherapy only affects the eyes, while PBT affects all cells, including those in internal organs and the brain. In phototherapy, light is directed at light-perceiving cells (phototransduction), resulting in the generation of electrical signals that are carried to the brain. PBT, on the other hand, stimulates chemical transformations and ion, electron, and heat transport within treated cells and tissues without the need for brain signaling. The effects are local and systemic, without the brain's assistance. For example, blind patients respond to PBT but not to phototherapy. Figure 4B illustrates another distinction between phototherapy and PBT. In the case of vision, i.e., photooptical stimulation or vision, the combination of red light 15A and blue light 15B emitted by a light source 14, when received by the eye 11, sends an electrical signal 9 to the brain 5, causing the brain to perceive the color of the impinging light as purple. In reality, violet/violet light has a much shorter wavelength than blue or red light and therefore contains photons with higher energy than red light 15a or blue light 15b. In the case of PBT, cells 16 and the mitochondria 17 contained therein will photochemically respond to the light source 14 as if it were emitting red light 15a and blue light 15b, and will not respond as if violet light were present. Only true short-wavelength violet light emitted from a violet or ultraviolet light source can generate a photobiomodulation response to violet light. In other words, mitochondria and cells cannot be "fooled" by blending different colors of light, as the eye and brain can. In conclusion, photo-optical stimulation is very different from photobiomodulation. Therefore, photo-optical therapy technologies and developments cannot be considered applicable or relevant to PBT.

Etymologically, nomenclature ambiguity led researchers to abandon the term "phototherapy" or PBT in favor of the more specific term "photobiomodulation" or PBT. The terms phototherapy (i) phototherapy involving visual stimulation, (ii) photobiomodulation therapy or PBT involving cell modulation, and (iii) photodynamic therapy or PDT, in which light activates injected chemicals or applied ointments to promote a chemical reaction, were commonly used to refer to therapeutic applications. The similarly broad term "photochemistry," a chemical reaction stimulated by light, also vaguely refers to all of the aforementioned treatments. Thus, while photochemistry and phototherapy have broad meanings today, PBT, PDT, and phototherapy have specific, non-overlapping interpretations.

Another source of confusion is that the term LLLT was originally intended to mean "low-level laser therapy" to distinguish lasers operating at low power levels (sometimes referred to in the popular press as "cold" lasers) from lasers operating at higher powers for tissue ablation and surgery. With the advent of LED-based therapies, some authors have confused the nomenclature of laser- and LED-based therapies into "low-level light therapy," both of which share the same acronym: LLLT. This unfortunate behavior has caused much confusion in the published art and indiscriminately blurred the distinction between two very different photonic delivery systems. "Low-level" lasers are safe for eyes and burns simply because they operate at low levels. Cold lasers can cause severe burns and blindness within milliseconds of a laser being intentionally or accidentally powered up to a high level and no longer "cold." In contrast, LEDs always operate at low levels and cannot operate at high optical power densities. Without high power levels, LEDs can cause blindness. Also, although LEDs can overheat if exposed to too much current for too long, they cannot cause instantaneous burns or tissue ablation as in the last example. Therefore, the term low-level light has no meaning when it comes to LEDs. Therefore, throughout this application, the acronym LLLT will refer only to Laser PBT, which stands for Low Level Laser Therapy, and will never be used to refer to LED PBT.

Current photobiomodulation therapy systems

A current state-of-the-art photobiomodulation therapy system, system 50, shown in FIG. 5, includes a controller 51 electrically connected to two sets of LED pads. Specifically, output A of controller 51 is connected by cable 53a to a first LED pad set including electrically interconnected LED pads 52b. LED pads 52a and 52f are optionally connected to LED pad 52b by electrical jumpers 54a and 54b, creating a first LED pad set operating as a single LED pad including over 600 LEDs. The pad set covers a treatment area of over 600 ^cm2 . Similarly, output B of controller 51 is connected by cable 53b to a second LED pad set including electrically interconnected LED pads 52e. LED pads 52d and 52f are optionally connected to LED pad 52d by electrical jumpers 54c and 54d, creating a second LED pad set operating as a single LED pad including over 600 LEDs and covering a treatment area of over 600 ^cm2 .

In the system shown, controller 51 not only generates signals to control the LEDs in the pads, but also provides power to drive the LEDs. The power delivered to the LED pads by controller 51 is significant, typically 12 W for two sets of three pads each. An exemplary electrical schematic of the system is shown in Figure 6A. The controller 61 includes a switched-mode power supply 220 VAC power supply 64, at least two regulated DC voltage sources: 5 V for control and logic, and 65 SMPS used to convert power to 120 V for a higher voltage source; +VLED is used to power the strings of LEDs in the LED pads. Typical voltages for +VLED range from 24 V to 40 V, depending on the number of LEDs connected in series. To facilitate algorithmic control, microcontroller (μC) 67 executes dedicated software in response to user commands entered into touchscreen LCD panel 66. The resulting alternating pulse sequence at Output A from logic buffers 68a and 68b is used independently to control the red and near-infrared (NIR) LEDs in the LED pads connected to Output A. A similar arrangement is included for output B, using its own dedicated logic buffer, but the μC67 can manage and control both the A and B outputs simultaneously.

The signal on Output A is then routed to one or more LED pads 62 via a shielded cable 63 including a high-current power line ground GND 69a, a 5V supply line 69b, and a +VLED supply line 69c, as well as an LED control signal line 70a. This also includes an LED control signal line 70a for controlling the conduction of NIR LEDs 71a-71m and an LED control signal line 70b for controlling the conduction of red LEDs 72a-72m. Each drives the base terminal of a bipolar junction transistor 73a or 73b, which acts as a switch to pulse the corresponding string of LEDs on and off. If the input to either bipolar transistor is low, i.e., biased to ground, no base or collector current flows, and the LED string remains dark. If the input to either bipolar transistor is high, i.e., biased to 5V, the base current flows and corresponding collector current flows, lighting up the LEDs in the corresponding LED string. The LED current flow is set by the LED turn-on voltage and current-limiting resistor 74a or 74b. Using a resistor to set the LED brightness is undesirable because variations in LED voltage due to manufacturing stochastic variations or temperature changes during operation result in variations in LED brightness. This results in poor uniformity of LED brightness across the LED pad, from LED pad to LED pad, and from one manufacturing batch to the next. Improvements in maintaining LED brightness uniformity can be achieved by replacing resistors 74a and 74b with fixed-value constant current sources or sinks 75a and 75b, as shown in FIG. 6B.

The physical connection between the PBT controller 61 and the LED pad 62 via the shielded cable 63 can also be described as two interacting communication stacks in terms of the Seven-Layer Open Source Initiative, or seven-layer OSI model. As shown in Figure 7, the PBT controller 61 can be represented as a stack 80 that includes an application layer, and the PBT controller's operating system is referred to as LightOSv1. In operation, the application layer transfers data to the Layer 1 physical layer, or PHY layer, which includes logic buffers. The stack 80 then transmits electrical signals 82 unidirectionally to the PHY layer 1, i.e., the LED string driver, within the passive LED pad 62's communication stack 81.

Because the electrical signal contains simple digital pulses, parasitic impedances in the cable 63 can affect the integrity of the communication signal and the operation of the LED pads. As shown in FIG. 8, the transmitted square-wave electrical signal 82 can be significantly distorted into a received waveform 83, including reduced amplitude and duration 84a, slow rise time 84b, voltage spikes 84c, oscillations 84d, and ground loops 89 that affect the signal's ground bounce 84e. Parasitic elements in the cable that contribute to these distortions include power line series resistance 87a-87c, inductance 86a-86c, and inter-conductor capacitance 85a-85e. Other effects include ground loop conduction 89 and antenna effects 88.

Another drawback of using a simple electrical signal connection between the PBT controller 61 and the LED pads is that the PBT system cannot verify whether the peripheral device connected to cable 63 is actually a qualified LED pad or an invalid load. For example, an improper LED configuration will not match the PBT controller. The icons in Figure 9 are intended to represent classes of electrical loads but should not be considered specific circuits. In contrast, connecting too few LEDs in series, as shown in icon 92, can result in overcurrent, overheating, and the risk of patient burns.

Powering non-LED loads from the PBT controller 61 can damage ineffective peripherals, the controller, or both. This is particularly problematic because one pin on the PBT controller's output supplies high voltages (over 20V), exceeding the 5V rating of most semiconductors and causing permanent damage to the integrated circuit. Inductive loads, represented by icon 94, can cause overvoltage spikes that can damage the controller. Loads including motors such as disk drives and fans can lead to excessive and damaging inrush currents. A shorted cable or shorted electrical load, as shown by icon 93, can cause a fire. Connecting a battery to the PBT controller 61, as shown by icon 96, can result in excessive current and fire hazards. Overcharging or overvoltageing chemical cells can also cause severe fires or explosions. Unknown electrical loads, represented by icon 95, represent unspecified risks. Of particular concern is the connection between the PBT controller 61 and a power source such as a generator, car battery, or UPS, which could result in complete destruction of the system and extreme fire hazard. In Figure 9, the icons are intended to represent classes of electrical loads but should not be considered specific circuits.

Other problems arise when mismatched LED pads are connected to the same output. For example, in FIG. 10, two different LED pads 62 and 79, powered by a common cable 63, share connections to ground 69a, a 5V power supply 69b, a high-voltage +V LED power supply 69c, a visible LED (v) control signal 70a, and a near-infrared LED ( _nir) control signal 70b. As shown, LED pad 62 includes current sinks 75a and 75b and switches 73a and 73b that drive corresponding LEDs 71a-71m having a visible wavelength _λv and LEDs 72a-72m having a near-infrared wavelength _λnir . Alternatively, LED pad 79 includes the same current sinks 75a and 75b and switches 73a and 73b, but drives LEDs of different wavelengths, specifically, LEDs 76a-76m having a visible wavelength _λv2 and LEDs 77a-77m having a near-infrared wavelength _λnir2 . Similarly, the parallel connection of 810 nm and 880 nm LEDs driven by LED _nir signal 70a means that processing NIR LEDs of one wavelength may inadvertently drive a different wavelength. In operation, the parallel connection of red and blue LEDs driven by LEDv signal 70a means that processing red light may inadvertently drive blue light. Similarly, the parallel connection of 810 nm and 880 nm LEDs driven by LED _nir signal 70a means that processing NIR LEDs of one wavelength may inadvertently drive a different wavelength.

Another problem arises when two or more LED pads are connected to both LED outputs simultaneously, as shown in FIG. 11A. As shown, PBT controller 51 has two outputs, Output A and Output B. These outputs are intended to drive separate sets of LED pads. As shown, Output A connects to LED pad 52d via cable 53a. Output B connects to LED pad 52e via cable 53b and also to LED pad 52f via jumper 54d. However, by coincidence, jumper 54c connects LED pad 52e to LED 52d, thereby shorting Output A to Output B. The electrical impact of shorting Outputs A and B together depends on the treatment program being executed. FIG. 11B illustrates the case where both Outputs A and B of buffer 100 are driving the red/visible light output, specifically when buffers 101a and 101c are simultaneously active. It is connected to LED pad 105a via conductor 102a, to LED pad 105b via connector 104a, and finally shorted via connector 103a. During operation, the frequencies and pulse patterns of the two outputs are asynchronous, meaning any combination of high and low output bias can occur. If the pull-up transistor is too strong, one output buffer could destroy another. Otherwise, the alternating on signals could cause the LED to remain on at a high duty factor, causing overheating and risking burns to the patient.

In FIG. 11C, buffer 101a at output A powers the red LEDs on LED pads 105a and 105b, while buffer 101d at output B also powers the NIR LEDs on LED pads 105a and 105b. While independent operation of both the red and NIR LEDs does not represent an electrical problem, simultaneous conduction of both the red and NIR LEDs can cause the LED pads to overheat, potentially damaging the pads and potentially causing burns to the patient. This overpower condition is illustrated by the waveforms shown in FIG. 11D, where the power _Pv of the conductive visible LEDs, represented by waveform 110, has an average power Pave 113, and the power _Pnir of the NIR LEDs, represented by waveform 111, has an average power _Pave 114. Combined, the total power waveform 112 has an average power 115 of magnitude _2Pave .

With today's LED pads, overheating for any reason is an issue because they lack thermal protection. As shown in Figure 12, even if the LED pad 109 has temperature sensing, the unidirectional data flow 82 in the cable 63 does not provide a way for the LED pad 109 to notify the PBT controller 61 of an overheating condition or suspend operation.

As discussed above, there are numerous limitations to today's PBT systems that impact PBT system utility, functionality, safety, and scalability. These limitations include the following issues:
Electrical "signal" communication to the LED pads - The signals from the PBT controller to the LED pads are simple digital pulses, not differential communication between bus transceiver pairs. These signals are susceptible to common-mode noise and ground loops, which affect the magnitude and duration of the pulses that control the LED's operation. As simple electrical pulses, the system also has no error checking capabilities, so malfunctions cannot be corrected or detected.
Unidirectional signal flow from the PBT controller to the LED pads - Unidirectional data flow prevents the PBT controller from authenticating the LED pads connected to its outputs, and from monitoring the operational status of the pads once connected. Unidirectional data also prevents feedback of the LED pad status or reporting of other pad information to the host PBT controller.
- Inability to detect multi-pad misconnection/short circuit - Due to user error, two outputs of the PBT controller are misconnected to the same LED pad, meaning that if the two outputs are accidentally shorted together, both outputs will be driving the same LED string. This misconnection error can damage the LED driver circuit, cause the LEDs to overheat, risking burns to the patient and potentially causing a fire.
Inability to Identify Approved LED Pads or Certified Manufacturers - Lacking the ability to identify the lineage of LED pads, PBT systems unknowingly drive any connected LED, including illegal, counterfeit, or imitation LED pads. Driven pads not manufactured or certified by the system specifier or manufacturer can have unknown consequences, ranging from loss of function or reduced effectiveness to safety risks. Commercially, the commercialization and sale of counterfeit and imitation LED pads also deprives IP-licensed PBT device sellers of legal revenue.
LED Pads Cannot Identify Connected Devices - Connecting unauthorized electrical loads to the outputs of a PBT system without the ability to determine if the device connected to the PBT control output is an LED pad (rather than something completely unrelated to the surroundings, such as a speaker, battery, motor, etc.) will inevitably damage the accessory, the PBT controller, or both. Driving unknown electrical loads can also pose a fire hazard if high voltage is present on the controller's output pins during operation.
Inability to Identify Power Sources - The inability of a PBT controller to identify its connection to an output power source (such as an AC power adapter, battery, car power, or generator) represents a real safety risk. The PBT controller will compete with the external power source. Interconnecting two different power sources can result in excessive current, voltage, power loss, or uncontrolled oscillations, potentially damaging the external power source, the PBT controller, or both.
Inability to control or limit the driver's output current - Damaged pads, shorted wires, or connection of shorted loads such as motors with high inrush currents can create a high current risk and a fire hazard. Inductive loads such as solenoids can also momentarily generate excessive voltages that can damage low-voltage components.
Inability to detect batteries connected to the output of a PBT system - Connecting a battery pack to the output of a PBT system can damage the battery pack and accidentally charge the battery under incorrect charging conditions, resulting in overvoltage, overcurrent, or overheating of electrochemical cells. Improper charging of wet-chemistry or acid batteries can result in acid or electrolyte leakage. Improper charging of lithium-ion batteries can cause overheating, fire, or even explosion.
Inability to detect LED pad overheating - Overheating of the LED pad can cause patient discomfort, burns, damage to the pad, and in extreme cases, the risk of fire.
Inability to Identify LED Configuration within an LED Pad - If the series-parallel array configuration of LEDs within an LED pad cannot be identified, the PBT controller cannot determine whether the pad is compatible with the PBT system or whether LED operation is possible. For example, too few LEDs connected in series can damage the LEDs by drawing too much voltage. Too many LEDs connected in series can result in dim lighting or no illumination at all. Too many parallel strings of LEDs can not only cause excessive total current through the pad, resulting in overheating, but also increase voltage drop across the interconnects, reducing light uniformity across the LED pad and potentially damaging the PCB's conductive traces.
Inability to identify the type of LEDs contained in the LED pad - Because the PBT system cannot detect which wavelength LEDs are in the pad, it has no way to match the treatment program to the LED array or select the appropriate wavelength LED for each specific waveform in the treatment protocol.
PBT controller outputs are limited to a fixed number of control signals each - with only one or two control signals per output, today's PBT controllers cannot drive three, four, or more different wavelength LEDs with different excitation patterns within the same pad.
Limited mobility - Current medical-grade PBT systems require a cable connection to connect the central PBT controller to the LED pads. While such tethered PBT systems are generally acceptable in hospital applications (and sometimes clinical settings), in consumer, emergency care, and military applications, restricting movement with cables or wires is not useful.
Inability to Synthesize Waveforms - PBT systems lack the technology to drive LEDs with waveforms other than square-wave pulses. Square-wave pulse operation limits LED illumination patterns to one frequency of operation at a time. Because pulse frequency affects energy coupling to specific tissue types, single-frequency PBT systems can only optimally treat one tissue type at a time, extending the required treatment time and patient/insurance costs. Analysis also shows that square-wave pulses waste energy and produce off-harmonics that are not necessarily beneficial to treatment. LED drives using sine waves, chords, triangle waves, sawtooth waves, noise bursts, or audio samples require complex waveform synthesis within the LED pad. The host PBT controller would need sufficient computational power to synthesize such waveforms, but this capability is not beneficial because signals cannot be distributed over long cables without incurring significant waveform distortion. Unfortunately, LED pads are not up to the task. Using inexpensive discrete components, current LED pads cannot perform waveform synthesis, not to mention the lack of the communication protocols necessary to remotely select or change synthesized waveforms.
Distribution of new LED driver algorithms - Current PBT systems do not have the ability to download software updates from a database or server to fix software bugs or install new treatment algorithms.
Inability to capture and record real-time patient biometric data – Current PBT systems lack the capability to collect biometric data such as EEG, blood pressure, blood glucose, blood oxygen, and other biometrics during treatment, or to embed this collected data into the treatment file record.
Inability to collect real-time images of the treatment area – Current PBT systems do not provide a means to measure or create images of the tissue during treatment. The systems also lack the ability to store still and video images or match images to the treatment time of the PBT session.
Users (physicians) cannot create new treatment algorithms – Current PBT systems lack the ability for users (physicians, researchers, etc.) to create new algorithms or string together existing treatments to form complex, treatment-specific treatments, such as optimizing the excitation sequence for activation of infused stem cells (which helps promote stem cell differentiation while reducing the risk of rejection).
Electronic Distribution of Documentation - Current PBT systems are not capable of distributing and updating documents electronically. It would be beneficial to provide electronic distribution of FDA recommendations or rulings, as well as errors and updates to PBT operating and treatment manuals, treatment guides, and other documentation, to all PBT system users. Such functionality is not currently available in any medical device.
Treatment Tracking - Current PBT systems cannot track treatment usage history, record system usage in a treatment log, or upload the treatment log to a server. The lack of real-time treatment logs via a network connection poses challenges to widespread commercial adoption of PBT systems by doctors, hospitals, clinics, and spas. Without uploaded usage logs, lessors cannot verify lessees' system usage, making current PBT systems unable to support revenue-sharing leasing business models. Similarly, hospitals and clinics cannot verify PBT system usage for insurance audits or fraud prevention. In the pay-to-use SaaS (Software as a Service) payment model, PBT service agents cannot verify clients' usage history.
Electronic Prescriptions - Today's physical therapy devices, including PBT systems, are unable to securely transfer and distribute doctor prescriptions to medical devices.
Remote Disablement - Current PBT systems do not allow for device operation to be disabled to stop black market transactions in the event of non-payment or theft.
Location Tracking - Today's PBT systems do not allow you to track the location of a stolen PBT system to track the thief.
Secure Communications - Today's PBT systems use electrical signals rather than packet-based communications to control the LED pads, making communication between the host PBT system and the LED pads easy to hack and directly measure, and completely lacking security. Furthermore, today's PBT systems lack the necessary internet communications and security measures to prevent content hacking and comply with HEPA regulations to thwart identity theft. In the future, encryption alone will likely be insufficient to protect data communications over the internet. In these cases, connection to a private hyper-secure network will also be necessary.

In summary, current PBT system architectures are completely outdated, and entirely new system architectures, new control methods, and new communication protocols are needed to facilitate effective, flexible, versatile, and safe solutions for delivering photobiomodulation therapy.

In the photobiomodulation therapy (PBT) process of the present invention, a defined pattern (e.g., a sequence of square wave pulses, sine waves, or combinations thereof) of electromagnetic radiation (EMR) having one or more wavelengths, or spectral bands of wavelengths, is introduced into a living organism (e.g., a human or animal) using a distributed system including two or more distributed components or "nodes" that communicate using a bus or transceiver to transmit instructions or files between or among the constituent components. The radiation is typically in the infrared or visible portion of the EMR spectrum, but may also include ultraviolet radiation.

A single wavelength of EMR can be used, or the pattern may include EMR with two, three, or more wavelengths. Rather than being composed of radiation at a single wavelength, EMR may include a spectral band of radiation. This is often expressed as a wavelength range centered around a central wavelength, e.g., λ±Δλ. Pulses or waveforms may be separated by gaps where no radiation is generated, or the falling edge of one pulse or waveform may coincide in time with the rising edge of the next pulse, or pulses may overlap, resulting in radiation at two or more wavelengths (or spectral bands of wavelengths) being generated simultaneously.

In one embodiment, the components of the distributed PBT system consist of a PBT controller and one or more intelligent LED pads that communicate using a unidirectional serial data bus that transmits data, files, instructions, or executable code from the PBT controller to the intelligent LED pads. In a second embodiment, the components of the distributed PBT system consist of a PBT controller and one or more intelligent LED pads that communicate using a bidirectional data bus or transceiver, allowing the PBT controller to transmit data, files, instructions, or executable code to the intelligent LEDs. Conversely, the intelligent LED pads can return data to the PBT controller, including pad operational status or patient status, including LED pad configuration data, program status, fault conditions, skin temperature, or other sensor data. Other sensors may include two-dimensional temperature maps, two-dimensional or three-dimensional ultrasound images, or biometric data such as pH, humidity, blood oxygen, blood glucose, or skin impedance, which optionally alter treatment conditions, i.e., operate in a closed biofeedback loop.

In one embodiment, EMR is generated by light-emitting diodes (LEDs) arranged in a series "string" connected to a common power source. Each LED string may contain LEDs designed to generate radiation at a single wavelength or band of wavelengths in response to a defined, constant, or time-varying current. The LEDs are embedded in flexible pads designed to fit snugly against the skin surface of the human body, exposing the target tissue or organ to a uniform pattern of radiation. Power is supplied to each intelligent pad through a cable connecting the LED pad to a PBT controller, or to the LEDs from a separate power source. In an alternative embodiment, semiconductor laser diodes may be used in place of the arrayed LEDs to create a uniform pattern of radiation, or may be attached to a handheld wand to create a spot or small area of focused radiation.

In the distributed PBT system disclosed herein, each LED string is controlled by an LED driver, which is in turn controlled by a microcontroller contained within the intelligent LED pad. The LED pad microsystem communicates with another microcomputer or computer comprising the PBT controller via a communication bus. The communication bus may include a wired connection such as USB, RS232, HDMI, I2C, SMB, Ethernet, or a proprietary format and communication protocol. Alternatively, it may include wireless media and protocols, including cellular radio using Bluetooth, Wi-Fi, WiMax, 2G, 3G, 4G/LTE, or 5G protocols, or other proprietary communication methods.

Using a display, keyboard, or other input device connected to the PBT controller, a physician or clinician can select a specific algorithm (process sequence) appropriate for the condition or disease being treated. Instructions are then transmitted from the PBT controller via a wired or wireless data bus to one or more intelligent LED pads, instructing the pad's microcontroller when to start or pause PBT treatment and which actions to perform.

In one embodiment, called data streaming, the PBT controller sends a stream of data packets specifying the LED drive waveforms, including when the LEDs are instructed to conduct current and the magnitude of the current being conducted. The streaming instructions sent by the controller are selected from a "pattern library" of algorithms. Each algorithm defines a specific process sequence for the EMR pulses or waveforms produced by the LED string. Upon receiving the data packets over the data bus, the intelligent LED pad stores the instructions in memory and begins "playing" the streaming data file—that is, driving the LEDs according to the received instructions. During streaming playback, bus communication from the PBT controller to the intelligent LED pad is interrupted to accommodate system safety checks and to allow the intelligent LED pad to report its status or upload sensor data to the PBT controller.

Unlike prior art PBT systems, the disclosed distributed PBT-based PBT controller does not constantly send instructions to the intelligent LED pads. While the PBT controller is silent, such as listening to the bus or receiving data from the intelligent LED pads, each intelligent LED pad must operate autonomously and independently from the PBT controller and other LED pads connected to the same data bus or communication. In other words, the PBT controller must send enough data to the intelligent LED pads to store in the pad's memory buffer and support uninterrupted LED playback operation until the next data file is delivered.

In another embodiment, the PBT controller distributes a complete playback file to the intelligent LED pad that defines the entire execution sequence of a PBT treatment or session. In this method, the file is distributed before starting playback, i.e., before executing the treatment. As soon as the file is loaded into the memory of the intelligent LED pad, the local microcontroller in the pad can execute the playback performed according to the file's instructions. The transmitted playback file can be (i) an executable code file containing the entire set of all LED driving waveform instructions that define the treatment duration and settings interpreted by executable code including LED player software; (ii) a passive playback file that defines the treatment duration and settings interpreted by executable code including LED player software; or (iii) a data file containing waveform primitives that control LED lighting patterns and are then combined in a predetermined manner by the LED pad's microcontroller to execute the PBT treatment or session.

In the latter two examples, the executable code required to interpret the playback file, i.e., the LED player, must be loaded into the intelligent LED pad before playback can begin. This LED player can be loaded into the intelligent LED pad when the user instructs the PBT controller to begin treatment. Or, it can be loaded into the intelligent LED pad the day before, such as during manufacturing or if the LED pad was programmed at the time of manufacture. Ensure the PBT controller is turned on and the intelligent LED pad is connected to the controller's local area network. If the LED player file has previously been loaded into the intelligent LED pad and stored in non-volatile memory for an extended period of time, the distributed PBT system must include provisions to check whether the loaded software is still up-to-date or has become obsolete. If the system detects that the LED player is up-to-date, it can immediately begin LED playback. Alternatively, if the PBT controller detects that the LED player is obsolete, out-of-date, or simply not up-to-date, the PBT controller can download new LED player executable code, either immediately or with first obtaining user approval. In some cases, running a treatment using discarded LED player executable code may result in improper playback or system malfunction. In such cases, the Intelligent Pad LED Player may be forced to suspend operation by the PBT Controller until a software download and update is performed.

The ability of an LED pad to function independently and autonomously for a defined period of time distinguishes it as "intelligent" compared to passive LED pads. In contrast, passive LED pads are limited to responding only to real-time signals sent from the PBT controller; any interruption in communication immediately interrupts the operation of the LED pad, affecting the LED pulse train or waveform. In other words, the bus communication between the PBT controller and one or more intelligent LED pads can be considered a packet-switched local area network (LAN).

Another key feature of the disclosed distributed PBT system is its autonomous safety system—protection and safety functions operate in each intelligent LED pad independently of the PBT controller. Particularly in networked specialized medical devices, the safety system must continue to operate reliably even if network connectivity is lost. As a key feature of the invention, during operation, each intelligent LED pad periodically executes safety-related subroutines to ensure that the software is operating correctly and that no hazardous conditions exist. Embedded protection features in the intelligent LED pad include a "blink timer-related software" subroutine, a watchdog timer, overvoltage protection, LED current balancing, and overtemperature protection. The autonomous safety features include firmware that constitutes the intelligent LED pad's local operating system (referred to herein as LightPadOS), stored in non-volatile memory and executed by an embedded microcontroller residing within each intelligent LED pad.

Upon receiving an instruction to begin treatment, the LightPadOS for a particular pad starts a software timer, simultaneously resetting and starting the microcontroller's hardware counter. The LightPadOS then launches executable code to execute PBT processing, either as a streaming data file or as an LED player (playing a specific playback file), synchronized with the ongoing program counter. The program counter advances at a frequency defined by either a shared system clock or a precise time reference specific to one or more intelligent LED pads. Such a time reference can be established using an RC relaxation oscillator, an RLC resonant tank oscillator, a crystal oscillator, or a micromechanical machine-based oscillator. In this way, nanosecond-accurate pulses can be used to synthesize square wave pulses, sine waves, and other waveforms of varying frequency and duration. The synthesized waveform is then used to drive a string of different waveform LEDs in a selected pattern according to a defined algorithm.

While the program is running, both the software blink timer and the hardware-based watchdog timer continue counting, synchronized with the program counter's time base. When the blink timer reaches a certain predefined time (referred to here as the blink interval), for example 30 seconds, the software timer generates an interrupt signal that is sent to the pad's local control, LightPadOS. This pauses the treatment's program counter and initiates an interrupt service routine, or ISR. The ISR then performs housekeeping functions, including reading the temperature of one or more sensors within the intelligent LED pad, transmitting the temperature data via a transceiver to the PBT controller, and simultaneously comparing the maximum measured temperature with a defined range. If the temperature exceeds a warning level, a warning flag is also generated and communicated to the PBT controller to request the system to take some action, such as reducing the LED duty factor (time per cycle) to reduce pad temperature or interrupting treatment.

However, if the maximum measured temperature exceeds a predetermined safety threshold, the Intelligent LED Pad immediately pauses the execution of the therapy program and simultaneously sends a message via the transceiver to the PBT Controller. Unless the PBT restarts the program, the overheating Intelligent LED Pad will remain off indefinitely. Thus, if an overheating condition occurs while the PBT Controller is unavailable or malfunctioning, or if the network or communication bus is busy or unavailable, the default state is to stop therapy.

During an ISR, the Intelligent LED Pad can perform other safety tests. For example, it can check for excessive input voltage due to a power supply failure, excessive current due to an internal pad short, or excessive moisture due to sweat or water coming into contact with the Intelligent LED Pad. This may indicate a missing or improperly applied hygienic barrier between the patient and the LED Pad. In either case, the malfunctioning Intelligent LED Pad first suspends operation and then sends a message to the PBT Controller notifying the distributed system of the fault. In such a case, other LED Pads can continue to operate independently (even if one pad has stopped operating), or all Intelligent LED Pads can be shut down simultaneously (via the PBT Controller or direct pad-to-pad communication). Once the ISR is complete, control returns to PBT processing execution by restarting the program counter, restarting the software blink timer, and restarting the watchdog timer.

If a software execution error occurs in either the LED regeneration executable code or the ISR subroutine, the program counter will not resume operation and the blink timer will not be reset and restarted. If the watchdog timer reaches its full count (for example, in 31 seconds) without being reset, it indicates a software execution failure. A watchdog timer timeout immediately generates an interrupt flag, suspends program execution on the problematic LED pad, and sends a fault message to the PBT controller and, optionally, other LED pads. Therefore, a software failure will always default to a non-operational state for the malfunctioning LED pad to ensure patient safety, even in the absence of network connectivity.

Aside from autonomous safety features, in another embodiment, the disclosed distributed PBT system includes centralized protection of the networked components managed by the PBT controller. Specifically, the PBT operating system running on the PBT controller, referred to herein as LightOS, includes several protection features, such as the ability to detect whether a component connected to a network or communication bus is authorized or unauthorized. If a user attempts to connect a light pad or other component to the PBT controller's network that fails to pass the required authentication process, the component is denied access to the network. The PBT controller's LightOS operating system can enforce various methods of tamper prevention, such as shutting down the entire distributed system until the offending device is removed, not sending data packets to the IP address of the unauthorized device, or encrypting commands so that unauthorized components cannot see them.

To achieve multi-layered secure communication in the disclosed distributed PBT system, the operating system of the PBT controller (LightOS) and the operating system of the intelligent LED pad (LightPadOS) are composed of parallel communication stacks using consistent protocols and shared secrets that are unrecognizable to device operators, hackers, or unauthorized developers. As such, the distributed PBT system operates as a protected communication network with the ability to enforce security at any number of communication layers, including data link layer 2, network layer 3, transport layer 4, session layer 5, presentation layer 6, or application layer 7.

For example, a cryptographically hidden numeric code, or shared secret, installed in both the PBT controller and the intelligent LED pad can be used to verify the authenticity of a network-connected intelligent LED pad without revealing the key itself. In one method of LED pad verification, performed at Data Link Layer 2, the PBT controller passes a random number to the intelligent LED pad over the network or communications bus. In response, the microcontroller within the LED pad decrypts its copy of the shared secret (numeric code), merges it with the received random number, and then performs a cryptographic hash operation on the concatenated number. The intelligent LED pad then openly returns the cryptographic hash value over the same transceiver link.

