JPS6014540B2 - Receiving machine - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field of the Invention The present invention relates generally to receivers with digital data modems, and more particularly to modems with the ability to rapidly adapt to the characteristics of a transmission channel. be.

Description of the Prior Art Multipoint communication networks typically include a master station that selectively communicates with one of a plurality of slave stations.

The number of slave stations may range from 5 to 12 for a particular device, for example.
Varies between stations. The transmitter and receiver of the master station are coupled by telephone lines to the respective transmitters and receivers of the slave stations. In this type of communication network, the transmitter of the master station and the receiver of the slave station are always coupled.

The master station queries each of the slave stations in turn to determine whether there is data to be transmitted. If a particular slave station has data to send to the master station, that station responds to the inquiry signal by sending an "accept" signal followed by the data to be transmitted. Before the receiver of the master station can accurately receive data from one of the slave stations, it must adapt its characteristics to the characteristics of the telephone line. 0 The characteristics of these telephone lines vary for each slave station, primarily due to the varying distance of each slave station from the master station.

Once the receiver has adapted to the telephone line, relatively slow prior art methods may be sufficient to track the changing characteristics of the telephone line. However, when the receiver is first connected, its characteristics may differ significantly from those of the telephone line. Under these starting conditions,
The relatively slow tracking methods of the prior art required as long as two seconds to adapt the receiver to the telephone line. Since this adaptation time is multiplied by the number of slaves for a single interrogation of the network, this significantly reduced the data yield of prior art systems. Various attempts have been made to provide special start-up procedures to facilitate rapid adaptation of the receiver to the telephone line.

Typically, special permutations of data are transmitted by a transmitter to provide a signal with known characteristics to a receiver. Timing and phase error signals were generated in response to this particular permutation. These error signals were highly dependent on the accuracy of the automatic gain control in the receiver. As a result, there was a high probability of incorrect timing and phase learning. Although these adaptation techniques in the prior art have been effective to some extent, it has always been desirable to reduce receiver adaptation time.

Any such reduction in time is multiplied by the number of stations in the multipoint communication network, significantly reducing the adaptation time associated with a single interrogation of the network. It would also be desirable to provide a signal whose phase, timing, and equalization are independent of the signal magnitude and adjust to current conditions. <Summary of the Invention> The present invention has been made for the purpose of eliminating the above-mentioned conventional problems, and in order to achieve this purpose, the present invention corrects the phase error of the first signal. In a receiver having adapted first and second channels, the first as described above for providing a phase modified signal having first and second components in each first and second channel. a phase correction means for correcting the phase of the signal, and an angle of the phase error of the phase-corrected signal, the angle depending on the quotient of the component of the first signal divided by the component of the second signal; means for providing a phase error signal representative of the magnitude of an angle corresponding to a trigonometric function about which the phase is modified; means responsive to said phase error signal for adjusting said phase error signal and a fixed predetermined adjustment of said phase modifying means for reducing the phase error of said phase modified signal; and means for responding to said first signal component having a specific value.

According to an embodiment of the invention, the receiver includes means for sampling and non-interferentially demodulating the received signal to provide a digital code to each of the orthogonal first and second channels. include.

The equalization circuitry equalizes the incoming signal and the phase modification circuitry modifies the phase of the equalized signal. In response to the transmission of the prologue in the first channel, a quotient is formed by dividing the signal in the first channel by the signal in a second channel that is orthogonal to the first channel. A fast learning network derives the phase error by determining the arctangent of . A large phase jump is made to prevent the arctangent from becoming larger than a reference number. The timing error of the sampling IJ means is
is derived by determining the arctangent of the quotient formed by dividing the current digital symbol in a channel by the preceding digital symbol in that channel. In response to the introductory impulse, the introductory samples in the impulse response are stored while the phase error is calculated from the main samples. The phase of the introduction sample is then modified. Equalization coefficients are derived from samples of the phase corrected impulse response. Therefore, the modem, which is an embodiment of the present invention, has the property of adapting not only the phase and timing characteristics but also the equalization characteristics to a specific communication channel. This is achieved in less than 30 milliseconds.As 20 slave stations are interrogated in a period of less than 1 second, this very short time spans the interrogation period of a multipoint communication network. This is a great advantage.

A further advantage is that the specific timing and phase error signals are yielded regardless of the amplitude of the received signal. This independent relationship is desirable because the adaptation is less dependent on the accuracy of the automatic gain control. the result,
The probability of incorrect timing and learning is reduced. This quick learning technique is particularly suited for use with transmitters and receivers that both have first and second channels.

According to an embodiment of the invention, a prologue of data is transmitted on the first channel, while nothing is transmitted on the second channel. This prologue includes a carrier wave, a modulated tone and a pair of impulses. At the receiver, the carrier waves in the preamble are used to detect transmissions and adjust automatic gain control. During the modulated tone, the phase error is calculated as a function of the arctangent of the quotient formed by dividing the signal in the second channel by the signal in the first channel. To ensure that this quotient does not approach infinity, fast learning circuitry can determine early on whether the phase error is greater than 4y. If it is larger, a larger phase jump of 600 is provided so that the bridge is not larger than the reference number. Also, in response to a modulated tone, the timing is adjusted as a function of the arctangent of the quotient formed by dividing the current signal in one channel by the immediately preceding symbol in that channel. be done.

A corresponding phase jump is calculated from the timing jump. To make final timing adjustments, a first impulse is transmitted to provide signal components over a spectrum of significant frequencies.

Following this final timing correction, a second impulse is transmitted to facilitate equalization adaptation. A final phase correction is desirable before this equalization adjustment is made. This phase modification is calculated from the main samples of the impulse response, which corrects the phase of the leading samples of the impulse response by storing the leading samples and making subsequent phase corrections of these samples. used for Equalization adjustments are then made by calculating brushing coefficients from the impulse response samples.

