JPH025347B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a noise margin measurement method and an apparatus for error rate prediction in a digital transmission system.

In digital transmission systems, the transmitted signal is bound to a set of possible states. That is, the digital value represented by the transmitted signal corresponds to these possible states or a combination of these possible states. For example, when the transmitted signal is a binary pulse train and the amplitude of the pulse is 0 to 0.2V, the state is 0,
If 4.5 Noeda 5V is set as state 1, state 0 can be made to correspond to digital value 0, and state 1 can be made to correspond to digital value 1.

These conditions can be distinguished by differences in pulse amplitude, position or width, or by differences in carrier amplitude, phase or frequency, or some combination thereof. What all these methods have in common is their reliance on a set of distinct transmit stages. In order to decode the signals, the system receiver must accurately recognize each reception condition. This is made more difficult by the combination of signal attenuation along the transmission path and the noise added to the signal. Noise arises from a variety of potential sources, such as thermal noise or crosstalk noise from other transmission systems. This causes the received value of the received signal to be perturbed from the ideal value by the disturbance, and the receiver must determine which of a set of acceptable transmission conditions is closer to the one that was transmitted. Large perturbations result in erroneous decisions and the errors are incorporated into the receiver's digital output.

In any transmission system, the transmitted signal is modulated by the transmission channel characteristics as it propagates through the transmission channel. When the information content of a signal is in digital format, degradation of channel transmission quality appears as ISI (intersymbol interference). That is, the energy transmitted during one baud period, when received, is no longer limited to the same period, but becomes blurred over several periods. Moderate ISI levels only reduce the noise margin of the received baud, but at higher ISI levels, the signal becomes completely unreadable without equalization. Complete equalization consists of passing the received signal through a network whose transfer function is the inverse of the channel transfer function. However, perfect equalization is neither attempted nor desirable. What has been attempted is to reduce the ISI value at the decision point to zero at the sampling point.

The currently accepted method of measuring the noise tolerance of a digital transmission system is to add a known amount of noise of a desired spectral density and amplitude probability distribution to the signal. This is usually either a standard line signal or a statistical pattern that approximates a normal signal. Once noise is added, the received output is monitored and all errors are counted. Repeating this for a series of different noise powers produces a plot of error rate versus noise power or error rate versus signal-to-noise ratio.

This method has several disadvantages. 1st
However, it is not easy to generate noise with desired spectral density and amplitude characteristics. Second, it takes an extremely long time to measure the low error rates that occur at signal-to-noise ratios in typical systems. Therefore, in the normal case,
Performance checks are performed at much higher noise levels and the results are extrapolated. The associated large noise power has the potential to disturb the system under test, making the results questionable.

Australia Post Research Institute's "Prototype Primary Level PCM (Pulse Code Modulation) Regenerator Threshold Level Tester" (GJ
Primal
A prototype test device for measuring the determination threshold level of a PCM playback device is disclosed. The device is also said to be useful for measuring ISI (intersymbol interference) caused by equalization pulses at decision points in playback devices under test.

The device disclosed in the report includes a manually programmed 16-bit word generator and a
A device is included for varying the pulse height at a selected position of the first word generated by the generator of a sequence of 4, 8 or 16 words.
Variation of the pulse height can only be done by manual control of the pulse height variation device. It is possible to detect and indicate errors in the device disclosed in the report or in the PCM generator during pulse fluctuations,
As a result, it is possible to sense that the value is other than the initial value.

However, with 16-bit words, even if the words are repeated up to 16 times, the standard traffic in digital transmission systems is such that the combination of "0" and "1" (+ or -) varies more than can be expressed in 16-bit words. I can't express it. Therefore, the playback device under test cannot be tested under normal operating conditions, but only under conditions of 16-bit language usage.

Degeneracy of transmitted information in a digital transmission system is caused by noise, and the tolerance of the system itself to noise is extremely important. Therefore,
It is important to be able to measure the noise margin in the system and to compile the noise margin distribution of the system from these measurements using, for example, a histogram.

The equipment described in the above report could not be used to make noise measurements in this manner because it cannot simulate standard traffic conditions using a 16-bit word generator. Given the context of the report, the suggestion that sequences of 16 bits or more would be used is objectionable, since the problem addressed by the report did not require the use of sequences of 16 bits or more.

When measuring noise margin and noise margin distribution, the variable pulse position is varied over a large number of pulse positions (typically 16 or more) so that the variable pulse can be tested in an environment with different sequences of pre- and post-pulses. What you can do is important.

The purpose of the invention is to use a pulse sequence that simulates the normal operation of a digital transmission system.
It is an object of the present invention to provide a noise margin measurement method that easily provides an accurate noise margin probability distribution in a short time.