At the same time, the PBT controller performs an identical operation to decrypt its own copy of the shared secret (a numerical code), merge it with the generated random number sent to the LED pad, and then perform a cryptographic hash operation on the concatenated number. The PBT controller then compares the received hash value with a locally generated hash value. If the two numbers match, the pad is authentic, i.e., it is "authorized" to connect to the network. The aforementioned authentication algorithm can be performed over any PHY Layer 1 and/or Data Link Layer 2 connection over any data bus or packet-switched network, including USB, Ethernet, Wi-Fi, or cellular wireless connections. In the case of a Wi-Fi connection, the data link can also be established using the Wi-Fi Protected Access Protocol (WPA2).

For "management" purposes and security tracking, the authorization date and time (and GPS location, if available) of authenticated components are stored in non-volatile memory and uploaded to a server as needed. The benefits of employing secure communication and AAA (Authentication, Authorization, and Administration) verification of all connected components in a distributed PBT system are critical to ensuring safety and protection from the intentional connection of unauthorized and potentially insecure imposter devices. In this way, imposter devices cannot be powered by the distributed PBT system. AAA verification also protects against the accidental connection of devices not intended to operate as part of the PBT system, such as lithium-ion battery packs, unauthorized power supplies, speakers, disk drives, motor drivers, high-power Class III and Class IV lasers, and other potential hazards unrelated to the PBT system.

Security for distributed PBT systems using packet-switched networks (such as Ethernet or Wi-Fi) can also be enhanced through the use of dynamic addressing at Network Layer 3 and dynamic port assignment at Data Transport Layer 4. When an unconnected PBT controller is operating, it generates a dynamic IP address and dynamic port address for the Internet or other local area network, then broadcasts the address to other network-connected devices, which respond with their own dynamic IP address and unique dynamic port address. When a distributed PBT system is connected to a router or the Internet, it assigns dynamic IP addresses using Dynamic Host Configuration Processor (DHCP). Similarly, it assigns dynamic port numbers using Remote Procedure Call (RPC). Because the dynamic IP address and dynamic port change each time a device connects to the network, the cyber attack surface is reduced. Additional Layer 4 security can be added using TLS "Transport Layer Security," IPSec security protocols, or other protocols.

Once the components of a distributed PBT system are established through Layer 2 authentication and Layer 3 and Layer 4 network and port address assignment, the distributed PBT system is ready to perform processing. When the PBT Controller receives the user's "Start" command, PBT processing begins by establishing a Layer 5 session between the PBT Controller and the networked Intelligent LED Pads through the exchange of encryption keys or digital certificates. Once the session is opened, the PBT Controller and Intelligent LED Pads maintain a secure link during the exchange of files and commands until treatment is complete or terminated. Additional network security can be achieved using encryption at Presentation Layer 6 or Application Layer 7.

As disclosed, the networked, distributed PBT system functions as a single, unified virtual machine (VM) capable of reliably and safely delivering photobiomodulation therapy using multiple intelligent LED pads.
- No waveform distortion due to cable parasitics - Bidirectional communication between the PBT controller and the intelligent LED pad - Capability - Detects short circuits in multi-pad misconnections - Ability to identify approved LED pads or certified manufacturers - Ability to identify connected devices as intelligent LED pads - Ability to identify power sources and control their operating voltage - Ability to control and limit the LED current in the driver - Ability to detect batteries and prevent their connection to the output of the PBT system - Ability to detect overheating conditions in the LED pad - Ability to identify the configuration of LEDs in the LED pad - Ability to identify the type and structure of LEDs contained in the intelligent LED pad - Ability to control multiple outputs independently - Ability to perform distortion-free waveform synthesis within the intelligent LED pad - Ability to distribute new LED driver algorithms to the intelligent LED pad - Ability to capture and record real-time patient biometric data - Ability to collect real-time images of the treatment area - Ability for users (doctors) to create new treatment algorithms - Ability to support electronic distribution of documents - Ability to perform treatment tracking - Ability to manage the distribution of electronic prescriptions - Ability to support networked remote control - Ability to perform location tracking of the PBT system - Ability to perform secure communication between components

In another embodiment, the disclosed distributed PBT system includes three-stage waveform generation including digital waveform synthesis, PWM pulse generation, and a dynamically multiplexed multi-channel LED driver capable of generating square, triangular, sawtooth, and sine waveforms. The waveforms may consist of a single periodic function or a chord of multiple frequency components.

In another embodiment, the disclosed waveform generator can generate chords based on a predetermined key and frequency scale, for example, chords containing two, three, or four different frequencies with noise filtering. LED drive waveforms can also be generated from audio samples or by combining codes of scalable audio primitive waveforms of various resolutions and frequencies. Waveforms can be stored in libraries based on waveform synthesizer parametric, PWM waveforms, and PWM codes such as major, minor, diminished, augmented chords, octave, and inversion. Software-controlled LED drivers include I/O mapping (multiplexing), dynamic current control, and various dynamically programmable current references.

In another embodiment, a decentralized PBT system includes multiple sets of intelligent LED pads controlled from a centralized multi-channel PBT control station. An optional WiFi PBT remote control is included to facilitate local start-start and pause control. In yet another embodiment, the PBT controller includes an application running on a mobile device or smartphone that controls the intelligent LED pads. The mobile application includes intuitive UI/UX controls and biofeedback display. The app can also connect to the internet or a PBT server as a treatment database. In another embodiment, the PBT system includes a fully autonomous set of LED pads programmed over the network.

Distributed PBT systems can also be used to control LEDs attached to a mouthpiece to combat gingivitis and periodontal disease, or to drive individual LEDs attached to earbuds inserted in the nose or ear to kill bacterial infections in the sinuses. A variation of the individual LED buds can be used as "spots" to be placed on acupuncture points.

The aforementioned distributed PBT system is not limited to driving LEDs, but can be used to drive energy emitters placed adjacent to a patient to inject energy into biological tissue, such as the emission of coherent light from a laser or time-varying magnetic fields (magnetic therapy), microcurrents (electrotherapy), ultrasound energy, infrared, far-infrared electromagnetic radiation, or any combination thereof.

In one such embodiment, an LED or laser handheld wand includes a large-area head unit and pivoting handle, an integrated temperature sensor, a battery charger, a step-up (boost) voltage regulator, and an integrated safety system such as a proximity detector. In yet another embodiment, the magnetic therapy device includes a multilayer printed circuit board-mounted coil used to generate a time-varying magnetic field. The magnetic therapy device can be mounted on a pad or wand. Magnetic therapy, used to reduce inflammation and joint pain, can be performed independently or in combination with PBT.

Another handheld wand version contains a modulated voice coil that acts as a vibrator to apply pressure to muscles and tissues at ultrasonic frequencies, i.e., less than 10 Hz, similar to massage therapy but with deeper penetration. Used to reduce muscle relaxation and improve flexibility and range of motion, infrasound therapy can be used independently or in combination with PBT.

In another embodiment, an ultrasound therapy device includes a flexible PCB with one or more piezoelectric transducers modulated in the ultrasound band from 20 kHz to 4 MHz. The pads with piezoelectric transducers may also include LEDs modulated with pulses in the audio spectrum. One application of an ultrasound-LED combination device uses ultrasound to improve circulation and destroy scar tissue with PBT, which is then used to remove dead cells.

1 shows a PBT system operating under the control of a therapist. Demonstrating mitochondrial photobiomodulation. 1 shows the optical absorption spectra of various biomaterials. Contrasts the differences between photooptical therapy and photobiomodulation therapy. 1 shows photochemical stimulation of intracellular organelles, mitochondria, with blended wavelengths. 1 shows a distributed PBT system with active LED pads. FIG. 1 is a schematic diagram of a PBT system with passive LED pads using current-limiting resistors. FIG. 1 is a schematic diagram of a PBT system with passive LED pads using current control. 1 is a network description of a PBT system with active LED pads that uses only physical (PHY) Layer 1 communication. The equivalent circuit of a communication cable and its effect on electrical signals. This is a representation of the interconnection of a photobiomodulation therapy system with unqualified or inappropriate electrical accessories or LED pads. 1 illustrates a photobiomodulation therapy system in which a common set of electrical signals drives different LED pads. Shows an improper "short output" connection of one of the two LED system outputs to a common LED pad. 1 shows a shorted output connection driving a red LED string with multiple competing control signals. A shorted output connection is shown driving both NIR and red LEDs into the same LED pad simultaneously with overlapping or simultaneous control signals. 10 shows the power supply output waveforms for an output short connection driving both NIR and red LEDs to the same LED pad simultaneously with overlapping or simultaneous control signals. It is a PBT system that lacks temperature sensing, protection, or feedback. 1 shows a distributed PBT system with active LED pads. FIG. 1 is a schematic diagram of a distributed PBT system with intelligent (active) LED pads. FIG. 1 is a network diagram of a PBT system with intelligent (active) LED pads using a three-layer OSI stack. 10 is a flowchart of an LED pad authentication sequence. 1 shows a block diagram of an active LED pad with an identification data register. FIG. 1 shows a block diagram of an active LED pad with LED configuration registers. FIG. 1 is a schematic diagram of an exemplary LED array including three wavelength LEDs and movable electronics. FIG. 1 is a schematic diagram of a current sink type switched low-side LED driver with an N-channel MOSFET and a current sensing gate bias circuit with a reference current input I _ref . FIG. 1 is a schematic diagram of a current sink type switched low-side LED driver with an N-channel MOSFET and a current sensing gate bias circuit with a reference current input I _ref . FIG. 1 is a schematic diagram of an exemplary current sink type low-side switched LED driver implementation including a current mirror sensor, a transconductance amplifier bias circuit with a reference current input I _ref , and a transmission gate with a digital input. FIG. 1 is a schematic diagram of an example multi-channel current reference generator with DAC resistor current trim. FIG. 1 is a schematic diagram of an example multi-channel current reference generator with DACMOSFET gate width current trim. FIG. 1 is a schematic diagram of an exemplary multi-channel current reference generator with a DAC and an arithmetic logic unit calculation input including current calibration and target reference input currents. FIG. 1 is a schematic diagram of a high-side switch current control element or “current source” driving a string of LEDs including “m” LEDs. Schematic diagram of a current source type switched high-side LED driver with a P-channel MOSFET and a current-sensing gate bias circuit with a reference current input (-I _ref ). FIG. 1 is a schematic diagram of an exemplary current source type switched high-side LED driver implementation including a current mirror sensor, a transconductance amplifier bias circuit with a reference current input (−I _ref ), and a transmission gate with a digital input. 1 is a schematic diagram of a high-side current control element or "current source" driving a series string of LEDs including an effective low-side N-channel MOSFET digital and "m" LEDs. Schematic diagram of a current source type high-side LED driver with a P-channel MOSFET and current sensing gate bias circuit with a reference current input (-I _ref ) driving a string of LEDs in series with a low-side N-channel MOSFET digital enable. 1 is a schematic diagram of an exemplary current source type high-side LED driver implementation including a current mirror sensor, transconductance amplifier bias circuit with a reference current input (−I _ref ) driving a string of series-connected LEDs with a low-side N-channel MOSFET digital enable. 1 is a flowchart illustrating a master-slave, data streaming based LED drive. 1 shows real-time streaming data transfer to an LED pad using packet transfer over USB. 1 illustrates a just-in-time or "JIT" sequential data transfer method for a stream-based LED drive. This shows the destination shifting method for stream-based LED driving. Compare JIT to the forward-and-forward method of LED driving. 10 is a flowchart of LED pad autonomous pad playback using unencrypted files. 1 illustrates the storage of executable code files in active LED pads. An exemplary treatment protocol is shown that includes three PBT "sessions," each comprising three sequential treatment algorithms. An exemplary process is shown, each showing an LED control sequence for on and off indications and durations. 1 shows the Arndt-Schultz biphasic dose-response model of PBT. 1 shows a four-layer serial bus-based LightOS communication protocol stack. 10 shows the encrypted packet preparation of a PBT processing file. 10 shows the encrypted packet preparation of a PBT session file. 13 shows active LED pad decoding and storage of an incoming encrypted packet. 10 is a flowchart of LED pad autonomous pad playback using post-transfer file decoding. 10 shows the storage of a ciphertext file in an active LED pad. 10 is a flowchart of LED pad autonomous pad playback using on-the-fly decoding during playback. File comparison of bulk file decoding before playback and on-the-fly decoding during playback. Illustrates downloading a file from an LED player to an LED pad. 10 is a flowchart illustrating the operation of the "Waveform Synthesizer" module. 10 is a flowchart illustrating the operation of the "PWM player" module. 10 is a flowchart illustrating the operation of the "LED driver" module. FIG. 1 is a block diagram illustrating waveform generation using a waveform synthesizer, a PWM player, and an LED driver module. FIG. 2 is a block diagram showing details of the waveform synthesizer operation, including synthesis by either a unit function generator or a primitive processor. Examples of waveforms generated by identity functions are shown, including a constant, a sawtooth wave, a trigonometric function, a sine wave, and a sine-sine wave. 1 is a functional description of the synthesizer summing nodes and auto-ranging operation used in waveform synthesis. Shows examples of sine waves of various frequencies and their blended chords. A counter-based sine wave synthesis system capable of blending chords across 10 octaves with independent weighting and auto-ranging is presented. We demonstrate the synthesis of a two-sine wave code using a counter-based sine wave synthesis system. Demonstrates the synthesis of three sine wave codes employing a counter-based sine wave synthesis method. FIG. 1 is a block diagram of a counter-based sinusoidal code synthesizer using a single sine primitive with 24-point angular resolution. An example of two sinusoidal code synthesis using a single fixed resolution primitive. An example of three sine wave code synthesis using a single fixed resolution sine primitive. Shows an example sine wave and blended code using a single fixed-resolution sine primitive to highlight quantization noise. Here is an example of three sine wave code synthesis using multiple scaled resolution sine primitives. It shows an example sine wave and blended code using sine primitives of multiple scaled resolutions to completely eliminate quantization noise. Comparison of fixed resolution and scaled resolution sine wave synthesis for three sine wave mixing codes. FIG. 1 is a block diagram of a counter-based sine wave code synthesizer using scaled resolution sine primitives and four clock scale ranges. FIG. 1 is a block diagram of a universal primitive sinusoidal code synthesizer applicable to sinusoidal primitives of any resolution. Shows the UI/UX interface for setting global keys for synthesizing signs and chords based on even-toned scales and four-octave scale-based keys. It shows a global key for synthesizing signs and chords based on other scales, and a UI/UX interface for setting four octave note-based keys. 10 shows a UI/UX interface for setting global keys for sine and string synthesis based on customized frequencies. Block diagram of an algorithmic chord builder for triad/quad synthesis of musical chords (with optional +1 octave note), including major, minor, augmented triads and diminished seventh chords. Shows the UI/UX interface for a custom triad chord builder with optional +1 octave note. Shows compression of a three sine sum composite signal without autoranging. Compare with the waveform synthesized by adding three sine waves without auto-ranging amplification. FIG. 1 is a functional diagram of a PWM generator function used in a waveform synthesizer. 1 shows examples of non-sinusoidally generated waveforms and their corresponding PWM representations. This is a diagram showing the operation of the chopping function of the PWM player. Shows the rough functional equivalent of the pulse width modulator used in a PWM player. 1 shows a block diagram of the LED driver operation. The waveforms shown are for a PWM player configuration that produced a square wave with a 50% duty factor and 10mA average LED current. The waveforms shown are for a PWM player configuration that generated a square wave with a 20% duty factor and 10mA average LED current. The waveforms shown are for a PWM player configuration that produced a square wave with a 95% duty factor and 10mA average LED current. 1 shows the constituent waveforms of a square wave generated by a PWM player with a 50% duty factor and an average LED current of 10 mA, which is then stepped up to 13 mA. The waveforms shown are for a square wave configuration generated by an LED driver with a 50% duty factor and 10mA average LED current. The constituent waveforms of a sine wave generated by the LED driver ADC (analog-to-digital converter) at an average LED current of 10 mA are shown. It shows the constituent waveforms of audio samples generated by an LED driver ADC (analog-to-digital converter) plucking a guitar string with an average LED current of 10mA. Shown are waveforms of an LED driver ADC (analog-to-digital converter) configuration that generated 10mA average LED current and an audio sample of a cymbal crash. The PWM waveform is a composite sine wave that generates an average LED current of 10mA. Then we will show the waveform of the 10mA average LED current boosted to 13mA and the PWM composite sine wave configuration. The waveforms shown are for a sample PWM synthesized audio composition containing a sine wave code for a 10mA average LED current. This shows the waveform of a PWM composite triangular wave with an average LED current of 10mA. The waveforms shown are for a PWM synthesized audio example containing a guitar string being plucked at 10mA average LED current. The waveforms shown are for a sample PWM synthesized audio composition containing a cymbal crash at 10mA average LED current. The component waveforms of a PWM synthesized sine wave are shown. The average LED current is 10mA, and then stepped up to 13mA by the PWM player. Illustrates downloading of playback files to LED pads. 1 shows the LED playback data file including the playback file ID, synthesizer parameter file, primitive file, PWM player file, LED driver file, and their components. FIG. 10 is a schematic analog diagram of the firmware used to control the PWM player clock Φ _ref . It consists of a communication stack for an Ethernet-based distributed PBT system. This paper configures the communication stack for a WiFi-based distributed PBT system. FIG. 1 is a block diagram of a WiFi communication-enabled PBT controller for a distributed PBT system. This is a block diagram of a Wi-Fi communication-enabled LED pad for a distributed PBT system. A multi-user distributed PBT system and communication network. It consists of a communication stack for a mobile phone-based distributed PBT system. 1 shows a distributed PBT system using a mobile phone app and WiFi-based control. UI/UX menu for PBT control using mobile device application programs. 1A-1C are cross-sectional, top, and bottom views of a handheld PBT wand for LED or laser treatment. FIG. 1 is a block diagram of a handheld PBT wand for LED or laser treatment. 1A-1C are cross-sectional and bottom views of a PBT wand eye safety system for laser PBT utilizing capacitive touch sensing. FIG. 1 is a schematic diagram of an eye safety system for laser PBT utilizing capacitive touch sensing. FIG. 1 is a schematic diagram of a distributed systems laser PBT driver circuit. 1A and 1B show cross-sectional, top, and side views of an intelligent LED pad with an autonomous integrated switch. 1 is a flowchart illustrating a program switch sequence of an autonomous intelligent LED pad. This is a cross section of a rigid-flex PCB. FIG. 1 is an exploded view of the planar magnet used in the magnetic therapy pad. FIG. 1 is a side view of a magnetic therapy pad with planar magnets. FIG. 1 is a top view of a magnetic therapy pad with planar magnets. FIG. 1 is a schematic diagram of a distributed system magnetic therapy drive circuit. FIG. 1 is a cross-sectional view of a magnetic therapy pad using individual magnets. A magnetic therapy pad that includes an array of electromagnets. A magnetic therapy pad that includes an array of electromagnets and permanent magnets. It is a magnetic therapy pad that contains an array of electromagnets and stacked hybrid electromagnet-permanent magnets. It is a magnetic therapy pad that consists of an array of electromagnets and stacked hybrid permanent magnet electromagnets. A distributed system compatible handheld magnetic therapy device. 1A and 1B are plan and cross-sectional views of a U-shaped PBT periodontal mouthpiece. 1A-1C are side views of the manufacturing process for producing a U-shaped PBT periodontal mouthpiece. 1A-1C are side views of the manufacturing process for producing an H-shaped PBT periodontal mouthpiece. This is a side view of the manufactured H-shaped periodontal PBT mouthpiece. 1 shows the bonding process in the manufacture of an H-shaped PBT periodontal mouthpiece. 1 shows a schematic diagram of a periodontal PBT mouthpiece. 1 shows a circuit diagram of an ultrasonic PBT pad combination with an H-bridge drive. 1 shows a circuit diagram of an ultrasonic PBT pad combination with a current sink drive. FIG. 1 is a perspective view of an ultrasonic PBT pad combination.

To overcome the aforementioned limitations faced by existing-generation PBT systems, an entirely new system architecture is required. Specifically, the generation of the combined sine wave and code must occur very close to the driven LEDs to avoid significant waveform distortion due to cabling. Such design criteria require waveform synthesis to be relocated and moved from the PBT controller to the LED pad. Achieving this seemingly minor repartitioning of functionality is actually a significant design change, transforming the LED pad from a passive component into an active system or "intelligent" LED pad. While a passive LED pad contains only an array of LEDs, current sources, and switches, an intelligent LED pad must integrate a microcontroller, volatile and nonvolatile memory, a communications transceiver or interface, LED drive electronics, and an LED array. Due to the need for long cabling or wireless operation, the microcontroller's time base must also be relocated to the LED pad. Essentially, each intelligent LED pad becomes a small computer that, when instructed, can independently generate LED excitation patterns.

Therefore, rather than using a centralized PBT controller that generates and distributes electrical signals to passive LED pads, the new architecture is "distributed," consisting of a network of autonomously operating electronic components lacking centralized real-time control. The first distributed PBT system of its kind requires the invention of an intelligent LED pad—a therapeutic light delivery system in which the LED pad generates dynamic LED excitation patterns and performs all the necessary calculations to safely drive the LEDs accordingly. With distributed PBT operation, the role of the PBT controller is dramatically reduced to that of a UI/UX interface, allowing the user to select a treatment or session from an available protocol library and start, pause, or terminate treatment. This lack of central hardware control is virtually unprecedented in medical devices, as ISO 13485, IEC, and FDA regulations have always required hardware controllability for safety reasons. Therefore, implementing effective safety systems in distributed hardware medical devices requires new and innovative approaches, as safety functions must be executed locally and communicated throughout the system. Such safety protocols must be specified, designed, verified, validated, and documented in accordance with FDA design rules and international safety standards.

Another implication of a distributed PBT system with intelligent LED pads is the replacement of electrical signal communication with command-based instructions containing data packets. Such command-based communication involves the design and development of a packet-switched private communications network between components of the distributed system, adapting digital communications to meet the unique and stringent requirements of medical device control. Packet routing, security, and data payload must be designed to prevent hacking and system malfunctions and must convey all the information necessary to perform all required PBT operations.

Implementing a distributed PBT system using intelligent LED pads requires two interrelated innovations. This application discloses the operation of an intelligent LED pad, including time-based LED excitation patterns delivered via streaming or file transfer. This disclosure also considers in-pad generation of waveforms using a three-stage process of waveform synthesis, PWM player operation, and dynamic LED drive, as well as necessary safety features. Related U.S. application Ser. No. 16/377,192, entitled "Distributed Photobiomodulation Therapy Apparatus and Method, Biofeedback, and Communication Protocol Therefor," discloses a data communication hierarchy stack and control protocol.

In the distributed PBT system disclosed herein, LED regeneration can be controlled using either a time-based instruction sequence (called streaming) or by command-based waveform generation and synthesis. In either case, data packets digitally carry the LED excitation patterns in their payload. In operation, a user or therapist selects a PBT treatment or therapy session through a graphical interface and agrees to begin treatment. The commands are then packetized—that is, prepared, formatted, compressed, and packed into communication packets—and delivered to one or more intelligent LED pads via a serial peripheral communication bus, LAN, broadband connection, Wi-Fi, fiber, or other media. While the payload data carried in each data packet is digital, containing bits organized as octets or hexadecimal words, the actual communication medium is analog and includes differential analog signals, radio waves, or modulated light.

In wired communications, communication buses typically use electrical signals consisting of analog differential waveforms modulated at a specified rate, called the symbol rate or baud rate (https://en.wikipedia.org/wiki/symbol_rate). Each symbol can contain a frequency or code of a defined duration. Detection of each sequential symbol is unaffected by distortions caused by reactive parasitic elements or noise sources in the cable, thus overcoming all problems associated with digital pulse signal transmission in traditional PBT implementations. In Wi-Fi communications, incoming serial data is divided and transmitted in small packets across multiple frequency subbands, known as OFDM. This technique, orthogonal frequency division multiplexing, achieves high symbol rates and low bit error rates. Similar frequency division methods are used in Fibre Channel and DOCSIS communications to achieve high symbol rates. Because each transmitted symbol can represent multiple digital states, the data rate of the serial bus bits is higher than the symbol rate of the medium. Below is a summary of effective bit data rates (https://en.wikipedia.org/wiki/wiki/list_of_device_bit_rates) for some of the most common serial and wireless communication protocols that exceed 50MB/s:

In response to user commands, the PBT controller converts instructions into communication data packets that are then sent to all connected and authorized LED pads. The LED pads receive the commands and responses, and initiate a therapy session or perform other tasks accordingly. Due to high-bandwidth communication, the user experience of the PBT system is instantaneous. This means that the user perceives real-time UI/UX responses, even though system operations are actually performed as a series of device-to-device communications.

The disclosed distributed PBT system includes multiple interacting components, each of which performs a dedicated function within the distributed system. The number of unique components integrated into the system affects the overall complexity of the system and influences the sophistication of the communication protocol, i.e., the "language" used for device-to-device communication. Various components of the disclosed distributed PBT system may include:
A user interface consisting of a central PBT controller or mobile application used to execute UI/UX based commands and dispatch instructions over a communications network.
• Intelligent LED pads delivering dynamic photobiomodulation therapy treatments with localized in-pad excitation pattern generation and waveform synthesis, and optional integrated sensor or imaging capabilities.
A computer server, or private communications network used for accessing or maintaining over the Internet, PBT treatments, sessions, and protocols, or for distributing or uploading patient response, case study, or clinical trial data and related files (e.g., MRIs, X-rays, blood tests).
-Optional treatment accessories such as laser wands and ultrasound therapy pads.
- Optional biometric sensors used to capture and upload patient samples or real-time data (e.g., EEG sensors, ECG monitors, blood oxygen, blood pressure, blood glucose, etc.).
- Computer peripherals, including high-resolution displays and touchscreens, keyboards, mice, speakers, headphones, etc.

By combining or excluding various components of the PBT system, different performance and system costs can be tailored to suit a wide range of users covering hospitals and clinics, as well as individual users and consumers, spas, aestheticians, sports trainers, athletes, etc., as well as professional mobile applications for emergency medical, police, or military field doctors. Because PBT components use voltages above 5V, care has been taken in the disclosed design to prevent users from accidentally connecting USB peripherals to high-voltage (12V-42V) connections or buses.

LED control for distributed PBT systems

One basic implementation of a distributed PBT system, shown in Figure 13, includes three components: a PBT controller 120-component, a power supply 121, and a single intelligent LED pad 123 with an intervening USB cable 122. Figure 14 shows a block diagram of an exemplary distributed PBT system implementation, including a PBT controller and bus transceiver 131, one or more intelligent LED pads 337, a USB cable 136, and an external power "brick" 132. While the power brick 132 is shown as a separate component in the figure, in systems where the PBT controller and bus transceiver 131 use a wired connection to the intelligent LED pad 337, the power supply can be included internal to the PBT controller and transceiver rather than being a separate power supply. As shown, the PBT controller and bus transceiver 131 includes a main microcontroller μC or MPU 134, a touchscreen LCD 133, non-volatile memory 128, volatile memory 129, an interface bus 135, and a clock 124 running on a system clock 197 at a rate Φ _sys . The clock and memory elements are shown separately from the main MPU 134 to illustrate their functionality and are not intended to illustrate any particular implementation or division of components. An RTC real-time clock (not shown) may also be included in the PBT controller 131. The RTC is very low power, runs continuously, and synchronizes to an international time standard or network time whenever possible.

The main MPU 134 architecture may include a fully integrated single-chip microcontroller or microprocessor-based module, optionally including the main system clock 124, communications interface 135, and portions of nonvolatile memory 128 and volatile memory 129. Any number of partitions are possible, including multiple silicon integrated circuits (ICs), system-on-chip (SOC) integration, system-in-package (SIP), or modular use. For example, the volatile memory 129 may include dynamic random access memory (DRAM) or static random access memory (SRAM). This memory may be integrated in whole or in part within the main MPU 134 or realized by a separate integrated circuit. Similarly, the nonvolatile memory 128 may include electrically erasable programmable random access memory ( ^E2PROM ) or "flash" memory, which may be integrated in whole or in part within the MPU 111. High-capacity nonvolatile memory within the PBT controller 131. Data storage can also be achieved using mobile media storage such as optical discs (CD/DVD), by magnetic hard disk drives (HDD), or even via network connections to cloud storage.

The role of non-volatile data storage 128 in the PBT controller 131 is multipurpose, including storage for the main operating system, referred to herein as LightOS, and holding PBT treatment and session program libraries, which are typically stored in encrypted form for security reasons. Non-volatile memory 128 may also be used to capture treatment logs, upload sensor data, and possibly hold treatment metadata. In contrast to its non-volatile counterpart, the role of volatile memory 129 in the PBT controller 131 is primarily that of scratchpad memory, temporarily holding data while calculations are performed. For example, when preparing a PBT session containing a series of individual PBT processes, encrypted processing algorithms must first be decrypted, assembled into a PBT session, re-encrypted, and then assembled into communication packets ready for network transfer. The volatile memory holds the data content during the communication packet assembly process.

Another consideration with a distributed PBT system is the power distribution required to power the PBT controller and LED pads. Options include:
Uses an internal power supply to power the PBT controller, which in turn powers the LED pads via the communication bus.
Use an external power supply (brick) to power the PBT controller, which in turn powers the LED pads via the communication bus.
Use the internal power supply to power the PBT controller and a dedicated external power supply or power supply (brick) to power the LED pads.
Use an external power supply (brick) to power the PBT controller and a dedicated external power supply (brick) to power the LED pads.

In the example shown, an external power brick 132 powers the entire PBT system, providing 5V to the integrated circuits and +VLED to the LED strings. A USB cable 136 carries transceiver symbol data from the PBT controller's bus transceiver 135 and bus transceiver 131 to the communication interface 338 of the LED pads 337. The USB cable 136 also provides power. Specifically, it connects ground (GND), 5V, and +VLED to the intelligent LED pads 337. These are typically carried by low-resistance copper conductors thicker than the cable's signal lines. Each LED pad 337 is connected to a μC pad 339, a communication interface 338, a RAM volatile memory (e.g., SRAM or DRAM) 334a, a NV-RAM nonvolatile memory (e.g., EEPROM or flash) 334b, a time reference clock 333, an LED driver 335, and an LED array 140. The LED driver includes switched current sinks 140, 141, and others (not shown), typically one current sink for each string of LEDs. LED array 140 includes a series of series-connected LEDs 142a through 142m for producing light of wavelength λ ₁ , and a series of series-connected LEDs 143a through 143m for producing light of wavelength λ ₂ , and typically other series of LEDs (not shown).

The memory within the LED pad 337, including both volatile memory 334a and volatile memory 334b, is similar to the semiconductor memory used in the PBT controller 131, except for its smaller overall capacity. Because integrating fragile data storage into the LED pad 337 poses risks of mechanical shock and damage to moving media storage, the memory within the LED pad 337 must be a semiconductor solution. Specifically, the volatile memory 334a (labeled RAM) within the LED pad 337 may include dynamic random access memory (DRAM) or static random access memory (SRAM), which may be fully or partially integrated within the pad μC 339. In the LED pad, volatile memory is useful for storing data that does not need to be retained unless it is in use, such as LED streaming files, LED player files, and LED playback files. The advantage of temporarily storing only the executable code required to execute the current PBT treatment (rather than the entire treatment library) is that the memory capacity and cost within the LED pad 337 can be significantly reduced compared to that of the PBT controller 131. Another benefit is that if power is removed from LED Pad 337, all data is lost, making reverse engineering and copying the treatment program more difficult.

Nonvolatile memory 334b includes electrically erasable programmable random access memory ( ^E2PROM ) or "flash" memory, which can be fully or partially integrated within the Pad μC 339. Nonvolatile memory 334b (designated NV-RAM) is preferably used to hold firmware that does not change frequently, pad identification data, manufacturing data including an LED Pad ID register, and the operating system for the LED pad, referred to herein as LightPadOS, along with manufacturing-related LED configuration data. Nonvolatile memory 334b can also be used to maintain a user log of what treatments have been performed. Because a single PBT controller is often sold with multiple LED pads (up to six or eight per system), low-cost design of the LED pad is another important economic consideration. To lower overall memory costs, it is beneficial to centralize memory, especially nonvolatile memory, in the PBT controller, where there is only one device, and minimize the memory contained within each LED pad, which occurs in multiple instances per system.