With these phase, timing and equalization adjustments associated with the apparatus and method of the present invention, the receiver is ready for successful data reception. These and other features and advantages of the invention will become more apparent from the description of the preferred embodiments taken in conjunction with the accompanying drawings. DESCRIPTION OF THE PREFERRED EMBODIMENTS A multipoint communication network is shown in FIG.
Generally indicated by

The communication network 11 includes a main station 13 and a plurality of slave stations 15 and 17.
, and 19, which transmit data via telephone lines 21 and 23. Main station 13 typically includes an office machine 25 and a modem characterized by a transmitter 27 and a receiver 29. The office machine 25 is connected to the telephone line 21 via a transmitter 27 and to the telephone line 23 via a receiver 29. In this type of communication network 11, each slave station 15, 17 and 19 includes an office machine 31 and a modem characterized by a transmitter 33 and a receiver 35.

Office machine 3 in each of slave stations 15, 17, and 19
1 is connected to the telephone line 23 via an associated transmitter 33;
It is connected to the telephone line 21 via an associated receiver 35. Although only three slaves 15, 17 and 19 are shown in FIG. 1, it will be appreciated that communications network 11 can include any number of slaves. Telephone lines 21 and 23 may be those designated by the Bell System as unconditional telephone channels. These channels are leased telephone lines and include a pair of wires for each telephone line 21 and 23. The procedure for transmitting data in such a communication network 11 is such that the master station 13 selectively sends each of the slave stations 15, 17 and 19 to determine whether there is any data to be transmitted. It is started when a query is made to .

If there is no data to send to office machine 31 in slave stations 15, 17 and 19, it typically responds to the inquiry by sending a "disapproval" signal. In the American standard code for information exchange, such a signal is coded as NAK. Although this communication is very short, the receiver 29 of the main station 13 must nevertheless adapt to the characteristics of the fin line 23 before it can accurately receive this signal or any data. The characteristics of these telephone channels 23 will be different for each of the slave stations 15, 17 and 19. In the past, each adaptation of the receiver 29 to the characteristics of the telephone line 23 took a considerable amount of time, for example 2 seconds.

In a network containing as many as 20 slave stations, the total adaptation time was on the order of 4 min. No data was transmitted during this time, and as a result, the production value of prior art systems was significantly reduced. Particular advantageous features for the modem of the present invention include a short 30 millisecond
During the period, there is the ability to adapt to the characteristics of the telephone line.

This feature is based on the receiver 29 of the main station 13 and the slave stations 15, 17.
and 19 are of particular advantage. Because receiver 29 must communicate with each transmitter 33 individually, it must be individually adapted to each telephone line associated with each transmitter 33. These quick learning characteristics are less advantageous in the transmitter 27 of the master station 13 and the receivers 35 of the slave stations 15, 17 and 19.

Since these transmitters and receivers are always on-line, they only need to adapt to each other when they are first switched on. The transmitter 27 is the receiver 3
Communicate with all 5 at the same time. Therefore, each receiver 35
adapts to the associated telephone line until it is switched off.
Maintain its adaptation. The receiver of the present invention has a "quick learning" capability that allows the modem to rapidly adapt to the characteristics of the telephone line 23.

FIG. 2 shows the transmitter 33 in one of the slave stations 15, 17 and 19 in more detail.

The office machine 31 supplies a plurality of data words here indicated by the symbol dj. Here, the subscript j indicates the j-th data word. Following the associated office machine 31, the transmitter 33 includes an encoder 37, which has a speed of 4800 baud per second (bps).
Data digitized from the office machine 31 at a specific rate such as
It is suitable for receiving Tadj. In the encoder 37, the incoming data is randomized and encoded separately to provide D1j codes in phase matched or in one channel and DQj codes in phase 90° different or in Q channels. A suitable randomizer for use in encoder 37 is described in Doelz, US Pat. No. 3,196. A separate encoding device suitable for use in encoder 37 is that described in McAull Ships US Pat. No. 3,701,9481. Data codes ○1j and DQj represent one of a plurality of data levels, the number of which depends on the speed of operation.

For 4800 bps operation, one of three levels (e.g.
32, 0), and one of five levels (eg, 2, 1, 0) in the other 1 and Q channels.

Digital codes are typically generated at a sample rate of 1600 samples per second. Then, the time interval between adjacent codes is 1/1600 second. This interval is usually called the baud interval, and the frequency with which the codes occur is called the baud rate. At a baud rate of 160 equal signs per second, the output of the modem is 160 equal signs per second times -3 bits per sign, or 4800 bits per second. 0 When correctly encoded, the input data symbol is the inhibitor 39
(described in detail below) and are passed to a pair of digital low-pass filters 41 and 43 in the 1 and Q channels, respectively.

These filters 41 and 43, in conjunction with similar filters in receiver filter 29, are tuned to give the system ideal impulse characteristics. After being properly combed,
The signals in the 1 and Q channels are passed to multipliers 42 and 44, respectively, where they are multiplied by digital quantities from a sine-sine ROM 45 at a carrier frequency, such as 1600 hertz.

These modulated signals are combined in an adder 47 and converted to analog form in a digital to analog converter 49. It is smoothed by a low-pass filter 51. In its analog form, the signal is conveyed to telephone line 23. The transmitter 33 also includes a starting permutation generator 55, which is controlled by a timer 57.

Timer 57 is typically clocked at a rate. In response to the inquiry signal from the transmitter 27,
Inhibitor 39 prevents data from encoder 37 from passing to digital low pass filters 41 and 43.

At the same time, the starting permutation generator 55 conveys data in a known order to one channel. No data is sent during the Q channel. As explained in detail below, the generator
The channel may also be connected to convey data of a known nine-fold order. In this case, no data is transmitted during one channel. Data in a known order in one channel is
It is used by receiver 29 to facilitate its adaptation to telephone line 23.

Both the inhibitor 39 and the starting permutation generator 55:
A timer 57 activated via an ENABLE gate 59 is activated in response to the interrogation signal.

In the first state, the enabling gate 59 closes the circuit;
In the second state it operates as a switch providing an open circuit. Once the starting permutation is complete, the timer 57
changes the state of enabling gate 59 to inhibit generator 55 and allow data to be sent from encoder 37 to each of the 1 and Q channels.
The signal transmitted via the Z telephone line 23 is more or less distorted depending on the quality of the telephone line 23. For example, telephone line 23 may transport an entire spectrum of data. This is typically referred to as a frequency offset. Poor telephone line 2 Z3 quality typically results in phase jitter, where the phase of the carrier of the received signal changes in a sinusoidal manner with respect to the phase of the carrier of the transmitted signal. There will also be an initial difference of 2 between the phase of the carrier of the transmitter 33 and the carrier of the receiver 29. This is commonly referred to as a phase offset. Telephone line 23 typically exhibits symmetric as well as asymmetric delay and attenuation distortions.