The present invention provides a method for measuring the noise margin (according to the definition given below) in a digital transmission system consisting of at least one receiver, and this measurement method comprises: (a) pulses in a specific state (after and allowing at least one parameter of the pulses to vary within a sequence of pulse trains substantially representative of the normal traffic along a constant transmission path of the system. , (b) until the above pulses during fluctuations can be detected by the detection circuit of the system.
(c) establishing that the detection circuit has detected the pulse prior to error correction by a system-incorporated error correction device; d) steps (a), (b) and with said pulses at different predetermined positions within different said sequences;
repeating (c); and (e) accumulating, for each of the plurality of pulses, the number of occurrences of a variation value from a standard value of the at least one parameter that is a deviation, in order to determine a probability distribution of the deviation; It consists of the process of The deviation or variation value is the noise margin and is preferably classified into a number of categories, as described below on page 19 of the specification. Noise margins that fall into the same category are considered to be substantially equal.

The term "noise margin", as used herein in relation to pulses (subject to the definition below), refers to the amount of noise that such pulses have until they are detected by the transmission system's detection circuitry to be in a state different from the standard state. refers to the variation of at least one parameter from the value in the standard state.

As used herein, the term "pulse" shall include the interval at which the transmission system applies one signal from a predetermined set of signals; one parameter (e.g. voltage level, duration or timing or phase variation during the above interval, carrier frequency or amplitude, or a combination thereof)
They are distinguished by the variation in

When carrying out the method as defined in the claims, the particular state of the pulse is one of n possible states, and steps (a) to (e) are performed in the n repeated for each possible state,
The variation of said at least one parameter is such as to allow the detected state to be one of said n possible states, and the method also comprises varying the detected state prior to the accumulation step and said n further classifying the parameter variations for each of the above conditions. By repeating steps (a) to (e) a necessary number of times, a noise margin distribution histogram can be created for the relevant state of the pulse. What we are mainly interested in in digital transmission systems are “0” and “+1”.
Since it is a ternary signal consisting of a sequence of and "-1", it can be seen that n is 3.

Preferably, the step of providing a pulse comprises providing a pulse train of the sequence and inserting the pulse into the pulse train of the sequence in place of a pulse at a predetermined position in the train of pulses of the sequence. It is believed that erasing a pulse from this sequence and then replacing this erasing pulse with a variable pulse is more advantageous than adjusting the existing pulse by superimposing another variable pulse. ing. The reason for this is that the latter method introduces inaccuracies due to misalignment of the two pulses.

The pulse train of the above sequence is
A first amplifier responsive to an input signal from a sequence generator generates the pulses, and the pulses are generated by a second amplifier. The two amplifiers are controlled such that the pulses replace the pulses of the sequence at the predetermined positions. The presence of two amplifiers makes it possible for pulse parameters to vary rapidly from one pulse to the next, for example during voltage regulation.

The initial value of at least one parameter above is
It can be set to a selected level corresponding to the predicted probability detector threshold between two adjacent states. In this way, when used as a voltage parameter, the pulse height can be set as close as possible to the sum of the expected noise margin and the standard value, and the measurement time can be shortened.

The data accumulated by step (e) is used to adjust the initial value of the at least one parameter, if necessary. This allows for continuous updating and re-evaluation of parameters.

Step (c) comprises generating pulses of a reference sequence as an output from a pseudo-random sequence generator, and synchronously comparing this reference sequence with the sequence of the system, and detecting the difference between both sequences. It may happen. When implementing the method in this way, it is possible to detect errors at variable pulse positions in a sequence, or it is possible to detect errors at any other position of the sequence. However, this latter is not important for noise margin measurements. However, by being able to detect errors at positions other than the variable pulse positions, it is also possible to detect out-of-synchronization. Alternatively, step (c) may consist of testing the sequence of the system, in the case of a coded sequence, to find that a pulse train in the sequence matches the sign of said coded sequence. This alternative method requires less hardware.

Although the method according to the invention as described above is primarily useful for measuring noise margins and noise margin distributions in PCM (Pulse Code Modulation) reproducing devices such as those used in digital transmission systems, the method also It can also be applied to make similar measurements on equipment.

Furthermore, the invention may also be implemented in an apparatus used for measuring the noise margin (according to the above definition) in a digital transmission system consisting of at least one receiver, the apparatus comprising: (a) a receiver of the system; (a) apparatus for providing a sequence of pulses (as defined above) substantially representative of standard traffic to the apparatus; (b) apparatus for so inserting variable pulses at predetermined positions in the pulses of said sequence; and (c) (d) means for varying at least one of said variable pulses; and (e) said variable pulses also include: (f) apparatus for testing the output output from the detector of the system to which said sequence is sent, including;
(g) a state determining device for determining which of a number of possible output states of said variable pulse is indicated by said output value of said detector; and (g) during testing of said output from said detector. , a monitoring device for monitoring the value of the at least one parameter; and (h) a deviation of the at least one parameter from a standard value for each variable pulse of the plurality of variable pulses to determine a probability distribution of the deviation. and an accumulator that accumulates values.