During operation, user commands entered into the touchscreen LCD 133 of the PBT controller 131 are interpreted by the main MPU 134, which in response retrieves operation files stored in the nonvolatile memory 128 and transfers them via the USB communication interface 135 to the communication interface 338 within the intelligent LED pad 337 via a USB cable 136. Once transferred, the operation files are temporarily stored in volatile memory 338a. The pad μC 339 operates according to the LightPadOS operating system stored in nonvolatile memory 334b, which then interprets the operations stored in the RAM volatile memory 334a and controls the LED driver 335 according to the selected operation's LED excitation pattern. The array 336 illuminates a string of LEDs of various wavelengths in the desired manner. Since the PBT controller 131 and the LED pad 337 operate using their own dedicated clocks 297 and 299, the distributed PBT system operates asynchronously at two different clock frequencies, specifically Φ _sys and Φ _pad , respectively.

Because the two systems operate at different clock rates, communication between the PBT controller 131 and the LED pads 337 is asynchronous, that is, without a common synchronous clock. Asynchronous communication is compatible with a wide range of serial bus communication protocols, including USB 136 as shown, or Ethernet, WiFi, 3G/LTE, 4G, and DOCSIS-3. While a synchronous-clock version of a distributed PBT system, i.e., one with a shared clock, is technically possible, synchronous operation offers no performance or effectiveness advantages over asynchronous operation. Furthermore, distribution of a high-frequency clock over long cables is subject to issues such as clock skew, phase delay, and pulse distortion.

Figure 14, which includes a distributed PBT system with two or more microcontrollers or computer "brains," represents a fundamental architectural variation of PBT systems that otherwise typically include either an all-in-one pad with an integrated controller or an active PBT controller driving passive LED pads. Those skilled in the art should recognize that instead of being a separate hardware device, the PBT controller may alternatively include an application program running on a notebook or desktop personal computer, a computer server, a mobile device such as a tablet or smartphone, or other host device capable of running computer software such as a video game console, and/or an IoT device. Examples of such alternative embodiments are provided throughout this application.

As shown in Figure 15, PBT operations can be interpreted as a series of communications used to control hardware operations. Using the Open Systems Implementation or OSI representation, the PBT controller 120 includes a communications stack 147, including an application layer 7, a data link layer 2, and a physical layer 1. Within the PBT controller 120, the application layer 7 is implemented using an operating system customized for photobiomodulation, referred to herein as LightOS. LightOS user-supplied commands are passed to the data link layer at layer 2, which, along with PHY layer 1, communicates using USB differential signaling 332 to the corresponding PHY layer 1 of the communications stack 148 within the intelligent LED pad 123 using the USB protocol. Thus, while electrical signals constitute Layer 1 communications, the USB data structure operates as if the PBT controller and intelligent LED pad are communicating at Layer 2 with packets arranged in time as USB data "frames." When communications stack 148 receives a USB packet, the information is transferred up to application layer 7, which is executed by the LED pad's resident operating system, referred to herein as LightPadOS. If the PBT controller's LightOS and the intelligent LED pad's operating system, LightPadOS, are designed to communicate and execute instructions in a self-consistent manner, the bidirectional link between communications stacks 147 and 148 acts as a virtual machine at the application layer. The devices behave as if they were a single piece of hardware.

It is important that components can exchange information and execute instructions at a high level of abstraction, and that at some application layer, the two operating systems, LightOS and LightPadOS, are developed in parallel, using the same encryption and security methods and protocols at any layer. These criteria include adopting a common shared secret, executing a predefined verification sequence (required for components to join the system's private network), and running a common encryption algorithm.

Before the two components can begin communicating and perform tasks, the PBT controller must first establish whether the LED pad is, in fact, a manufacturer-approved, system-verified component. This test, called "authentication," is illustrated in the flowchart of Figure 16. It is shown in two parallel sequences: one occurring within LightOS, acting as the "host," and the other occurring within LightPadOS, acting as the "client." As shown, once the physical USB connection is established, i.e., insertion 150, the controller's LightOS operating system initiates subroutine 151a called "LightPad Placement," while the LED pad's LightPadOS operating system simultaneously initiates subroutine 151b. In the first step 152a, used to determine whether the client is a power source (and rejecting it if it is), the PBT controller performs check 158 to determine whether the USB D+ and D- pins are shorted. If these data pins are shorted, according to the USB standard, the peripheral is a power source, not an LED pad, and the system will reject the connection, terminate authentication, and notify the user that the LightPadOS peripheral is not a valid component. Unplug it immediately. If the pins are not shorted, the LightPadOS installation and authorization process may continue.

In steps 153a and 153b, the two devices negotiate the maximum data rate that each understands and can reliably communicate at. Once the communication data rate is established, the symmetric authentication process 154a and 154b begins. During symmetric authentication, in step 154a, LightOS first queries LightPadOS to determine whether the LED Pad 123 is a valid manufacturer-approved device by checking the data stored in the LED Pad ID data register 144. In the mirror authentication process of step 154b, the LED Pad 123 verifies that the PBT Controller is a valid device with a valid manufacturer ID approved for use with the LED Pad 123. This exchange involves the exchange of certain encrypted security credentials and manufacturer identification data, such as serial number, manufacturing code, and GUDID number, ensuring that both the PBT Controller 120 and the intelligent LED Pad 123 are from (or are licensed as approved by) the same manufacturer. If authentication fails, the host LightOS notifies the user that the LED Pad is not authorized for use with the system and instructs them to remove the LED Pad. If LightOS cannot authenticate the LED Pad 123, the PBT Controller 120 stops communicating with the peripheral. Conversely, if the peripheral LightPad OS cannot determine the authenticity of the PBT Controller 120, the LED Pad 123 ignores the PBT Controller 120's commands. Operation can only continue if symmetric authentication is confirmed.

Any number of authentication methods can be implemented to establish a private network and authorize device connection to it. These methods include symmetric or asymmetric encryption and key exchange, employing "certificate authority"-based identity verification through the exchange of digital CA certificates, or exchanging cryptographic hash data to verify that devices possess the same shared secret—i.e., were manufactured by a certified manufacturer. For example, a cryptographically hidden numeric code—i.e., a shared secret—installed in both the PBT controller and the intelligent LED pad can be used to verify the authenticity of an intelligent LED pad connected to a network without revealing the key itself. In one such method of LED pad verification, performed at Data Link Layer 2, the PBT controller passes a random number to the intelligent LED pad over the network or communications bus. In response, the microcontroller within the LED pad decrypts its copy of the shared secret (the numeric code), merges it with the received random number, and then performs a cryptographic hash operation on the concatenated number. The intelligent LED pad then openly returns the cryptographic hash value over the same transceiver link.

At the same time, the PBT controller performs an identical operation to decrypt its own copy of the shared secret (a numerical code), merge it with the generated random number sent to the LED pad, and then perform a cryptographic hash operation on the concatenated number. The PBT controller then compares the received hash value with the locally generated hash value. If the two numbers match, the pad is authentic, i.e., it is "authorized" to connect to the network. The aforementioned authentication algorithm can be performed over any PHY Layer 1 and/or Data Link 2 connection over any data bus or packet-switched network, including USB, Ethernet, WiFi, or cellular wireless connections. In the case of a WiFi connection, the data link can also be established using the WiFi Protected Access Protocol (WPA2).

For "management" purposes and security tracking, the authentication date and time (and GPS location, if available) of authenticated components are stored in non-volatile memory and optionally uploaded to a server. The benefits of employing secure communication and AAA (Authentication, Authorization, and Administration) verification of all connected components in a distributed PBT system are critical to ensure safety and protection from the intentional connection of unauthorized and potentially insecure fraudster devices. In this way, fraudster devices cannot be powered by the distributed PBT system. AAA verification also protects against the accidental connection of devices not intended to operate as part of the PBT system, such as lithium-ion battery packs, unauthorized power supplies, speakers, disk drives, motor drivers, and high-power Class III and Class IV lasers, which pose potential hazards unrelated to the PBT system.

Security for distributed PBT systems using packet-switched networks (such as Ethernet or Wi-Fi) can also be enhanced using dynamic addressing at Network Layer 3 and dynamic port assignment at Data Transport Layer 4. When the PBT Controller operates without being connected to the Internet or other local area network, it generates a dynamic IP address and a dynamic port address and broadcasts the address to other network-connected devices, to which the intelligent LED pad responds with its own dynamic IP address and its own dynamic port address. When the distributed PBT system is connected to a router or the Internet, it assigns dynamic IP addresses using Dynamic Host Configuration Processor (DHCP). Similarly, it uses Remote Procedure Call (RPC) to perform dynamic port assignment. The dynamic IP address and dynamic port change each time the device connects to the network, reducing the cyber attack surface. Additional Layer 4 security can be added using TLS Transport Layer Security, IPSec security protocol, or other protocols. When the intelligent LED pad is connected to a network, additional information, such as LED configuration data, can be exchanged to allow the component to operate as part of the distributed PBT system.

In step 155a, LightOS requests information about the LED configuration of the LED pad. In step 155b, LightPadOS responds by relaying the information in the configuration register 145 of the LED pad 123 to the PBT controller 120. The configuration file not only contains a detailed description of the LED array, but also specifies the manufacturer's specifications for the maximum, minimum, and target voltages required to power the LED strings in the array. The configuration file also specifies the minimum current required to drive the LEDs. If multiple pads are connected to an output, the LightOS solicitation receives the same information from all connected LED pads, meaning it analyzes the entire network of connected devices.

In step 156a, LightOS checks the voltage requirement of each pad and compares it to the output voltage range of the high-voltage power supply. For PBT controllers using high-voltage power supplies capable of a fixed output voltage +V, the LightOS operating system verifies that this voltage is within each LED pad's specified voltage range, from _Vmin to _Vmax . The system also verifies that the total current required for all "n" LED strings does not exceed the power supply's rated current (this is generally not an issue, but the current check is included to support power-limited, low-cost consumer PBT device designs).

In step 156a, if the power supply output meets the operating range of all connected LED pads, i.e., _Vmin ≤ + _VLED ≤ _Vmax , the PBT controller 120 enables the high-voltage supply +VLED. Optionally, in step 156b, the PBT controller 120 notifies the LED pad 123 of the selected supply voltage, which is stored in non-volatile memory 334b, allowing for final documentation of the last power supply voltage applied to the LED pad (useful for quality issues and inspection). If the PBT controller 120 employs a programmable voltage supply, the LightOS operating system selects the optimal voltage based on the LED pad 123's operating _Vtarget , which is stored in the pad's LED configuration register 145. If the target voltages do not match, the LightOS operating system selects a voltage for +VLED as a compromise between the various reported target voltages. The term "high voltage" in this context means a voltage between a minimum of 19.5V and a maximum of 42V. Common power supply voltages include 20V, 24V, or 36V. Even after +VLED is enabled, this high voltage is not connected to the output socket or applied to the LED pads until a therapy is selected and therapy begins.

During the authentication process and in case of user inquiries, the PBT controller 120 may need to request information regarding the manufacture of the LED pad. This data can be useful for complying with medical device regulations regarding traceability, debugging quality or field failures, or processing return material authorizations (RMAs). Figure 17 shows an example of the type of product manufacturing information contained in the "LED Pad Identification Data Register" 144 stored in the LED pad's non-volatile memory 334b. This data may include the manufacturer's part number, the manufacturer's name, the unit's serial number, a manufacturing code linked to a specific description of the manufacturing history or pedigree, and the USFDA-designated Global Unique Device Identification Database (GUDID) number. This data may include the manufacturer's part number, the manufacturer's name, the unit's serial number, a manufacturing code linked to a specific description of the manufacturing history or pedigree, and the USFDA-designated Global Unique Device Identification Database (GUDID) number [https://accessgudid.nlm.nih. gov/about-gudid], and associated 510(k) numbers, if applicable. The register may also optionally contain a country-specific code for importing the device and other customs-related information, such as an export license number or free trade certificate. This register is stored in non-volatile memory 334b during manufacturing. LED pad identification data register 144 also contains security credentials (e.g., encryption keys) used in the authentication process. The security credentials may be static, remaining installed at manufacturing, dynamically rewritten each time the LED pad is authenticated, or rewritten after a specified number of valid authentications.

As described, during the authentication process, the PBT controller 120 collects information regarding the LED configuration of all connected LED pads. As shown in FIG. 18, the pad's LED configuration information is stored in the LED pad's non-volatile memory 334b in an "LED configuration register" 145, written during the pad's manufacturing process. The register contains a number of LED strings "n," including specific information about the LEDs in the string, including the wavelength λ of the LEDs and the number of series-connected LEDs in each string, "m." During operation, this LED string information is used to match LED processing to a specific type of LED pad. For example, processing designed specifically to drive red LEDs will not function if an LED pad containing blue or green LEDs is installed. The user's UI/UX, i.e., menu selections on the PBT controller's touchscreen, are adjusted according to the LED pads connected to the system. If the corresponding LED pads are not installed, menu selections requiring that type of pad will be hidden or grayed out.

The LED configuration register 145 is essentially a tabular description of the circuit diagram of the LED pad. Referring to the schematic diagram of Figure 19, which shows a portion of an LED pad including an LED driver 335 with an LED controller circuit 160 and current sinks 161a-161c,
String #1 LED configuration register 145 describes string 162a including six series-connected near-infrared LEDs of wavelength λ ₁ =810 nm driven by current sink 161a carrying current I _LED1 .
String #2 LED configuration register 145 describes string 163a including four series-connected red LEDs of wavelength λ ₂ =635 nm driven by current sink 161b carrying a current I _LED2 .
String #3 LED configuration register 145 describes string 164a including four series-connected blue LEDs of wavelength λ ₃ =450 nm driven by current sink 161c carrying a current I _LED2 .
String #4 The LED configuration register 145 describes a string 164b including six series-connected near-infrared LEDs of wavelength λ ₁ =810 nm driven by a current sink 161a carrying a current I _LED4 =I _LED1 .
String #5 The LED configuration register 145 describes a string 164b including four series-connected red LEDs of wavelength λ ₂ =635 nm driven by a current sink 161b carrying a current I _LED5 =I _LED2 .
String #6 The LED configuration register 145 describes a string 164b including four series-connected blue LEDs of wavelength λ ₃ =450 nm driven by a current sink 161b carrying a current I _LED6 =I _LED3 .

The above is not intended to represent a specific design, but rather to illustrate, but not limit, the data format of the LED configuration register 145 and its corresponding equivalent circuit diagram. In particular, the number of LED strings "n" and the number of LEDs connected in series to a particular string "m" contained within an LED pad will likely exceed the numbers shown in this example. In practice, the number of LEDs in various strings may be the same or may vary from string to string. For example, an LED pad may contain 15 strings of 14 LEDs in series, or 210 LEDs. These LEDs can be arranged in three groups of five LED strings each: one-third NIR, one-third red, and one-third blue. Each LED type can be configured with five parallel strings and 14 series-connected LEDs, i.e., three 14s5p arrays.

The LED configuration register 18 also contains the minimum and maximum operating voltages for the LED pad. For proper LED operation, the power supply voltage +VLED must exceed the LED pad's minimum voltage specification, _Vmin , to ensure uniform illumination, but the power supply voltage must not exceed the specified maximum voltage, _Vmin , to avoid damage from excessive voltage or heat. In other words, the acceptable supply voltage value for powering the LED pad must satisfy the criterion _Vmin < + _VLED ≤ _Vmin . The manufacturer-specified _Vmin value stored in the LED configuration register 145 must statistically exceed the highest voltage string of LEDs in the LED pad, ensuring that the highest voltage string of the pad remains fully illuminated during operation as long as the Vmin < +VLED criterion is maintained. Specifying a _Vmin voltage that is too low may result in individual LED strings in some LED pads being dimmer than others during treatment. Poor brightness uniformity limits the peak and average power of PBT treatment, reducing the total energy (dose) of the treatment and negatively impacting treatment efficacy.

The highest voltage string in an LED pad is determined by both design and stochastic voltage variations in LED manufacturing. Each LED string consists of m series-connected LEDs, each with its own forward conduction voltage, _Vfx , where x varies from 1 to m, and the total string voltage is the sum of these individual LED voltages, _ΣVfx . The highest voltage may occur in a string with a small number of higher-voltage series-connected LEDs, or in a string with a large number of lower-forward-voltage LEDs. LED pad manufacturers must use statistical sampling data of LED forward voltages for each lot to ensure that LED string voltages are not manufactured to exceed the specified _Vmin value.

Although less accurate, the power supply must be able to provide the minimum average current, I _min , required to light all LEDs of a particular color (wavelength) at once. Generally, in a two-wavelength LED pad, 50% of the n LED strings may be conducting simultaneously. In a three-color LED pad, only one of the three LED wavelengths may be on at a time to avoid overheating, but the worst-case assumption of 2/3 or 67% of the n strings can be used to calculate the maximum current. In continuous operation, the peak current of the conducting LEDs will not exceed 30 mA per string in the worst case, i.e., I _LED ≤ 30 mA. Using this worst-case assumption, a pad with n = 30, 2/3 of the strings on at once, and I _LED ≤ 30 mA would require a value of I _min = 30(2/3) * (30 mA) = 600 mA.

The I _min value specified in LED configuration register 145 does not describe the maximum current that can flow through the LED, but rather the maximum safe current at a 50% duty factor for the pad's conductive trace. This current is the current flowing through the LED pad itself plus any current bused through the LED pad to another LED. This specification is included to prevent operation of the pad that could cause a significant voltage drop in the LED pad's power line, resulting in overheating, malfunction, electromigration, or metal fusion. One possible design guideline for the LED pad's printed circuit board (PCB) is to utilize copper conductors capable of carrying at least twice the rated current. This means that the pad can safely carry its own current and the current of another LED simultaneously. An additional design guard band of δ = 25% is included as a safety margin. For example, if I _min = 600 mA, using a 25% guard band results in I _min = 2I _min (1 + δ) = 1,500 mA. Configuration register 145 also contains the mirror ratio α used to convert the reference current I _ref to the LED string current I _LED (and vice versa) according to the relationship I _LED = αI _ref . If different ratios are used for different channels, the table can be modified accordingly to include α ₁ , α ₂ , α ₃ , etc., so that I _LED1 = α ₁ I _ref1 , I _LED2 = α ₂ I _ref2 , etc.

Referring again to FIG. 19, the current I _LED1 in each NIR LED string is controlled by a dedicated series-connected current sink 161a and conducts an on-state current proportional to I _ref1 . The current I _LED2 in each red LED string is controlled by a dedicated series-connected current sink 161b and conducts an on-state current proportional to I _ref2 . The current I _LED3 in each blue LED string is controlled by a dedicated series-connected current sink 161c and conducts an on-state current proportional to I _ref3 . The current control device connected in series with each LED string can be connected either as a current “sink” on the cathode side (as shown in FIG. 20A) or as a current “source” on the anode side of the LED string (as shown in FIG. 22A). In both the current sink 161a and current source 200a implementations, the current I _LED flowing through the current control device and LED string 165 or 201, respectively, is controlled by an analog reference current I _ref and a digital enable pulse En. The origin of these two signals in a distributed PBT system will be discussed later in this application. (Note: The terms "current source" and "current sink" are well known in the art to refer to components that provide or receive ("sink") a current whose magnitude is relatively unaffected by the magnitude of the voltage across the component.)

FIG. 20B shows a block diagram of an idealized current sink 161a, showing current sense and control element 166 driving the gate of N-channel MOSFET 167. The MOSFET (or bipolar junction transistor) maintains a controlled current while maintaining a voltage across its drain-source terminals. Gate bias is provided by current sense and control element 166 to maintain a constant current despite variations in the drain-source voltage. FIG. 20C shows one implementation of the described low current sink, in which N-channel current mirror MOSFETs 168a and 168b sense the current I _LED . The ratio β of the gate width of MOSFET 168b to the gate width of MOSFET 168a is less than 1, meaning that the current in current mirror MOSFET 168b is a small, but significant, fraction of the load current in current mirror MOSFET 168a (I _LED ). This measured current is reflected by a unity current mirror, consisting of P-channel MOSFETs 169a and 169b with matched gate widths Wp, converting the sense current from a ground-referenced current to a 5V supply reference current with a large βI _LED . A differential "error" signal, ΔI _err , containing the difference between I _ref and βI _LED is then amplified by transconductance amplifier 170, converted proportionally to voltage VG, and fed to the gate of the current control element, i.e., MOSFET 167, forming a closed-loop feedback path. During operation, the transconductance gain Gm creates a gate bias, V, that nulls the error signal ΔI _err and forces I _ref = βI _LED . For convenience, we redefine β = 1/α, which allows us to express the current source transfer function as I _LED = αI _ref . The same reference current is distributed to all LED strings in the same LED pad, ensuring uniform brightness across all LEDs.

An analog transmission gate, including digital inverter 171, P-channel MOSFET 172, and grounded N-channel MOSFET 173, controls the gate of N-channel current sink MOSFET 167, which performs the function of the digital E n input. Specifically, when enable signal E n is high, the output of inverter 171 is grounded, turning on P-channel MOSFET transmission gate 172 and turning off N-channel MOSFET 173. Because the P-channel has its gate grounded, it is biased fully on, i.e., in its linear region, and acts like a resistor, passing the analog voltage V g from the output of transconductance amplifier 170 to the gate of N-channel current sink 167. Conversely, when enable signal E n is low (digital 0), the output of inverter 171, connected to P-channel transmission gate MOSFET 172, is biased to 5 V, turning off the P-channel, disconnecting the gate of N-channel current sink MOSFET 167 from the output of transconductance amplifier 170. At the same time, N-channel MOSFET 172 turns on, pulling the gate of current sink MOSFET 167 to ground and turning current sink MOSFET 167 off. That is, I _LED = 0. In conclusion, the circuit in Figure 20C represents one circuit for implementing a switch-controlled current sink. When the current sink is enabled (En digital = 1), the current sink conducts and passes a controlled current, I _LED = αI _ref . When the current sink is disabled (En digital = 0), the current sink is off and I _LED = 0.

In a similar manner, current source 200a of FIG. 22A can be realized by using a P-channel current mirror MOSFET to provide a controlled current to the anode of LED string 201 from a +5V supply. FIG. 22B shows a block diagram of this ideal current source 200a, showing a P-channel MOSFET 203 (or bipolar junction transistor) gate drive current sense and control element 202 that maintains a controlled current while maintaining a constant voltage across its drain-to-source terminals. Gate bias is provided by current sense and control element 202 to maintain a constant current despite variations in drain-to-source voltage.

FIG. 22C shows one implementation of the described constant current source, in which P-channel current mirror MOSFETs 204a and 204b sense the load current I _LED . The ratio of the gate width of MOSFET 204b to the gate width of MOSFET 204a is β, where β<1 means that the current in mirror MOSFET 204b is a small fraction, but an accurate fraction, of the LED load current. This measured current, representing a +V LED high-voltage supply reference current of magnitude βI _LED , is then input to a differential transconductance amplifier 206 and compared with a reference current I _ref , which is also mirrored to the +V LED high-voltage supply rail. A differential “error” signal ΔI _err , containing the difference between I _ref and βI _LED , is then amplified by transconductance amplifier 206, proportionally converted to a voltage −VG, and fed to the gate of the current control element, P-channel current source MOSFET 203, forming a closed-loop feedback path. In operation, the gain Gm of the transconductance amplifier 206 provides a gate bias −VG that drives its error signal _ΔIerr to zero, thereby forcing _Iref = _βILED . For convenience, we redefine β = 1/α, which allows us to express the current source transfer function as _ILED = _αIref . The same reference current is distributed to all LED strings in the same LED pad, ensuring uniform brightness across all LEDs.

In the switched current source implementation shown, digital inverters 211a and 211b, as well as an analog transmission gate including P-channel MOSFET 207 and +VLED-connected P-channel MOSFET 208, perform the digital enable function of the En input and control the gate of P-channel current source MOSFET 203. Specifically, when enable signal En is high, the output of inverter 211a is at ground and the output of inverter 211b is at 5 V, turning on high-voltage level-shifting N-channel MOSFET 210a and turning off high-voltage level-shifting N-channel MOSFET 210b. When high-voltage level-shifting N-channel MOSFET 210a is on, current is conducted through resistor 209a, pulling the gate of P-channel MOSFET transmission gate 207 to a voltage near ground and turning the transistor on. Because the gate of P-channel MOSFET 207 is biased near ground, the device operates in its linear region. That is, it is fully on, acting like a resistor and passing the analog voltage -VG from the output of transconductance amplifier 206 to the gate of P-channel current-source MOSFET 203. At the same time, subsequent high-voltage level-shift N-channel MOSFET 210b is off, no current flows through resistor 209b, and the voltage at the gate of P-channel pull-up MOSFET 208, connected to its source, is +VLED, and the transistor is off. Therefore, whenever P-channel current-source MOSFET 203 is on, P-channel pull-up MOSFET 208 is off and has no effect on the gate voltage of P-channel MOSFET current source 203.

Conversely, when enable signal En is low (digital 0), the output of inverter 211b is biased to ground, turning off high-voltage level-shifting N-channel MOSFET 210a. Because high-voltage level-shifting N-channel MOSFET 210a is off, no current flows through resistor 209a, and the voltage at the gate of P-channel transmission gate MOSFET 207 is biased to +VLED, turning the P-channel transmission gate off. This disconnects MOSFET 207 and the output of transconductance amplifier 205 from the gate of P-channel current source 203. Simultaneously, N-channel MOSFET 210b turns on, conducting current through resistor 209b and pulling the gate of P-channel pullup MOSFET 208 down near ground, turning MOSFET 208 on. With P-channel pullup MOSFET 208 on, the gate of P-channel current source 203 is biased to +VLED, thereby biasing the current source off and _ILED = 0. In conclusion, the circuit of Figure 22C represents one circuit for implementing a switch-controlled current source. When the current sink is enabled (En digital = 1), it conducts and draws a controlled current I _LED = αI _ref . When the current sink is disabled (En digital = 0), it is off and I _LED = 0.

Note that the current sink circuit implementation of Figure 20C is inherently a low-voltage circuit. The only component that requires specifications capable of withstanding the high-voltage LED supply +V _LED is N-channel current sink MOSFET 167. This requires MOSFETs with high off-state drain-to-source blocking capability, which is not the case with the current source circuit of Figure 22C, particularly P-channel current source MOSFET 203, which must conduct a controlled current while simultaneously sustaining a high voltage. It exhibits a wide safe operating region without secondary failure (snapback) or hot carrier reliability concerns. Of particular concern is the maximum gate-to-source voltage rating, or V _GSp (max), of P-channel MOSFETs 207 and 208. To avoid gate oxide damage to these devices, the values of resistors 209a and 209b must be carefully selected to avoid generating on-state gate drive that exceeds the V _GSp (max) of the devices. As a precaution, Zener diodes may be included across the gate to source terminals of MOSFETs 207 and 208, respectively, to clamp the maximum gate bias to a safe level. In some integrated circuit processes, fabricated high voltage P-channel transistors may optionally utilize thicker "high voltage" gates, but this option varies depending on the wafer foundry used to fabricate the IC.

FIG. 23A shows another way to achieve a switched current source. In this case, the analog current control circuitry is separated from the digital enable function, whereby LED string 201 is connected in series between controlled current source 200a and grounded N-channel enable MOSFET 212. A block diagram of this circuit, shown in FIG. 23B, illustrates that an ideal current source realization includes current sense and control circuitry 202 and high-voltage P-channel current source MOSFET 203. The circuit implementation of a "low-side switched" current source is significantly simpler than that of the fully integrated switched current source of FIG. 22C. Current sensing remains unchanged; in this embodiment, amplifier 206 uses P-channel MOSFETs 204a and 204b, a current sense mirror including the differential input transconductance of P-channel MOSFETs 205a and 205b, and a current reference mirror. All high-voltage level shifting, transmission gate, and gate pull-up circuitry has been completely eliminated and replaced with a single grounded N-channel MOSFET 212 driven by low-voltage gate drive inverters 221a and 221b.

In both the high-voltage current source circuits of Figures 22C and 23C, the required reference current is a ground-referenced current sink current, −I _ref . Because most currents are referenced to the source current rather than the sink, a source-to-sink current mirror is required. This mirror is represented by threshold-connected N-channel MOSFET 213a with the current reference input, I _ref , mirrored by N-channel MOSFET 213b, which generates the current-sink reference current, −I _ref , to power +VLED-referenced P-channel current mirror MOSFET 205b. Note that this is the inverse of the circuit shown in Figure 23C, using a high-voltage P-channel MOSFET and level shift circuit for the enable function and a grounded current sink for current control. However, high-side switched current sinks generally offer no special advantages over the fully integrated switched current sink shown in Figure 20C and are therefore not discussed in this application.

In all of the circuits described above, LED current control relies on a common reference current. To achieve the precision necessary to control the LED brightness, the reference current I _ref requires active trimming during manufacturing. One method of trimming the reference current using a resistor is shown in Figure 21A. The reference current I _ref0 is determined by a threshold-connected p-channel MOSFET 180a in series with resistor 181. Threshold-connected refers to a MOSFET with its gate connected to its drain to create a two-terminal device where VGS = VDS. The term "threshold" is used to describe the voltage at which a rapid increase in drain current occurs near the device's threshold voltage, _Vtp , i.e., VGS = VDS ^~ Vt (~ denotes an approximation; the same applies below). Therefore, the current of P-channel MOSFET 180a is approximately I _ref0 ^~ (5V - _Vtp )/ _R0 . This reference current is mirrored by shared gate connections to other MOSFETs 180b-180e of the same structure and gate width to generate multiple matched reference currents _Iref1 , _Iref2 , _Iref3 , _Iref4 , etc. Mismatches in gate width W, such as _Wp0 = _Wp1 = _Wp2 = _Wp3 = _Wp4 , are an insignificant source of variability compared to the variability in the comparative resistance _R0 of integrated circuit resistor 181. To electronically trim the circuit to correct for manufacturing variations, _Iref resistor trim circuit 182 includes an array of switched resistors 185a, 185b...185n with corresponding resistances _R1 , _R2 ... _Rn , which may be electrically connected (or not) to resistor 181 depending on whether N-channel MOSFETs 184a, 184b...184n are selected. Each transistor is biased into a conducting state by gate drivers 185a, 185b,...185n. For each activated transistor, a corresponding resistor is placed in parallel with resistor 181, decreasing the effective resistance _R0 and increasing the magnitude of current _Iref0 . This trimming method trims the resistance in one direction and increases the current; that is, the initial values are maximum resistance and minimum current. During manufacturing, the LED current is measured, and the combination of trim MOSFETs that are turned on and off is adjusted by changing digital value calibration register 186 until the target current is reached, the contents of which are written to nonvolatile memory. While this method of accounting for switched parallel resistors represents one resistor trimming method, another method involves series-connected resistors shorted out by conducting MOSFETs. In this series trimming method, the resistance value starts at its maximum with all MOSFETs off and at minimum current, and as the trim progresses and MOSFETs are turned on, more resistance is shorted out.

FIG. 21B illustrates another trimming method using gates of scaled-width MOSFETs. As shown in the resistive reference circuit of FIG. 21A, in this reference circuit, a reference current I _ref0 conducted by threshold-connected P-channel MOSFET 180a is mirrored to multiple outputs via identically sized MOSFETs 180b through 180e. Unlike the previous case, however, a bandgap reference circuit 190 with output V _bandgap generates the reference current. The bandgap voltage is converted to a current by a series resistor and mirrored by threshold-connected current-mirror N-channel MOSFET 192a with gate width Wn, which mirrors MOSFET 192b with gate width γWn to generate the reference current I _ref0 . The temperature-dependent output voltage _Vbandgap (T) of bandgap voltage reference 190 can be designed to largely offset the temperature variation of resistor 191, so that γ[ _Vbandgap (T)/ _R0 (T)] = _Iref0 , and _Iref0 remains constant over temperature. Trimming occurs by varying the effective gate width of P-channel MOSFET 180a by paralleling any number of threshold-connected MOSFETs _193a , _193b , ... 193n. P-channel MOSFETs 194a, 194b, ... _194n , each with a gate width W, are digitally switched on and off by digital inverters 195a, 195b, ... 195n. For example, when MOSFET 194b is turned on by inverter 195b, MOSFET 193b is essentially in parallel with P-channel MOSFET 180a, and the gate width of the current mirror changes from Wp0 to a larger value ( _Wp0 + _Wpx2 ). The larger gate width of the threshold-connected MOSFET pair means that less voltage is required to drive the same reference current, resulting in a smaller output reference current. In other words, the current mirror ratio between _Iref0 and _Iref3 , for example, changes from [ _Wp3 / _Wp0 ] to a smaller ratio [ _Wp3 /( _Wp0 + _Wpx2 )]. This means that the output current decreases due to active trimming. Therefore, the trim is unidirectional, starting with a maximum output current when the trim MOSFET is off and decreasing as more transistors are connected in parallel. In production, the LED current is measured and the contents of the adjustment register 186 are written non-volatilely and the combination in which the trim MOSFETs are turned on and off is adjusted by changing the digital value calibration register 186 until the target current is reached.