This distortion is based on the treatment that telephone line 23 imparts to each particular frequency in the subek2tram with respect to the frequency of the carrier wave. For example, some frequencies in the spectrum are subject to greater delays than other frequencies in the spectrum. Similarly, telephone line 23 attenuates some frequencies more than others. Of course, it is desirable that the receiver 29 be able to compensate for all of these undesirable characteristics of the telephone line 23 in order to minimize the difference between the received signal 3 and the transmitted signal. Furthermore, in order to minimize the time used for operations other than data transmission, it is desirable that the receiver 29 be as short as possible and able to accommodate the telephone line 23 in a period of time. Receiver 2
9 is shown in more detail in FIG.

This receiver 29 has the property of adapting to a telephone line, such as line 23, in 30 milliseconds or less to minimize delays before data is accurately received. The receiver 29 is
An input analog signal is typically received from telephone line 23 and passed through analog filter 63 and automatic gain control 65. Phil evening 63
selects the desired band and automatic gain control 65 provides the desired signal level. An analog-to-digital converter 67, which may include a sampler, samples and digitizes the incoming analog signal at a rate of, for example, 6400 bits per second, corresponding to a signed multiple or the baud rate of the transmitter 33. It is set up to do so. Converter 67 is controlled in a dependent manner by a timing signal, which is characterized by a stream of clock pulses.

Converter 67 responds to each pulse by taking a single sample of the analog signal and representing the amplitude of that sample as a digital number. Pulses in a timing signal are separated by a timing interval,
It can be adjusted to change the sampling rate of converter 67. It will be seen that in this embodiment, throughout the remainder of the receiver 29, the signal is in digital form.

The digital signal from converter 67 is sent to a pair of multipliers 6
9 and 71, sine-cosine ROM7
multiplied separately by the amount from 3.

In this way, the signal is non-interferentially demodulated and separated into one channel and a Q channel at receiver 29.

To select the desired baseband from the demodulated signal, the signals are passed from multipliers 69 and 71 to a pair of digital low pass filters 75 and 77 in the 1 and Q channels, respectively.

It was previously mentioned that filters 75 and 77 in the receiver are adjusted along with filters 41 and 43 in transmitter 33 to provide an ideal impulse response.

In the preferred embodiment, filters 41, 43 and 75, 77 are adjusted to provide reduced cosine signaling. This type of signaling is generally preferred over partial response signaling because it provides a faster response time.

Partial response signaling provides a narrow data spectrum, which may be advantageous for transmitting data over communication lines having narrow, medium-bandwidth. On the other hand, raised. Cosine signaling provides a wider data spectrum and faster response time, which is particularly desirable for providing "quick learning" characteristics to modems. The signal at the output of filters 75 and 77 is
Generally designated as XI and XQ for the 1 and Q channels, respectively. Following filters 75 and 77, the two baseband signals in the 1 and Q channels are passed to equalization network 79. This circuitry 79 corrects for symmetrical as well as asymmetrical delay and attenuation distortions produced by telephone line 23. At the output of equalization network 79, the equalized signals in the 1 and Q channels are generally designated as YI' and YQ'. Equalized signal Y
I' and YQ' are then passed to phase correction network 81, which compensates for frequency offsets, phase offsets, and phase jitter caused by telephone line 23. At the output of phase modification network 81, the equalized and phase modified signals in the 1 and Q channels are generally
Denoted as Q. Following phase correction, the signals in the 1 and Q channels are detected in detection circuitry 83.

At the output of detection circuitry 83, the detected signals are generally designated as DI and DQ. The codes forming the DI and DQ signals are expressed in relation to the data dj as shown in the following table. From the output of detection network 83, signals DI and DQ
is communicated to error calculator 87 along with signals YI and YQ from network 81.

In the preferred embodiment, error calculator 87 provides an error signal to adjust the sampling rate of converter 67, the equalization of circuitry 79, and the phase correction of circuitry 81 to current conditions. By providing a single error calculator 87 that depends only on the incoming data, the receiver 2
It is particularly advantageous for receiver 29 to compensate for imperfections in the timing, equalization, and phase modified incoming signal. With the modification of these characteristics, the detected data from the detection circuitry 83 is conveyed to a decoder 89 where the signals are separately decoded, restoring the randomization to the This will be communicated to machine 25. Error calculator 87 provides error signals EI and EQ for the 1 and Q channels, respectively. These signals are expressed as follows. EI: YI-DI's.

EQ=YQ-DQ2.

where YI and YQ are the equalized phase corrected signals in the 1 and Q channels, respectively.

DI and DQ are the detected signals, and 〇 is a current-conditioned estimate of the ideal signal impulse response value 〇. Equalization control circuitry 91 is configured to provide an equalization error signal that controls equalization circuitry 79.
Error signals EI and EQ and detected signals DI and D
Respond to Q.

These equalization error signals from control circuitry 91 are also communicated to a decision value control 93 which generates a quantity value.

This quantity is used to set the threshold of the detection network 83 and is used to determine the error term EI in the manner described above.
and is also used to calculate EQ. Timing control circuitry 94 is made responsive to the EI and EQ signals from error calculator 87 to control the timing of sampler 67.

Timing control circuitry 94 may be of a type responsive to specific signals in equalization circuitry 79. The error calculator 87 also provides a signal 8FLL derived from the error terms EI and EQ.

This signal EPLL is conveyed to phase control circuitry 95, which provides a signal to phase modification circuitry 81. receiver 29
The device for rapidly adapting the telephone line 23 to the telephone line 23 is primarily controlled by the rapid learning circuitry 97. This circuit network 97 is
It is shown in block form in FIG. 3 and in more detail in FIG. Quick learning circuit 97 is adapted to receive signals YI and YQ from phase modification circuitry 81.
These signals YI and YQ each include a plurality of digital symbols generated as a result of sampling by converter 67. For example, the YI signal is the current symbol Y1 occurring in the current baud interval. and the code YI-, which occurs in the previous baud interval. In the same machine, the YQ signal includes the current symbol Y occurring in the current baud interval and the symbol YQ- occurring in the previous baud interval.
Prior to transmitting data onto telephone line 23, a transmitter, such as transmitter 33, sends a special quick learning prologue to a receiver, such as receiver 29. This introductory section may be similar to that shown in FIG. 5 and designated by reference numeral 99. The prologue 99 for the preferred embodiment of the invention has five periods each equal to one solid baud interval. These periods are defined by time mo to T7 shown in FIG. During a first period h-T, an unmodulated carrier wave 101 is transmitted.