Among them, the condition determining device, the monitoring device and the accumulating device are provided by a data processing device programmed to control the pulse position, the variation of the at least one parameter and the data accumulation within the sequence. In this way, the design of the device implementing the invention and the program of the data processing device can, for example, when used for testing a PCM playback device, in accordance with the invention, indicate whether the playback device is satisfactory or unsatisfactory in each case. The technician may only receive a "pass" or "fail" instruction from the device performing the procedure. Possibly as an alternative method suitable for use in a laboratory environment, a plotter or visual display device could be employed, through the use of suitably programmed data processing equipment, e.g. noise margin distribution histograms when characterizing playback equipment. research becomes possible.

Preferably, said apparatus for providing said sequence of pulses comprises a first amplifier arranged to provide said sequence of pulses as an output responsive to an input from a pseudo-random sequence generator.

The device for providing the variable pulse may comprise a second amplifier having an output connected to an output of the first amplifier.

Preferably, a device is provided for generating a predetermined subsequence of said sequence, and said device for providing said variable pulse is triggered by a coincidence of said subsequence with a subsequence portion of said sequence. For example, by applying variable pulses in subsequences of 10 bit length, the pulses generated by the sequence applying device (which can use a pseudo-random binary generator) can be divided into two ¹⁰ -pulse sequences. It is possible to uniquely position within the

Apparatus may be provided for suppressing pulse generation in the sequence at the predetermined location in response to the match between the subsequence and the subsequence portion within the sequence.

The device for varying the at least one parameter of the variable pulse is a DA (Digital Analog) device controlled by a data processing device.
It may consist of a transducer.

Further, the system includes a sequence generator having an output, the output and the first;
It may consist of a transmitter that itself consists of a trigger detector connected to a second amplifier.

The monitoring device for monitoring the value of the at least one parameter includes a sequence generator and a trigger detector at least substantially identical to the sequence generator and the trigger detector of the transmitter for generating a reference sequence. and a trigger detector, wherein the trigger detection is performed at substantially the same reference sequence position as occurs in the transmit sequence.

Preferably, a device is provided for synchronizing the sequence generator output of the monitoring device with the output from the transmission system under test.

Apparatus may be provided for comparing the output from the digital transmission system under test with the reference sequence.

At a time position corresponding to the output from the trigger detector of the monitoring device, a device may be provided for classifying the difference between the output and the reference sequence. In this way it is possible to determine whether the difference between the output and the reference sequence is due to variable pulses or to synchronization losses.

Apparatus embodying the present invention is most useful for testing PCM playback equipment on PCM transmission lines where measurements of noise margin and noise margin distribution are important for predicting performance of the playback equipment prior to installation.

Thus, in an apparatus implementing the invention, it is desirable to provide a transmission line from a transmitter capable of transmitting variable pulses to the receiver. Alternatively, the transmission path to the receiver may include free space.

Furthermore, the present invention provides an error rate prediction method in a digital transmission system, which method includes (i) performing steps (a) to (e) of the method according to the present invention as necessary; (ii) for each state and each distribution for that state, quantize the deviations into ranges; (iii) quantize the deviations into ranges for each range; (iv) summing the error rates of each range within each distribution for each state, by all values within each range, determining the error probability that contributes to the total error probability of the system. .

Similar to the method according to the invention, the particular state of the pulse is one of n possible states;
and the steps (a) to (e) are repeated for each of the n possible states, and the variation in the at least one parameter is such that the detected state is one of the n possible states. Preferably, the method also comprises the step of further classifying the parameter variations for each of the detected states and said n states prior to the accumulation step. By repeating steps (a) to (e) a necessary number of times, a noise margin distribution histogram can be created for the relevant state of the pulse. Since in digital transmission systems we are primarily interested in ternary signals consisting of sequences of "0", "+1" and "-1", n may be 3. The method we have developed can be used to predict the error rate of playback devices down to very low levels. This method measures the system under standard (ie, low noise) operating conditions and does not require expensive, difficult to calibrate hardware noise sources. Access to internal circuit connection points is not required and the results obtained allow considerable diagnostics regarding malfunctions.

BRIEF DESCRIPTION OF THE DRAWINGS A detailed description is therefore provided below, which should be read with reference to the accompanying drawings of the method according to the invention and the apparatus implementing it. However, these methods and apparatus were selected to illustrate and explain the invention and are not intended to limit the invention.