To dynamically vary the reference current and thereby the LED current, the value of the reference current can be digitally altered by overriding the calibration register 186 to adjust or regulate the LED brightness. However, doing so is disadvantageous because the precision achieved by the calibration reference trim during manufacturing is lost. This problem is overcome by the dynamically programmable reference circuit of FIG. 21C, which includes two reference current registers—the aforementioned I _ref calibration register 186, and a separate dynamic target reference current register 199a specific to a particular PBT treatment. While the dynamic target reference current 199a changes over time, the calibration table does not. In this regard, the data in the calibration table 186 can be considered a fixed offset to the data in the dynamic target reference current register 199a. The two registers are easily combined using a simple subtraction performed by the arithmetic logic unit (ALU) 198 to generate the compensated dynamic drive current register, specifically, the "I _ref input word" 199b. This digital word is used to drive digital-to-analog (D/A) converter 197, which outputs an analog voltage as a function of the digital input. Precision ranges from 8-bit to 24-bit resolution, but 16-bit DACs commonly available in many microcontrollers produce 1024 combinations, sufficient resolution for the required waveform synthesis. As shown, the D/A converter's output voltage, V _DAC , is converted to a current by resistor 191 and mirrored by N-channel MOSFETs 192a and 192b to generate reference current I _ref1 , where I _ref1 ^∼ β[(V _DAC -V _tn )/R ₀ ]. This reference current is mirrored by threshold-connected P-channel MOSFET 180a and matching MOSFETs 180b, 180c, 180d, and 180e to generate corresponding current reference outputs I _ref1 , I _ref2 , I ref3 , I _{ref4 ,} etc. D/A converter 197 may also include a current-output D/A converter that produces an analog current instead of a voltage. In such cases, the value of resistor 191 is not critical and may even be eliminated.

Once the components of the distributed PBT system have been established through Layer 2 authentication, Layer 3 and Layer 4 network and port address assignment, and LED pad configuration data has been exchanged, the distributed PBT system is ready to perform processing. When the PBT Controller receives the user's "Start" command, PBT processing begins by establishing a Layer 5 session between the PBT Controller and the networked intelligent LED Pads through the exchange of encryption keys or digital certificates. Once the session is opened, the PBT Controller and the intelligent LED Pads maintain a secure link during the exchange of files and commands until treatment is complete or terminated. Additional network security can be achieved using encryption at Presentation Layer 6 or Application Layer 7. PBT processing execution begins using either the data streaming or file playback methods described below.

Data streaming in a distributed PBT system

By incorporating all LED drive circuitry as shown in Figure 18, the PBT controller in a distributed PBT system does not need to be concerned with how the pads select specific LED strings, how they control LED current, or how they are used to pulse or modulate LED conduction. Instead, the PBT controller performs user interface tasks and prepares drive instructions for the selected treatment. These drive instructions can be transferred from the PBT controller to the LED pads in two ways. In one method, software called an LED player is first installed on the pad and later used to interpret and execute the treatment. Then, an instruction set called a play file is transferred, instructing the LED player's executable code on what to do. Another approach is for the PBT to send a streaming file.

In master-slave data streaming, a series of LED instructions are sent sequentially, instructing the LEDs when to turn on and off. Similar to audio streaming files, data transfer from the PBT Controller to the intelligent LED Pad must occur before a specific step can be executed. The sequentially sent incoming instruction packets must stay ahead of the therapy execution, or the therapy will stall due to lack of instructions. This process, illustrated in the flowchart of FIG. 24, shows LightOS operations occurring on the PBT Controller host and LightPadOS operations occurring in parallel on the intelligent LED Pad client. Specifically, after selecting a therapy session 250, both the controller and pad operating systems begin execution 251a and 251b of the selected session 250. Next, in step 252a and at time _t1 , LightOS transfers the first therapy segment to the LED Pad, after which, in step 252b, LightOS executes the first therapy segment. In step 253a and at time _t2 , LightOS transfers the second treatment segment to the LED pad, after which in step 253b the LightPad OS performs the second treatment segment. In step 254a and at time _t3 , LightOS transfers the third treatment segment to the LED pad, after which in step 254b the LightPad OS performs the third treatment segment. Finally, in step 256a, at time tn, LightOS transfers the nth treatment segment to the LED pad, after which in step 256b the LightPad OS performs the nth treatment segment, after which both sessions 257a and 257b end.

An example of USB data packet transfer and command execution during master-slave streaming is shown in Figure 25. Preparation of treatment command 260a occurs while the red LED is off and begins with LED command 261, represented by hexadecimal code representing a sample "turn on LED" command. Command 261 is then embedded as the payload into a USB packet, combining the payload, command 261, with header 262. Next, in step 263, the packet is sent from the PBT controller to the LED pad. Command 261 is then extracted and decoded into bits 264 describing which LEDs are turned on and which are not. The bits are then loaded into LED register 265, which is executed at time 266 when the red LED current changes from off to on, starting a timer to prepare and load the next command to turn all LEDs off. The red LED switching is indicated in the bottom graph of Figure 25 by off-to-on transition 267a and on-to-off transition 267b.

Streaming instruction execution can be performed using two techniques: just-in-time (JIT) sequential transfer and destination shifting. In the JIT sequential transfer method shown in Figure 26A, a serial packet data stream 272 sent from the PBT controller to the intelligent LED pad is interpreted by a decoder 270 according to a decoding table 271, resulting in two outputs to color shift register 279a and time shift register 279b, respectively. Each successive interval includes the on and off time of the interval. The elapsed time is calculated one interval at a time as the shift register advances sequentially. For example, _t5 = _t4 + ( _ton4 + _toff4 ). This process is performed using a first-in, first-out algorithm, with only the first-in, first-out shift register data frame 277 driving the LED driver 278. All subsequent frames and all previous frames waiting in the queue are discarded once executed. The corresponding color shift register in data frame 277 specifies which LEDs are illuminated by LED driver 278. For example, register [|Blue|Red NIR1|NIR2] has the bit string 0100, 1000 drives only the red and blue LEDs, and 0011 drives both the NIR1 and NIR2 LEDs. The resulting light output includes a red pulse 275a, a blue pulse 275b, an NIR1 pulse 275c, and an NIR2 pulse 275d, as well as both simultaneous NIR1 and NIR2 pulses 275e. In this manner, the shift register advances at a variable rate, speeding up or slowing down based on the values of _{t_on} and _{t_off} .

In the shift-destination method shown in FIG. 26B, decoder 270 simultaneously outputs four separate bit strings 275a, 275b, 275c, and 275d to drive the red, blue, NIR1, and NIR2 LEDs clocked relative to a fixed-rate clock. To extend the duration of the LED illumination, the on-state bit is repeated for the entire on-duration. In the shift-destination method, a file containing the lighting pattern is transferred to the LED pad and decoded before LED playback.

Figure 26C contrasts the JIT sequential transfer method with the destination shifting method. The JIT method decodes the four LED color registers 279 and drives them at specified intervals until the color register changes, while the destination shifting method continuously decodes the transfer into a 4-bit sequence, stores it, and then plays it back sequentially from memory. Either way, data streaming has the advantage that the LED pad does not require significant memory to store the treatment data. Streaming has the disadvantage of requiring a steady flow of data from the PBT controller to the LED pad.

Another approach is to transfer the entire playback file from the PBT controller to the intelligent LED pad before initiating LED therapy. This operation, shown in the flowchart of Figure 27, involves two parallel operations: one executed by the LightOS operating system in the PBT controller host and the other executed by LightPadOS in the LED pad client. As shown, after file transfer, execution occurs autonomously within the LED pad without intervention from the PBT controller. After a program is selected in step 300, the playback file for driving the LED sequence is transferred from the host to the client. The LED pad receives the file transfer in step 302 and then decompresses the file in step 303, removing the file's Layer 2 MAC data, such as the header and checksum bits. The payload data is extracted and loaded into volatile memory, such as static RAM. This process is illustrated graphically in Figure 28, where an incoming USB packet 310 is sent over a physical medium, such as USB, to the communication interface 338 of the intelligent LED pad 337. Upon receipt, payload 311 is extracted and then unpacked (step 312), creating the necessary decompression or file formatting to create executable code 313. Executable code 313 is then stored in volatile memory 334a. Executable code 313 is sufficient to run on the LightPadOS operating system without requiring any other files or subroutines beyond the LED pad's operating system, and includes hard-coded data for algorithms 314 used in PBT therapy, either for a single treatment or for an entire PBT session. This code can be implemented, for example, in C++ or other common programming languages.

Returning to FIG. 27, once the playback file has been unpacked and stored in RAM in step 303, the LightPad OS notifies the host PBT controller in step 304b that it is ready to begin a session. Once the user confirms readiness by selecting the Start Therapy button 309, the session execution instructions are put into effect in step 304a, beginning with step 305a, in which a start session command is sent to the LED pad. As therapy progresses, the LED pad occasionally reports its status (step 306b) to the host PBT controller, including time, temperature, or other relevant program status information, which the PBT controller can display in step 306a. If a fault condition occurs in the LED pad, the LightPad OS's interrupt service routine 307b and the LightOS's 307a may communicate and negotiate what to do about the condition that caused the interrupt. For example, if the LED pad is unplugged and accidentally reconnected during a session, the session pauses, notifies the user of the connection error, and informs the user how to correct the fault. Once the fault is corrected, the interrupt routine closes and therapy resumes until, in step 308b, the LED pad notifies the host PBT controller that the therapy program is complete. In response, in session termination step 308a, the PBT controller notifies the user that the session or therapy is complete.

For purposes of this discussion, the term "treatment" is defined as a single treatment procedure, typically 20 minutes in duration, designed to invoke photobiomodulation in a specific tissue type or organ. Furthermore, a "session" consists of a series of treatments. For example, as shown in FIG. 29, a treatment protocol for injury recovery (e.g., treating a sprained and severed ankle from a bicycle accident) might involve three "injury" sessions 315a, 315b, and 315c, conducted consecutively on alternate days, with each session involving three consecutive treatment sequences containing different algorithms that vary the light wavelength, power level, modulation frequency, and duration. For example, PBT session 315a, called "Inflammation," aims to promote healing by accelerating (but not eliminating) the inflammatory phase of the healing process. Session 315a includes a sequence of three steps 314a, 314f, and 314b, each containing algorithms 23, 43, and 17, respectively. Session 315b, titled "Infection," shown in FIG. 29, includes a sequence of three steps, 314c, 314b, and 314g, each including algorithms 49, 17, and 66, respectively. Note that treatment 314b, including algorithm 17, was utilized in both the inflammation and infection sessions. Session 315c, titled "Healing," includes a sequence of three steps, 314g, 314h, and 314g, each including algorithms 66, 12, and 66, respectively. Note that treatment algorithm 66 was utilized once in infection session 315b and twice in healing session 315c.

The step sequence of implementing sessions for inflammation, infection, and healing together creates Injury Protocol 316 by first accelerating the inflammatory phase of healing, including fibroblast and collagen scaffolding, cellular apoptosis, and phagocytosis, and then combating opportunistic colonizing secondary microbial infections. Finally, after inflammation subsides and all infection is eliminated, the final step of Injury Protocol 316 promotes wound healing by improving the thermodynamics and energy supply necessary to provide healthy tissue regeneration. Injury Protocol 316 does not employ daily treatment sessions but purposefully spreads the first three sessions over five days. The need for intervening periods of rest rather than daily treatment is illustrated by graph 317, shown in Figure 30, which describes a generalized biphasic dose-response model following the work of Arndt-Schultz [https://en.wikipedia.org/wiki/Arndt%E2%80%93Schulz_rule]. According to Wikipedia, "The Arndt-Schultz rule or Schultz's law is a law concerning the effects of various observed concentrations in pharmacology. It states for all substances: small doses stimulate; moderate doses inhibit; large doses kill. There are many exceptions in pharmacology - for example, when small doses of a drug do nothing - and the theory has evolved into its modern counterpart, "hormesis", but the underlying principle remains the same: in medicine, there is an optimum therapeutic dose, beyond which the therapeutic effect may be reduced or recovery may actually be inhibited."

Despite controversy surrounding the results of pharmacological studies, the biphasic model of "energy medicine" has been reaffirmed by numerous studies, from cancer radiotherapy to photobiomodulation. For example, in cancer treatment, small doses of radiation may not adequately kill cancer cells, while large doses are toxic and can kill the patient much faster than leaving the cancer untreated. Applying the biphasic model to photobiomodulation, graph 317 presents a pseudo-3D representation of the PBT state, with the axes representing treatment time. The orthogonal y-axis describes the power density of the PBT treatment scale in W/ ^cm² , and the vertical axis measures the amount of energy available in J/ ^cm² or eV/ ^cm² , or electron volts, scaled by the product of power and time and the observed magnitude of photobiomodulation, otherwise known as the observed treatment effect. Topographically, the graph appears as two coasts, a mountain range, and an interior valley. As demonstrated by low-dose treatments, known as subthreshold doses, the treatment has insufficient power, or rate of energy delivery, to do anything. Similarly, very short periods of time will not provide enough energy to invoke photobiomodulation, regardless of power level. In other words, applying too little energy too quickly will not trigger photobiomodulation.

Moderate power density and duration combinations result in irritation, resulting in a peak response curve for power densities or total energy doses above this level, where beneficial PBT responses and treatment effects rapidly decline and may even inhibit healing. Of course, excessively powerful laser levels can cause burns, tissue damage, and ablation (cutting). And while LEDs are not compatible with laser power densities, they can still be driven at high currents and cause overheating. However, these treatment conditions occur far beyond the power levels and energy doses shown in the graph. The graph on the right of the case study [1] confirms that the dose (fluence) dependence of PBT efficacy is indeed biphasic, with a minimal response at 1 J/ ^cm² , a peak response at ² J/ ^cm² , diminishing benefits at 10 J/ ^cm² , and suppression at 50 J/cm². Suppression means that the effects of PBT treatment were worse than doing nothing. Therefore, for this reason, along with concerns about safety and patient comfort, PBT treatments should be spread out over time and limited in power and duration.

Data security in distributed PBT systems

To achieve multi-layered secure communication in the disclosed distributed PBT system, the PBT controller's operating system (LightOS) and the intelligent LED pad's operating system (LightPadOS) use parallel communication stacks with consistent protocols and shared secrets that are imperceptible to device operators, hackers, or unauthorized developers. As such, the distributed PBT system operates as a protected communications network, enforcing security at any number of communication layers, including data link layer 2, network layer 3, transport layer 4 during setup, and session layer 5, presentation layer 6, or application layer 7 during operation.

As disclosed, "treatment, session, and protocol" define sequences of light excitation patterns and operating parameters, such as LED wavelength, modulation pattern and frequency, treatment duration, and LED intensity (brightness), which determine instantaneous power, average power, treatment dose (total energy), and ultimately treatment efficacy. To prevent copying and duplication, these sequences must be stored and communicated securely using encryption and other methods. While several data security methods and associated security certificates can be implemented as part of an application, LightOS and LightPadOS provide an additional level of security by including a "presentation" Layer 5 in the communication stack of the PBT controller host and networked intelligent LED pad clients.

The presentation layer is represented diagrammatically in Figure 31. The PBT controller 120 includes an OSI communications stack 330, which includes an application layer 6, a presentation layer 5, a data link layer 2, and a physical layer 1. As previously mentioned, in the PBT controller 120, application layer 6 is implemented using a PBT-specific operating system called LightOS. During operation, Layer 6 LightOS programs execute to generate actions that require communication to the intelligent LED pad. These actions are encrypted in presentation layer 5 and then passed to the lower-level communications layers in encrypted form, or ciphertext. Specifically, the ciphertext passed to the Layer 2 data link layer is packetized—that is, converted into a series of communications packets containing an unencrypted header and ciphertext payload according to a specific communications protocol, such as USB, I2C, or FireWire, and communicated to the LED pad via the physical PHY layer 1. For example, PHY Layer 1 may communicate with a corresponding PHY Layer 1 of the communications stack 331 residing within the intelligent LED pad 123 using the USB protocol using UV differential signaling 332. Thus, while the electrical signals constitute Layer 1 communications, the USB data structure operates as if the PBT controller and intelligent LED pad are communicating at Layer 2 with packets arranged in time as USB data "frames."

When the communications stack 331 receives a USB packet, it forwards the extracted ciphertext payload to the presentation layer 5, where it is decrypted and converted to plaintext. The plaintext file is then passed to the application layer 6 for execution by the LED pad's operating system, LightPadOS. Designed to communicate and execute instructions in a self-consistent manner between the PBT controller's LightOS and the intelligent LED pad's operating system, the bidirectional link between communications stacks 330 and 331 functions as a virtual machine at the application layer 7, meaning the distributed devices behave as if they were a single piece of hardware, performing encryption and decryption bidirectionally at the presentation layer. In this way, data can be transferred between the PBT controller and the intelligent LED pad. However, to prevent source code copying, the processing library is stored in encrypted form. For added security, the encryption key is separate from the communication key used to store the algorithm. Therefore, the treatment file must first be decrypted before it can be securely communicated.

The process for preparing, communicating, and executing an encrypted operation is represented diagrammatically in FIG. 32 via a graphical UI 341. A user selects an operation 342 from a library encryption algorithm 340. The encrypted algorithm 17 is then decrypted using a system key 343, which converts the ciphertext to plaintext and recovers the unencrypted operation 344. In an encryption process 345, the plaintext file of the algorithm 17 is re-encrypted using an encryption key 346 exchanged with the intelligent LED pad client. The resulting ciphertext 347, containing the re-encrypted algorithm 17, is then packetized 348 and transmitted 349 using UV or another suitable communication medium.

In addition to therapy data, the same method can be used to prepare and transfer PBT session data from the PBT controller to the LED pad. This process is shown in the schematic diagram of Figure 33, where, through a graphical UI 351, the user selects a session 352 constructed from a library of encrypted algorithms 340, which in this example includes three encrypted algorithms. The ciphertext is then decrypted and converted to plaintext using a system encryption key. The three plaintext files are then merged 354 and then encrypted using an encryption key 356 exchanged with the intelligent LED pad client. The resulting ciphertext 357, containing the encrypted and merged algorithms, is then transmitted packetized 358 and 359 using USB or another suitable communication medium.

As shown in FIG. 34, incoming data packets 359 received by the LED pad's 337 communications interface are first processed to remove the packet header, extracting the payload 360. The pad μC 339 then decompresses 361 to extract the encrypted merge algorithm 362. The ciphertext is then decrypted using key exchange to extract a plaintext file 364 containing the processing algorithm, or, in the case of a session file, the merged algorithm. The algorithm 366 containing executable code or the merged executable code 365 is stored in volatile memory 334a. The processing is stored in RAM, so that the file is erased upon power interruption, making copying the unencrypted executable code difficult. Autonomous pad playback of PBT sequences using bulk decryption after transfer (before playback), as shown in FIG. 35, involves user selection of a session 300. This involves decrypting 390 the encrypted file 302 received by the LED pad and loading it into RAM. In step 304b, the LightPad OS notifies the host PBT controller that it is ready to begin a session. Once the user confirms readiness by selecting the Start Therapy button 309, the session execution instructions are enabled in step 304a, beginning with step 305a, in which a Start Session command is sent to the LED Pad. The LightPad OS responds in step 305b by initiating treatment by executing a treatment algorithm 314. As treatment progresses, the LED Pad displays its status 306b, including time, temperature, or other relevant program state information, and which PBT controllers are available, to the host PBT controller in step 305a. If a fault condition occurs in the LED Pad, the LightPad OS's interrupt service routine 307b and the LightOS's 307a may communicate and negotiate what to do about the condition that caused the interrupt. Once the fault is corrected, the interrupt routine is closed and therapy resumes until the LED pad notifies the host PBT controller that the therapy program is complete in step 308b. In response, in session end step 308a, the PBT controller notifies the user that the session or therapy is complete.

Even greater security can be achieved by storing the algorithm in encrypted form in the LED pad. As shown in FIG. 36, incoming packets 359 received by the communications interface 338 in the LED pad 337 are processed to extract the payload 360, which is subsequently decompressed 361 and then stored as ciphertext 368 in volatile memory 334a. Files are played back when a user initiates a session, at file run-time, i.e., during autonomous playback, by decrypting the file as it plays. This process, known as "on-the-fly" decryption playback, is illustrated in the flowchart of FIG. 37. This process is identical to that of the bulk decryption process flow shown in FIG. 35, except that after the LED pad receives the sequence file 302, the next step simply decompresses and, if necessary, decompresses the file 303, but does not decompress it. During playback in step 391, the ciphertext is read from the SRAM volatile memory and executed on the fly, i.e., as playback progresses.

Figure 38 contrasts the bulk decryption and on-the-fly playback methods. In bulk decryption, the entire playback file 368 stored in ciphertext is read from volatile memory and decrypted to extract a set of plaintext instructions 365 that are executed to play the entire file. In contrast, in on-the-fly playback decryption, a portion 368a of the stored playback file is read and decrypted 365a, then executed 392a by adding new plaintext instructions to the playback buffer. Meanwhile, another section of ciphertext 368a is read from volatile memory and decrypted 363 to recover plaintext executable file 165b, then executed 392b by adding this file to the end of the playlist.

Distributed PBT system with LED pad player

While JIT or shift-to-source-based data streaming for LED drive control can be used to control LED pads in a distributed PBT system, delivering real-time data over the communications network connecting the PBT controller and one or more LED pads becomes problematic when more advanced algorithms are required. Even when high-bandwidth communications are available, streaming clock signals or multi-MHz digital data represents a questionable command and control method, especially in safety-critical applications such as medical devices. An alternative enabled by the disclosed distributed PBT system is to employ a two-stage process for driving LEDs: first, downloading an "LED player" to the LED pad, and then downloading an "LED playback file" that defines the specific PBT operation or PBT session to be executed. With this disclosed method, LED drive execution is performed autonomously within the intelligent pad based on commands from the PBT controller. Because the LED driver is local within the LED pad, advanced functionality such as waveform synthesis and sine wave drive can be realized. Running multiple operations or sessions requires only a new "LED playback" file to be newly downloaded; the original LED player can be retained.

The first step in intelligent LED pad playback is downloading the LED player from the PBT controller to the LED pad. In a manner similar to the streaming file transfer process shown in FIG. 36, the download process shown in FIG. 39 involves transferring the encrypted playback file 480 shown in FIG. 39 from the PBT controller to the intelligent LED pad. The download process involves the encrypted LED player file 480a being decrypted 363 with the system key and then re-encrypted 370 with the LED pad (client) key 356 to create the encrypted LED player file 480b. This ciphertext is then sent to the intelligent LED pad, where the payload is extracted and decompressed 361, then decrypted 363, and stored in volatile memory 482. The downloaded LED player content includes a waveform synthesizer 483, a PWM player 484, and an LED driver 485.

Waveform synthesis algorithmically generates excitation patterns such as sine waves and sinusoidal chords, but can also generate triangle and sawtooth waves or play audio samples. The operation of the waveform synthesizer 483, shown in Figure 40, involves the waveform synthesizer 483 converting its input, a waveform parametric file 486, with a system clock Φ _sys to generate a synthesized waveform f(t), represented as a synth output data table 489, i.e., a function table f(t) paired with respect to elapsed time t. The PWM generator 555 then converts the function table into a high-frequency PWM pulse train 490 to generate a synthesis file 488, which contains a synthesized waveform 491 embedded within the PWM output 490. Depending on the algorithm, the waveform synthesizer 483 may also utilize waveform primitives 487. While the synthesizer can be realized in hardware, waveforms up to 20 kHz, i.e., within the audio range, can be easily implemented using software. For example, for the period between 0.5 and 1.0 ms, the value of f(t) = 0.6545. The process ΨP[f(t)] converts the function f(t) into a PWM pulse train with on and off times, where the output is in the high (on) state 65.45% of the specified interval (0.500 - 0.827 ms) and the low (off) state 0.827 - 1.000 ms. Therefore, the duration t _on = 0.827 - 0.500 ms = 0.327 ms, and the off duration t _off = 0.500 - 0.327 ms = 0.173 ms. In other words, the value f(t) is the duty factor during the period D = t _on /T _PWM , where T _PWM = t _on + t _off .

Because the duty factor D is an analog value limited between 0% and 100%, for convenience, f(t) is limited to any value between 0.0000 and 1.0000. If f(t) is allowed to exceed 1.000, the value must be scaled by the function's maximum value, i.e., f(t) = [f(t) _unscaled / f(t) _max ]. Otherwise, the waveform will be clipped to the value 1.000ΨP[f(t)] by the process. The PWM clock frequency, called the symbol rate clock _Φsym , is given by _Φsym = 1/T _PWM . The symbol rate is derived from the system clock _Φsys and must exceed the highest frequency waveform f(t) to be synthesized, or written mathematically as _Φsys > _Φsym > f(t). The following table shows the time interval _tx = (x-1)T _PWM divides each 500 ms interval into start times _tx (on) and _tx (off).

The second process in the LED player is PWM player function 484, shown in Figure 41, which processes synth output data file 488 in response to its input PWM parametric 491 and reference clock Φ _ref to generate PWM player outputs 493a and 493b. In operation, PWM player 484 generates a pulse-width modulated (PWM) pulse train 492, _Gpulse (t), which comprises the algebraic product _Gsynth (t) · _Gpulse (t). The waveform of _Gpulse (t) consists of repeating pulses made up of on duration t _on = DT _PWM and off duration t _off = (1-D)T _PWM .

The PWM player function can be performed in hardware, but is easily implemented in software. Written in logical pseudocode in terms of a high-speed counter and x (incremented each loop), it looks like this:

This means that for each cycle of duration T _PWM from time xT _PWM ≦ t < (xT _PWM + DT _PWM ), the magnitude of the PWM player's output is equal to the input (on state), and for the interval (xT _PWM + DT _PWM ) ≦ t < (x + 1)T _PWM , the PWM player's output is grounded and is a digital "0." By chopping the input G _synth (t) with the PWM pulse G _pulse (t), the waveform at output 493a is digitized with a value equal to G _synth (t) · G _pulse (t). The underlying waveform is superimposed on top of PWM signal 494. Typically, PWM player 484 outputs only a single digital waveform, but it can generate multiple outputs if desired. For instance, in the example shown, output 493a contains the multiplicative combination of two PWM pulses, while output 493b is the same as G _pulse (t), meaning G _synth (t) = 1. PWM player 484 can also output a constant time-invariant value: G _synth (t) · G _pulse (t) = 1.

The third step in LED player operation is the LED driver 485. As shown in FIG. 42, the LED driver 485, synchronized to a reference clock Φ _ref , combines a driver parametric 495 with the output of the PWM player 484 to generate an LED drive stream 497. Unlike the waveform synthesizer 483 and PWM player 484, which output digital signals, the output of the LED driver 485 is analog. Using the driver parametric 495, a programmable reference current 496 is generated with magnitude αI _ref (t) and multiplied by the output of the PWM player 484. Specifically, G _synth (t) · G _pulse (t) generates an output 497 composed of αI _ref (t) · G _synth (t) · G _pulse (t). The output waveform I shown in graph 498 _is a time-varying waveform, specifically a sine wave, a digital pulse, and a current that varies over time. The PWM player 484 can output a single output as input to the LED driver 485, but it can also provide two or more different outputs if desired. This could be useful, for example, in a large PBT system where many zones are required to uniquely target each body part, thus increasing tissue specificity.

The entire LED regeneration process, summarized in the example of FIG. 43, sequentially utilizes a waveform synthesizer 483, a PWM player 484, and an LED driver 485 to generate an LED drive stream 497. Unlike prior art methods, the LED drive of the disclosed distributed PBT system is generated entirely within the LED pad, advantageously keeping all therapy libraries and PBT system control in a common PBT controller separate and distinct from the LED pad. The waveform generation process utilizes a system clock with a frequency Φ _sys generated within the LED pad to perform its tasks, eliminating the need for high-speed clocks distributed over long lines. To ensure synchronization between the PWM player 484 and LED driver 485 and the waveform synthesizer 483, the system clock Φ _sys is divided using software or a hardware counter to generate a reference clock Φ _ref . Therefore, the regeneration of LEDs within a particular LED pad is fully synchronized. While the waveform synthesizer 493 and PWM player 484 both output digital PWM signals containing repeated transitions between digital 0 and 1 states of varying duration, the LED driver output is analog and can drive LED brightness with any waveform, including, but not limited to, sine waves, sinusoidal chords, triangle waves, sawtooth waves, audio samples of acoustic or electronic music, audio samples of cymbal crashes and other noise sources, and any frequency within the audio spectrum from 20 Hz to 20 kHz, from the 0th to the 9th musical octaves. Additionally, LED conduction can be tailored to a range of frequencies, including the -1st and -2nd octaves, down to 0.1 Hz for example, or by driving the LED with direct current (0 Hz) to provide continuous wave (CW) operation.

Note that strictly speaking, the disclosed distributed PBT is an asynchronous system, since each pad independently communicates asynchronously with the PBT controller and each LED pad generates its own internal time reference for LED playback. That said, due to the high clock rate, precise time reference, and high-speed communication network, timing discrepancies between LED pads are in the microsecond range, are imperceptible to UI control and UX response, and do not affect the effectiveness of the PBT.

Waveform synthesis in distributed PBT systems

In a distributed PBT system, one PBT controller controls many intelligent LED pads, such as three, six, or even more. Due to the number of intelligent LED pads required, economic considerations mandate limiting the complexity of the LED pads, particularly the cost and processing power of the μP339 pads. Similarly, to control product costs, the total memory within the LED pads must also be limited. Due to limited computing power and memory, several criteria must be met to synthesize waveforms within the LED pads of a distributed PBT system.
- The amount of data transferred or stored on the LED pad should be limited.
Calculations performed on the LED pad should preferably involve simple arithmetic operations such as addition and subtraction, avoiding complex repetitive processes such as function or matrix operations, unless absolutely unavoidable or rare.
Calculations must be done in real time with minimal power consumption or heating.

The input file synthesis method 550, the detailed operation of which is shown in FIG. 44, loads 483 into the waveform synthesizer used to calculate the function f(t) 553. Upon selection, any utilization unit, function generator 551, or primitive processor 487, including the waveform synthesizer parametric 486, all executes synchronously to the system clock Φ _sys . For waveform synthesis, the primitive processor 487 requires access to a detailed waveform description, specifically waveform primitives 487. The resulting function f(t) 553 contains a Cartesian pair of f(t) versus time t, shown diagrammatically in function table 554. Function table 554 then generates the synth output file 488 using the process ΨP[f(t)], which is converted to time-varying digital data by PWM generator 555. The synthesizer output 488 is represented graphically as G _synth (t) 490 which contains the synth output table 489 and the numerically equivalent digital PWM file.

Waveform synthesis using a unit function generator

The operation of the unit function generator 551, shown in Figure 45, involves selecting a mathematical function and calculating the function's values over a series of time periods to generate the function table 554. These functions are called "unit" functions because they have analog values limited to real numbers between 0.0000 and 1.0000. An example of a time-varying function, f(t) = 1, or a "constant" unit function, is shown in graph 560. Another function, the unit sawtooth, shown in graph 561, is represented by the formula f(t) = MOD(tf,1), where (tf) is the argument of the modulus function and 1 is the base. That is, the function is a linear decimal fraction between 0 and 1. For any number greater than a multiple of 1, the modulus function remains constant; for example, if (tf) = 2.4, then MOD(2,4) = 0.4. In a sawtooth, the function rises to 1, then returns to 0 and repeats. The ramp-up function is one, and the ramp is symmetrical to back down to zero. The equation f is a triangular wave shown in graph 562, given by f(t) = 1-2*ABS[MOD(tf,1)-0.5].

The composition of a single sine wave or a code of three or more sine waves, each of frequencies _f , _f , and _f and relative magnitudes _A , _Ab , and A _, is expressed by the formula f(t) = _Aα (0.5 + 0.5[ _Aα sin(2πtfα ₎ + _Ab sin(2πtfβ ₎ + _Ac sin(2πtfβ ₎ ]/[( _Aα + _Ab + _Ac )]) + 0.5(1- _Aα ). This mathematical process, shown in Figure 46, mixes three sine waves 564, 565, and 566 with gains 580, 581, and 582, respectively, and is summed in digital mixer 583 using linear summation of digital words.