This carrier wave may be the same as the data transmission carrier wave and has a frequency of 1600 hertz in the preferred embodiment. During a second period T, -T4, the carrier wave is modulated with a particular frequency to generate a modulated tone 103. The particular frequency may be one quarter of the carrier frequency, which in the preferred embodiment is 400 hertz. At the end of the third period L-T5, the first impulse 105 is transmitted.

Similarly, at the end of the fourth period T5-T6, a second impulse 107 is transmitted. After a delay associated with the fifth period T7, normal data transmission can begin. The time from To to T7 corresponds to the actions performed in receiver 29 according to the flow chart shown in FIG.

During the first period To-T, the receiver 29 detects the carrier wave and causes the automatic gain control to adjust the received signal to a predetermined value. This operation is illustrated by block 109 in the flow chart of FIG.
After a period, such as 10 baud, the signal should be sufficiently adjusted so that receiver 29 better adapts to the timing, phase and equalization characteristics of telephone line 23. During the second period T, -T4, the modulated tone 103 (fifth
) is transmitted to the receiver 29.

In response to this modulated tone 103, a large phase jump occurs between times T and L, as shown in FIG. 6 from block 11. Block 11
2, a third phase jump takes place between times L and T3. From time T3 to L, both timing and phase jumps are made, as indicated by block 114. At the end of the third period L-T5, impulse 105 is transmitted and the timing and phase of receiver 29 jump together according to block 116 shown in FIG.

A phase correction is also made at the end of the fourth period in response to the second impulse 107. This phase correction is the sixth
This is indicated by block 118 in the figure. block 1
The equalization adjustment indicated by 20 is also made in the fourth period after the phase correction associated with block 118. After an appropriate delay associated with a fifth period L-T7, normal operation of receiver 29 begins, as indicated by block 122 of FIG. To describe the second periods T and T4 in more detail, the phase adjustment in the quadrature modulation method is
It has been found that this can be accomplished most quickly by transmitting a tone on one side of the channel and then measuring the phase error.

A particularly accurate and easy-to-perform phase error measurement is
It has been discovered that this can be done by taking the arc tangent of the quantity formed by dividing the signal in one channel by the signal in the other channel.
For example, when the quick learning prologue 99 is transmitted on the Q channel, the phase error is measured by the following equation. M=Shuhada N(
(Equation 1) where Y1o is the code generated in one channel at the output of the phase correction network 81 during the current baud interval.

YQ is the phase correction network 8 during the current baud interval.
is the code generated in the Q channel at the output of 1.

Assuming that the prologue 99 is transmitted during the Q channel and there is no phase error, the sign Y1o will be equal to zero. Thus the quotient Y
L/YQo will be equal to 0, and the arctangent of its quotient will also be 0. Under this condition, the amount △? is equal to 0, indicating no phase error. Quotient Y1o
It has been found that the arcsine of /YQo gives similar results.

Thus, there is no phase error and the signal Y1. If is equal to 0, the arcsine of the quotient will also be ○. However, in certain embodiments it may be more desirable to use an arctangent function. This is because this function goes from 0 to 1 in the range 00 and 450. On the other hand, the sine function ranges from oo to 45o, and is only 0.707 from the magnitude of 0. Thus, the arctangent function provides greater accuracy for more typical low value phase errors. On the other hand, there may be difficulties in obtaining arctangent functions for angles between 450 and 900.

In this range, the arctangent function goes from a value of 1 to infinity. Since it is difficult to maintain accuracy over such a wide range of values, it is desirable in the first place to ensure that the phase error is less than 450.
This is accomplished in the preferred embodiment, which provides a constant phase jump of about 600 when the magnitude of Y1o is greater than or equal to the receiver reference level lo. This is more easily understood with reference to the 400 hertz demodulated tone shown in FIG. If no timing errors existed, this 400 hertz tone would be on line 11.
It will be sampled at a time corresponding to 1. If the timing error is a maximum of 1'2 baud, then a sample is taken at a time corresponding to line 113. This maximum timing error, which corresponds to 1/2 baud in the preferred embodiment, produces a sign, one of the alternating signals having a magnitude of 2.49o. If there were no timing errors, the sample would have a value of 1.7321o. These sizes are shown in Figure 8 as a pair of circles 1 1
5 and 117, respectively. Yen 11
A pair of vectors 119 and 12 related to 5 and 117
1 is equal to lo, respectively, Y1.

The ingredients are described. Q axis and these vectors 119
The angular difference between and 121 is the phase error △? to be corrected. It is. If, corresponding to vector 119, there is no timing error, and Y1.

If the component is less than lo, the maximum phase error is 35.
It turns out that it is 3o. In comparison, if, corresponding to vector 121, the timing error is maximum, and Y1. If the component is less than lo, the maximum phase error is 24.10. In either case, if the Y1o component is less than lo, the phase error (regardless of timing error) will be less than 450. The arc-tangent function is thus particularly desirable for determining the magnitude of the phase error if YL is less than lo.
If Y1. If the magnitude of the component is greater than lo, the reference phase of the receiver 29 is jumped by 600 to make the phase error 4. greater than 450. For example, if the YQ components are equal to 0 indicating a phase error of 90o,
Jumping by 600 reduces the phase error to 300 regardless of the timing error. Y1. In the other extreme case where is equal to lo, a 600 jump of timing error 0 corresponding to vector 119 would result in a phase error of 24.70. A 60o jump with maximum timing error corresponding to vector 121 would result in a phase error of 35.90. These values are tabulated in FIG. Y1. is smaller than lo, the phase error is 45
It turns out that it is smaller than 0. If Y1. is greater than or equal to lo, the reference phase is jumped by 600 to make the phase error smaller than 4y. Under these circumstances, an arctangent function for determining the phase error according to Equation 1 is particularly desirable. The 600 phase jump associated with block 1101 of FIG. 6 is determined according to the following equation.

