DE1762361C - Adaptive equalizer for a digital data transmission channel - Google Patents

Description

The invention relates to an adaptive equalizer. However, this method has the disadvantage that

For a digital data transmission channel with a known cable, the characteristics of the transmission channel

Shift registers, with correlators and with one and must remain constant so in the after

Subtraction circuit, following permanently programmed switching network which he

In the past few years, extraordinary assignment and evaluation have been carried out

can be neatly large expenses telephone systems.

created primarily for speech connections are set up. With the increasing interest in delay deformation in a transmission line Transversal filters are used for the transmission of digital data. A trans Development of facilities necessary with which io versalfilter consists of a tapped delay This can be used with the existing telephone line and several multipliers, vcndener gen can be sent out. In order to achieve this, one tap on the delay line to each had to achieve a series of difficulties is ordered. The multipliers represent the conquered Am of which the greatest is the amplitude and the polarity of the line obtained from the delay status that the typical telephone transmission line at the tap in question a significant delay time distortion. The output signals of these multipliers meadows, d. That is, signal components of certain frequencies are then summed and form the transversal within the filter output in the listening area. By setting the tempering band appropriately are subject to longer delays than tapping intervals and the multiplication factors Components of other frequencies. Although this is ao for each tap, the amplitude characteristics run-time distortion the recognizability of the multipliers on the ristics set so that they Line transmitted speech-frequency signals do not correspond to the pulse characteristics of the transmission line, Digital signals are significantly impaired, which then the filter effectively eliminates the decay the line transferred, processes significantly behind those transferred via the line tugs. as digital pulses. In addition, the transversal filter should

So far, different methods have been used so that the impulse behavior of the det, to avoid the distortion of digital data, which corresponds to the line, and this measure also requires such a transmission wheel, either changes a laborious setting correct by hand. Is z. B. d ^: .e Characteristic of the over circuit or a complicated circuit. Although transmission lines are known, equalization can be achieved in which the VYt is reduced by predistortion of the transmitted signal distortion interference for lines, ie the signal to be transmitted is predistorted its impulse behavior in a certain range, and the other distortions in the varies, this is generally not so satisfactory. Then change the signal in such a way that a signal is similar to a setting in which compensation for the desired waveform is received. This is done specifically for a single line.
The method is of course limited to those cases in which the transversal filters are not restricted to one in which the delay characteristics are adapted to the particular line, they do not result in full-line constant and known. constant access. Such filters can generally

In general, however, the impulse behavior of mine is not used for changes in the line properties Transmission line unknown and 40 th to be adapted. Furthermore, such transcann moreover, change over time. With older universal filters the considerable disadvantage. that sic not Transmission systems. where a compensation work digitally, but the use of a Such unknown features were intended to require analog delay line. Although verden Compensation networks on the receiving cables were sought to digitize such transversal filters, used. These networks lead into the transmission 45 so this requires the use of complicated Imgungspfad an additional delay in the sequence code modulation method and an extended one At frequencies one that the transmission line at the circuit, moreover, the delay of the transmission must be the least delayed. This means that the signal component universal filters are much longer than the decayed ones, which are received first are made of the grain duration of the impulses. Furthermore, the filter can do that peirtationsnetzwerk delayed by a period of time. 50 ratio of signal to noise in the system the delay time of the transmission over the line as a result of the addition of the noise components corresponds to other frequencies. Though ten at every tap.

Such compensating devices are largely counterbalanced. Such a device is in "The Bell System

however, they refer to the substantial Technical Journal, Vol. XLV, 1966, No. 2, p. 255

Disadvantage) that they have to be reset every time, see up to 286. Although this adaptive digitate

must if the line delay changes. Equalizer using a transversal filter «

These settings are tedious and time-consuming and come close to a fundamental solution to the problem,

and usually have to be done by hand, it has an extremely high level of lag

will. is on the order of several seconds.