As mentioned above, the transmitted signal of a given transmission system is affected by various characteristics of the transmission line as the signal travels down the line. When the information content of the signal is converted into digital format (Figures 1a and 1d), the effects of such characteristics appear as ISI (intersymbol interference), as shown in Figures 1b and 1e. The result is that a single pulse effectively extends beyond its original pulse width and affects adjacent pulses.
When ISI is low, noise in the receiving band
Although the margin is only reduced, at high levels the signal may become completely unreadable without signal equalization. Complete equalization, as described above, consists of passing the received signal through a network whose transfer function is the inverse of the channel transfer function. Complete equalization is not desirable, and it is conceivable to reduce the value of ISI to zero at the sampling time at the decision point (Fig. 1c).
and FIG. 1f) The general principles governing the invention are based on the considerations discussed above. A repeating sequence of states is transmitted over a digital transmission system (DTS) and one state within the sequence is considered. It is assumed that the transmission conditions are fixed. The received value in this state differs from the ideal value due to noise and ISI (intersymbol interference). Noise is a random perturbation and varies with each pattern iteration, whereas ISI is constant and depends on neighboring states. If the perturbation of the received value in the relevant state can be measured, the perturbation necessary for error occurrence can be found. This prerequisite requires knowledge of the receiver's decision threshold, ie, the point at which the receiver circuit changes the classification of the received value from one state to another. The value of the added perturbation is called the noise margin. Given n possible states, there are n-1 ranges of perturbation values or noise margins. Each corresponds to a different inaccurate state being detected. Noise margin is not necessarily a unidimensional quantity. For example, the noise margin of a combined amplitude/phase modulation scheme is expressed as a combined function of amplitude and phase. From the noise margin and knowledge of the amplitude characteristics of the noise source, the error rate at that noise source for that condition can be calculated. If the noise margin of every state of a sequence is known, the error probability of the sequence can also be calculated. These things can be calculated for any value or type of noise.

Our method uses a non-contact method to measure the noise margin, so there is no need to access the inside of the receiver. A method that has been applied to a transmission system with only two binary lines in the transmitting state is described below. The same basic method can be extended to systems with multiple states carried by any type of digital transmission system.

Consider a simple binary system with two transmit states, 0 volts and V volts, represented by "0" and "1", respectively (Figure 2). In this case, assume that there is a fixed repeating pattern, e.g. a pseudorandom binary sequence (PRBS), transmitted over the system. Select one bit of this sequence. To measure the noise margin of this bit at the receiver, the transmitted value of this pulse is perturbed by adding a signal.
It is desirable to do this with minimal disturbance to the received values of adjacent states, and it is desirable to be able to calculate the effect of the applied signal on the receiver decision points without accessing the receiver decision points. The optimal signal is one that is identical in format to the normally transmitted signal, but with a different amplitude. In this way, the applied signal undergoes exactly the same attenuation, delay, amplification, etc. as the standard signal, but appears at the decision point identical in form to the standard signal, no matter how distorted. In this way, if the standard signal height is 10% at the transmitter, the signal height will also be 10% when it appears at the decision point (Figure 3a, Figure 3c, Figure 3e). .

The noise margin of a selected bit can be measured by applying a variable pulse of increasing amplitude to the bit in question, and checking the receiver output state as the transmitted value causes a gradual shift from one state to the other. do. At some value of the variable pulse, the receiver begins to detect the bit as the other state. If there is no noise in the system, this change will occur at a strictly defined level. In the presence of noise, this change occurs over a range of variable signal amplitudes. If the probability of receiving the bit incorrectly is plotted against variable pulse amplitude, some information about the amplitude probability density function and standard deviation of the noise can be deduced from its width and shape. This type of noise is called intrinsic noise. The variable pulse amplitude value with a probability of 50% gives the average noise margin for that bit and contains information about the ISI and the accuracy of the receiver's decision threshold.

Repeating the above procedure for other bits in the sequence will give different results since the ISI depends on the neighboring state. The noise components must be the same. These results can be used to create a histogram of the average noise margin for each transmission state. In this case, there are two states. One is a "0" state, ie, P0(v), and one is a "1" state, ie, P1(v). Here, v varies from 0 when no signal is applied to + or -v when a full height signal is applied. Both distributions are each normalized to unit area, so that both distributions become probability density functions. The probability of a state having a noise margin Vm is P1 (Vm) or P0 (Vm) for 1 or 0 (see Figures 4a and 4b). In this way, the influence of a decision point can be measured by the influence of the rising pulse at that point. Therefore, the noise margin of 0.4V represents the noise margin of 40% of the signal generated by the rising pulse at the determination point of the sampling time. Ideal sampling must be performed at a time corresponding to the peak value. This is because it has the highest noise margin. The noise margin is then measured relative to the peak signal generated by the rising pulse at the decision point. This is not the same as the actual peak at the decision point caused by transmitting a complex sequence. Due to ISI, individual pulses have different wave heights, and the signal peak may be larger or smaller than the peak of the rising pulse (see Figures 3b, 3d, and 3e). ). The error probability of the system can now be estimated by knowing the probability density function (pdf) of the noise signal amplitude at the decision point. Pn this
(v). Consider one state "0" of a binary sequence. Noise margin Vm
For _a particular value ^of
(c)) (1) In this way, for all “0” Pe=∫ ^V ₀ P0(Vm) ∫ ^∞ _Vn Pn(v)dv dVm (2) For complete generalization , the limits of integration for Vm must be +, -∞. This must also be the measuring range. However, in an actual system, the error probability is zero when a pulse that is not modified is transmitted. That is, Vm
→0 or V and P0(Vm)→0.