Δ0=-600SGN(Y1.

YQ) (Equation 2) The direction of the 60o jump is determined by the sign of A■ determined according to Equation 2. When the sign of △◇ is negative, the referenced phase is adjusted clockwise in FIG. When the sign is positive, the referenced phase is the 8th
Adjusted counterclockwise in the figure. The signal ΔI calculated according to Equation 2 is guided by the quick learning network shown in detail in FIG.

This quick learning network 97 each has a pair of conductors 12
3 and 125, receive signals Y1o and YQo.
These signals are both multiplied in a multiplier 127,
A comparator 129 is provided with the resulting product. Comparator 12
9 conveys the sign of this product to the multiplier 131, and this multiplier 1
31 also receives element -60o. Multiplier 131
At the output section of , the signal △◇=-600SGN (Y1.
YQ) is supplied to the conductor 133. Vector component Y1
. is greater than or equal to lo, this Δ0 signal on conductor 133 is used to jump the reference phase. Y1
The size of o is Y1 on the conductor 123. The signal is transmitted to the large Z-knot extractor 135, and the resulting signal is transmitted to the comparator 137.
is determined by comparing with the quantity lo. The output of comparator 137 provides a signal on conductor 139 if the magnitude of the YL component is at least as large as the magnitude of lo. Z on conductor 139 This signal is Δ on conductor 133
◇Used to transfer the signal to the phase control network 95. Since the phase error is guaranteed to be less than 45o, Equation 1
The phase correction described in 1.2 is more easily yielded and determines the exact phase error.

This calculation is done according to the structure shown in FIG. In this structure, the circuit network 97 has terminals 143, 144 and 1
45 includes a switch 141 having a first position, a second position and a third position corresponding to 45. Second switch 147 has first and second positions corresponding to a pair of terminals 149 and 151. When these switches 141 and 147 are in their first position, conductor 123
is connected to terminal 143 so as to transmit signal Y1o to multiplier 153 via switch 141. The conductor 125 passes the signal YQ through the switch 147 to the inverse ROM.
155 is connected to terminal 149. R.O.
The signal 1/YQo at the output of M155 is also connected to the multiplier 15.
Supply the elements in 3. The resulting product Y1o/YQo
is communicated to ARC TAN ROM 157 to supply the quantity ARC TAN YL/YQ. By transmitting this signal through the inverter 159, a signal △small=-ARCTAN(Y1o/YQo) is applied to the conductor 161.
supply. This signal on conductor 161 is communicated to phase control network 95, as shown in FIG. In the phase control network 95, the reference phase angle is adjusted by Δ0 degrees,
This corresponds to the phase angle of the received signal. In the method described above, both large and fine phase jumps as shown in blocks 110 and 112 of FIG. 6 are made using the modulated tone 103 shown in FIG.

Once the phase of receiver 29 is accurately adjusted, the timing is also adjusted using modulated tone 103 according to the following equation: △T=-〇. 50zu ARC side (Y device) (
Equation 3) Here, Y is the signal YQ-, in the Q channel in the current baud interval, and the magnitude of the signal YQo in the Q channel in the previous baud interval is YQ-,
Calculation of the timing error according to Equation 3 is greater than or equal to the magnitude of the timing error shown in FIG.
It will be more easily understood with reference to the diagram.

In this timing diagram, the sample times are divided into evenly spaced vertical lines 163, 165,
167 and 169. Each adjacent pair of these lines 163-169 defines a bow interval. Signal 171 is shown to show the precise timing by t converter 67.

If the timing is accurate, samples taken at times corresponding to lines 163-169 will be equal in magnitude. Signal 173 is also shown;
It is 1/4 baud ahead of 1. A sample of signal 173 taken at a time corresponding to line 165 is
It will be noted that the size is larger than the sample taken at the time corresponding to . Thus, the condition that the magnitude of YQo is greater than or equal to the magnitude of YQ-, is satisfied, and the timing error according to Equation 3 is equal to line 16
3 and 165. In contrast, signal 175 is shown lagging behind signal 171 by 1'4 baud.

A sample of signal 175 taken at a time corresponding to line 165 will have a smaller magnitude than the previous signal taken at a time corresponding to line 163. Under these conditions, calculating the timing error according to Equation 3 is less desirable. This is because the tangent function will be larger than the number of criteria. However, timing errors occur on lines 165 and 167.
is easily and accurately determined in the bow interval defined by . It will be noted that the sample taken at the time corresponding to line 167 has a larger magnitude than the sample taken at the time corresponding to line 165. Thus, the condition YQoZYQ- is satisfied. As previously noted, sampling a correctly timed signal, such as signal 171, provides equal magnitude samples YQo and YQ-.

Thus, the quotient in Equation 3 has a nominal magnitude, and the arctangent of this quotient represents an angle of 45°. 9 in equation 3
When divided by the quantity 00, this yields a quotient of 0.5, and in Equation 3 it is shown that when this quantity is added to the constant -0.5, the timing error ΔT becomes 0. 0.25 as the timing of 29 is shown in 173
Leading the baud signal 171, the tangent of YQ-,/YQ is equal to 22.50.

Dividing this amount by 900 gives a quotient of 0.25. When this quotient is added to the constant -0.5, a timing error of -0.25 is indicated. This error instructs receiver 29 to retard its timing to correct the timing error. If receiver 29 is 0.25 baud behind signal 171, as shown in signal 175, receiver 29 is instructed to retard its timing by 0.75 baud.

This is equivalent to an advance of 0.25 baud for a 400 hertz modulated tone 103. Equation 3 is implemented according to the configuration of circuitry 97, as shown in FIG. Conductor 125 carries signal YQ to delay network 177 whose output is connected to terminal 145 of switch 141 . Then switch 141 is placed in the third position and switch 147 is placed in the third position.
With ARCTANYQ- and /YQ in the first position, signals ARCTANYQ- and /YQ are supplied to the output of the ROM 157. This signal is sent to the multiplier 1 along with a signal having a magnitude equal to 1/90.
79 will be informed. The resulting product is passed to the positive input terminal of summation network 181. This network receives the signal 0.5 at its negative input terminal. The resulting signal ΔT, shown in advance Equation 3, is at the output of summation network 181:
Supplied on conductor I88. The size of YQo is YQ−,
A signal having characteristics corresponding to the condition that the magnitude is equal to or greater than the magnitude of (Fig. 3).