The French patent specification 1422118 proposes white ββ Furthermore, the setting of the multi-

I propose to apply the received distorted signal to a multiplication factor at the taps of the delay

Amplitude discriminator to be fed to the transmission line by means of step-by-step adjustable damping. The presence of several discrete levels examines elements that are very quickly complex,

and, if one of these discrete ones in the received signal, if increased setting accuracy is required. Level is present, on one of this associated Aus * 6j Despite this effort, the readjustment accuracy

output line emits a binary signal, the correction arrangement generated in this way is still not so good,

iügilaldatcn are in a subsequent switching * that on the transmission of defined test signals for

net / linked in such a way that an equalized digital signal checking the transmission channel and

3 4

There is no need to set the trensversal filter in a precise manner, whereas with one it is stored in this one

den could, level stored binary "0" the sign of the

In contrast, it is the object of the invention to oppose the multiplier output signal,

to avoid these disadvantages of the prior art- The invention will now be described in detail-

den, by creating an adaptive equalizer, 5 ben, Tn the drawings is the

which moves automatically and continuously without any essential- 'F i g. 1 is a block diagram of a preferred embodiment

Great delay in fluctuations in the characteristics of the device according to the invention,

adapts the transmission channel and the received F i g, 2 a simplified block diagram of a

genes distorted signal correspondingly equalized. Data transmission channel in which the facility

This object is achieved by an adaptive io according to FIG. 1 can be used,

Equalizer for a digital data transmission "- Fig. 3a, 3b each a graphic representation of ty-

channel with a shift register, correlators with pische delay characteristics of two types

and with a subtraction circuit, which is characterized by common telephone lines,

It is shown that the corrected Fig. 4a to 4f received last are each a graphic representation of

digital data signal fed to the shift register 15 waveforms in digital data transfers

and the uncorrected received signal is transmitted via a transmission channel according to FIG. 2, where-

Correlators for the purpose of correlating with the corrective in the mode of operation of the correction device

ten signal] is supplied to a corrected signal of FIG. 1, we show _v , in particular

to generate the b: nary data sequences in a

received signal is subtracted, on the corrected pulse shape with and without direct current component,

Derive signal. which are sent out via a transmission channel

The device according to the invention works to be able to adapt, while FIG. 4c shows the typical imsend, by changing the pulse behavior of a transmission channel according to the of the transmission channel is constantly determined and grain- Fig. 2 shows. Fig. 4d shows the shape of a Sipensicrt. The device according to the invention requires as gnals. that the data sequence according to Fig. 4a or 4b does not change any setting by hand and can therefore be included in, and that from a transmission channel operated without maintenance. The catch was that the impulse behavior of Fig. 4c Device can together with quadrature and has, while the F i g. 4e a correction signal or multi-level modulation systems are used who shows the grounding of the device according to the invention, whereby the transmission of more than 3 ° is testified when the signal is received according to FIG. 4d Data bit is enabled. Also will catch distortion. 4f shows the data signal. cLis by channel coupling, which is obtained in such a system when the correction signal after the can occur, are also adaptively cor- Fig. 4e with the received signal after the greed. The impulse-responsive correction in Fig. 4d is combined,

Direction according to the invention enables the transmission 35 Fig. 5 a simplified schematic representation

supply of digital data over a telephone line of a signal sampling circuit. those in the correction ciii

at speeds that can be used either above or in the direction of Fig. 1,

below the Nyquist speed for this line Fi g. ή a simplified schematic diagram

lie. a digital multiplier and an integrator, dvv

Preferably, each of the stages of the sliding 40 in the correction device is nath the I · ig 1 vcrwcn-

register in which the last data received can be dct.

bit is stored, a Kcrrelator from a multi Fig. 6a a truth table for a circuit

plikator and an integrator assigned to the element of the digital multiplier according to the Fi g. 6th

signal received from the transmission channel with the in Fig. 7 a simplified block diagram of a

the size contained in each stage multiply 45 two-channel transmitter, in which the facility after the

and the product; integrate. Invention can be used.