For state "1", the applied noise signal must be negative to cause an error.

Pe=∫ ^-∞ _Vn Pn(v)dv. 0=>Vm＞=-V (3) In this way, for all “1”, Pe=∫ ^-V ₀ P1(Vm) ∫ ^{- ∞} _vn Pn(v)dv dVm (4) To find the total probability of error in the entire sequence, each probability P(1) and P
Multiply (0) by two equations related to “1” and “0”.

Pe=P(0)∫ ^V ₀ P0(Vm) ∫ ^∞ _Vn Pn(v) dv
dVm＋P(1)∫ ^-V ₀ P1(Vm) ∫ ^-∞ _Vn Pn(v)dv dVm(5)
This allows the
pdf: The probability of an error caused by a noise source in Pn(v) is obtained. If the statistics of the sequence are similar to the standard transmitted signal of the transmission channel, the calculated error rate is also applied to that signal. The noise in the above equation is the sum of two components. The first is
Even if there is a completely noise-free signal at the receiver input, there is noise at the sampling point, that is, inherent noise. The second is the noise added to the receiver input signal by external noise sources. For any receiver input condition, the pdf of the inherent noise is
It can be estimated from measuring the error rate for variable pulse heights. For linear systems, this noise is independent of the selected state. For other systems,
This is not necessary; separate estimates need to be made for each state.

Let R(v) be the error rate as a function of variable pulse height.
Let K be the pulse height for an error rate of 50%. Therefore, the cumulative inherent noise
The pdf is Q _i =R(v+k) (Fig.4(d)) (6) And the pdf of the unique noise is the derivative of the above equation, P _i (v)=dR(v+k)/dv( Fig.4(e)). (7) This distribution is measured relative to the peak value at the sampling point. To determine the combined effect of external and intrinsic noise sources, each of the two pdfs must be convolved.

Pn(v)=P _i (v) ^* Pe(v) (8) Here, the “*” mark (only in this case) means “convolution”, and Pe(v) is the externally added This is the PDF at the noise sampling point. This is also scaled by the "peak" value.

PCM (Pulse Code Modulation) Regenerator Now, a PCM (Pulse Code Modulation) regenerator, such as the one used in the UK Post's 2.048Mb/s digital line system.
Let's consider the case of a PCM line regenerator.
(See Figure 5).

The line signal is a return-to-zero pulse train that has three states: "+1", "-1" and "0" with +V, -V and 0 volts, returning to a half-width "0". V is usually 3 volts and the lines are 120 ohm twisted pairs. Extending the above, P0+(v) is the noise margin distribution for 0 detected as +1.

P0-(v) is the noise margin distribution for 0 detected as -1.

P+0(v) is the noise margin distribution for +1 detected as 0.

P-0(v) is the noise margin distribution for -1 detected as 0.

P+-(v) is the noise margin distribution for +1 detected as -1.

P-+(v) is the noise margin distribution for -1 detected as +1.

There are three types of transmission states, and for each transmission state,
One correct detection state corresponds to two incorrect detection states. This results in a total of six noise margin distributions. Looking at each possible error in turn: Case of 0→+1 error Let P(+/0) be the probability of receiving +1 when 0 is sent.

P(+/0)=∫ ^V ₀ P0+(Vm) ∫ ^∞ _Vn Pn(v)
dv dVm(9) In case of 0→-1 error P(-/0)=∫ ^-V ₀ P0-(Vm) ∫ ^-∞ _Vn (v)
dv dVm(10) -1→+1 error P(+/-)=∫ ^2V ₀ P-+(Vm) ∫ ^∞ _Vn Pn(
v) dv dVm(11) -1→0 error case This is more complicated because the large noise voltage will generate a -1→+1 error rather than a -1→0 error. Therefore, the total error rate is determined and the one due to -1→+1 errors is subtracted.

P(0/-)=∫ ^V ₀ P-0(Vm) ∫ ^∞ _Vn Pn(v
)dv dVm-P(+/-)(12) In case of +1→-1 error P(-/+)=∫ ^-2V ₀ P+-(Vm) ∫ ^-∞ _Vn Pn
(v) dv dVm(13) +1→0 error P(0/+)=∫ ^-V ₀ P+0(Vm) ∫ ^-∞ _Vn Pn(
v) dv dVm-P(-/+) (14) Here again, the limits of integration are chosen for practical reasons. For complete generalization, the limits of integration with respect to Vm must be +∞, −∞. As explained in the simple binary case, we can limit the range of Vm. As the transmit state approaches state x or y, Pxy(Vm) must all approach zero. Therefore, integration only needs to be performed over this range.