A signal equal to the amount lo' is maintained at terminal 187. Conductor 125 conveys signal YQo to magnitude extractor 189, which provides signal IYQol to the positive input terminal of comparator 191. This signal is also passed through a delay network 193 to provide a signal YQ-, at terminal 185, which in turn passes through switch 183 in the first position to the negative input terminal of comparator 191. It will be done. On conductor 195 at the output of comparator 191, signals IYQI and IYQ
-,l can be used in timing control circuit 94 (FIG. 3) to activate a timing error signal on conductor 188. It is well known that when the timing is adjusted, the phase must also be modified to some extent in response to the timing deviation.

This corresponding carrier phase jump is derived from the following equation. Reference carrier phase error △? =8(ΔT) (Equation 4) where ΔT is the change in timing due to baud fraction derived from Equation 3, and a' is the number of reference carrier angles per baud.

For the specific example given, the carrier wave is 1600Hz
, IHZ per boat, and a equals 3600, the reference carrier phase jump corresponding to the timing jump in this particular example is calculated as follows.

ΔT=−360ΔT (Equation 5) A signal corresponding to Equation 5 is induced by passing the signal ΔT on conductor 188 to multiplier 197, along with a signal having a magnitude corresponding to -360°.

The resulting product provides the signal Δ0 of Equation 5 on conductor 199. This signal on conductor 199 is connected to phase control network 95 (
(Fig. 3), and together with the timing adjustment, the phase that is the reference standard of the receiver is adjusted to the current state. During the third expectation L-T5, the first impulse 105 is the prologue 99
sent inside.

In general, impulses such as impulse 105 are particularly useful for making fine timing and phase corrections. The carrier modulated tone 103 provides a typical duffle sideband signal, which represents information only in the high and low sideband frequencies, whereas the impulses have frequencies that span the entire frequency spectrum. ing. The resulting impulse allows the adaptation of the receiver 29 to make more accurate channel measurements. A typical response of receiver 29 to an impulse such as first impulse 105 is shown in FIG. 11 and designated by reference numeral 196.

Telephone line impulse response 196 by converter 67
The sampling of is indicated by a plurality of equally spaced vertical lines 198. These lines 198 are also designated by the associated symbol YQj. These signs Y
Qj is the main code YQo, the subordinate codes YQ-, , YQ-2 and YQ-3 (generally called introduction codes), and YQ
, , YQ2 and YQ3 (generally called tracking codes). It has been found that the timing error ΔT is calculated using these sign values according to either of the following two equations: YQ, (Formula 6) △T=-rQ-tenNQ, l △T;-〇. Equations 6 and 7 both give good approximations of the timing error for the impulse response across the various channels.

For various timing errors between 0.25 and -0.25, typical values for YQ and YQo are
Shown in the figure.

For each timing error, the values of ΔT calculated according to equations 6 and 7 are also tabulated in FIG. It can be seen that Equation 6 probably gives the most accurate approximation of the timing error ΔT. On the other hand, Equation 7 gives a better approximation of the timing error ΔT and is easier to derive. ΔT in Equation 7 is induced by placing switches 141 and 147 in their respective second positions.

The product YQ, /YQ at the output of multiplier 153 is passed to multiplier 201 together with the element 0.67. The resulting mirror provides the timing error signal ΔT described in Equation 7. This signal is conveyed on conductor 203 to converter 67. When the first impulse 105 is received, it may also be desirable to correct any accumulated phase errors due to the frequency representation.

This error correction is accomplished in a manner similar to that discussed with reference to the method step indicated by block 112 of FIG. A signal equal to that expressed by Equation 1 is thus induced and transmitted on conductor 161 to phase control network 95. As previously mentioned, phase corrections are also conveniently performed based on timing jumps made according to equations 6 or 7.

For the preferred embodiment, this corresponding phase jump is △? =-360ΔT. If necessary, the phase jump associated with the new phase measurement and the phase jump associated with the timing jump are calculated according to the following equations. M = side △T - wind C side (total) (Equation 8) The signal representing Equation 8 is the signal △T on the conductor 203 as element -
360o to multiplier 205.

This product, along with the signal △◇ on the conductor 161, is
7 can be conveyed. The resulting signal, given by Equation 8 above, is conveyed on conductor 209 to phase control network 95. By modifying the phase and timing according to block 116 of FIG.
7, proceed to calculate the equalization coefficients. In order to correctly calculate the equalization coefficient from the impulse, the impulse must be
A period of zero transmitted energy must be provided both before and after the impulse with sufficient protection time. Since the receiver 29 is not phase locked during this zero energy interval, a phase error occurs due to frequency changes in the channel. If the protection time is 10 baud, the frequency change will be 10 hertz and the baud time will be 625 baud.
It is 0rs. The 2.250 phase errors per baud would cumulatively give a total of 22.50 phase errors.

Such a phase error causes a large error in calculating the equalization coefficient. If there is a phase error, it is the main code Y1.

as well as introduction samples YQ-, , YQ-2 and Y
Minor codes including Q-3 will also be affected. All these signs are important in calculating the equalization coefficients. However, the detection and calculation of the phase error according to Equation 1 is performed after the introduction pulse has passed. Therefore, the phase correction at that time is for the main code and the secondary tracking codes YQ, , YQ2, and Y
It will only affect Q3. In the preferred embodiment of the invention, this problem is solved by accumulating previously received introductory samples until the main imbalus sample YQ occurs.

By measuring the overall phase error according to Equation 1,
The receiver phase reference is adjusted to the current state in the manner described above. If this phase modification occurs during the current time interval, only the main sample and the Z-tracked samples will be affected by this phase modification. To modify the phase of the introduced samples, the accumulated values Y1j and YQj are replaced by the corresponding quantities YI↑ and YQ◇ calculated according to the following equations.

Y↑jhiY1jCOS△○±YQjSm△○Y li=
YQjC gourd△◇driY1ISM△◇ (Formula 9) where,
j is −3 or more and 0 or less.

After phase correction of both introduction and tracking samples,
These signals, referred to as collective jobs and Yruj, are
It is normalized as related to the change in vibration according to the following equation: Small = Cape Mizo Fence Here, j is 3 or more and 4 or less.