To the further processing of the signal voltages Fig. Fs a block diagram of the receiving part of the

to facilitate, is according to an embodiment of the invention two-channel data transmission system, in which an aii-

Application a scanning circuit is provided, which is based on the received embodiment of the correction device

Signal is used in a series of discrete voltage steps vcr- 5 ° of the invention,

converts by periodically sampling the signal and Fig. 9a shows a representation of Cires typical impulses

provides an output signal having an amplitude equal to the behavior of a transmission line measured in

that of the signal at the previous sampling point the in-phase channel of the data transmission

possesses, the multiplier the output signal plant pach the F i g. 7 and 8.

digital with that contained in the corresponding stage. FIG. 9b shows a representation of the pulse ratio

size multiplied. measured in the in-phase channel of the data transmission

To the impulse response of the transmission signal so transmission system according to FIGS. 7 and 8, in the

to be able to capture as accurately as possible, after simultaneous data transmission via the quadrature

a further embodiment of the invention the channel,

Shift register at least as many stages as the number 60 Flg. 10a and 10b each show a representation of the ty pider during the decay of an impulse transmission see impulse behavior in the different Kanägenen BiM. len when using shaping filters in the data

According to a further embodiment of the invention, the transmission system is used.

the multiplier supplies an output signal whose The Fig. 1 shows a block diagram for a be-amplitude glcicii of the output signal of the Ab- 6j preferred embodiment of the equalizer. The Eiritastkreises and its sign that of the output direction receives from a transmission channel signals of the Abtantkrcises correspond if in the signal that the originally sent digital corresponding stage of the shift register a binary data in a Dutch the Obcrtragunpicigenscniiniiii

of the channel contains distorted shape. Distortion caused by the bin direction leads to the transmission after the pig. 1 determined from the received data at high speed to the greatest Signal digital the impulse response of the associated difficulty.

transmission channels. Hereinafter, a correction 'h The Oesamtimpulsantwort generated over a typical signal in association with the s tragungskanals is illustrated the recovery of the aüsge * IS 4c in the received signal by the curve. The curve U therefore represents the form of digit data sent in untrusted form from the transmission channel 12 (FIG. 2). The facility is adaptive, ie. Changes captured signal, whereby the input information of the impulse response in the transmission channel consists of a single impulse 1. It is then averaged and compensated. to pointed out that curve 20 is positive

Since in a typical transmission channel the same maximum value A is reached at time r (as shown by the vertical line 30 shown as a block diagram in FI g positive line can be transmitted because they can have value for the * tive value. The amplitudes of the handling of direct current signals are not cinge- 13 curve 25 in the following data transmission direction, a modulator 13 is used. The times r, and f _t are at A, (line 31) and at h _t modulator 13 generates a low-frequency output (Line 32).

signal, which amplitudes from the input data, fre- Are the transmission channel 12 on

is frequency or phase modulated. These digital subsequent data bits are supplied slowly enough, see above The input data can be as shown in FIGS. 4a and 4b have component. If desired, curve 25 can have the general appearance. In one Input data before the modulation by the filter In such cases, the individual impulses can be shaped by one, and their function can be differentiated later, since there is enough time between them is written. 45 2 to fade away. Such a plant would be a

The output signal of the modulator 13 runs via error-free data transmission, however the transmission path 15 to a remote location. have the disadvantage that the minimum time between at which the output signal from demodulator 16 to successive data bits of the decay time converted to a received signal. This must correspond to the impulse response curve 25 - a The received signal contains the input data at a considerable disadvantage, which is not compatible with the currently distorted form as a result of the overall impulse response required high speeds of the data over-A, the transmission channel 12. This impulse response transmission.

A, represents the totality of the destructive effects of the die transmitted over a transmission channel

Shaping filter 14 as well as the conversion by input data can be received either the modulator 13 and the demodulator 16 and 35 in a form with constant components (Fig. 4a) Delay and amplitude distortion in or in the form of positive and negative pulses Transmission path 15 (Fig. 4b). In any case, the period between