Digital addition—binary, octal, and hexadecimal arithmetic—is the same as adding octal numbers, except that the numbers involve the binary or binary equivalent representation of the numbers: base 2 (b2), base 8 (b8), or base 16 (b16) rather than base 10 (b10). While digital addition can be performed using dedicated devices, the arithmetic logic unit (ALU) within the LED Pad's microcontroller functions can easily perform tasks required in binary mathematics. Converting the numbers s to another base, then adding them to the alternative base and converting them back to base 10, produces the same result. This principle of equivalence is illustrated in the example table below for adding three numbers in different bases. In the context of waveform synthesis, the numbers being added represent the instantaneous values of three sine waves at any given time and are summed to produce the digital sum of the three numbers. For illustrative purposes, if the value of a sine wave is scaled by a factor of 10, say A _x f _x (t ₁ ) where A _x = 10, and x = 1-3 at a particular time t, the function values f _a (t ₁ ) = 1, f _b (t ₁ ) = 0.5, and f _c (t ₁ ) = 0.5. If the gain factors are weighted equally, i.e., where A _a = 10, A _b = 10, and A _c = 10, then the sum 10(Σf _x (t ₁ )) = 20. To convert this number to a unit function, the resulting sum must be scaled to a decimal between 0.000 and 1.000. This is the task performed by auto-range function 584.

For each time instant _tx , dividing _Ax ( _Σfx ( _tx )) by the sum of the gain multipliers ( _Aa + _Ab + _Ac ) yields the average of the blended code. For equal weighting, i.e., Ax = 10, the sum of these gain factors ( _Aa + _Ab + _Ac ) = 30. When applied to the above sum, auto-range scaling transforms the sum of 20 into an auto-range scaled number of 20/30 = 0.666, the same number obtained by averaging three numbers with instantaneous values of 1.0, 0.5, and 0.5. The auto-range feature also works when sinusoids are blended with uneven weighting. In this case, one or more sinusoidal frequency components dominate the mix. For example, a blend of _Aa = 20%, _Ab = 40%, and _Ac = 40% produces the following signal combination:

In this case, (A _a +A _b +A _c ) = 100, g(t) = 70, so the output of the autorange function is 0.7. The autorange function uses a positive multiplier; _{A α} > 0 is used to scale the signal to compensate for the amplitude compression. Because we not only shift the scalar A _α function, but also its shifted average value, a DC offset correction term, 0.5(1 - A _α ), is added to the sine wave sum to bring the function back to a mean of 0.5.

Figure 47 shows several sine waves and sinusoidal chords created according to the identity function generator. In the example shown, three sine waves, each an octave apart (i.e., _fc = _2fb = _4fa ), are generated with various gain factors to produce various complex functions. The gain factors [ _Aa , _Ab , _Ac ] control the mixing or "blending" of the frequency components. Because the components are averaged, the gain factors can be any positive real number. However, for convenience, the three components can be scaled by percentages. In some cases, the weighting factor is zero, meaning that a sine wave of a particular frequency is not present in the mix. For example, in graph 564, [ _Aa , _Ab , _Ac ] = [1, 0, 0], so only the sine wave fa is present. Similarly, in graph 565 where [A _a , A _b , A _c ] = [0, 1, 0], only the middle octave sine wave f _b exists, and in graph 566 where [A _a , A _b , A _c ] = [0, 0, 1], only the highest octave sine wave exists.

This diagram also shows various mixed blend chords. Graph 567 shows an evenly weighted mixed blend of sinusoids of frequency f, f _a and f _b . Graph 568 shows an evenly weighted mixed blend of sinusoids of frequencies f _b and f _c . Graph 569 shows an evenly weighted mixed blend of sinusoids of frequencies f _b and f _c . A non-uniform mixed blend of two sinusoids with 2/3rd weighting of frequency f and a sinusoid of 1/3rd frequency f, f _b , is shown in graph 570. A three-sinusoid mix includes an evenly weighted chord 572 and an unevenly weighted sinusoidal chord 571, where [A _a , A _b , A _c ] = [0.2, 0.4, 0.4]. The algebraic calculation of sin(θ), where θ = f _x t, for x = a, b, c..., requires the calculation of a series for each sin(θ) evaluation. [http://www2.clarku.edu/~djoyce/trig/compute.html]

Here, n = n (n-1) (n-2)...3 2 1. Note that the same method can be used to generate a cosine waveform, but with a 90° phase shift. Generating the three-sine code Ax ( _Σfx ( _tx )), with the ninth and highest frequency sine wave, requires approximately 20 kHz, 360-degree precision, and all the aforementioned calculations along with PWM generation. ΨP[f(t)] must occur at a rate of 7.2 MHz, or within 138 ns. This approach is computationally intensive, wasting computing cycles and power consumption, especially when synthesizing high frequencies.

Waveform synthesis using primitive processors

An alternative method that is much less computationally intensive and better matches the limited computing power of the LED Pad μP339 is to use a table lookup to evaluate the function. For periodic functions, the values of the function at regular increments of the period, e.g., at a fixed angle or a fixed percentage, can be pre-calculated and loaded into a table, referred to herein as function "primitives." For example, the value of sin(θ) depends on its argument, the angle θ, so here

Because the sine function is periodic, there is no reason to recalculate the same value every time you need to evaluate sin(θ).

However, lookup tables face some fundamental hurdles. For example, a table can only return the value of a function under the same input conditions—i.e., the same arguments—as previously calculated. Just because a table contains the value of sin(45°) does not necessarily mean it knows the value of sin(22°). In a subroutine call to a lookup table, it is unlikely to guarantee that the input arguments match the available arguments unless the two are co-developed to ensure they use the same values. Another issue with using lookup tables is the stiff equation problem of performing high-resolution waveform synthesis across several orders of magnitude of frequency. For example, if a 20 kHz sine wave (9th octave) is synthesized using a PWM method with 16-bit precision, the required sample rate is (20,000 Hz)(162) = 1,310,726,000 Hz, or approximately 1.3 GHz. If, in the same simulation, a 0.1 Hz (-2nd octave) infrasound excitation pattern is added to the string, the period of the low-frequency component is T = 1/f = 1/(0.1 Hz) = 10 seconds. This means that to synthesize a single 10-second infrasound tone while maintaining the required resolution at 9 octaves, a table of (1.3 GHz) (10 seconds) = 13 billion data points would be required. Such a large data table would not only take too long to transfer from the PBT controller to the intelligent LED pad, but would also require a lot of memory.

To solve stiff equation problems while ensuring matching arguments between subroutine calls and lookup tables, the inventive method disclosed herein uses predefined periodic waveform primitives, such as sine waves or linear (scalar) functions, in combination with a set of counters that share a common numeric base, e.g., base 2. As used herein, the term "primitive" means a tabular, time-independent description of a waveform. The waveform is described using arguments specified relative to the function's period T, rather than absolute time. For example, for a linear function such as a sawtooth wave, entering a linear (Cartesian) argument into a lookup table will return a unique value. For a linear unit sawtooth that ramps from 0 to 1 over period T, the input p is unitless, so that at 25% of T the function "saw(p)" has a value of 0.25, at 78% of T the function "saw(p)" has a value of 0.78, and so on. To accommodate repeating cycles, it is convenient to represent the argument input "p" using the modulus function MOD(argument, limit). For a positive input, MOD(p,1) returns a value. It is bounded between 0 and 1, i.e., the remainder when divided by the largest integer multiple of the limit. For example, for any value of z, MOD(0.78,1) = 0.78, MOD(5.78,1) = 0.78, and MOD(z.78,1) = 0.78. Therefore, to describe a repeating waveform, only data covering one period T is needed.

The same function applies to polar coordinates. Evaluating sin(MOD(θ,360°) produces a repeating sequence of values between sin(0°) and sin(359.99...°). At 360°, the entire cycle repeats because sin(MOD(360°,360°)) = sin(0°). Note that in actual code or spreadsheets, the angle argument θ to sin or other trigonometric functions would be expressed in radians rather than degrees, but the principles of the modulus function and its application remain the same. Using the modulus function in the disclosed manner can limit the size of a lookup table for any periodic function to a single period, dramatically reducing the size of the table. Thus, the number of data pairs in each lookup table is equal to the principal resolution _ξ , providing a one-to-one correspondence between the input to the lookup table _Φx and its output _fx . Here, for any octave x, the relationship _Φx = _ξxfx represents the transformation performed by the lookup table subroutine call.

While these function primitives consist of a collection of time-independent states that describe mathematical functions, waveform synthesis requires them to be combined with oscillators, including either digital or analog clocks, to generate time-varying waveforms. In particular, for linear functions with period T, such as triangle waves or sawtooth waves, the argument x can be expressed as x = t/T, while for sine waves, sinusoidal chords, and other trigonometric identity functions, θ = tf. In either case, a time source is required to convert a time-independent waveform primitive into a time-varying function. One such implementation for generating a range of time sources, algorithmically represented in Figure 48A, combines a series of binary (÷2) digital counters 590 through 598 to generate ten synchronous clock frequencies _Φ9 through _Φ0 from a common clock—specifically, a symbol clock rate _Φsym with programmable frequency. The clocks are then used to synthesize periodic functions, such as audio-spectral sine waves with corresponding frequencies from 9F in the octave to _f0 in the octave zero, mixed in various combinations as desired. The same method, not shown, can be used to generate infrasound, i.e., vibration waveforms below 20 Hz, and ultrasound, including frequencies above 20 kHz (if a suitable transducer is used).

During synthesis, each clock is converted into a time-varying waveform f(t) using a periodic function lookup table. For example, sine waves, sinusoidal codes, triangle waves, sawtooth waves, etc. Each clock is paired with the waveform it creates. For example, Φ ₈ uses sine wave lookup table 618 with primitive resolution ξ ₈ to generate sine wave frequency f ₈ , Φ ₃ uses sine wave lookup table 613 with primitive resolution ξ ₃ to generate sine wave frequency f ₃ , and sine wave lookup table 611 with primitive resolution ξ ₁ to generate sine wave frequency f ₁ .

And generally, f _x =Φ _x /ξ _x . Thus, in operation, the 10-octave waveform summation implementation primitive processor 552 uses nine binary counters 598-590 to generate ten clock frequencies, including input Φ ₉ =Φ _sym and clocks Φ ₈ through Φ ₀ , to drive corresponding sine wave lookup tables 619-610 to synthesize sine waves f ₉ through f ₀ .

The mixing process involves selecting various combinations of sine waves using octave data switches 609-600, mixing the selected sine wave components in digital mixer summing node 630, and weighting the components by various percentages using digital gain amplifiers 620-629. The blended sum is scaled to the range 0.000 to 1.000 by auto-range function 631. While the primitive processor can be implemented in hardware or firmware-controlled hardware, the functionality can be fully emulated using software, with mixer 630 being implemented digitally using binary addition, and auto-range function 631 being implemented using binary mathematics, performing one of several division algorithms (https://en.wikipedia.org/wiki/Division_algorithm). To avoid performing unnecessary operations, primitive processor 552 only performs operations on the selected octave switches 600-609.

Using the method shown in FIG. 48A, the implementation primitive processor 552 performs wide-bandwidth waveform synthesis and code construction at frequencies over 30 years, i.e., 10 octaves spanning a frequency range of 20 Hz to 20,000 Hz, using only lookup tables and a series of counters. The disclosed method is computationally efficient, requiring minimal memory and calculation power to execute, and, unlike the unit function generator 551 of FIG. 44, does not involve real-time evaluation of a power series. A key function of the synthesizer in wide-bandwidth algorithmic waveform generation is the role of the counter operation. Counters 599-500 together generate 10 octave clock frequencies used as inputs to corresponding lookup tables 619-610. Because each octave is fed by its own dedicated clock frequency, the number of points in the corresponding table and the table are limited to the precision required for that particular octave and do not include data used in other frequency bands. In this way, the disclosed combination of counters and lookup tables overcomes the stiff equation problem discussed above. To further minimize computational intensity and avoid unnecessary calculations, lookup table subroutine calls are restricted to only the tables selected by the octave switch.

To avoid aliasing and phase-shift distortion, counter cascades 698-590 are synchronized to a common clock, called _Φsym , output from tuner (counter) 599. For convenience, the symbol rate _Φsym is equivalent to the 9-octave waveform synthesis clock signal _Φ9 , but this relationship is arbitrary. A symbol rate higher than the PWM resolution of the highest synthesis frequency, _Φsym ≥ _{ξsymfmin ,} is sufficient. Counter cascades can be implemented using hardware or software. Ripple counters can also be used, but synchronous counters are recommended to prevent clock phase shift. A ripple counter is a counter cascade in which the output of each counter stage is available as soon as it enters the next stage. Due to propagation delays through each counter stage, the output of the high-frequency clock changes state before the low-frequency clock. Thus, states change "ripple" through the cascade, with the first clock Φ ₉ changing state, and some time later Φ ₈ , then Φ ₇ , Φ ₆ , Φ ₅ , etc. rippling like waves across the surface of a pond.

In contrast, synchronous counters operate synchronously. While the digital count takes time to ripple through the counter chain, the output changes only simultaneously with the synchronous clock pulse. In this way, signal ripple through the counter cascade is invisible to the user. More specifically, whether implemented in hardware or software, asynchronous counters operate like ripple counters, but with D-type flip-flops [https://en.wikipedia.org/wiki/flip-flop_(electronics)] with latching outputs. D flip-flops retain their previous state until enabled by a latch signal with a corresponding truth table. This means that the high or low state of the data input is copied to the latch output only when the synchronous clock goes high. The synchronous clock can then return low, and the flip-flop output will remain latched in the state it was at the D input at the time of the last synchronous clock pulse until the next synchronous pulse occurs. During the interval between clock pulses, the output of each counter stage can change without a transition appearing at the counter's output. To avoid cluttering the circuit diagram, counters 599 through 590 can represent synchronous counters without explicitly depicting D flip-flop latches or synchronous clock inputs. To ensure that the clock transitions ripple completely through the counter cascade before updating the state of clock outputs _Φ9 through _Φ0 , the synchronous clock pulses are derived from the state transitions of the lowest synthesized frequency clock, represented in this example as _Φ0 .

The symbol rate Φ _sym that feeds the counter cascade is generated from the system clock rate Φ _sys using a programmable counter "tuner" 599. The symbol clock rate Φ _sym is generated to produce a maximum output frequency f _min with a resolution ξ _sym . The value of the primitive resolution ξ _sym is a programmable input to tuner 599 that can be changed depending on the waveform synthesis being performed. The numeric variable ξ _sym , referred to herein as the "primitive symbol resolution," is defined as the resolution of the highest synthesis frequency, where ξ _sym = Φ _sym /f _min has a value ranging from 24 to 65,536 depending on the synthesis precision required. For example, if ξ _sym = 96 is selected for the sine wave synthesis means, the synthesizer's highest pitched sine wave is related to the symbol clock rate by the relationship Φ _sym = ξ _sym f _min = 96f _min , where 90° of arc uses 24 points, one point every 3.75°. In an operational configuration, tuner 599 generates an entire cascade of frequencies derived from and tuned to the symbol clock rate Φ _sym . The resolution of ξ _sym does not need to match the lower octave lookup table. Lookup tables 600 through 619 can be used for different precision levels of ξ _x , or alternatively, identical precision lookup tables may be used to generate some or all of the desired frequency components. Alternatively, the same lookup table can be used for all generated sine waves. In such a case, all sinusoidal frequencies f _x have the same precision ξ ₉ =ξ ₈ =ξ ₇ ...ξ ₁ =ξ ₀ .

Because the entire counter cascade is driven from a common symbol clock rate, _Φ , the exact frequency relationship of the synthesized waveform is precisely defined by the counter frequency, _Φ , and the corresponding lookup table resolution, _ξ . This relationship is illustrated using a binary (divide-by-two) counter, but the counter divisor is not limited. Dividing by two is convenient because it halves the frequency, which corresponds to one octave or 12 semitones in the musical scale. However, a cascade of counters, each with a different divisor, can be used. Alternatively, a programmable counter can be used, where the count is loaded into the counter. Furthermore, because the counters run at a fixed clock rate and complete one full oscillation period—that is, one full cycle of the lookup table—at every _ξ data point, the relative timing and phase of any two periodic functions are precisely known. For example, given two sine waves with frequencies _f and _f, the divisors are:

Then, the waveform frequency ratio is given by the following formula:

This ratio shows that frequency scaling can be achieved by changing the clock Φ _x or by changing the lookup table resolution ξ _x . For example, if a lookup table of constant resolution is used, with ξ _x = ξ _y = 24, the frequency ratio f _x /f _y of the synthesized sine wave depends only on the ratio of the clock rates Φ _x /Φ _y .

In such a case, the clock frequency ratio Φ _x /Φ _y = 4 results in two sine waves of the same note two octaves apart. For example, note A at 1,760 Hz at the sixth note and note A at 440 Hz at the fourth note. Figure 48B shows only the sixth and fourth octave switches 606 and 604 enabled and used to access data in sine wave lookup tables 616 and 614. Each waveform has a primitive resolution ξ ₆ = ξ ₄ = 24. The outputs are amplified by digital gain amplifiers 626 and 624 and then combined in digital summing node 630 to produce a combined waveform output. In operation, tuner (counter) 599 generates a symbol clock Φ _sym from the system clock Φ _sys . A cascade of ÷2 counters 598, 597, and 596 divides the symbol clock Φ _sym to generate the 6-octave clock Φ ₆ , which is divided by counters 595 and 594 to generate the 4-octave clock Φ ₄ .

The resulting two sine waves are given by their sum:

The multiplier [0.5 + 0.5 * (period)] is used to increase the magnitude of the sine wave peaks centered around the zero mean value from ±1 to ±0.5. The adder 0.5 shifts the curve up by +0.5 to span the positive range of 0.000 to 1.000. By enabling the octave switch 601 as shown in Figure 48C, component 1 of lookup table 611 driven by clock Φ is added to the code. Clock Φ ₁ is generated from clock Φ ₄ using counters 593, 592, and 591. The added one-octave frequency component is given by the following equation:

This is obtained by adding the three resulting sine wave codes.

As mentioned above, the above synthesis methods utilize a single waveform primitive to generate two or three sinusoidal codes simultaneously.

Additional details of the primitive processor operation are shown in the single primitive code synthesis shown in Figure 49. As shown, tuner 599 comprises two counters—a system clock counter 640 and a symbol clock counter 641. The system clock counter converts the μC system clock, at frequency Φ _sys , to a reference clock frequency Φ _ref of a convenient fixed frequency (e.g., 5 MHz). The symbol clock counter then converts Φ _ref to the symbol clock rate Φ _sym, which is used to define the reference frequency of the counter cascade for sine synthesis. In the example shown, counters 598 through 593 comprise binary counters, generating multiple sine wave frequencies, each an octave apart, as listed in the table above. Further inspection reveals that there is no binary counter cascade.
Φ _x is the clock rate, which is a multiple of 2 for all octaves and for the symbol rate Φ _sym .
All octave frequencies f _x are multiples of two of the maximum resultant frequency f _min , which is shown in, but not limited to, the ninth octave of the musical scale.
The relationship between the symbol clock rate Φ _sym and the maximum synthesis frequency f _min is determined by ξ _sym , the resolution of the highest frequency waveform to be synthesized. The F-power product f _min ξ _sym = Φ _sym sets the maximum clock rate in the counter cascade.
The relationship between the symbol clock rate Φ _x and the composite frequency f _x of each octave x is determined by ξ _x , which is the fundamental resolution of the waveform of that octave.

All relationships between clock rates and frequencies within a single primitive binary counter cascade include precise ratios to other frequencies present in the primitive processor, so setting the frequency and resolution of any one composite waveform at frequencies _fx and _ξx automatically determines the frequencies of all composite frequencies and clocks throughout the counter cascade, including the symbol rate _Φsym and maximum frequency _fmin . The frequency scaling of the primitive processes is summarized in the following table:

In this regard, the disclosed primitive processor represents a "tuning" system in which an entire multi-octave synthesizer is set to a single "key" frequency, similar to tuning a monophonic musical instrument to a single note or key. For example, an instrument tuned in the key of A. For this reason, an operational symbol clock counter 641, namely f _key , is set by two parameters, 642, and a lookup table 645, which selects a key with a primitive resolution ξ _sym . As shown, the lookup table 645 is stored in either volatile or non-volatile memory within the LED pad, which is selected by an identifier, such as a hexadecimal code 643, or its binary equivalent 644.

Since the entire synthesizer is tuned to multiples of an octave, the f _key selection input 642 is arbitrary. For convenience, digital tuning can conform to an international frequency standard for pitch. For example, pitch "A" has a frequency of 440 Hz, located within the fourth octave above middle C. This 440 Hz tone is considered a common tuning standard for musical pitches [https://en.wikipedia.org/wiki/A440_(pitch_standard)]. Also known as A440, A4, or Stuttgart pitch, the International Organization for Standardization classifies it as ISO-16. Adapting this standard to a primitive processor allows the disclosed synthesizer to be tuned to a specific key by selecting notes or frequencies within the four octaves.

Specifically, the "Key Select" input 642 sets the note or frequency in the fourth octave to which the entire synthesizer is tuned. If four octaves are arbitrarily selected as the frequency input range for tuning the synthesizer, so that the maximum synthesis frequency is in the ninth octave of the audio spectrum, then the ninth and fourth octaves are five octaves apart. Since ²⁵ = 32, _fmax = _f9 = _32f4 , which means that Key Select 642 _sets the maximum frequency _fmax = 32fkey. Given _Φsym = _ξsym and _fmin , then _Φsym = _ξsym ( _32fkey ). For example, if you set Select Key to 440 Hz (standard A above middle C), then _f4 = 440 Hz and _fmax = _32fkey = 32 (440 Hz) = 14,080 Hz, automatically scaling the entire available synthesis frequency spectrum to: _f9 = 14,080 Hz, _f8 = 7,040 Hz, _f7 = 3,520 Hz, _f6 = 1,760 Hz, _f5 = 880 Hz, _f4 = 4400 Hz, _f3 = 220 Hz, _f2 = 110 Hz, _f1 = 55 Hz, _f0 = 22.5 Hz, and f-1 = 11.25 Hz. If _{the fkey} is set to middle D, all synthesis frequencies _fx will also be multiples of D. Or if the f key is set to middle A ^# , all binary synthesis frequencies will also be A ^# multiples. Synthesis of sine waves other than octave multiples will be discussed later in this disclosure.

Referring again to the implementation of the primitive processor in Figure 49, as shown, lookup table 645 contains an exemplary primitive description of a sine wave with a resolution of 24 points. This tabular primitive description of a sine wave is time-independent, based solely on the argument θ of sin(θ) as input. The _key f primitive processor is selected with selection key 642, and the table 645 is populated with primitives established by selecting a waveform ξ _sym = 24, e.g., 440 Hz, and a resolution ξ _sym , where the symbol clock rate Φ _sym and corresponding period T _sym are given by:

This symbol rate corresponds to a synthetic maximum frequency f _min of 9 octaves, where f _max = f ₉ = Φ _sym /ξ _sym = (337,920 Hz)/24 = 14,080 Hz, corresponding to a period T ₉ = 1/f ₉ = 71.02 μs, which also corresponds to T _sym ξ _sym = (2.9592...μs)(24) = 71.02 μs.

By establishing a time reference using a binary counter cascade, the time-independent sine primitive table 645 is converted into a time-based description of a function 646a, specifically g(t). The same clock symbol clock Φ _sym is used in time space to generate clocks Φ ₆ and Φ ₄ , which are used to synthesize sixth and fourth octave sine waves 647a and 648a.

These clocks have frequencies _f6 and _f4 and are used to synthesize two synchronous sine waves with the following frequencies:

In a given way, sine waves of equal resolution but different frequencies can be synthesized using a common clock and a single waveform primitive. In other words, the primitive table sets the shape of the waveform, while the resolution ξ and counter clock determine the frequency of the generated sine wave. The following example table shows the relationship between the argument of the sine function θ measured in degrees (or radians), the normalized unit sine wave function 0.5 + 0.5 sin(θ), and the corresponding times for states of a sine wave oscillating at frequencies _fmax in the 9th octave, _f6 in the 6th octave, and _f4 in the 4th octave.

This table shows the detailed pattern between 0° and 90°, but for brevity, the detailed 15° descriptions for the other three quadrants are redundant and omitted (since sine waves are symmetrical functions, all four quadrants are shown in one quadrant). The time required to complete a 360° cycle of a sine wave, or its period, T, varies with the frequency of the sine wave. For example, consistent with the previous calculation, sine waves with frequencies _f9 , _f6 , and _f4 have periods of 71 μs, 568 μs, and 2,273 μs, respectively. Specifically, the value of the function 0.5 + 0.5 sin(θ) = 1 when the argument θ = 90° = π/2. The period of the sine wave, T, occurs in four times this duration when θ = 360° = 2π. For example, a six-octave sine wave tuned to the key of A requires 142 μs to complete one-quarter of its cycle, so its period is T ₆ = 4(142.05) = 569.2 μs.

Figure 50 illustrates a code synthesis using a binary cascade counter to generate a single waveform primitive with independent time bases, f(t) = 1,168 Hz and f(t) = 292 Hz, respectively, which are converted to time-based sine wave tables 647 and 648 in a key of D with a resolution of ξ _sym = _{ξ x} = 24 (not shown). The component sine waves are then increased or decreased in amplitude by digital _gain amplifiers 626 and 624 with gain multipliers A6 and A4, which are implemented arithmetically using digital multiplication operations. The two sine waves are then mixed by digital summing node 630 to produce the sum g(t), where

Using weighted average with divisor (A ₆ +A ₄ ), we get:

During averaging, the term [A6 + A4] appears in both the numerator and denominator of a fraction that changes the mean value of the function, and therefore does not affect the 0.5 offset. A secondary objective of the auto-range function is to maximize the sinusoidal component _Aα to full scale, without any actual change in the mean of the function. To avoid a 0.5 mean value shift, the auto-range function disclosed herein uses an additive correction factor of 0.5(1- _Aα ).

As explained, the sum g(t) is scaled by auto-ranging function 631 by the scalar [ _Aα /(A6+A4)], which performs a weighted average of the sinusoidal components along with digital multiplication by gain factor _Aα . The resulting time-varying waveform f(t) 553, shown in tabular form 649, represents two sinusoidal chords 655 at frequencies _f6 and _f4 , with a mean value of 0.5 and the ability to maximize the amplitude of the periodic function over the range 0.000 to 1.000, without clipping or distortion of the signal. PWM generator 555 then processes f(t) through the PWM transformation ΨP[f(t)], producing synth output data 488, which includes a PWM string of data 499 referred to as G _synth (t).

One issue arising from the disclosed synthesis method is quantization noise. While a single sine wave does not present this problem, adding two or more sine waves introduces noise into the waveform. The source of this noise is shown in Figure 51A. Here, a cascade of binary counters 596-593 is used to generate three clocks, _Φ6 , _Φ5 , and _Φ4 , at half the input frequency. Using a fixed primitive resolution of ξ=24, the resulting sine wave data for _f6 , _f5 , and _f4 are tabulated in table 651. Upon inspection, we see that the data for frequency _f6 corresponds uniquely, one-to-one, to clock time _Φ6 , while the other frequencies do not change as rapidly. For example, at both t=0.1727 and t=0.1784, the data value for sine wave _f5 remains constant at 0.7500 as sine wave _f6 changes. Similarly, for the low frequency sine wave _f4 , the data output for the interval t = 0.1427 to 0.2497 remains constant at 0.6294 even though the _f6 data changes four times.

The impact of using fixed-resolution primitives at different clock rates is illustrated in Figure 3. In Figure 51B, various curves are contrasted over a fixed time interval. For the duration shown, graph 652 shows no digitization noise for a sine wave of frequency _f6 . A contrasting sine wave of frequency _f5 , generated by _Φ6 /2ξ, graph 653, exhibits a small but noticeable degree of noise. For the _f4 sine wave of graph 654, two octaves below _f6 , i.e., _f4 = _Φ6 /4ξ at ξ = 24, significant noise is evident. The noise problem is evident in the two-sine code of graph 655, combining _f6 and _f5 , and is further exaggerated in graph 656, showing the sum of sine waves of frequencies _f5 and _f4 .

One solution to this problem is shown in Figure 52A. Three different frequencies _f6 , _f5 , and _f4 are generated from a common clock frequency _Φ6 . Rather than scaling the clock frequency, the resolution is scaled and higher resolution primitives are used to generate the lower sinusoidal frequencies. Specifically, in lookup table 616, _ξ6 = 24, and in lookup table 615, the primitive resolution is doubled to _ξ5 = _2ξ6 = 48; similarly, in lookup table 614, _ξ4 = _4ξ6 = 96.

Therefore, the sinusoidal frequencies f ₆ , f ₅ , and f ₄ generated from the common clock Φ ₆
are all factors of two relative to one another, as shown in Table 661. Thus, the time step is constant for all frequencies generated. The resulting curves shown in Figure 52B, including sine waves 662, 623, and 624, as well as codes 665 and 666, show no signs of quantization error at this resolution. The frequency ratio of any two sine waves using this method remains accurate due to the previously defined criteria.

If Φ _x =Φ _y it holds.

This method, referred to herein as scaled primitive summation 660, contrasts with the single primitive summation 650 of chords blending three composite sine waves in FIG. 52C. In this, the block diagram of the sine wave lookup tables 616, 615, and 614 are identical at their resolution ξ = 24, but generated from Φ ₆ , Φ ₅ = Φ ₆ /2, and Φ ₄ = Φ ₆ /4 binary cascade counters fed by three different clocks. The time graph of the resulting code 659 shows significant digitization noise. In contrast, scaled primitive summation 660 uses a common clock Φ ₆ to drive three different resolution lookup tables 616, 615, and 614 at increasing resolution ξ _x = 24, 48, and 96 for x = 6, 5, and 4, respectively. The resulting waveform 669 shows no signs of digitization noise at this resolution.

To limit the maximum size of the primitive lookup table, the audio spectrum can be divided into bands. For example, the upper, middle, and lower musical scales, and the infrasonic band (i.e., below 20 Hz) of the zero and negative octaves. Such an approach is employed in the four-range scaled primitive synthesis block diagram shown in FIG. 53. The exemplary primitive processor in this example, tuner 599, includes a system clock counter 640 and a symbol clock counter 641 that converts the system clock Φ _sys to its fixed reference frequency Φ _ref , e.g., 5 MHz, where the symbol counter generates the Φ _sym clock frequency defined by the ratio Φ _sym /Φ _ref = (32ξf _key )/(5 MHz) in response to a key selection input 642, a note or key in the octave. In tuner 590, a cascade of counters, 3 divide-by-8 counters 672, 673, and 674, co-generate four frequency clocks: Φ _sym , Φ ₆ = Φ _sym /8, Φ ₃ = Φ _sym /64, and Φ ₀ = Φ _sym /512. Counters 672 through 673 each contain three-stage binary cascade counters, shown as single divide-by-8 counters for simplicity.

The highest frequency clock in the cascade, the symbol clock Φ _sym , is then used to synthesize sine waves for the four bands. In the upper bands, Φ _sym is used to generate sine waves _f9 , _f8 , and _f7 , respectively, according to selectors 609, 608, and 607. When the selector switches are enabled, the clock pulses of Φ _sym are passed to the corresponding sine wave lookup tables 699, 698, or 697 to generate sine waves _f9 , _f8 , and _f7 as needed.

Specifically, activating sine wave 699 with resolution ξ ₉ = 24 generates sine wave f ₉ with frequency f 9 = Φ _sym /ξ _9. The frequency of this sine wave is 32 times the selected frequency of the f _key and 1/24th of the symbol frequency Φ _sym . For the same upper acoustic, activating sine wave 698 with resolution ξ ₈ = 48 generates sine wave f ₈ with frequency f ₈ = Φ _sym /ξ ₈ = Φ _sym / (2ξ ₉ ). The frequency of this sine wave is 16 times the selected frequency of the f _key and 1/48th of the symbol frequency Φ _sym . Similarly, activating sine wave 697 with resolution _ξ7 = 96 generates sine wave _f7 with frequency _f7 = _Φsym / _ξ7 = _Φsym /( _4ξ9 ). The frequency of this sine wave is eight times the _fkey selection frequency and 1/96th of the symbol frequency _Φsym . Because the sine waves of frequencies _f9 , _f8 , and _f7 are generated from the same clock frequency _Φsym , their waveform synthesis uses the same time increments, thereby avoiding the problem of digitization errors in the upper acoustics mentioned above.