The phase correction of the introduced sample is performed using the rapid learning network 9 in Figure 4.
The equalization coefficient calculator 211 is shown in block diagram form in FIG.

This is shown in detail in FIG. Rapid learning network 97
, conductors 1 23 and 125 convey signals Y1i and YQj to calculator 211. To be more specific,
These signals are passed to summation networks 213 and 215, as shown in FIG. Circuit networks 213 and 21
The signals from 5 are stored in storage registers 217 and 219, respectively. The code Y1j from the register 217 is sequentially transmitted to a pair of calculators 221 and 223. Similarly, the YQj code is passed to a pair of multipliers 225 and 227. Sine △? supplied by sine/cosine ROM A signal equal to is also passed to multipliers 223 and 225.

Similarly, a signal equal to cosine △◇ provides an additional element to multipliers 221 and 227. Multipliers 223 and 22
The products from 7 are added in adder 229 to provide signal Y8j of equation 9 on conductor 231. The products provided by multipliers 221 and 225 are combined in differential adder 233 to produce the signal Y? of Equation 9. j is supplied to the conductor 235. The signals on conductors 231 and 235 are coupled to provide additional inputs to respective summation networks 215 and 213.
can be returned. Once the phase of both the lead and track samples have been corrected, both the Y↑j and Yaj signals are passed to multipliers 237 and 239, respectively. In these multipliers 237 and 239, the signal is divided into elements 1/Y6 according to equation 10. Multiplied by The resulting mirrors Y moxibustion and Y■ are stored in a pair of storage registers 241 and 243.
are stored in each. After the introduction and tracking samples have been phase corrected and normalized according to block 118 of FIG.
Equalization coefficients are calculated.

The outputs of the 1-channel and Q-channel LPFs 157 and 77 are input to the equalization network 79 (FIG. 3),
j and XQj are expressed by the following formula. ×lj=D1
jhl-, ten D1j-, ten DQjhQ-, XQ=DQj
hl-, 10DQj-, 1D1jhQ-, (Formula 11)
Here, D1j and DQj are the distorted data symbols in the i-th baud interval.

D1j-, and DQj-, are the distorted data symbols in the baud interval immediately before the i-th baud. hL and hQ-, respectively, are the values at the sample rate of the impulse response of the raised cosine characteristic of the distortion of the in-phase channel and the 900 phase different channel of the introduction, multiplied by the aftertone carrier, and the value of the phase correction. Therefore, it is a value obtained by multiplying the sine component and cosine component of Δ0.

In equalization network 79, a transversal equalizer (not shown) is provided with multiple taps with variable coefficients CIj and CQj and multiplies by the codes X1j and XQj.

In the preferred embodiment, the transversal equalizer in equalization network 79 has one fixed tap and a corresponding number of equalization coefficients. For purposes of this discussion, these transversal equalizers have three taps in one channel: CI-, , CI-2 and CI-3 and Q
Three taps in the channel CQ-, , CQ-2 and CQ
Let's assume we only have -3. Let us further assume that the center tap CIo in one channel is equal to 1.0 and that the center tap CQo in the Q channel is equal to 0. Under these assumptions, at the output of equalization network 79, The signal is expressed as follows.

Yrj=X1j CI-30×1j-, CI-20×1j
-2CI-, 10X1j-3-XQjCQ-3-XQj-
, CQ-2-XQj-2CQ-, YQj=XQjCL3
10×Qj-, CI-20×Qj-2CI-, 10×Qj-
3-×1jCQ-3×lj-, CQ-20-X1j-2
CQ-, (Equation 12) Equation 11 is substituted into Equation 12, and the coefficient for the inter-symbol term is set to 0 to minimize the influence of inter-symbol interference.
is solved to have the size of

Following this procedure for the assumed number of taps, coefficients CIj and CQj are assigned to the following calculated values to equalize the XD and XQ signals. CI-,=hl-,CI-
2=(CI-,)2-(CQ-,)2CI・3d(C1
-・)(C1-2)-(CQ-,)(CQ-2)
(Equation 13) The above equation shows that the main intersymbol interference is hl−,
and hQ-, are supplied.

A more general solution considering the introduction and channel distortions hlk and hQk of 900, respectively, is given by the following general solution. CIk=-hlk{k>0} CI.

=1.0CI-,=1hl-2, CI-2=(CI-,)2-(CQ-,)2-hl-2
CI-3=CI-, CI-2-CQ-, CQ-2-hl
-3CIk=-hlk: {kku-3} where hlx=Y
Take k {all k master 0} CQ=hQk{k〉0}
I-, CQ-20CQ-, 10CI-20hQ-3CQ=
hQk{k<-3} (Formula 14) where h
Qk=Y winter k {all k≠0} These equations are satisfied by the portion of equalization coefficient calculator 211 shown in FIG.

The phase is corrected and the regular signal 1/Y valley -. The conductor 24
7 is transmitted to the inverter 245.
This signal is transmitted to multiplier 249 and multiplier 251
is squared by . Its output is connected to the positive input terminal of summation network 253. The phase corrected and normalized signal Y' is also passed to multiplier 249 and squared in multiplier 255, the output of which is connected to the negative input terminal of summation network 253. .

The product from multiplier 249 is multiplied by a base of 2 in multiplier 257 and its output is connected to the positive input terminal of summation network 259 . The quantity Y-2 from register 241 is passed to the positive input terminal of summation network 259. The amount Y taken from the same machine strain resistor 243 is transmitted to the negative terminal of the summation network 253. The signal Y, is the register 24
1 to the register 261 directly, and the equalization coefficient CQ-
, is supplied. Symbol 1/Y valley on conductor 247 -. is transmitted to register 263 and provides equalization coefficient CI-,.
The output of the summation network 253 is connected to the register 263 and supplies the equalization coefficient CI-2. Similarly, summation circuit network 2
The output of 59 is connected to the register 261 and the equalization coefficient CQ
-2 is supplied. The output of summation network 259 is also communicated to multiplier 265 and multiplier 267.