The total error rate is the sum of the six individual probabilities. Some simplifications are possible. The first is +- and -
It is to ignore the + error. For moderate error rates, the probabilities of these errors are relatively insignificant. On average, these noise margins are V/2 for 0-, +0, 0+, 0- errors.
It is about 3V/2 compared to Actually, Gauss
Noise with pdf is normally specified. That is, Pn(v)=1/√2πe ^-(v/ 〓 ^)2/2 (15) When V=V/2 and an error rate of about 0.002, the noise standard deviation (σ) must be V/6. It won't happen. V
increases from V/2 to 3V/2, Pn(v)
The value of is reduced to 1/e ³⁶ or 1/(4×10 ¹⁵ ). For the 2 Mb/s line system under consideration, the +→− (or −→+) error is, on average,
Occurs every 548,000 years.

Setting P(+/-) and P(-/+) to 0 has the effect of causing P(0/-) and P(0/+) to contain any of these errors. Therefore, the error is still (albeit small) relative to the calculated error rate.
You will make a contribution. These errors may be misclassified. This simplification eliminates the need to measure P-+(Vm) and P+-(Vm).

Therefore, the formula for the total error rate is: Pe=P(0)[P+(+/0)+P(-/0)]
+P(-1)P(0/-)+P(+1)P(0/+)(
16) =P(0)∫ ^V ₀ P0+(Vm) ∫ ^∞ _Vn Pn(v)dv dVm +P(0)∫ ^-V ₀ P0−(Vm) ∫ ^-∞ _Vn Pn(v)dv
dVm +P(-1)∫ ^V ₀ P-0(Vm) ∫ ^∞ _Vn Pn(v)dv
dVm +P(+1)∫ ^-V ₀ P+0(Vm) ∫ ^-∞ _Vn Pn(v)
dv dVm (17) P(0), P(-) and P(+) are the probabilities of 0, -1, +1 in the transmitted signal.

Another simplification is to assume that P(+1)=P(-1)=P(1). This is a necessary property for a line code that does not have a DC component in its signal. This is usually the case since the transformer coupling commonly used in line regenerators does not transmit direct current.

It can also be assumed that the pdf of the noise P(y) is symmetric with respect to zero mean. (Pn
(v)=Pn(-v)). In reality, this may not be true. However, we do not want to find special cases, but instead want to find a symmetrical collective average.

So the integral becomes:

Pe=P(0)∫ ^V ₀ (P0+(Vm)+P0-(-Vm))・∫ ^∞ _Vn Pn(v)dv dVm +P(1)∫ ^V ₀ (P-0(Vm)+P+0(-Vm) )・ ∫ ^∞ _Vn Pn(v)dv dVm (18) =∫ ^V ₀ P(0)(P0+(Vm)+P0-(-Vm)) +P(1)(P0-(Vm)+P+0(-Vm)) )・∫ ^∞ _Vn Pn(v)dv dVm (19) The second integral of the above double integral is the complementary cumulative Gaussian distribution G(x) defined as follows.

G(x)=√2π∫ ^∞ _x e ^-t2/2 dt (20) Naturally, this only applies when the noise pdf: Pn(v) is Gaussian. This is not a panacea, but all currently specified exams use Gaussian noise, and the most common pdf you will encounter in practice
is definitely this.

Taking out the second integral and substituting it into Pn(v), I=1/√2πσ∫ ^∞ _Vn e ^(-x2/2 〓 ^)dx (21) Substituting t=x/σ and normalizing it, I= 1/√2π∫ ^∞ _Vn/ 〓e ^(-t2/2) dt=G(Vm/σ)(2
2) The value of I is easily determined either from a table or from a polynomial approximation to G(x).

In this way, the integral becomes the following equation.

Pe=∫ ^V ₀ (P(0)(P0+(Vm)+P0−(−Vm)+P
(1) (P0−(Vm)+P+0(−−Vm)))G(Vm/σ)
dVm (23) In actual measurement, Pxy (Vm) (x, y = 0, +,
−) is determined not as a continuous function but as a histogram. In the current case these are:
100 of Pxy(n) (n=1, 2, 3,...99, 100)
Stored in a point array. When the measurement is finished, Pxy(n)
The absolute value of the noise margin is V*(n-
1) It is determined as the number of times between /100 and V*n/100 volts. The average noise margin exhibited by this position is V*(n-1/2)/100. These arrays can then be normalized to unit area so that each location contains a probability of noise margin within each limit. Therefore, the integral is replaced by the summation, and P(e)= ₁₀₀ 〓 〓 ⁿ⁼¹ (P/0)(P0−(n)+P0+n))+P(1)(P+
0(n)+P-0(n))G((n-1/2)/100*
V/σ) (24) An illustrative apparatus embodying the invention is shown in FIG. 6, and as can be seen from FIG. ing. The reproducing device includes an input transformer 14
receives input from cable simulator 10 on line 12 (power supply to which is provided by power feed 15) and transformer 18
is connected to provide an output to line 16 via.