The same clock Φ _sym is also divided by 8 in counter 672 to generate a low-frequency rate clock Φ _sym used to synthesize the sine waves f _{6 ,} f ₅ , and f ₄ at the mid-range scale. When any of selector switches 605, 606, and 604 is enabled, a clock pulse containing Φ _sym /8 is passed to the corresponding sine wave lookup table 696, 695, or 694 to generate the sine waves f ₆ , f ₅ , and f ₄ as needed. Specifically, enabling sine wave 696 with resolution ξ ₆ =24 generates sine wave f ₆ with frequency f ₆ =Φ _sym /ξ 6 =Φ _sym /(8ξ ₆ ). The frequency of this sine wave is four times the f _key selection frequency and 1/192 of the symbol frequency Φ _sym . For the same medium sound, activating sine wave 695 with resolution _ξ5 = 48 generates sine wave _f5 with frequency _f5 = _Φ6 / _ξ5 = _Φsym / ( _16ξ6 ). The frequency of this sine wave is twice the _fkey selection frequency and 1/384 of the symbol frequency _Φsym . Similarly, activating sine wave 694 with resolution _ξ4 = 96 generates sine wave _f4 with frequency _f4 = _Φ6 / _ξ4 = _Φsym / ( _32ξ6 ). The frequency of this sine wave is equal to the _fkey selection frequency and 1/768 of the symbol frequency _Φsym . To generate sine waves with frequencies _f6 , _f5 , and _f4 coming from the same clock frequency _Φ6 = ( _Φsym /8), the waveform synthesis employs identical time increments, thereby avoiding the aforementioned problem of digitization errors within the intermediate scale.

To generate sine waves _f3 , _f2 , and _f1 at a lower scale than clock _Φ6 , counter 673 divides the clock _Φ3 by 8 to generate a low-frequency rate clock. If optional switches 603, 602, and 601 are enabled, the clock pulse _Φ3 = ( _Φsym /64) is passed to the corresponding sine wave lookup table ₆₉₃ , 692, or 691 to generate the desired sine wave for _f3 , _f2 , and _f1 . Specifically, enabling sine wave 693 with resolution _ξ3 = 24 generates a sine wave _f3 with frequency _f3 = _Φ3 / _ξ3 = _Φsym /( _64ξ3 ). The frequency of this sine wave, _f3 , is 1/2 the selected frequency of the f _key , and 1/1,536 of the symbol frequency Φ _sym . At the same low scale, activating sine wave 692 with resolution ξ ₂ = 48 generates sine wave _f2 with frequency _f2 = Φ ₃ /ξ ₂ = Φ _sym / (128ξ ₃ ). The frequency of this sine wave is 1/4 the selected frequency of the f _key , and 1/3,072 of the symbol frequency Φ _sym . Similarly, activating sine wave 691 with resolution ξ ₁ = 96 generates sine wave _f1 with frequency _f1 = Φ ₃ /ξ ₁ = Φ _sym / (256ξ ₃ ). The frequency of this sine wave is 1/8 of the selected frequency of the f _key and 1/6,144 of the symbol frequency Φ _sym . Because the generation of sine waves of frequencies f ₃ , f ₂ , and f ₁ occurs from the same clock frequency Φ ₃ = Φ _sym /64, the waveform synthesis uses the same time increments, thereby resulting in low acoustics that avoids the digitization error problem mentioned above.

A counter cascade can also be used to generate infrasound excitation of an LED, i.e., a sine wave with a frequency less than 20 Hz. As shown, the output of a divide-by-8 counter 674 with a clock frequency Φ ₀ = Φ _sym /512, when selected by a selector 600, generates a sine wave f ₀ with a resolution ξ ₀ = 24. Here, the generated frequency is given by f ₀ = Φ ₀ /ξ ₀ = Φ _sym /(512ξ ₀ ). Using the above principle, the scaling concept can be extended to generate two lower infrasound frequencies f-1 and f-2 (if desired) by including two additional sine lookup tables with resolutions of ₄₈ and 96, respectively, driven by the clock Φ 0. As discussed above, using time increments consisting of fixed intervals minimizes quantization noise, but requires larger, higher-resolution lookup tables, increasing the memory required within the LED pad.

If your lookup table has enough data points, you can use a single table to generate multiple octaves of data from a single clock. For example, a 24,576-point table can be used to synthesize a sine wave spanning 11 octaves with an angular precision of 0.0146484375° per data point. Using a 337,920 Hz clock to synthesize an 11-octave universal primitive table, you can generate frequencies, for example, from the middle of the key range Af ₉ = Φ _sym /ξ _sym = 14,080 Hz, 13.75 Hz below the ninth octave (including the A at 440 Hz). This example is shown in the fourth column of the table below. Using the same symbol clock rate, i.e., the same table columns, if the number of synthesized frequencies is reduced to only seven octaves, the size of the universal primitive data table is reduced to 1,536 data points spanning the nine-octave range from 14,080 Hz to f ₃ = 220 Hz.

Alternatively, using the same seven-octave universal primitive table, the frequency band covered can be shifted by using a lower symbol clock rate. For example, as shown in item 5 of the following table, a symbol clock rate of Φ _sym = 168,960 Hz and a universal primitive with 1,536 data points can cover a range from 7,040 Hz for eight octaves to 110 Hz for two octaves. A compromise between sine wave frequency range and data table size can also be achieved by reducing the table size and the symbol clock. Referring to item 6 of the table below, a symbol clock rate of Φ _sym = 42,240 Hz can generate sine waves from 1,760 Hz for six octaves to 55 Hz for one octave using a lookup table with only 768 data points.

The process of waveform synthesis using universal primitive synthesis is shown in Figure 54. Here, a tuner 599 generates a programmable symbol clock _Φ = _Φ /(32ξf _key ) according to key selection 642 and converts the clock into one or more sine waves of varying frequency. For example, a universal primitive table 677 is used to convert from _f9 to _f0 , blended with a programmable gain Ax according to digital gain amplifier 678, and summed in mixer 630 to produce g(t). As shown for each synthesized sine wave, the conversion from clock _Φ to time-based sine wave table 679 depends on the "ξ Resolution Select" input 675 and the available resolution selection. Table 676 is shown to indicate available table resolutions, ranging from a minimum of 12 points to 16-bit resolution with 65,536 data points, but is not limited to this. The number of data points in the sine wave lookup table 677 determines the maximum available resolution.

Waveform synthesis using universal primitive tables uses the same table to generate sine waves with the same or lower precision than the table's precision. For example, if Table 677 has a resolution of 96 points, or 3.75° increments, the same table can be used to generate 48, 24, or 12 point sine waves. The higher the resolution, the lower the synthesized frequency.

Sine waves of various frequencies can be synthesized by searching through all angles or by systematically skipping angles. For example, in the following table, using a symbol clock with frequency Φ _sym = 224,256 Hz in rows 00, 04, 08, 0C, 10... results in a 5,672 Hz sine wave, while selecting all rows of the table results in a 1,168 Hz sine wave.

Key selection and custom waveform synthesis

As previously mentioned, periodic waveform generation involves cascaded counters with fixed frequency multiples, so the waveform synthesizer is essentially "tuned" to a specific key. The user interface (UI) and resulting operation (UX, or user experience) are shown in Figure 55A. When the user selects the "CHOOSEAKEY" menu 701, various "musical" scales, "physiological" (reported medical frequencies), "custom" scales including manual input, and "other" key selection are facilitated. Provision is also included for returning to the "default" scale setting. The "Enter A Key" menu 702 appears, allowing for predefined scale note selections to be loaded onto the LED pads. When the "Music" setting is selected, the "Enter A Key" menu 702 allows for the selection of a key ranging from middle C to middle B at 261.626 Hz in the f _key to 493.883 Hz. If middle A is selected, as stored in table 703, then 703 transfers the 440 Hz "A" value to symbol clock counter 642 in accordance with Φ _sym /Φ _ref = (32ξf _key ) / (5 MHz), generating Φ symbol rates. Sine waves of various frequencies are synthesized based on this scale, e.g., f ₉ = Φ _sym /ξ ₉ . A table of exemplary frequencies by octave is shown below for a variety of tunings, here shown below for the musical keys F through C (https://en.wikipedia.org/wiki/scientific_pitch_notation). The scale shown is called "equal tempered" tuning.

Below is a table of typical frequencies per octave for various tunings from F#/G♭ to B. The scales shown are called "equal temperament."

Another option in the UI menu 701 is the "Other" selection, which allows other scales to be used to modulate the LEDs. These scales, which include the Pythagorean, Just Major, Mean Tone, and Werckmeister scales shown in the table below, share the middle C frequency of 261.626 Hz with the even scales, but differ in the relative frequency relationships between the 12 semitones spanning the octave. For example, in the even tempered scale, the A tone 4 above middle C is set at 440 Hz, but varies from 436.05 Hz to 441.49 Hz in other scales.

In custom mode, the user interface (UI) and resulting operation (UX user experience) are shown in FIG. 55B, where the user selects the "Select Key" menu 701 and then selects "Other" to open the "Select Scale" menu 700. The user then selects an alternate tuning from the menu—Pythagorean, Just Major, Meantone, or Werckmeister—to open a submenu 702 titled "ENTERAKEY." Once a key (note) is selected, a frequency is selected from the tuning table below and loaded into the "f _key select" key register 641, which is then transferred to the LED pad and ultimately loaded into the symbol clock counter 642. For example, if the key "A" is selected from the Werckmeister scale, then the value for "A" of 437.05 Hz is loaded into the symbol clock counter 642 according to Φ _sym /Φ _ref = (32ξf _key )/(5 MHz). Thus, the symbol counter generates a symbol rate Φ _sym = (32ξf _key ), from which sine waves of various frequencies based on this scale are synthesized. For example, f ₉ = Φ _sym /ξ ₉ . Since the key frequency f _key is used to generate Φ _sym , the entire nine-octave scale is adjusted accordingly. For example, if f _key = f ₄ is set to 437.05 Hz, then f ₅ = 2f ₄ = 874.1 Hz, f ₆ = 4f ₄ = 1,748.2 Hz, etc.

While scales vary across octaves, they all agree on the C frequency. For example, as shown for comparison, the C5 frequencies across the five octaves shown in the table below all agree at _f5 = 525.25 Hz = _2f4 . The notation used in Pythagorean, Just Major, and Mean-tone scales differ slightly from the Werckmeister and Mean-tone scales in their use of sharp #s and flat bs. While the exact tuning differences in PBT effectiveness have not been fully characterized, scientific studies have confirmed that the therapeutic effects of PBT treatment are clearly frequency-dependent. For Physio, the UI menu 701, item "Physiology," is selected, and the frequency scale is used for the f _key values reported in these medical studies to be therapeutically beneficial. Otherwise, use the custom button shown in Figure 56 instead. When a selection is made in menu 701, a UX response appears containing a custom "Enter Key" menu 7704. Upon entering, pressing the keypad digits, e.g., 444 Hz as shown, and the DONE button, the f _key 641 is loaded with the selection register custom key value 444 Hz, which is then transferred to the symbol clock generator 642. This value is then used to calculate the symbol clock rate using the symbol clock counter 642 according to the relationship Φ _sym /Φ _ref = (32ξf _key ) / (5 MHz), producing the output Φ _sym = (32ξf _key ).

The disclosed PBT system can also generate excitation patterns containing chords of three frequencies within the same octave, i.e., triads, with an optional additional frequency of a seventh or an octave above the root note of the chord. A block diagram of the algorithmic chord builder is shown in Figure 57A. Here, a tuner 590 configured according to an f _key selection 642 generates a symbol clock of frequency Φ _sym = (32ξf _key ), which is fed to a chord construction algorithm 680. The chord builder then uses well-known mathematical relationships to generate frequency components for various common chord types according to an "octave, chord, blend selection" input 681 selected from a chord builder menu 688. Triad chords include the selection of the octave of the root note on which the chord is constructed and the type of chord implemented (major, minor, diminished, augmented, or custom). Quadruple chords include the aforementioned triads with an additional note added: a seventh, a minor seventh, a major seventh, or an octave above the root. The relative amplitudes or "blends" of the component frequencies are also specified in Table 688, including the volumes of the chord's root note, its third, fifth, and optionally the seventh or note one octave above the root.

The chord construction algorithm 680 in operation uses tables 682b, 684, 683, and 682a to synthesize four scaled fractional sine waves for the symbol clock Φ _sym , driving up to four look-ups: the fundamental root of frequency f♪f, the third in frequency f♪3, the frequency f♪5, and the seventh or one octave above the root of frequency f♪t (depending on selection). The three or four frequencies are then blended with gains A♪f, A♪3, A♪5, and A♪t, respectively, according to digital gain amplifiers 685a, 686, 687, and 685b, and mixed in summing node 630 to produce g(t) (where ♪ represents a quarter note).

The exact frequencies of the notes in a chord depend on the value of the selected octave 681 and the selected key for the f _key 642, i.e., the tuning or key of the binary cascade counter. These synthesizer settings determine the frequency or root note, also known as the fundamental of the chord. The remaining notes in the chord are calculated as ratios to the chord's fundamental frequency according to the following table (https://pages.mtu.edu/~suits/chords.html), which describes the frequency ratios of common musical chords.

While the Chord Builder can be a library element used in predefined treatments and sessions, chords can also be created using a UI menu, such as the example shown in Figure 57B. The "Select a Chord" menu 705 allows users to choose from major, minor, diminished, augmented, diminished, custom, and seventh chords, including minor seventh and major seventh chords. Selecting a custom chord opens the BUILDACHORD menu 706, allowing the user to select the chord's octave, root note, third note (i.e., the next highest note), fifth note (i.e., the third-highest note), and optionally, whether to include a note one octave above the root. Once the root note is selected, the third, fifth, and +1 octave notes are arranged monotonically in ascending frequency, even if the notes extend to the next higher octave. Chord inversions of the second and third chords must be entered as custom chords, using the lowest-pitched note as the chord's root. The volume of the notes is equally weighted unless adjusted using the up and down arrows. Once the parameters are entered, after a timeout period or signaled by other means such as a double screen tap, they are formatted into a data table 688 and transferred to the chord construction algorithm block 680 within the intelligent LED pad, where the sine wave ultimately resides. A lookup table 677, digital gain stage 678, and mixer 630 create g(t). If another menu item is selected from the "Choose Chord" menu 705, another submenu (not shown) opens, allowing the user to select the octave and relative amplitude mix of the constituent frequency components. However, the submenu does not allow the user to change the notes, as it precisely defines the relative frequencies present in the minor, major, diminished, etc. chord.

Returning to the synthesizer block diagram in Figure 44, regardless of the synthesized waveform or how it was created, for the PWM generator 555 to implement a PWM duty cycle, it must process the waveform g(t) to create f(t) 553 by limiting its range from 0.000 to 1.000. The coefficient transformation ΨP[f(t)] is necessary to create the synthesizer file 488 because the maximum duty cycle of a PWM-modulated pulse is 100%, and therefore, one per full clock cycle, PWM representation of data greater than 1.000 is impossible. Such a PWM transformation ensures that ΨP[f(t)] ≤ 100%, and therefore 0.000 ≤ f(t) ≤ 1.000. The auto-ranging operation 584 averages the function g(t) while limiting the range of the data and f(t) to the unit function range (i.e., between 0.000 and 1.000).

An example of this function is shown in Figure 58A. The sum of sine waves 662, 663, and 664 results in code 669. While each sine wave spans the full range from 0.000 to 1.000, the sum of the sine waves for code 669 does not span the full range of the identity function. Therefore, while the mathematical mean of the chord, specifically 0.5, remains constant, the periodic, time-varying function does not extend to the full range of 0.5 ± 0.5. As shown in Figure 58B, code 669 extends only from 0.13 to 0.87, reestablishing 74.4% of its full range. To increase the amplitude of the time-varying component, the averaging function is amplified by the scalar A _α . Setting _{A α} = 1.344 increases curve 669 to its full range, as shown in code 689. To prevent a shift in the mean value of the function, a correction term, 0.5 (1 - A _α ), is included to keep the function centered at 0.5 to prevent clipping. The result is a unit function f(t) with a mean value of 0.5, a full-scale periodic function with the same dynamic time-varying frequency content as the synthesized waveform g(t).

FIG. 59 illustrates the process by which PWM generator function 555 converts unit function f(t) 553 into synth output file 488, which describes PWM waveform G _synth (t) 490. As shown, function table 554 contains a description of the value of function f(t) versus time tΦ at each time increment. For example, at tΦ=5 μs, function f(t)=0.5 and remains at that value until tΦ=10 μs, where the function value changes to f(t)=0.8. The output of the conversion, ΨP[f(t)], is stored in PWM table 489 at time t. At _tΦ =5.00 μs, the LED goes on, and at time tΦ=5.10 μs, the LED goes off, and then turns on again. For a duration of 0.10 μs and T = 1/Φ _x the LED goes from 5.00 to 5.10, then from 5.00 to 5.20 until the LED is turned on again, or from 5.00 to 5.20, or 0.20 microsecond duration pulse duty ratio D = (ΔtΦ/T) = (10 μs/20 μs) = 0.50 or 50%, then the duty factor is equal to the function f(t) = 0.5 during this interval and until time tΦ = 10 μs the duty ratio switches to 0.8 or 80%. The resulting synth output file 488 is shown graphically by PWM pulse string 675.

Examples of PWM output 490 using the transform ΨP[f(t)] are shown for a variety of non-sinusoidal functions in Figure 60, including a PWM bit stream 670 for a constant function 560, a PWM bit stream 671 for a sawtooth function 561, and a PWM bit stream 672 for a trigonometric function 562. The same PWM transform ΨP[f(t)] can be used to encode audio samples of any kind, including simple tones like triangles, strings like guitars or violins, complex tones like symbol crashes, and music.

PWM player operation

Reconsidering the block diagram of Figure 43, the output G _synth (t) = ΨP[f(t)] of the waveform synthesizer 483 is input to the PWM player 484. The PWM player then combines G _synth (t) with the waveform G _pulse (t) to produce a pulse train 492. The function of the PWM player 493 is twofold.
Dynamically controlled duty cycle DPWM to generate the audio spectrum PWM pulse train G _pulse (t).
Performing dynamic "gating", ie, blocking or passing the contents of G _pulse (t) based on the state of G _synth (t).

The truth table for the above function can be written as logical pseudocode as follows:

Because G _pulse (t) is composed of pulses in a PWM string, the waveform alternates between high and low logic states. Specifically, whenever function G _pulse (t) = 1, i.e., whenever PWM pulse 492 is in its high or logic "1" state, the digital state of G _synth (t) is exactly reproduced at the output of PWM player 484. For example, when G _pulse (t) = 1, the output of PWM player 484 is high, and when G _synth (t) = 0, the output of PWM player 484 is low. However, whenever function G _pulse (t) = 0, i.e., whenever PWM pulse 492 is _in its low or logic "0" state, the digital state of G _synth (t) is forced to zero, and the next state is ignored. Input G _synth (t). Logically, this function is the same as an "AND" gate. Mathematically, it is equivalent to a digital multiplication where the output of PWM player 492 is given by the product G _synth (t) · G _pulse (t). The actual implementation of PWM player 492 may be achieved in hardware, software/firmware, or some combination thereof.

As shown schematically in FIG. 61A, PWM player 484 comprises a PWM clock counter 710, a pulse width modulator 711, digital inverters 712a and 712b, and an AND gate 713. Inputs to PWM player 491 include clock reference Φ _ref , synth output 488, and PWM player parametric 491. In operation, reference clock Φ _ref =5 MHz provides a time reference with period T _ref =0.20 μs as input to PWM counter 710, generating PWM clock Φ PWM =20 kHz. Pulse width modulator 711, with period T _PWM =5 μs and 250 times longer than the reference clock Φ _ref period, generates a sequence of PWM pulses 492 of varying duration t _on =D _PWM T _PWM created according to table 714 defined in PWM player parametric input 491. For example, from 0 to 180 seconds in table 714, G _pulse (t) is pulsed at a frequency of 2,836 Hz with a 60% duty factor, after which the pulse frequency changes to 584 Hz. At time t = 360 seconds, the pulse frequency returns to 2,836 Hz. From the perspective of pulse train 492, the period T _PWM during the interval from 0 to 180 seconds = 0.43 ms on-time; the portion of the period during which the pulse is in the high state is given by T, t _on = D _PWM T _PWM = (60%)(0.43 ms) = 0.26 ms.

The off portion of the pulse is given by t _off = T _PWM - t _on = (0.43 ms) - (0.26 ms) = 17 ms. If the pulse frequency is changed to 584 Hz, the period increases to 1.712 ms and the on time becomes 1.027 ms. Thus, pulse string 492 is dynamically generated by pulse width modulator 711 according to the dynamic conditions specified in table 491. The output of PWM player 484, shown as gated PWM pulse string 493, has waveform 494 output from the waveform synthesizer embedded within it.

The operation of pulse width modulator 711 essentially involves two successive counters, one to count the on time and one to count the off time, with G _pulse (t) = 1 for the t _on interval and G _pulse (t) = 0 for the t _off interval. In logical pseudocode, the operation of pulse width modulator 711 can be described by defining the following subroutine:

The above subroutine titled "Pulse Width Modulator" is a software pseudocode description that performs the same function as block 711, i.e., a loop of interval Δt containing alternating digital pulses at a logic 1 state for durations t _on and a. It repeats a logic 0 state for a duration (T _PWM -t _on ) until the clock count T _ref = 1/Φ _ref exceeds Δt. The variables [Δt, T _PWM , t _on ] are loaded into the subroutine from a sequence defined in table 714 or PWM Player Parametric 49, as shown in the following exemplary executable pseudocode, where the table lookup is specified by the following values: (row, column) pair, i.e., (row, column), where row is a defined variable:

As explained above, the executable pseudocode above repeatedly reads table 714 and uses its duration Δt, PWM pulse period T PWM , and PWM pulse on time t _on arguments to load data into the pulse width modulator subroutine call, incrementing the row number after each loop. For example, if starting at row 0, Δt is calculated by the difference in time between the entries in the second and first rows of the first column of the table. That is, T PWM (2,1) = 180 seconds, T PWM (1,1) = 0, so for the first loop of the code, Δt = 180 seconds. Similarly, in the first row and fourth column, the PWM period data is T _PWM = T PWM (1,4) = 0.43 ms, and in the first row and fifth column, the PWM period data is t _on = T on (1,5) = 0.26 ms. At the end of the loop, the row number is incremented from 1 to 2, so new data is read from the second row. Here, Δt = [Hyper(3,1) - Hyper(2,1)] = [360 s - 180 s] = 180 s, T _PWM = Hyper(2,4) = 1.712 ms, and t _on = Hyper(2,5) = 1.027 ms. This process continues until a null entry for T _PWM is detected, that is, T _PWM = Hyper(row,4) = 0. At that point, program execution terminates. Thus, as shown, the functions of PWM player 484 and pulse width modulator 711 can be implemented using software or hardware, or some combination thereof.

For example, the functionality of PWM player 484 is represented schematically in FIG. 61B, which includes set/reset flip-flop or S/R latch 720, _{t on} and t _off counters 721 and 722, and gates 723 and 724, inverter 725, start-up resistor 733, and t _on and t _off registers 726 and 727. In operation, start-up resistor 733 pulls up the S input of S/R latch 720, which sets the Q output to a logic-on or "1" state. The rising edge of this 0-to-1 logic transition triggers the load function of t _on of counter 721, copying data from t _on register 726 into the counter. The logic-high state of the Q output is also an input to AND gate 723, and its inverse state, the output of inverter 725, presents a logic "0" input to AND gate 724.

As such, clock pulses from clock Φ _PWM are routed through AND gate 723 to ton counter 721 but are blocked from reaching toff counter 722 by AND gate 724. Thus, toff counter 721 counts down for duration t. During that countdown, the output of toff counter 721 remains in a logic "0" state and has no effect on S/R latch 720. Simultaneously, the lack of clock input activity of _toff counter 722 suspends it. Referring to the associated timing diagram, during this interval from _Tx to ( _Tx + _ton ), PWM clock Φ _PWM 728 continues counting, reset signal 729 comprising the R input to S/R latch 720 remains low, set signal 730 comprising the S input to S/R latch 720 remains low (activation pulse not shown), and output G _pulse (t) 731 remains high.

When t _on counter 721 completes counting down the interval t _on , the counter's output momentarily goes high, as indicated by reset pulse 734. Simultaneously, the falling edge of the Q output generates a rising edge at the output of inverter 725, triggering the loading of t _off register 727 data into tough counter 722. A logic high input to AND gate 724 allows the Φ _PWM clock to be routed to tough counter 722. Referring to the associated timing diagram, during this interval from (T _x + t _on ) to (T _x + T _PWM ), PWM clock Φ _PWM 728 continues counting, reset signal 729, which includes the R input to S/R latch 720, remains low (except for reset pulse 734 at the beginning of the interval), set signal 730, which includes the S input to S/R latch 720, remains low, and output G _pulse (t) 731 remains low. When the counter counts down to zero after an interval of t _off , its output generates a short set pulse 732 which switches the Q output of S/R latch 720 to a logic "1" state and loads the current value from t register 726 into t _on counter 721, restarting the whole process.

As shown, G _pulse output 731 switches from a logic-high state of duration t _on = D _PWM T _PWM to a logic-low state of duration t _off = (1-D _PWM ) T _PWM . Each time set pulse 732 is triggered, the current value of ton register 726 is loaded into ton counter 721. Similarly, each time reset pulse 734 is triggered, the current value of tough register 727 is loaded into t _off counter 722. In this way, PWM player parametric file 491 dynamically changes the PWM player frequency and duty factor to produce the same waveform as a software-equivalent implementation. Note that resistor 733, used to pull the S input to S/R latch 720 high during startup, has high resistance and is unable to overcome the logic-low state output from tough counter 722 once startup is complete and power to the circuit has stabilized.

In conclusion, in a PWM player, the frequency f _PWM and the corresponding duty factor D _PWM vary over time according to a specific playback file, thereby defining a PWM sequence of pulses with varying tone and duty durations. Note that the pulse frequency f _PWM = 1/T _PWM of the pulse width modulator is lower than the PWM clock Φ _PWM = 20 kHz used to drive the modulator. Furthermore, the PWM frequency f _PWM is much lower than the oversampled clock Φ _sym used by the PWM generator ΨP[f(t)] in the waveform synthesizer block. In other words, 1/Φ _sym >> 1/Φ _PWM ≥ f _PWM .

LED driver operation

The third stage of the LED player in a distributed PBT system is the LED driver circuit. Referring to Figure 43, this function is where the LED driver 485 converts its input G _synth (t) · G _pulse (t) along with an optional time-dependent reference current 496 into one or more analog control signals, i.e., LED drive stream 497. An aggregate signal equal to αI _ref (t) · G _synth (t) · G _pulse (t) is then used to control the current in multiple LED strings as shown in exemplary waveform 498.

Details of the LED driver's operation are shown in the block diagram of LED driver 485 in Figure 62. While this diagram shows only two PWM pulse string inputs, IN1493 and IN2750, and two outputs for driving LED strings 743a and 743b, those skilled in the art of PBT will understand that any number of composite waveforms can be used. For example, 1 to 16 may be required, and the number of LED strings can vary from n = 1 to 36 strings (or even more for larger devices), while for smaller LED pads, the number of strings could range from 8 to 24. It is also understood that the number of series-connected LEDs, "m," can vary from string to string, as long as the entire series connection does not require a voltage greater than +VLED to operate properly.

The LED driver 485 includes two buffers per input, e.g., IN1 inverters 744a and 744b, and IN2 inverters 745a and 745b, as well as a PWM clock counter 710, an LED pad controller 747, and multiple channels of outputs, _ILED1 , _ILED4 , ..., each channel including a controlled current source or sink, and optionally a D/A converter and associated _Iref data register. For example, as shown, the _ILED1 output includes a controlled current sink 740 driving LED string 743a, a D/A converter 741 generating reference current _Iref1 , and an associated _Iref1 data register 742a. Similarly, the _ILED4 output includes a controlled current sink 740d driving LED string 743d, a D/A converter 741d generating reference current _Iref4 , and an associated _Iref4 data register 742d. An optional crosspoint matrix 746 is used to dynamically assign, i.e., assign inputs IN1 and IN2 etc. to outputs _ILED1 , _ILED2 , _ILED3 , _ILED4 , _ILED5 , ... as needed. Apart from the PWM waveform inputs, the _Gsynth (t) and _Gpulse (t) LED drivers 485 also require an LED driver parametric file 749 and a reference clock _Φref .

In operation, input waveforms are mapped to output channels that dynamically control the current in their assigned LED strings. For example, waveform 493 is input to IN1 and then mapped via crosspoint switch 746 to the digital En1 input to current sink 740a and other channels (not shown). As detailed in the accompanying legend, a filled circle on the crosspoint switch indicates a closed switch, or connection, while a hollow circle indicates no connection, or open circuit. Similarly, waveform 750 is input to IN2 and then mapped via crosspoint switch 746 to the digital En2 input to current sink 740d and other channels (not shown). Simultaneously, analog signal _Iref1 is provided to current sink 740a and analog signal _Iref4 is provided to current sink 740d, synchronized by PWM clock Φ _PWM . The currents _Iref1 and _Iref4 are loaded into digital values set by _Iref1 and _Iref4 in registers 742a and 742d, respectively, by D/A converters 741a and 741d. The resulting waveforms 748a and 748d control currents _ILED1 = _αIref1 and _ILED4 = _αIref4 . The design, implementation, and operation of current sinks (or alternatively current sources) are described in the examples of Figures 20A through 23C. The LED driver function can also be specified and executed in two steps using software. For example, first map the inputs to the outputs.

While it is possible to dynamically change this mapping, it is likely that the mapping will be performed only once per treatment and will remain unchanged throughout the treatment. Often, only a single input is used. The current executable code for each channel can be fixed to a constant value.

During manufacturing calibration, the error terms or curves Icalib are stored in non-volatile memory for each channel. For example, Icalib1 = 1.04 mA, Icalib4 = -0.10 mA, and Icalib5 = 0.90 mA. The LED pad also stores the value of the mirror ratio α. For example, if α = 1/β = 1,000,1,000, then an output current of mA requires a corresponding reference current of microamperes. Before playback begins, the pad μC calculates and stores the value of _Iref for each channel.

The I _ref value is stored in digital equivalent form in I _ref registers 742a, 742d, 742e, etc. in volatile memory prior to program execution. If the value of the target LED current changes, the register value can be overwritten before program execution or dynamically overwritten "on the fly" as treatment progresses. For example, using executable pseudocode, a dynamic LED drive might include the following:

During runtime, the value of _Iref for each channel is set by [ _ILED + _Icalib ]α, where _ILED1 = 'Drive' (row, 2), _ILED4 = 'Drive' (row, 5), etc. Column 2 contains the LED current drive data for _ILED2 , column 5 contains the _ILED4 data, etc. If column 2 contains the LED drive current data for ILED2 and ILED2, column 5 contains the ILED4 data. The column values for ILED4 and ILED3 contain the data used to define various intervals for treatment, e.g., delivering 20 mA for up to 540 seconds and then delivering 23 mA.

If all channels have the same current, you can remove the channel-specific columns from the table and replace them with a single column, as shown below.

A program can also call a function instead of a table, as in the Treat Headache example.

In the above example, a 20 mA sine wave is generated by a mathematical function of a reference current _ILED (t) with a defined frequency, e.g., 5.5 Hz, using the _Φref clock (or an optional multiple thereof). At each instance of the desired output current _ILED (t), the reference current _Iref1 , corresponding to α, is converted to equality by the mirror ratio α, corrected for each channel using the calibration table data, in registers 742a, 742b, and 742e, as instructed "set t = t + (1/ _Φref )." Each loop at time t increments the time (1/ _Φref ) and stores the sum back in variable t, thereby overwriting the previous value. Thus, variable t acts as a clock that is incremented with each loop of the program. The clock continues counting, repeatedly generating a sine wave with a fixed period of TLED = 1/fLED until the terminal condition t ≥ tend is met.

Distributed PBT system LED player

43, the arrangement of waveform synthesizer 483, PWM player 484, and LED driver 485 generates LED drive stream 497 during playback operation, but waveform synthesis is performed at a clock frequency Φ _sym , which is significantly above the audio frequency spectrum, where Φ _sym >> 20 kHz, while the PWM clock Φ _PWM used by PWM player 484 and the LED clock Φ LED used by LED player 485 operate in the audio spectrum Φ PWM ≦ 20 kHz and Φ LED ≦ 20 kHz. In summary, the operation of the LED player includes the following:
Generate a time-dependent analog unit function f(t) mathematically using an unit function generator or using an oversampled lookup table-based primitive processor.
Transform the unit function f(t) into a PWM pulse stream using the transformation G: G _synth (t) = ΨP[f(t)].
Generate an audible spectrum PWM pulse string G _pulse (t).
• Gating, or logical ANDing, G _synth (t) with the PWM pulse string G _pulse (t) to produce the multiplicative identity function output G _synth (t)·G _pulse (t).
I _LED =αI _ref (t)·G _synth (t)·G _pulse (t) is generated by the unit function output of the driven LED player along with the time change of the analog current αI _ref (t).