The sign Y-, supplies an additional element to multiplier 267 and the resulting product is passed to the negative terminal of summation network 269. This network 269 also receives the Y valley at its negative input terminal. The outputs of summation network 253 and conductor 247 are connected to multiplier 271 and the resulting product is communicated to the positive input terminal of summation network 269. This circuit network 26
The output of 9 is transmitted to the register 263 and is equalized by the equalization coefficient CI-3.
supply v. The signal on conductor 247 is conveyed to multiplier 265 and the resulting product is conveyed to the positive input terminal of summation network 273.

This network 273 also receives the code Y-3 at its positive input terminal. The output of network 253 is combined with the sign Y in multiplier 225 and the resulting product is applied to the positive input terminal of summation network 273. circuit network 273
is connected to the resistor 261 and supplies an equalization coefficient CQ-3. As described above, Equation 14 is satisfied, providing equalization coefficients to registers 261 and 263 that are passed to equalization network 79 on respective conductors 277 and 279.

These coefficients are combined with signals XI and XQ as shown in Equation 12 to provide equalized signals YI' and YQ'. With the rapid adjustment of converter 67, equalization network 79, and phase correction network 81, receiver 29 is ready to receive data.

It will be appreciated that the rapid learning circuitry 97 and method of the present invention is particularly advantageous because it adapts the receiver 29 to the characteristics of the telephone line 23 in a minimum time of less than 30 milliseconds. This extremely short time period is particularly advantageous during interrogation of multipoint communication systems, since as many as 20 slave stations can be interrogated in a period of less than one second.

This significantly increases the data yield of the system.
Although the invention has been disclosed with respect to particular method steps and apparatus for accomplishing the method, the method may include other steps and the apparatus may be embodied in other ways. This will be clear to those skilled in the art.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a multipoint communications network including a master station communicating via telephone lines with a plurality of slave stations, each including a transmitter and receiver having features of the present invention. . FIG. 2 shows one block of the transmitter shown in FIG.
This is a diagram. FIG. 3 is a block diagram of the receiver shown in FIG. 1, including quick learning circuitry to quickly adapt the receiver timing, phasing and equalization to the characteristics of the telephone line; I'm here. Figure 4 shows
The block of the rapid learning network shown in FIG. This is a block diagram. FIG. 5 is a rapid learning prologue for use with the rapid learning network of the present invention. FIG. 6 is a flowchart illustrating the preferred method of the present invention. Figure 7 shows
3 illustrates the modulated tones used with the initial phase and timing corrections of the preferred method; Figure 8 shows 45o
Figure 2 is a phase diagram showing the effect of a 60o phase jump in response to larger phase errors. Figure 9 shows 450
FIG. 6 shows the maximum phase error when a 60o jump is made in response to a larger phase error. FIG. 10 is a timing diagram illustrating the effects of timing modifications in accordance with the preferred method of the present invention. Figure 11 shows
2 is an impulse response showing large and small samples generated by a converter in the receiver. FIG. 12 is a diagram illustrating timing errors calculated according to the preferred method of the present invention. FIG. 13 shows an equalizer coefficient calculator for adjusting equalization coefficients according to the preferred method of the present invention. 1 1...Communication network between ponds, 13...
...Main station, 15, 17, 19...Slave station, 21,
23...telephone line, 25...office machine, 27...
...Transmitter, 29...Receiver, 33...
...Transmitter, 35...Receiver, 37...Encoder, 39...Inhibitor, 41, 43...Digital low-pass filter, 45...Sine/cosine RO
M, 47...Adder, 49...Converter, 51...Analog low-pass filter, 55...
...start permutation generator, 57...timer,
59...Enabling gate, 63...Analog filter, 651...Automatic gain control, 67...
Converter, 69, 71... Multiplier, 73...
...Sine/cosine ROM, 75, 77...
...Low pass. Filter filter, 79...Equalization circuit network, 81...・
Phase correction circuit network, 83...Detection circuit network, 87.
...Error calculator, 89...Decoder, 91...
... Equalization control circuit network, 93 ... Decision threshold control, 94 ... Timing control circuit network, 95 ...
・Phase control network, 97... Rapid learning circuit network, 99
...Rapid learning prologue, 101...Unmodulated carrier wave, 103...Tone, 10
5,107...invalus, 109...automatic gain control, 110...large phase jump, 1 1
2...Minute phase jump, 1 14,116...
...Timing and phase jump, 118...Phase correction, 120...Equalization adjustment, 123, 125... Conductor, 127... Multiplier, 129... Comparator, 131... Multiplier, 133... Conductor, 1
35...Size extractor, 137...Comparator, 139
......Conductor "141, 147...Switch, 153...Multiplier, 155...Inverse ROM, 1
57...ARCTANROM, 1 59...
... Inverter, 161 ... Conductor, 179
．．・．．・．． ．． Multiplier, 181...Summing circuit network, 183...
...Switch, 189...Size extractor, 191
...Comparator, 193 ...Delay circuit network, 19
7,201,205... Multiplier, 207...
...Summator, 211...Calculator, 213,215...
Summation circuit network, 217...Register, 221, 223, 2
25, 227... Multiplier, 229...
Adder, 233... Differential adder, 237, 239...
... Multiplier, 241, 243 ... Accumulation register, 245 ... Inverter, 249, 25
1... Multiplier, 253... Sum seeker circuit network, 257... Multiplier, 259...... Sum seeker circuit network, 261, 263... Register ,265,267
...Accumulator, 269...Summing circuit network, 271...
...Multiplier, 273...Summing circuit network. Figure 1 Figure 2 Figure 6 Figure 3 Figure 4 Figure 5 Figure 7 Figure 8 Figure 9 Figure 10 Figure n Figure 12 Figure 13

Claims (1)

Claims: 1. A receiver having first and second channels adapted to provide phase error correction of a first signal comprising: (a) phase modification means for modifying the phase of said first signal to provide a phase modified signal having first and second components in respective first and second channels; (b) the angle of the phase error of the phase-corrected signal, which angle corresponds to a trigonometric function that varies depending on the quotient of the component of the first signal divided by the component of the second signal; means for providing a phase error signal representative of magnitude; (c) means responsive to said phase error signal for adjusting said phase modifying means so as to reduce the phase error of the phase modified signal; (d) responsive to said first signal component having a magnitude of at least a reference level for making a fixed predetermined adjustment of the phase modifying means to reduce a phase error of the phase modified signal; means.