The means of an apparatus embodying the invention for providing multiple sequences of pulses is provided by a pseudorandom binary sequence (PRBS) generator 20 (which may be a Hewlett Packard Model 3762A data generator). Sequence generator 21 in it
provides a sequence of pulses (e.g., 2 ¹⁰ pulses) representing typical traffic over the transmission cable as shown in FIG. The output from the PRBS generator 20 is high density bibolar (HDB)
It is sent to an encoder 22 which converts the incoming input into an HDB3 code (in this case the maximum number of consecutive zeros is 3). Encoder 22 has positive and negative outputs to a combined output amplifier and pulse canceller 24,37.

Output amplifier 24 becomes the first amplifier of the apparatus embodying the invention and provides a sequence of encoded pulses as an output in response to the input from HDB3 encoder 22.

Furthermore, the device embodying the invention consists of a variable pulse 26 which is itself constituted by a combined output amplifier 24 and a second amplifier having an output 28 connected to the output 30 of a pulse canceller 37. Pulse remover 37 suppresses the occurrence of pulse(s) in the sequence of pulses at predetermined positions as described below for replacement by pulse(s) from variable pulse generator 26. It is a device that does

The generator 26 itself is controlled by a computer 50 that varies at least one parameter (i.e., voltage) of the variable pulses such that the initial value of the parameter is set at a level selected to correspond to the expected threshold of two adjacent states. It consists of a digital-to-analog converter controlled by

The variable pulse generator 26 and the pulse deleter 37 are adapted to control the sequence of pulses by the subsequence generator 23 of the PRBS generator 20 in order to establish a match in the comparator 25 between the subsequence and the subsequence portion of the sequence of pulses. Both are activated by a trigger circuit 34 responsive to the occurrence of a predetermined subsequence. (See also Figure 7).

The suppression of pulse generation in a sequence of pulses is carried out according to the coincidence of the subsequence with the subsequence part of the sequence, such that in the sequence provided on the combined output line 31, the deletion pulse is a variable pulse from the generator 26. can be replaced. Depending on the parameters of the pulse to be monitored, for example the voltage level or timing, duration, phase, frequency or amplitude can be varied.

If desired, a sequence of pulses containing variable pulses can be sent to the cable simulator 10 (or the cable simulator 10, for example, when testing a memory or storage device, or when the transmission medium is free space). When not needed, connect the cable to the drip as shown in 33.
simulator can be bypassed). When this sequence is sent through the simulator 10, the output sequence is attenuated and the same intersymbol interference (ISI) added to the output sequence is added to it by the transmission cable itself. This sequence of pulses passes through the generator under test, via the first transformer 14, the second transformer 18 and line 16 to the combined data input amplifier clock recovery circuit 3.
Reach 6. In this circuit, the clock signal is recovered, amplification of the signal provided by the pulse sequence is performed, and the positive and negative outputs are sent to the HDB3 decoder 3.
Supply to 8.

Power to the regenerator under test is provided by power sources connected to the center taps of the secondary and primary windings of the first and second transformers 14 and 18, respectively.
Made from feed. Of course, if the regenerator is at the end of the line, there is no output transformer, but instead there is a single wire output.

The output of the HDE3 decoder 38 is, if there is no error,
It must be identical to the pattern generated by PRBS reproducing device 20. Reference pattern generator 40 controlled by computer 50
can be commanded to synchronize itself to the output from demultiplexer 38. This output then goes to line 4
6 patterns preceded by a fixed number of clock periods, so that after encoder 42 the two patterns are synchronized and the output from encoder 42 is compared with the output on line 46. Therefore, errors are detected. Reference pattern generator 40
A trigger detector 48 connected to generates a trigger at a position in the reference sequence that is substantially the same as a variable pulse position in the pattern of output 31. This timing allows for the classification of errors detected in the error detector 44 according to their position within the sequence (ie, that the error occurs at the position of the variable pulse or at another position). Errors occurring at other positions are taken as a signal of loss of synchronization.

To compare two HDB3 encoded signals, two lines are required for each signal. One for positive pulses and one for negative pulses. Therefore also two error detectors or comparators are required. By "ORing" two lines for each signal into one, it can be simplified without sacrificing accuracy. In that case only one comparator is required.
This is line 4 from amplifier 36 to detector 44.
This is the reason why there is only one 6 and why the encoder 42 has a single output.

Amplifier/clock recovery circuit 36, combiner 38 and reference pattern generator 40
For this purpose, a modified Hewlett-Pacard 3763A error detector could be used. Trigger detector 48 is not required since this detector incorporates its own synchronization logic. Line 46 requires access to internal circuit connections to obtain an output equivalent to that from generator 40.
These outputs with appropriate timing delays are
It is connected to an encoder 42 and a detector 44.