Figures 63 to 65 show examples demonstrating the versatility of the disclosed LED player with a variety of waveforms.

Figure 63A shows a constant f(t) = 1 function 761, resulting in a constant time-invariant G _synth waveform 762, where ΨP[f(t)] = 100%. The constant ΨP[f(t)] is then multiplied by a PWM pulse string 773a with D = 50% to generate a pulse string 774a containing G _synth (t) · G _pulse (t). When multiplied by a constant reference 781a to generate 20 mA, the resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) consists of a 20 mA peak square wave 802a with a 50% duty factor and an average current of 10 mA.

Figure 63B shows a constant f(t) = 1 function 761, resulting in a constant time-invariant G _synth waveform 762, where ΨP[f(t)] = 100%. The constant ΨP[f(t)] is then multiplied by a PWM pulse string 773b with D = 20% to produce a pulse string 774b with a value G _synth (t) · G _pulse (t). When multiplied by a constant reference 781b to produce 50 mA, the resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) consists of a 20 mA peak square wave 802b with a 20% duty factor and an average current of 10 mA.

Figure 63C shows a constant f(t) = 1 function 761, resulting in a constant time-invariant G _synth waveform 762, where ΨP[f(t)] = 100%. The constant ΨP[f(t)] is then multiplied by a PWM pulse string 773c, with D = 95%, to produce a pulse string 774c containing G _synth (t) · G _pulse (t). Multiplying by a constant reference 781c to produce 10.6 mA, the resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) consists of a 10.6 mA peak square wave 802c with a 95% duty factor and an average current of 10 mA.

Figure 63D shows a constant f(t) = 1 function 761, resulting in a constant time-invariant G _synth waveform 762, where ΨP[f(t)] = 100%. The constant ΨP[f(t)] is then multiplied by a PWM pulse string 773a with D = 50% to produce a pulse string 774a with value G _synth (t) · G _pulse (t). When multiplied by a step reference 781d to produce a 20 mA step up from 25% to 25 mA, the resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) consists of a 20 mA peak square wave 802c with a 50% duty factor. The 10mA average current is stepped up to a 25mA peak square wave with a 50% duty factor and an average current of 112.5mA.

Figure 63E shows a constant f(t) = 1 function 761, resulting in a constant time-invariant G _synth waveform 762, where ΨP[f(t)] = 100%. The constant ΨP[f(t)] is then multiplied by a constant value 771 to produce a constant value 772 with D = 100%, where G _synth (t) · G _pulse (t) = 100%. It is then multiplied by a pulse reference 782 to produce a 20 mA square wave. The resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) consists of a 20 mA peak square wave 802a with a 50% duty factor and an average current of 10 mA.

Figure 63F shows a constant f(t) = 1 function 761, resulting in a constant, time-invariant G _synth waveform 762, where ΨP[f(t)] = 100%. The constant ΨP[f(t)] is then multiplied by a constant value 771 to produce a constant value 772 with D = 100%, where G _synth (t) · G _pulse (t) = 100%. This is then multiplied by a sine wave reference 783 to produce a 20 mA sine wave. The resulting waveform, I _LED = αI _ref (t) · G _synth (t) · G _pulse (t), consists of a 20 mA sine wave 803a with an average current of 10 mA.

Figure 63G shows a constant f(t) = 1 function 761, resulting in a constant, time-invariant G _synth waveform 762, where ΨP[f(t)] = 100%. The constant ΨP[f(t)] is then multiplied by a constant value 771, producing a constant value 772 with D = 100%, where G _synth (t) · G _pulse (t) = 100%. This is multiplied by analog-to-digital samples 784a to produce a plucked string with a peak value of 20 mA. The resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) consists of 20 mA samples 804a with an average current of 10 mA.

Figure 63H shows a constant f(t) = 1 function 761, resulting in a constant, time-invariant G _synth waveform 762, where ΨP[f(t)] = 100%. The constant ΨP[f(t)] is then multiplied by a constant value 771, producing a constant value 772 with D = 100%, where G _synth (t) · G _pulse (t) = 100%. This is multiplied by analog-to-digital samples 784b to produce a cymbal crash with a peak value of 20 mA. The resulting waveform, I _LED = αI _ref (t) · G _synth (t) · G _pulse (t), consists of 20 mA samples 804b with an average current of 10 mA.

Figure 64A shows a sine function 763, f(t) = sin(ft), resulting in G _synth = ΨP[f(t)] as a continuously varying PWM pulse string waveform 764 with a defined period T _synth . The PWM string ΨP[f(t)] is then multiplied by a constant value 771, with D = 100%, to generate a digital pulse string containing G _synth (t) · G _pulse (t), which contains a PWM representation of a sine wave 775. When multiplied by a constant reference 781a to generate 20 mA, the resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) consists of a 20 mA peak sine wave 803a at 50% of the 10 mA average current.

Figure 64B shows a sine function 763, where f(t) = sin(ft), resulting in G _synth = ΨP[[f(t)] as a continuously varying PWM pulse string waveform 764 with a defined period T _synth . The PWM string ΨP[[f(t)] is then multiplied by a constant value 771, with D = 100%, to produce a digital pulse string G _synth (t) · G _pulse (t), which contains a PWM representation of a sine wave 775. Multiplied by step reference 781d to produce 20mA stepped up by 25% to 25mA, the resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) comprises a 20mA peak sine wave 803b at 50% average current of 10mA, stepped up to a 25mA peak sine wave at 50% average current of 112.5mA.

Figure 64C shows a sine function 763, where f(t) = sin(ft), resulting in G _synth = ΨP[f(t)] as a continuously varying PWM pulse string waveform 764 with a defined period T _synth . The PWM string ΨP[f(t)] is then multiplied by a constant value 771, with D = 100%, to produce a digital pulse string G _synth (t) · G _pulse (t), which contains a PWM representation of a sine wave 776. When multiplied by a constant reference 781a to produce 20 mA, the resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) consists of a 20 mA code of a sine wave 803c with a 50% average current of 10 mA.

Figure 64D shows a sawtooth wave 763 converted into a periodically varying PWM pulse string waveform 767 having a period T _synth defined by G _synth = ΨP[f(t)]. The PWM string ΨP[f(t)] was then multiplied by a constant reference 781a to generate a 20 mA containing PWM representation 777 of the G _synth (t) · G _pulse (t) sawtooth wave, multiplied by a constant value 771 with D = 100% production digital pulse train; the resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) contains a 20 mA sawtooth wave 804 with an average current of 50% of 10 mA.

Figure 64E shows an audio sample of a periodically varying PWM pulse sine waveform 768a with a period T _synth defined by G _synth = ΨP[f(t)]. The PWM string ΨP[f(t)] is then multiplied by a constant value 771 to produce a digital pulse string G _synth (t) · G _pulse (t) containing a PWM representation of a sawtooth wave 779a with D = 100%. When multiplied by a constant reference 781a to produce 20 mA, the resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) consists of a 20 mA audio sample 805a with a 50% average current of 10 mA.

Figure 64F shows an audio sample of a guitar string 768a converted into a PWM pulsed string waveform 769a that varies periodically with a defined duration by G _synth = ΨP[f(t)]. The PWM string ΨP[f(t)] is then multiplied by a constant value 771 of D=100% to generate a digital pulse string G _synth (t) · G _pulse (t), which consists of a PWM representation 779a of the guitar string. When multiplied by a constant reference 781a to generate 20 mA, the resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) consists of a 20 mA audio sample 805a with a 50% average current of 10 mA.

Figure 65 shows a sine function 763, where 763f(t) = sin(ft), resulting in G _synth = ΨP[f(t)] as a continuously varying PWM pulse string waveform 764 with a defined period T _synth . The PWM string ΨP[f(t)] is then multiplied by a PWM pulse 771d of fixed period D = 67% to produce a digital pulse string G _synth (t) · G _pulse (t), which comprises a chopped PWM representation 778 of a sine wave gated by a low frequency PWM pulse. Multiplied by a constant reference 781a to produce 30 mA, the resulting waveform I _LED = αI _ref (t) · G _synth (t) · G _pulse (t) consists of a 30 mA code of a sine wave 803e with an average current of 10 mA.

To perform a PBT procedure, a specific LED playback file is first downloaded from the PBT controller to the LED pad, followed by the execution of the LED player. Once the LED player is downloaded, the LED player does not need to be reloaded each time a new treatment is selected. As long as the player remains in the LED pad's volatile memory, new playback files can be repeatedly loaded to perform new treatments or sessions. However, turning off the PBT system or disconnecting the LED pad from the PBT controller erases the LED player software from the LED pad's volatile memory and requires reinstallation on the pad before an LED playback file can be executed to perform a treatment or session. While storing LED player files in non-volatile memory avoids the program wipe issue, for security reasons, it is recommended that programs be written to volatile memory such as SRAM or DRAM rather than non-volatile EEPROM or flash. In this way, any attempt by a hacker to extract the program is thwarted by the immediate loss of executable code when power is shut off, as any attempt to reverse engineer the program's contents is thwarted by the immediate loss of executable code.

As shown in Figure 66, an LED playback file 830 containing payload data 831 is transferred to volatile memory 832 where the payload is then extracted into waveform primitives 487, uncompressed, PWM waveform synthesizer parametrics 486, loaded into waveform synthesizer 833, PWM player parametrics 491, and loaded into LED driver 835. An example of the contents of payload data 831 is shown in Figure 67. The contents of the waveform primitives 487, waveform synthesis parametrics 486, PWM player parametrics 491, and LED driver parametrics 749 are included in the waveform sizer parametrics 486, which include the information required to execute a particular treatment or session, which is an instruction file. A typical instruction file for waveform synthesis contains the following:
- The waveform synthesis method employed in the file, either functional synthesis or primitive synthesis.
Program tuning (key), i.e., setting the f _key register for synthesis. Available keys for PBT synthesis consist of predefined binary multiples of notes in the fourth octave, and generated harmonics spanning the audio spectrum from the ninth octave to the first octave. By default, the scale is uniformly tuned. The "Other" submenu includes alternative tunings such as Werckmeister, Pythagorean, Just Major, and Mean Square Tone scales. The physiological scale "Physio" is based on empirically derived scales derived from observation. The "Custom" UI/UX allows users to manually set the f _key value as a fourth octave frequency (entered in Hertz, not notes) and pass this frequency to the key register.
The waveform sequence to be synthesized, including the duration of each waveform "step" in the synthesis. The program exits with an exit code indicating the process or session is complete.
If using Function Synthesis, the mathematical formula for each function and its frequency f. Periodic waveforms available using Function Synthesis include constant, sawtooth, triangle, and single-frequency sine waves.
If primitive synthesis is used, each primitive subroutine calls X with frequency f and resolution ξ _x . The primitive's playback subroutine. Available primitive-based waveform subroutine calls include constant, sawtooth, triangle, sine, or audio sample. Primitive-based synthesis of sine wave chords is also available using the "Chord Builder" subroutine.
Chord Builder subroutines include how to create chords and specify the octaves and notes that exist. Chord Builder algorithms include "octave" synthesis and "3/4" chord synthesis.
With octave synthesis, any chord can be described by its component octave "octave" numbers (a number between -1 and 9 representing the frequency _fx created according to the _fx register setting) and the corresponding primitive resolution _ξx and blend _Ax for each octave. The three-quarter chord builder allows blending of three or four fixed-resolution sinusoidal notes across an octave with adjustable amplitude set by the gain Ax. Available chord triads include major, minor, diminished, and augmented, and each triad includes an optional fourth note +1 octave above the chord's root note. Alternatively, a fourth note can be added to form a seventh chord, specifically a quad-note chord with a seventh, major, and minor seventh structure. "Custom" chords allow you to generate three-note chords that span an octave, even dissonant ones, and choose a fourth note +1 octave above the chord root note. All Chord Builder outputs can be scaled to increase the chord's periodic amplitude by a digital Gauss A _alpha , without shifting the 0.5 mean value of the unit function.
All outputs of the waveform synthesizer represent the identity function, i.e., analog values between 0.000 and 1.000 are converted to a PWM pulse string with a duty factor between 0% and 100%. Synthesized waveforms outside this range are truncated.

During operation, only the waveform primitives 486 required by the playback file specified by the waveform synthesizer parametric 487 are downloaded to the LED pads. The downloadable primitive library 487 includes a selection of sine wave primitives at various resolutions ξ, using, for example, 24, 46, 96, 198, or 360 points or 16-bit resolution. The exemplar library also includes 24-point descriptions of triangle and sawtooth waves, although other resolutions may be included without limitation. Other library components, for example for ξ = 96, include code containing two sine waves spaced one octave apart by f and 2f, two octaves apart by f and 4f, or four octaves apart by f and 16f, or five octaves apart by f and 32f.

Other options include three-octave chords, such as [f, 2f, 4f], which span two octaves; [f, 2f, 8f] or [f, 4f, 8f], which span three octaves; and [f, 2f, 16f], [f, 4f, 16f], or [f, 8f, 16f], which span four octaves. Other triads include major, minor, diminished, and augmented chords, such as [f, 1.25f, 1.5f], [f, 1.2f, 1.5f], or [f, 1.2f, 1.444f]. Triads can be changed into quad chords by including a note one octave above the root.

The PWM player parametric file 491 contains settings for either constant or pulse mode. In pulse mode, the playback file consists of a series of PWM frequencies f _PWM and corresponding duty factors D _PWM versus playback time, thereby defining a PWM sequence of pulses of varying tone and duty duration. Note that the pulse width modulator pulse frequency f _PWM is lower in frequency than the PWM clock Φ _PWM = 20 kHz used to drive the modulator. Consequently, in PWM player operation, the PWM frequency f _PWM is not fixed; it is varied by the playback program specified in the PWM parametric file 491. While the frequency f _PWM can be as high as the clock Φ _PWM , in most cases the frequency f _PWM will be lower, so that f _PWM ≤ Φ _PWM . Furthermore, the frequency f _PWM is in the audible spectrum, far below the supersonic range of the oversampled clock Φ _sym used by the PWM generator ΨP[f(t)] in the Waveform Synthesizer block. This means that mathematically, f _PWM ≦Φ _PWM
<< 1/Φ _sym .

In the LED Driver Parametric 749, the unit function digital PWM inputs INx are mapped to current sink enables Eny. For example, input IN1 maps to current sink enable En4 for channel 4, and input IN2 maps to current sink enables En1 and En5 (not shown) for channels 1 and 5. The value of I _ref for each channel is set by the output of its corresponding D/A converter. This can include a constant, a periodic function, or audio samples. Alternatively, a single D/A converter can be used to supply the same function or constant value for the reference current of all output channels.

P layback begins in decentralized PBT system

After downloading the LED player and LED playback files to the LED pad, playback is enabled by a start signal 840 and PBT system timing controls, which can be implemented in software or using the example circuit of Figure 68. Figure 68 includes a start/stop latch 842, which includes a set/reset or S/R type flip-flop, an interrupt latch 843, a PBT system clock counter 640, a startup one-shot 848, logic AND gates 845 and 846, and logic OR gates 846 and 847. The two-input AND gate 845 serves as a system clock enable for the oscillator Φosc to the LED player and is gated by the start and control signals 840 and 841, as well as various interrupts, specifically the blink timer timeout 844, watchdog timer timeout 845, or overtemperature flag 846.

Upon activation, one-shot 848 generates a pulse that immediately drives the output of OR gate 846 high. Simultaneously, the one-shot signal triggers the set input S of interrupt latch 843 and its output Q high. When user input "start" 840 is selected, it generates a positive-going pulse that sets the output Q of start/stop latch 846 high. When the Q outputs of both start/stop latch 846 and interrupt latch 843 are set high, AND gate 845 is enabled. Therefore, oscillator Φosc is sent to the PWM player as clock _Φsys and divided by counter 640 as reference clock _Φref .

Selecting "Pause" 841 generates a pulse that resets the output of Start/Stop Latch 842 to zero, pausing playback. Playback remains latched off until "Start" 840 is selected, which cancels the pause command. As such, Start/Stop Latch 842 starts and stops program execution. If an interrupt occurs for any reason, i.e., if either input to OR Gate 647 goes high, the output of the OR Gate also goes high, resetting the output Q of Interrupt Latch 843 to zero. If the Q output is low and the output of AND Gate 846 Download 845 is low, the clock Φosc is disconnected from the LED output, pausing therapy. This situation continues until the cause of the interrupt is corrected, the input to OR Gate 647 is reset low, and a system restore pulse is sent to the S input of Interrupt Latch 843. For example, if an overtemperature condition occurs, the Temperature Flag 846 goes high, disabling operation of the LED pad until normal temperature returns and the fault flag is reset.

A unique safety feature of the disclosed distributed PBT system is the blink timer. This timer runs within the intelligent LED pad itself and is independent of the PBT controller. At regular intervals within the pad μC, for example, every 20 or 30 seconds, a program counter suspends operation to execute an interrupt service routine (ISR). During this interval, the blink timeout flag is set to logic 1 while the LightPadOS software performs safety checks on the LED pad's electrical connections, priority message or file updates, file parity checks, and so on. Upon completion of the blink interrupt routine, the blink timeout is reset to zero, the hardware watchdog timer is reset, and program execution returns to the main routine. After the ISR is completed, the pad μC generates a system restore pulse to interrupt latch 843, resuming program operation. If the software freezes for any reason, the program does not resume operation and the pad's LED string remains off. Otherwise, the LED pad resumes operation after a defined interval (for example, 2 seconds).

Another failure mode occurs when the software freezes while the LED is on and emitting light. If the condition persists, the LED may overheat, posing a risk of burns to the patient. To prevent a dangerous condition from occurring, a hardware watchdog timer (whose operation is independent of the software) counts down in parallel with the software program counter. If the software timer freezes in the on state, it will not reset the watchdog timer, which will time out and generate a blink timeout interrupt 844, halting operation of the PBT system until the fault condition is resolved.

In this manner, the disclosed distributed PBT system can be used to remotely control the operation of LED pads. Furthermore, the methods disclosed herein can be adapted to control multiple intelligent LED pads simultaneously from a common PBT controller.

Component communication over PBT distributed systems

Implementing the necessary communications between components in a distributed PBT system requires complex communications networks and proprietary protocols designed to accommodate a combination of real-time and file-based data transfer, some of which are linked to safety systems. Pursuant to FDA regulations, safety is a primary design consideration for medical devices. In a distributed system, this concern is exacerbated by the autonomous operation of components. Even if inter-device communications in a distributed PBT fail or are interrupted, safety systems cannot malfunction. The topics of communications, safety, sensing, and biofeedback are described in detail in a related patent, titled "Distributed Photobiomodulation Therapy Device, Method, and Communication Protocol," which was filed simultaneously as a continuation-in-part (CIP) application to this patent.

As explained, delivery of LightOS data packets in a distributed PBT system can be achieved using a four-layer communication protocol running over a wired bus such as USB, I2C, SMBus, FireWire, Lightning, or other wired communication media. However, if distributed PBT system communication is performed over a cellular network (e.g., 3G/LTE/4G or 5G), Ethernet, or wireless LAN by telephone, or if the data passes through a public router, communication cannot be performed exclusively using MAC addresses. In other words, Layer 1 and Layer 2 communication stacks are not sufficient to perform data routing through the network.

For example, Figure 69 shows a PBT controller 1000 communicating with an intelligent LED pad 1003 over Ethernet 1002 using a seven-layer OSI-compliant communications stack. Specifically, the communications stack 1005 of the PBT controller 1000 includes PHY Layer 1 and Data Link Layer 2, which implement the Ethernet communications protocol; Ethernet differential signaling 1004; Network Layer 3 and Transport Layer 4, which implement network communications according to TCP/IP (a communications protocol transmitted over an Internet Protocol network); and a LightPadOS operating system-defined application layer (encryption/decryption) including a session layer 5 for authentication, a presentation layer 6 for security, and an application layer 7 for PBT system control and management. The communications stack 1006 of the LED light pad 1006 includes the corresponding Layer 1 and Layer 2 protocols for Ethernet, Layers 3 and 4 for TCP/IP, and Layers 5-7 defined by LightPadOS. For point-to-point communication that does not involve IP routers, the Ethernet connection 1002 operates as a private network at network layer 3 or higher. The intelligent LED pad's operating system, LightPadOS, is a subset of LightOS, so even though they are physically separated, they can communicate with each other as a single virtual machine (VM).

Using the described seven-layer OSI communication stack, the network communication of the disclosed PBT system can be easily adapted to WiFi wireless communication. In the distributed PBT system shown in FIG. 70, a WiFi-enabled PBT controller 1010 powered by a power source 1011 communicates with an intelligent LED pad 1013 via WiFi signals 1012 using OFDM radio signals 1015 in accordance with the IEEE standard for 802.11. WiFi communication protocols include 802.11a, 802.11b, 802.11g, 802.11n, 802.11ac, or other related versions, depending on the chipset employed in the intelligent LED pad 1013. The PBT controller 1090 can support a superset of all standard WiFi protocols. Because WiFi cannot carry power, the intelligent LED pad 1093 must receive power via a USB cable 1014b, which is powered by either an AC/DC converter and DC power supply (brick) 1014a or a USB battery (not shown). WiFi communication occurs via a full 7-layer OSI communication stack 1016 present in the PBT controller 1010, which is connected to a communication stack 1017 present in the intelligent LED pad 1013.

In operation, the WiFi radio shown in FIG. 71A converts a wired communication link 1025 (e.g., PCI, USB, Ethernet) to microwave radio 1024 and converts MAC access 1020a to wireless access point 1020b using interface circuitry and associated firmware 1022. In operation, signals from communication link 1108 pass through communication stack 1021a as PHY signal 1119a, where the format is converted by interface 1022 to PHY signal 1119b, which becomes WiFi communication stack 1021b and is converted to radios 1026a-1026n operating at various radio frequencies transmitted over a multi-band antenna array 1024 for microwave communication. In operation, communications stack 1021a transfers data 1023a according to the link communications Data Link Layer-2 protocol, and interface circuitry and associated firmware 1022 converts this to WiFi data 1023b according to the Data Link Layer-2 protocol of communications stack 1021b, formatted for radios 1026a through 1026n. The WiFi radios connect to PBT controllers 131 through 135, which are also connected to Ethernet 2017 and USB 1028.

In FIG. 71B, the same WiFi radio 1024 communicates with an intelligent LED pad 337 via a wired data link 1030 using PCI, USB, or Ethernet protocols, and with an interface 338. This interface can also connect to other devices or sensors via USB 1033 and Ethernet 1032. An example of a distributed PBT communication network is shown in FIG. 72, where a WiFi router 1052 communicates with intelligent LED pads 1053, 1054, and 1055 by WiFi links 1012a, 1012b, and 1012c, and via WiFi link 1012b to a central control UI/UX LCD display 1050 with a system control window 1051a and a patient window 1051b. The system also includes a WiFi PBT remote control 1056, a component of the present invention that is useful for nurses to initiate treatment in the patient room without having to return to the central control UI/UXLCD display 1050.

Using wireless connectivity, the PBT controller can be replaced by an application program running on a mobile device such as a cell phone, tablet, or notebook computer. Figure 73 shows an example of a cell phone 1100 running PBT control application software (e.g., a PBT "light app") connected to a cell tower 1705 over a cellular network 1704, e.g., 3G/LTE, 4G, and 5G. The cell tower 1705 connects to the Internet 1706 and provides power via Ethernet, fiber, or other means. The cell phone 1700 running the aforementioned light app also connects to an intelligent LED pad 1701 using Wi-Fi 1702, which is powered by an AC adapter 1703a and cord 1703b. The seven-layer OSI communications stack 1714 of the radio tower 1707 uses a data packet communications stack 1709 for mobile network use. The light app, in turn, uses the seven-layer communications stack 1709 to connect to the intelligent LED pad 1701, which includes communications stack 1708. As shown, the PBT communications stack 1709 blends two seven-layer communications stacks: one for interacting with the cell phone tower 1705's communications stack 1707 and another for connecting to the Internet 1706 and a cloud-based server (not shown) via a router, which then connects to the intelligent LED pad 1701 and communications stack 1708. Here, only the application layer 7 bridges the two. In this way, the mobile phone 1700 running the aforementioned light app acts as a PBT controller, separately communicating with the cloud-based computer server (not shown) and the intelligent LED pad 1708 via the Internet 1706 without relinquishing local control.

Because PHY Layer 1 and Data Link Layer 2 are not shared for communication across Layers 1 through 6, the cell tower communication stack 1707 does not have direct access to the intelligent LED pad communication stack 1708. Instead, only Application Layer 7 within the communication stack 1709 bridges the two communication networks. Applications may include dedicated Light apps, such as LightPadOS, that operate as a reduced instruction set version of the LightOS operating system used in the dedicated hardware PBT controller described above. Essentially, the Light app emulates the operation of LightOS, facilitating PBT control functions and its UI/UX touchscreen-based control. The Light app is implemented as software designed to run on the operating system used in the corresponding mobile device. For example, on smartphones and tablets, Light apps are written to run on Android or iOS, while on notebooks, Light apps are written to run on MacOS, Windows, Linux, or UNIX. Converting a Light app's source code, which is its basic logic and functionality, into executable code adapted to run on a specific platform is a translation process called a "compiler."

Therefore, the conversion of source code to compiled code is platform-specific, meaning that multiple versions of the software must be distributed whenever a software revision, patch, or new release occurs. The operation of a mobile device-based distributed PBT system is shown in Figure 74. The mobile device 1102 also uses a mobile device control UI/UX 1100 and a host light app 1130 interface WiFi, which can connect to the Internet and cellular networks using a cellular link 1104, to control intelligent LED pads 1119a and 1119b using, for example, 3G/LTE, 4G, and 5G protocols.

An example of software control of PBT system operation is shown in the exemplary screen 1120 of FIG. 75. The UI/UX surface includes a treatment menu 1121 titled "Select Session," along with a button for "Extend Session" 1122, which increases the time of the PBT treatment. "Select LED Pad" 1122 is used to pair the mobile device to a specific intelligent LED pad. As shown, selecting a stress relief treatment opens a second screen, "Running" 1130, to monitor the ongoing treatment showing the treatment name 1131, cancel 1132, or pause 1133 the treatment. The window also shows the time remaining in treatment 1134, a step progress bar 1135, a treatment progress bar 1136, and biofeedback 1137.

Driving other distributed components

The PBT controller can be used to control other therapeutic devices besides LED pads. These peripheral components consist of laser PBT wands and systems, autonomous LED pads programmed on a distributed PBT system, magnetic therapy pads and wands, LED masks, LED caps, LED ear and nose buds, etc. The LED face masks, head caps, and LED beds are simple multi-zone PBT systems using a proprietary LED delivery system. Therefore, the electrical control is identical to the disclosed aforementioned PBT systems. In general, the aforementioned distributed PBT systems are not limited to driving LEDs but can also be used to drive energy emitters placed adjacent to the patient to inject energy into biological tissue, such as emitting coherent light from a laser or time-varying magnetic fields (magnetic therapy), microcurrents (electrotherapy), ultrasound energy, infrared, far-infrared electromagnetic radiation, or any combination thereof.

Because distributed therapy systems use different energy emitters for laser, PBT, hyperthermia, magnetic therapy, and ultrasound therapy, they require some modifications to drive the disclosed PBT controller than LEDs. Some examples of adapting the disclosed PBT system to alternative therapies are described below.

Laser PBT System

Figure 76 shows a handheld PBT device or "wand" useful for laser PBT treatment. As shown, the handheld wand 1150 includes a cylindrical arm 1153 with an LCD 1160 and control buttons 1161a and 1162b. The bottom of the cylinder handle also includes a USB port 1162 required for charging the battery 1166. The cylinder handle connects to the PBT head 1151 through a gimbal 1152 with a transparent faceplate 1154 containing a printed circuit board PCB 1155 with lasers 1156 and 1157 along with a sensor 1158. One inventive feature is a circular conductive blade 1159 used to sense contact with the skin to prevent illumination of the laser unless the unit is in contact with tissue.

The handheld PBT therapy device of Figure 77 includes a pad μC 1181, a clock 1183, volatile memory 1185, non-volatile memory 1184, a communication interface 1182, and Bluetooth 1190. The pad μC communicates via a data bus 1187, controls a UI 1177 with buttons 1161a and 1161b, controls a display driver UX 1176 with an LCD 1160, a laser driver 1174, and a safety system. As shown, the laser driver 1174 drives laser diodes 1156 and 1157. At the same time, a contact blade signal 1188 and a temperature sensor signal 1189 are used by the safety system interface 1175. The laser driver 1174 is powered by a laser power supply 1173 powered by a lithium-ion battery 1172 via a charger and regulator 1171 powered by a USB input 1186.

Details of the safety sensor are shown in Fig. 78A and include measurement of heat 1200 with PN diode 1202 (terminals A and K) and contact blade 1159 with capacitors 1201a and 1201b, which form a closed circuit that conducts AC current through the patient's tissue across terminals C and C'. Fig. 79 shows the laser PBT handheld safety system including oscillator 1220, contact sensor capacitors 1201a and 1201b, and sense resistor 1221 along with differential amplifier 1222, low pass filter 1223, comparator 1225 and voltage reference 1224. In operation, voltage Vosc oscillator 1220. At the switching frequency, fosc, a voltage divider formed between resistor 1221 and the series combination of capacitors 1201a and 1201b and resistor 1221 presents an equivalent impedance, Z, and the voltage drop across resistor 1221 is V, while the network voltage between nodes C and C' is V = ZC·Iave. The two equations for V are equal: VR = VoscR/(R + ZC). That is, when contact blade sensor 1159 is not in contact with the patient's skin, the value of ZC is large and VR approaches zero. In such a case, the output of the differential amplifier is lower than _Vref , the voltage of temperature-independent voltage reference 1224. Therefore, the output of eye-safety comparator 1225 is grounded, and the laser driver is inhibited. When the sensor blade contacts the skin, the AC impedance ZC drops significantly. After filtering out the AC signal with low-pass filter 1223, the average DC voltage across resistor 1221 becomes greater than _Vref . This causes the eye safety comparator output to be logic high, sending contact detection enable signal 1228 to the laser μC. Similarly, the temperature sensor 1202 is handled by temperature protection circuit 1231a. If an overtemperature condition occurs, an overtemperature flag 1232 is sent to the laser μC, causing the input to the logic gate to go low, disabling laser driver 1174. If there is no overtemperature condition, contact detection 1228 is confirmed. Logic gate 1226 passes the digital value of the output of PWM driver 493, i.e., laser driver 1174 is enabled.

Figure 80 shows an example schematic diagram of a dual-channel laser driver. As shown, the laser PBT control 1240 is similar to the previously described LED controller, including a laser μC 1181, a communications interface 1182, a clock 1183, non-volatile memory 1184, and volatile memory 1185. Protection features include eye protection 1131b and over-temperature protection 1131a with sensor 1202. The fault signal and PWM player output from the laser μC are input to logic gates 1228a and 1228b and buffered by two series inverter pairs 1247 and 1246. The outputs are fed to the digital inputs of the laser driver's digital current sinks 1256 and 1257. The 1174 dual-output D/A converter 1245 is also used to control the analog values of the currents the ILaser1 and ILaser2 current sinks are conducting.

Controlled current sink 1256 is used to drive an array of lasers 1156a through 1156n at wavelengths of λ _1. Controlled current sink 1257 is used to drive an array of lasers 1157a through 1157n at wavelengths of λ _2. The laser array is in laser array 1242 powered by power supply voltage +VHV, the output of boost switching regulator 1241 including input capacitor 1265, PWM controller 1260, low-side power DMOSFET 1262, inductor 1261, Schottky rectifier 1263, and output capacitor 1264 with voltage feedback to PWM controller 1260. The input to laser power supply 1241 is provided by a Li-Ion battery 1172 and battery charger 1171. USB power input. The 2.5-V voltage-regulated output can also be used to supply laser PBT control using a laser array after the power supply output boost converter used to drive the +VHV is operated if a high voltage output from the charger 1171 and filter capacitor 1266 is required to power the components of the laser PBT control circuit 1240.

Autonomous LED Pad for Photobiomodulation Therapy - Another peripheral compatible with the distributed PBT system is the autonomous LED pad, used in applications where a PBT controller or cell phone is unavailable or inconvenient to administer emergency treatment, such as on a battlefield or in a military setting. A plane crash in a mountainous area. During operation, a single button on the autonomous LED pad is used to select treatment. There is generally no UX display available for information. Also, while the autonomous LED pads operate "autonomously" (i.e., independently) during treatment, during manufacturing they are connected to a portion of the distributed PBT system and the appropriate program is loaded to ensure proper operation.

The PBT software programs loaded onto the LED pads vary depending on the target market and application. For example, a treatment program loaded onto an LED pad at a ski resort might include treatment for concussions (a common skiing injury), while one used by emergency medical personnel might focus on treating wounds such as lacerations and burns. In sports facilities or tennis clubs, autonomous LED pads for muscle and joint pain might be more common. In military applications, the primary field use is to slow or prevent the spread of infection in bullet or shrapnel wounds.

The electrical design of the intelligent LED 337 in Figure 14 is equally applicable to autonomous LED operation, except for the addition of pushbuttons for controlling on/off and program selection. During programming, the entire PBT system, including the power brick 132, PBT controller 131, USB cable 136, and autonomous intelligent LED pad 337, is present. During programming, the PBT controller loads manufacturing data and configures the LED pad by downloading the PBT player. If desired, preload LED regeneration files. The portable programming system can also be used to reprogram pads that have been sold or deployed in the field. This allows clients to reuse inventory to adapt to different types of disasters, such as winter frost bites, antiviral treatments during disease outbreaks, pandemics, lung damage caused by terrorist nerve agent releases, etc.

A key element of autonomous LED pads is the need to control costs by utilizing a standard design. This means using one common manufacturing flow and product BOM (build of materials) for all applications and markets, and using software downloads to convert the generic product into application-specific versions. An example of one generic pad is shown in Figure 81A, which includes a self-contained, pre-programmed intelligent LED pad shown in top view 1281, bottom view 1284, and a side view including a single USB socket 1198. Cross section 1280 includes rigid PCB 1288, flex PCB 1289, LEDs 1291 and 1292, sensor 1290, and control switch 1299. LED polymer pad cover 1281 includes opening 1295 and cavity 1296, thin section 1288 for switch 1298, and protective clear plastic 1287. LED pad 1280 includes a top cover flexible polymer 1281. Includes a lower flexible polymer 1284 with protrusions 1283 and 1285.

As discussed, the autonomous LED pad does not utilize a display, wireless link, or remote control, and therefore offers a limited number of preloaded treatment programs, generally selections 1 through 5, as shown in FIG. 81B. As shown, a single press of switch 1293 changes the autonomous LED pad in off state 1257a to state 1257b. Briefly selecting this state initiates treatment using program "Treatment 1." A second press of the button advances the program to state 1257c, which initiates "Treatment 2." In a similar manner, each subsequent button press advances the program to the next treatments 3, 4, and 5, shown as corresponding states 1257d, 1257e, and 1257f. Pressing switch 1293 a sixth time returns the autonomous LED pad to off state 1297a.

Pulsed LED hyperthermia—as well as visible and near-infrared photobiomodulation therapy—is the application of far-infrared light, typically involving wavelengths above 100 and below 1 μm. Examples of hyperthermia include spa treatments, heated compresses, and heated body wraps. According to Wikipedia, the therapeutic effects of heat include "increasing collagen tissue compliance; reducing joint stiffness; alleviating pain; relieving muscle spasms; reducing inflammation and edema, and assisting in the post-acute phase of healing; and increasing blood flow. Increased blood flow to the affected area provides protein, nutrients, and oxygen for better healing." It also promotes the delivery of metabolic waste and carbon dioxide. Hyperthermia can also help improve muscle spasms, muscle pain, fibromyalgia, contractures, and bursitis.

While its therapeutic claims overlap with those offered by PBT, the physical mechanism of hyperthermia is quite different. Unlike PBT, which provides molecules with absorbed photons to stimulate chemical reactions that would not otherwise occur—i.e., photobiomodulation—hyperthermia uses heat absorbed by tissue and water to accelerate molecular vibrational rates, driving ongoing chemical reactions. However, because photon energy is inversely proportional to its wavelength according to Einstein's relationship E=hc/λ, the energy of 3 μm far-infrared light is only 20% to 20% of the energy of red and NIRPBT. This energy difference is significant because lower energies are insufficient to break chemical bonds or transform molecular structure. Such hyperthermia is generally viewed as symptom relief without the accelerated healing associated with PBT. The penetration depth of far-infrared sources shorter than 3 μm (i.e., IR Type B) exhibits greater penetration depth than longer wavelengths and is therefore preferred over longer wavelength sources.

The aforementioned PBT system can be adapted for hyperthermia therapy by replacing the visible and near-infrared LEDs with LEDs in the far-infrared spectrum. LEDs are typically limited to wavelengths below 12 μm. "Far infrared radiation (FIR): its biological effects and medical applications," PhotonicsLasersMed., vol. 1, no. 4, November 2012, pp. 255-266: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3699878/ F. Vatansever, M. R. Hamblin. By tuning the crystal structure of III-V compound superlattice semiconductors to narrower bandwidths, LEDs operating in the far-infrared spectrum have been realized with wavelengths up to 8.6 μm ("Superlattice InAs/GaS Bright-Emitting Diode with Peak Emission at Wavelength of 8.6 μm," IEEE J. Quant. Elect., vol. 47, no. 1, January 2011, pp. 5-54). Therefore, the PBT system used to drive the NIRLEDs disclosed herein can be easily retrofitted to accommodate FIRLEDs by simply replacing the NIRLEDs with their longer-wavelength counterparts. The drive circuit can be used equally well using pulsed or sinusoidal waveforms. Due to the longer wavelength, drive frequencies below 100 Hz are more suitable for uniform irradiation of far-infrared light. At even lower frequencies, say below 10Hz, the FIR LEDs within the pads can be scanned row by row to create a wave-like massage across each pad, stimulating vasodilation sequentially in a systematic pattern throughout the treated tissue. Optionally, near-infrared LEDs for PBT and far-infrared LEDs for heat therapy can be combined in one intelligent pad and driven simultaneously or alternately.

Magnetic Therapy - Magnetic therapy (MT) is an alternative medical treatment in which damaged tissue is exposed to a magnetic field. The effects of fixed magnetic fields on tissue are questionable and it is generally considered quasi-medicine, pseudo-medicine, fringe medicine, or even pseudo-medicine. Several U.S. FDA studies have concluded that the medical claims of permanent magnet magnetic therapy are not fully supported by the results of scientific and clinical research. The FDA prohibits the sale of magnetic therapy products using medical claims (https://en.wikipedia.org/wiki/Magnet_therapy). Conflicting claims suggest that pulsed magnetic fields exhibit therapeutic effects because biological tissue contains numerous free ions and even electrically balanced molecules (such as water) that act as dipoles due to the direction of their electric charge. When exposed to an oscillating magnetic field, molecules are repelled and attracted according to their charge, in a manner similar to the imaging performed by magnetic resonance imaging (MRI), except the excitation occurs at a lower frequency. This type of magnetic therapy is commonly referred to as pulsed magnetic therapy or PMT.

The reported effects of PMT are primarily muscle relaxation, improved local blood circulation and vasodilation, anti-inflammatory effects, pain relief through the local release of endorphins, and beneficial effects on cell membrane action potentials. The mechanism of action is primarily believed to be electrochemical in nature rather than thermal, acting catalytically by accelerating the rate of ongoing chemical reactions. Reported PMT pulse frequencies span the entire audio and infrasound spectrum, from 20 kHz seeding to less than 1 Hz. It is not possible from the published literature to determine the accuracy of these reported claims or to confirm the therapeutic effects of pulsed magnetic therapy. Furthermore, PMT carries certain risks. In particular, PMT is contraindicated in the case of tumors and poses safety risks due to its potential impact on pacemaker operation.

According to the present invention, a pulsed magnetic therapy system can be realized by repurposing the disclosed PBT system by replacing the optical components with electromagnets and adapting the driver circuitry included in the intelligent pad or wand. Optionally, LEDs for the PBT can be driven simultaneously or alternately in time in combination with magnetic emitters. When driving an array of electromagnets, the electromagnet array must be mounted on a three-dimensional bendable printed circuit board (or 3D PCB) similar to the "3D bendable" used for the LED array herein and disclosed in USPTO Application No. 14/919,594. This printed circuit board features redundant interconnects, which are incorporated herein by reference. The rigid-flex PCB is required to adjust the orientation of multiple electromagnets at a 90° angle (i.e., a right angle) relative to the patient's tissue being treated without mechanically damaging the solder joints between the flex PCB and the rigid electromagnets. The rigid-flex PCB provides a perfect solution for achieving reliable 3D bendability.

Figure 82 shows a rigid-flex PCB with unprotected copper interconnects. As shown, the flex PCB includes an insulating layer 1303 sandwiched between metal layers 1301 and 1302, which typically include patterned copper. In some parts of the cross section shown, and in other parts (not shown in this particular cross section), this flex PCB is sandwiched in the middle of a rigid PCB including insulating layers 1304 and 1305, laminated with patterned metal layers 1311 and 1312. Generally, flex PCB metal layers 1301 and 1302 are thinner than rigid PCB metal layers 1311 and 1312. The cross section is for illustrative purposes only. The exact pattern of each layer in the cross section will vary depending on the location and the circuit being implemented. As shown, metal via 1307 is used to connect metal layers 1301 to 1311, and via 1308 is used to connect metal layers 1302 to 1312. A fully buried via 1306 is used to connect the flex metal layers 1301 and 1302.

Protective layers, including coatings of polyimide, silicone, or other scratch-resistant materials, are used to seal both the rigid and flex portions of the PCB. As shown, insulator 1304 protects metal layer 1301, which in turn protects metal layer 1302, completely sealing the flex PCB from moisture and the risk of mechanically induced scratches. In the rigid portion of the PCB, patterned insulating layer 1313 protects a portion of metal layer 1311, while unpatterned insulating layer 1314 completely protects metal layer 1312. Some portions of metal layer 1311 remain unprotected for the purpose of soldering components to the rigid PCB.

As shown, electrical interconnection of various metal layers within a given rigid PCB, between rigid PCBs, and within a flex PCB can be achieved without the need for wires, connectors, or solder joints using conductive vias 1306, 1307, and 1308. These conductive vias comprise conductive columns of metal or other low-resistivity material formed vertically in the various metal layers and can penetrate two or more metal layers to facilitate multi-level connectivity and non-planar electrical topologies, i.e., circuits in which conductors must cross each other to form an electrical short.

The role of the PMT pads in the rigid portion of the disclosed rigid-flex PCB can be utilized in a variety of ways. In some cases, separate electromagnets, permanent magnets, and permanent magnet/electromagnet stacks can be attached to the rigid portion of the rigid-flex PCB. Alternatively, PCB interconnects can be used to form toroids that, when combined with through-hole magnetic material, form a planar magnetic structure. One example layout of a planar magnetic toroid is shown in the exploded view of Figure 83, where metal conductive layers 1311, 1301, 1302, and 1312 form a circular toroid surrounding magnetic core 1316. Each circular conductor on a given layer is rotated relative to the metal layer below it, so that metal vias 1307, 1306, and 1308 can interconnect layers such that current flows counterclockwise on all layers located on each plane. This structure is illustrated in more detail in FIG. 84, where a rigid-flex PCB forms a layer of a toroid surrounding a magnetic core 1316. To prevent shorting between the conductive layers and the ferrous magnetic core, the magnetic core 1316 may be insulated from conductive layers 1311, 1301, 1302, and 1303 by insulator 1315. The resulting top view is shown in FIG. 85, in a plan cross-sectional view, intersecting rigid PCB 1320 and interconnecting flex PCB 1321. As shown, circular conductor 1302 surrounds magnetic core 1316, connecting to the overlying conductive layer through via 1306 and to the underlying conductive layer through via 1308.

An exemplary circuit used to drive a PMT is shown in FIG. 86. The PMT driver 1340 includes a PMT μC 1181, a clock 1183, non-volatile memory 1134, volatile memory 1135, a communications interface 1182, and a Bluetooth or WiFi wireless link 1190, as well as an intelligent LED pad or laser wand circuit consisting of a PMT driver 1340, an electromagnet driver 1341, an electromagnet power supply 1363, and an electromagnet array 1350, as well as a charger 1360, a lithium-ion battery 1361, and a USB connector. The digital pulse output of the PMT μC 1181 is gated by logic gates 1228a, 1228b, and optionally other gates (not shown) to facilitate over-temperature protection 1131a. The outputs of the gates are then buffered by dual inverter strings 1346 and 1347, which drive the digital inputs of programmable current sinks 1342 and 1343, respectively. Control current sinks 1342 and 1343 respond to the digital inputs to control the magnitude and waveform of electromagnet currents IEM1 and IEM2 flowing through electromagnets 1352 and 1353, and are also controlled by an analog reference current derived from the output of D/A converter 1345.

Whenever the current sink is switched off, the electromagnet stores energy by rapidly recirculating the inductor current until either EL = 0.5LI2 is consumed or the current sink is again conducting current. Freewheeling diodes 1354 and 1355 are included to prevent high voltage spikes. Capacitors 1356 and 1357 are intentionally used to filter noise or switching if necessary, to form a tank circuit that oscillates with the coil inductance at a resonant frequency of fLC = 1/(2πSQRT(LC)). Power to drive the electromagnet + VEM is derived from a switching power supply circuit, either a boost converter to step up the voltage, or a buck converter to step down the voltage. Alternatively, the voltage regulator can be eliminated anyway, since current sinks 1343 and 1343 control the inductor current.

While the operation of switching regulators is well known in the art, for illustrative purposes, an exemplary boost converter is included herein as electromagnet power supply 1363. In operation, PWM controller 1365 turns on power MOSFET 1366, causing the current in boost inductor 1369 to ramp up at a fixed rate over the switching period, after which power MOSFET 1366 is turned off. When the MOSFET is turned off, the drain voltage of power MOSFET 1366 rises instantaneously, forward biasing Schottky diode 1367 and charging capacitor 1368 to a voltage +VEM. A capacitor voltage feedback signal is then "fed back" to PWM controller 1365, allowing the controller to determine whether the output voltage is lower or higher than its target voltage.

If the voltage is below the target, the pulse width over time increases by a larger percentage (D = t _on / (t _on + t _off ) = (t _on / T _PWM )), where T _PWM is the time of the next clock cycle. As D increases, the average current in inductor 1369 increases, causing the output voltage + VEM to rise. On the other hand, if the output voltage is too high, the duty factor D, or the on-time of MOSFET 1366, decreases, gradually reducing the current in inductor 1369 over several switching cycles, thereby lowering the output voltage. By continuously adjusting the duty factor D and pulse width (the on-time of power MOSFET 1366), the output voltage is regulated to a constant value through voltage feedback. Therefore, the regulation process of a switching regulator that operates at a switching frequency and period T _PWM is called PWM, for pulse-width modulation. The role of the output capacitor 1368 is to filter the output voltage, and the input capacitor 1364 is used to prevent noise from being injected back into the power supply and stabilize the power network. As shown, the output voltage of a switching converter and regulator is higher than its input, i.e., +V > V, so the converter is called a boost converter. However, if the required electromagnet driver voltage is lower than the battery voltage, +V < V, a buck or step-down converter is required. Topologically, a buck converter can be realized by making only a slight modification to the boost converter circuit by rotating the three components connected at the common node to the right and rearranging the same components, i.e., replacing Schottky diode 1367 with inductor 1369 and power MOSFET 1366. The Schottky diode 1367 is used, and inductor 1369 is replaced with power MOSFET 1366.

Alternatively, a pre-assembled or discrete electromagnet module can be used, instead using planar magnetics to realize the electromagnet. As shown in FIG. 87, an individual surface-mounted electromagnet 1351, including a magnetic core 1376 and a wound coil 1375, is attached as a surface-mounted component to the rigid portion of a rigid-flex PCB by soldering metal legs 1359a and 1359b to two separate, electrically isolated conductive layer segments of the same copper conductor layer 1311A and 1311B. Isolated conductive segment 1311a then connects to the lower conductive layer 1312 through patterned vias 1309a, 1306a, and 1310a, as shown. In this manner, separate individual electromagnets can be placed on each rigid PCB to form an array as shown in the cross-sectional view of FIG. 88A, specifically where discrete electromagnet 1351a is attached to rigid PCB 1348a and connects to rigid PCB 1348b via flex PCB portion 1349a. Discrete electromagnet 1351b is attached to rigid PCB 1348b, which connects to rigid PCB 1348c via flex PCB portion 1349b. Discrete electromagnet 1351c is attached to rigid PCB 1348c and connects to another rigid PCB (not shown) via flex PCB portion 1349c.

In such a design, all magnets 1351a, 1351b, 1351c, etc. in the array are electromagnets that can be electronically controlled to change their magnetic field according to the previous PMT circuit in response to PMT playback generated from PMT driver 1340. The waveform may involve driving the electromagnets individually and in several sequences to form a specific pattern or magnetic wave across the PMT pads; alternatively, it may produce a continuous pulse or sinusoidal variation in the magnetic field of all electromagnets in the array, or may generate undulating magnetic field waves, for example, row by row across a pad or along the length of a series of pads. In other cases, some electromagnets are biased to generate a constant magnetic field, while others are modulated to generate a time-varying magnetic field.

In alternative embodiments, some electromagnets can be replaced with electromagnets to combine a mix of constant and time-varying magnetic fields. For example, in FIG. 88A, electromagnet 13511b (shown earlier in the figure) was previously replaced with permanent magnet 1370, while electromagnets 1351A and 1351c remain unchanged and are attached to rigid PCB 1348b. In FIG. 88C, rigid PCB 1348b drives a stack of electromagnet 1351d and underlying permanent magnet 1370b, or in FIG. 88D, rigid PCB 1348b drives a stack of electromagnet 1351e and underlying permanent magnet 1370c. In such cases, the operation of the electromagnets enhances (or alternatively reduces) the magnetic field generated by the stacked permanent magnets.

The PMT device may also be adapted for use as a handheld magnetic therapy device or wand 1450, as shown in FIG. 89. The PMT device includes a cylindrical handle with a display 1460 and 1458, pushbuttons 1461b for controlling operation and program selection, a movable gimbal 1452, a magnetic head on/off button 1461a via a battery 1643, and a USB connector 1462 on the cylindrical handle 1458 to connect the magnetic head unit 1453 to the magnetic head unit 1453. The unit 1453 includes an electromagnet 1455 including a ferrite core 1457 and coil 1556 mounted on a PCB 1454 along with control circuitry. When operated as part of a distributed system, the communication link of the handheld magnetic therapy wand 1450 to the PBT controller may be via USB, Wi-Fi, or possibly Bluetooth. As an autonomous device, the USB connector 1462 is used to program the wand during manufacturing by connecting it to the PBT controller.

Periodontal PBT LED Mouthpiece - While PBT can be performed to treat gum disease through the cheek, another option is to use a laser or LED to inject light directly into the patient's mouth in the near, infrared, and blue spectrums. The device must be small and fit comfortably in the mouth. As an autonomous treatment device, the device must use a lightweight software client capable of running only a few pre-programmed algorithms. Alternatively, the device may employ data streaming from a user control module using a wired connection, Bluetooth, or low-power Wi-Fi 802.11ah. The user control module communicates with a PBT controller, which operates similarly to the controller in an intelligent LED pad; however, its output does not drive LEDs in the pad; instead, it is streamed to the LED mouthpiece as a passive electrical signal, with no processing taking place within the mouthpiece.

An example of such a periodontal PBT device is shown in the three-dimensional perspective view of FIG. 90, which includes a molded mouthpiece 1500 with a horseshoe-shaped portion covering the teeth and gums 1503. Two different wavelength LEDs 1504 and 1505 (location 1506 identifies the location of the LEDs not visible in the 3D perspective view) line the horseshoe-shaped portion. An electrical cable 1501 and control unit 1502 include connectors for power and, optionally, bus communication. A corresponding cross section reveals a U-shaped cross section surrounding the tooth 1510, including a flex PCB 1513, a rigid PCB base 1515, and a rigid-flex PCB assembly with LEDs 1513. The mouthpiece is designed to position the LEDs 1513 so that they are positioned near the gums 1512 adjacent to the tooth 1511, rather than attempting to clean the teeth. The LEDs may include red, infrared, blue, or purple LEDs to combat inflammation and periodontal disease. The U-shaped assembly is contained within a thin silicone mouthpiece molded around a rigid-flex PCB.

Manufacture of a mouthpiece with a U-shaped cross section designed to cover and treat a single jaw (either the upper or lower jaw, but not both) is shown in Figure 91, which includes a rigid PCB portion 1513 and a flex PCB wing 1514. As shown immediately after SMT manufacturing, LED 1513a is attached to flex wing 1514, and optional LED 1513z is attached to rigid PCB 1515. During PCB surface mount technology (SMT) assembly, rigid-flex PCBs accommodate high-volume automated assembly requiring components. The solder temperature profile during reflow is selected, positioned, and uniform. It is important to keep the PCB firm and flat during SMT assembly. While the rigid and flex portions of the PCB are fixed in the same plane during pick-and-place, the rigid-flex PCB does not need to be straight and can be laid out in a gummy, horseshoe-shaped design, which can cause unnecessary bending of the flex PCB or add stress and potentially lead to subsequent failure. After surface mount assembly, the flex wings 1514 are bent into a U-shape perpendicular to the rigid PCB base 1515 and then molded into a clear silicone mouthpiece 1516 that covers the rigid-flex PCB.

The same process can be applied to fabricate an H-shaped mouthpiece, which is useful for using PBT on both the upper and lower jaws simultaneously. The method shown in Figure 92A utilizes the same manufacturing process as described for the U-shaped mouthpiece above, except that after PCB assembly, two separate components are electrically and physically bonded to produce the H-shaped mouthpiece. As shown, two PCBs are bonded together: one containing rigid PCB 1515a, flex PCB 1514a, LED 1513a, and optional LED 1513z, and the other containing rigid PCB 1515b, flex PCB 1514b, LED 1513b, and optional LED 1513y. In the bonding process, rigid PCBs 1515a and 1515b are soldered together to electrically and mechanically form a single multi-layer PCB 1517, as shown in Figure 92B. In this way, the mouthpiece can treat both the upper and lower gums simultaneously.

The joining of rigid PCBs 1515a and 1515b is shown in Figure 93. The diagram shows conductive surfaces 1518b and 1518d on rigid PCB 1515b soldered to corresponding conductive surfaces 1518a and 1518c under rigid PCB 1515a to establish electrical connection between the upper and lower PCBs and provide mechanical support and rigidity to the mouthpiece. Optionally, through-hole vias 1519a and 1519b filled with silver solder paste can be melted to form continuous through-holes that run through both upper and lower rigid PCBs 1515a and 1515b.

The circuit diagram for the periodontal PBT mouthpiece is shown in Figure 94. Because high voltages are not permitted in the patient's mouth, the input voltage +VIN must be stepped down. The low voltage +VLED is regulated by low-dropout linear regulator LDO 1520. Filter capacitors 1521 and 1522 are included to stabilize the regulator and filter input and output transients, respectively. Under the control of the unit's microcontroller 1535, which executes programs stored in volatile and non-volatile memories 1536a and 1526a according to clock 1534 and time base 1531, signals from the microcontroller are used to independently drive programmable current sources 1524a and 1524b with control signals 1537a and 1524b.

The signal can be used to digitally strobe LEDs on and off, program their conduction current, or synthesize periodic waveforms such as sine waves. The current from current source 1524a is mirrored by NPN bipolar transistor 1525a to control the current of NPN bipolar transistor 1526a, thus controlling the current of LEDs 1504a and 1504b, which in turn controls the current of LEDs 1504c and 1504d, all according to program execution by microcontroller 1535. Similarly, the current from current source 1524b is mirrored by NPN bipolar transistor 1525b to control the current of NPN bipolar transistor 1526b, thus controlling the current of LEDs 1505a and 1505b, and similarly LEDs 1505c and 1505d, according to program execution by microcontroller 1535. In this way, LED current can be controlled using a minimum of components to conserve space. Therefore, the miniaturized controller circuitry can be housed in the enclosure 1502 shown in FIG. 90.

Ultrasound Therapy - The disclosed distributed PBT system can also be applied to drive piezoelectric transducers to generate ultrasound waves in the frequency range of 100 kHz to 4 MHz. The primary therapeutic mechanism of action of ultrasound therapy is vibration, suitable for destroying scar tissue and causing heating with good depth of penetration. Driving algorithms can be similar to those used in the sinusoidal driving of the LEDs disclosed herein, including both digital (pulsed) and sinusoidal driving. The disclosed distributed PBT can perform ultrasound therapy independently or in combination with PBT. The disclosed system can also be used to combine an ultrasound transducer with an LED array to destroy scar tissue using ultrasound and carry it away using PBT-accelerated phagocytosis.

One implementation of a combined ultrasound PBT therapy system or USPBT pad is shown in Figure 95. It includes a microcontroller 1557 executing a program stored in volatile and non-volatile memories 1558a and 1558b according to a clock 1556 and a time reference 1553. Signals from the microcontroller are used to independently drive an H-bridge including low-side N-channel MOSFET 1563a. At the same time, high-side P-channel MOSFET 1564b turns off, and then low-side N-channel MOSFET 1563b turns on, causing Vy = 0, during which current flows from Vx to Vy. During the next half cycle, the current flow reverses from Vy to Vx.

High-side MOSFETs 1564a and 1564b are driven by level-shifting driver circuits 1566a and 1566b. Similarly, low-side MOSFETs 1563a and 1563b are driven by low-side buffers 1565a and 1565b. During operation, the half-bridge formed by low-side N-channel MOSFET 1564a and high-side P-channel MOSFET 1563a is driven out of phase with the half-bridge formed by low-side N-channel MOSFET 1564b and high-side P-channel MOSFET 1563b. Whenever high-side P-channel MOSFET 1564a is on and conducting, low-side N-channel MOSFET 1563a is off and V,X = +VPZ. At the same time, high-side P-channel MOSFET 1564b is turned off, then low-side N-channel MOSFET 1563b is turned on, causing Vy = 0, during which current flows from Vx to Vy. During the next half cycle, the current flow reverses from Vy to Vx. In operation, the two half-bridges are phase driven, so that the output of the 1567 half-bridge is bidirectional with an absolute value of ±VPZ in response to the output of the μC1557 pad. The output of the μC1557 pad is also used to drive the LED array 1560 via the previously disclosed LED driver 1560.

In an alternative embodiment shown in FIG. 96, a programmable array of current sinks replaces a half-bridge in driving multiple piezoelectric transducers. As shown, pad μC 1557 outputs a digital amplitude to D/A converter 1573, which is used to control the current conducted by current sinks 1576 and 1575 through corresponding piezoelectric transducers 1562a and 1562b, respectively. Piezoelectric currents IPZ1 and PZ2 are pulsed by inverters 1571 and 1572, which control the frequency of the digitally generated ultrasonic waves. An example of a USPBT pad is shown in FIG. 97, including an intelligent LED pad shown in top view 1581, bottom view 1584, and a side view including a single USB socket 1598. Flex PCB 1589, LED 1591, sensor 1590, piezoelectric transducers 1592a and 1592b. LED polymer pad cover 1581 includes opening 1595 and cavity 1596, and protective clear plastic 1587. LED pad 1580 includes an upper flexible polymer 1581c with protrusion 1583 and a lower flexible polymer 1584 with protrusion 1585.

Optionally, the LEDs for PBT can be combined with ultrasound piezoelectric emitters and driven simultaneously or alternately in time. Applications combining ultrasound and photobiomodulation therapy (herein referred to as USPBT) can be useful, using ultrasound to destroy scar tissue and PBT to accelerate the removal of dead cells.

Infrasound Therapy - Infrasound therapy is similar to tissue massage, except that it occurs at very low frequencies below the audio spectrum, typically 20 Hz to 1 Hz or less. The actuator for generating the low frequencies must be relatively large, e.g., 10 cm in diameter, and is therefore well suited for inclusion in a wand similar to that shown in Figure 89. The moving part is attached to a plunger or membrane, which presses on the treated tissue at very low frequencies, except the electromagnet is replaced by a voice coil driver similar to a speaker. Therefore, the disclosed PBT system is directly compatible for supporting ultrasound peripherals. Infrasound provides a deep massage to tissue, providing low-frequency sounds that help improve range of motion and muscle elasticity. Optionally, LEDs for PBT can be driven simultaneously or alternately in time in combination with an infrasound voice coil actuator.

PBT LED Buds for Nose/Ear Although PBT can be performed transcranially, another option is to inject light directly into the nose or ear using lasers or LEDs in the near-infrared and blue spectrum. The devices are small. As autonomous treatment devices, they require a lightweight software client capable of running only a few pre-programmed algorithms. Alternatively, the device may employ data streaming from a user control module using a wired connection, Bluetooth, or low-power Wi-Fi 802.11ah. The user control module communicates with a PBT controller, which operates similarly to the controller in an intelligent LED pad; however, its output does not drive the LEDs in the pad; instead, it is streamed to the LED buds as a passive electrical signal, so no processing is performed within the pad. Therefore, the disclosed PBT system is directly compatible to support PBT LED buds for nasal and ear treatment. Another advantage of intranasal and intraauricular (i.e., inside the ear) PBT is its ability to kill pathogens and bacteria infecting the sinuses.

PBT LED Spots for Acupuncture - Another small-sized LED source is a small LED or laser "spot." This is a coin-sized pad attached to the body over an acupuncture point. The device is small and does not have space for battery power. The device may employ data streaming from a user control module using a wired connection, Bluetooth, or low-power Wi-Fi 802.11ah. The user control module communicates with a PBT controller that operates similarly to the controller in an intelligent LED pad, but its output does not drive LEDs in the pad; instead, it is streamed to the LED/laser spot as a passive electrical signal, so processing occurs within the spot. Therefore, the disclosed PBT system is directly compatible to support PBT LED pads for acupuncture LED spots.

Bluetooth headphones - While not medically therapeutic, for relaxation applications, music can be broadcast to headphones via Bluetooth synchronized with PBT treatment waveforms. Given the waveform synthesis capabilities of the disclosed PBT system, it can support synchronized music and PBT treatments.

Claims (9)

1. A system including distributed photobiomodulation therapy (PBT), comprising:
a PBT controller;
a light emitting diode (LED) pad;
a communication link;
The PBT controller includes a main microcontroller, a communication interface, and a digital clock;
The LED pad is
a pad microcontroller;
a memory including a non-volatile memory;
an LED array including one or more LED strings, each LED string including a plurality of LEDs connected in series with one another;
one or more LED drivers connected to each of the LED strings to control the current through each of the LED strings, the LED drivers connected to the pad microcontroller to enable the pad microcontroller to control the current through each of the LED strings;
an LED pad ID data register stored in the non-volatile memory containing data used by the PBT controller and the LED pad in a symmetric authentication process to authenticate the ID of the PBT controller and the ID of the LED pad , wherein during the symmetric authentication process, the PBT controller queries the LED pad and determines whether the LED pad is a valid manufacturer approved device by checking the data stored in the LED pad ID data register, and then the LED pad verifies whether the PBT controller is a device with a valid manufacturer ID approved for use with the LED pad by checking the data stored in the PBT controller;
the communication link transmits digital data between the PBT controller and the LED pads to control the current and illumination of each of the LED strings;
Each of the LED drivers includes a MOSFET or a bipolar transistor and a current detection and control unit;
each of the MOSFET or the bipolar transistor and the current detection and control unit is connected in series with an LED in the LED string connected to each of the LED drivers;
the current detection and control unit is connected to a source of an enable pulse for turning the MOSFET or bipolar transistor on or off, and to a source of a reference current for controlling the level of current in the MOSFET or bipolar transistor when it is turned on;
system. The distributed PBT system of claim 1, wherein the LED array includes at least two LED strings, wherein the LEDs in a first LED string are capable of emitting light of a first wavelength and the LEDs in a second LED string are capable of emitting light of a second wavelength, the second wavelength being different from the first wavelength. The distributed PBT system of claim 1, wherein the reference current source for a first LED string includes a first reference MOSFET and the reference current source for a second LED string includes a second reference MOSFET. Each of the first and second reference MOSFETs is connected in a current mirror configuration with a third threshold-connected MOSFET. The distributed PBT system of claim 3 further includes a trim network connected to the third threshold-connected MOSFET. The distributed PBT system of claim 3 includes a digital-to-analog (D/A) converter, the output of which controls the magnitude of the current in the third threshold-connected MOSFET. The distributed PBT system of claim 1, wherein the communication link includes a USB cable or other wired communication medium. The distributed PBT system of claim 6, wherein the communication link includes a conductor for supplying power to the LED pad. The distributed PBT system of claim 1, wherein the PBT controller includes a first clock and the LED pad includes a second clock, the second clock not synchronized to the first clock. The distributed PBT system of claim 1, wherein the LED pad comprises a memory, the memory holding data for controlling the LED driver according to a selected PBT process.