Computer 50 and plotter 52 are devices that determine which of a number of possible output states of the variable pulses are indicated by the output value of detector 44, and plotter 52 provides a visual record of the cumulative value. give. The variable parameter values of the variable pulses can also be monitored and the difference between the output from the generator and the reference generator can be compared to the output from the trigger detector 48 according to the timely position.

The data accumulated by the computer 50 regarding variations in parameters of the variable pulses can be interpreted as deviations from a standard to determine a probability distribution of deviations, and this information can also be used to adjust the initial values of the parameters. The computer 50 is a Heuretsu Patscard Model 9825A desktop computer;
Patsu Card Interface Bus HP-IB (IEE
E488-1975) connected to a desktop computer via
It can be a Model 7225A graphics plotter.

Digital radio is another potentially rich application area for this technology. In the most common system,
2, 4 or 8 phase digital phase keying (PSK)
are using. In this case, there are n states (n=2, 4, 8), each state distinguished by a uniquely determined transmit carrier phase or phase shift from the previous state (differential phase modulation). case). With a suitable phase modulator and a suitable driving waveform, it is possible to derive a similar method of predicting the error rate by exactly the same process as described above. The transmit carrier can be phase modulated by a known sequence leading to a repeating sequence of transmit carrier phases. A selected state within the sequence can be changed by shifting its phase first in one direction and then in the other direction. This results in noise and noise as mentioned above.
A margin histogram can be created, but it is measured in degrees rather than volts. n
Each of the states has a histogram associated with it.
You will have your own. There is one noise margin for each neighboring state.

Therefore, according to the present invention, an accurate probability distribution of noise margin can be easily provided in a short time using a pulse sequence that simulates the normal operation of a digital transmission system. moreover,
In a digital transmission system in which the noise margin probability distribution is accurately given by the measurement method of the present invention, it is possible to easily calculate the error rate for a given noise source.

Furthermore, since the noise margin of each pulse in the pulse sequence is determined, it is also possible to analyze intersymbol interference.

[Brief explanation of drawings]

Figure 1 is a diagram showing the input pulses and the sequence of input pulses, the effects of attenuation and noise on the transmission line, and the subsequent equalization; Figure 2 shows the sequence of pulses for a simple binary line system. The diagram, Figure 3, shows a single pulse and a sequence pulse in a simple binary line system.
Figures 4a and 4b are diagrams showing the response of pulses to modified pulses; Figures 4a and 4b are graphs showing the probability density function of the noise margin for transmission states of "0" and "1", respectively; c is a graph showing the probability density function and error rate of the noise Pn(v) for the transmission state with the noise margin Vm; Fig. 4d
is a graph of the error rate against the pulse height added in the selected transmission state "0", Figure 4e is a graph of the probability density function of the inherent noise in the transmission line, and Figure 5a is a graph of the probability density function of the inherent noise in the transmission line.
shows a ternary line signal, FIG. 5b is a graph of the probability of the detected state in the receiver of the regenerator versus the transmitted pulse amplitude, FIGS. 5c and 5d.
are graphs showing the noise margin distribution for “0” and “1” states, respectively, and Fig. 5e is a graph showing “−”
1” shows the noise margin distribution for the state,
FIG. 6 is a configuration circuit diagram of a device implementing the present invention for measuring noise margin in a playback device under test. FIG. 7 is a configuration circuit diagram of the pseudorandom binary sequence generator of FIG. 6. Figure 8 is a configuration circuit diagram of the variable pulse generator shown in Figure 6, Figure 9 is a diagram of computer interface address information, and Figures 10A to 10F are diagrams using a Heuretsu Patscard Model 9825A desktop computer. 2 shows a flowchart of a routine for compiling a test pulse histogram. 10: Cable simulator, 15: Power feed, 20: Pseudo-random binary
Sequence, 21: Sequence generator,
22: encoder, 24: output amplifier, 26: variable pulse generator, 37: pulse delator,
38: Decoder, 44: Error detector, 4
8: Trigger detector, 50: Computer, 52:
Protsuta.

Claims (1)

[Claims] 1. A method for measuring noise margin in a digital transmission system comprising the following steps (a) to (e): (a) consisting of pulses that can take a plurality of states, simulating the normal traffic of the system; replacing said pulse having said predetermined state at a predetermined position in a repeating pulse sequence of said pulse with a variable pulse in which at least one parameter is variable; (b) varying the value of said at least one parameter such that a detection circuit of said system can detect a change of said variable pulse from said predetermined said state to said state different from said predetermined said state; Process. (c) establishing that said change is detected prior to error correction in said system; (d) (a) for said pulse at another said predetermined position within said repetitive pulse sequence;
A process in which steps (b) and (c) are repeated. (e) For a plurality of the pulses, a variation in the value of the at least one parameter from a standard value in the predetermined state until the change is detected is detected as a noise margin, and the noise margin is substantially equal. accumulating the number of occurrences of the noise margin to be able to determine the probability distribution of the noise margin;