In a particular system, the signal can see consecutive data bits (eg., Between distortion characteristics of the shaping filter 14, the mark 40 or 40a and the subsequent modulator 13 and the demodulator 16 known 40 Pause 41 or 41a) to be substantially smaller than the be so that compensation can be carried out without difficulty for the entire decay time of the impulse response curve 25. On the other hand, if the transmission channel 12 is supplied with such one-on-one distortion properties of the typical transmission data bits, then the excitation path 15 cannot be known and the resulting received signal can correspond to the signal shown in FIG. I'd change time. When the transmission path 15 has a shape shown with volt lines. It is included in the transmission, it is pointed out in the transfer that the curve 45 of the algebraic path introduces strong delay time distortions. ical sum of the individual impulse response curves

Commercial telephone lines, which corresponds primarily to 25, 26, 27 and 28 ..., corresponding to the data pulses transmitted 39, 40, 41, 42 ...
50 As can be seen, the typical received manner is shown in FIGS. 3a and 3b. Such signal 45 a significant distortion as a result of the lines can be used for transmitting data on the decay of the impulse response curves of the previous. Data pulses passing through the hatched areas. As a result, the amplitude is shown in FIG. 3a, such a line tude of the received signal 45 cannot be the same for signal components below 500 Hz and 55 the data sampling times r ", r" ( _s , /, ... For example above 2800 Hz a delay of up to causes the amplitude of the received signal 45 in 3 milliseconds, while the same line time t ₀ (corresponding to the characters 39 in the (hatched area 21) between 1000 and 2600 Hz Fig. 4a) has a height of -t 4, during in which a delay of less than 500 microseconds times /, and Λ, (corresponding to the characters 40 generated. Another line, as shown in FIG. 3 b 60 and 42) the received signal 45 amplitudes are also shown is, at frequencies between 1000 of the amount of +2 or i 7 may have. Likewise, and 2600 Hz have a delay of less than the amplitude of the received signal 45 in the time ¹ OO Nfikrosekunden, as shown at 22. t _t (ent During the pause 41) there is a little bit of a delay while at other frequencies there is a delay 5. Under exceptional conditions and

is longer. These delays lead to a result; depending on the impulse behavior of the transfer rate Distortion of a modulated digital channel. which is used for data transmission, signals. the sign of the received signal can be overridden via a message These two different running times can easily be wrong, whereby a "character" cmpfan

If a "pause" is actually sent out, the received signal is passed through the switch. Likewise, the received signal can be passed negatively to 68 and charges the capacitor 58 to one (reception of a pause) if it is positive Voltage with a polarity that corresponds to the voltage {generally a character was actually sent out). of the received signal 45 at the point in time when a multi-level transmission is used (where S trelens corresponds to the sampling pulse 46, Ger Ver * has both the amplitude and the sign of the stronger 59, which indicates a very high input impedance received signal information), takes effect Moreover, the voltage present at the capacitor 58 decreases as a result of the impulse behavior of the transfer voltage and, at a low impulse channel, generates an incorrect interpretation of the output voltage received, the level of which and the previous data. ic characters correspond to the span stored in the capacitor 58. The correction device described determines the voltage. Because of the high ring impedance impulse response of the transmission channel used danz of the amplifier 59, the capacitor 58 and generates a correction signal which does not discharge a back-sampling pulse 46 during the occurrence of the successive agreement with the received signal 45. When the following recovery of the originally transmitted data sampling pulse occurs, the capacitor signal discharges. via the switch 68 and is then switched back to Off in FIG. 1 it can be seen that the received loaded the new value of the received signal. Signal (e.g. the signal 45 in FIG. 4d) from a At the output of the amplifier 59 there is therefore a typical transmission channel 12 (FIG. 2) to a signal X , which in FIG ? daf impulse response determination device 50 and ao provides.

is immediately passed to the signal correction device 70. The amplitude X, of the received sampled

The impulse response determiner 50 can be represented by the following equation

is the impulse response of the used transmission correct:

transmission channel and generates a series of output _{χ = d} ^ _{+ dh} + _d u + .. ₀₎ signals A ₁ , A, ... It _n , which correspond to the amplitude of the pulse * s ^r ° '''' * * ^v ' a; itw, rtkurve 25 for the corresponding data - where / = 1, 2, 3 ... indicates which bit (the first corresponds to the sampling times. In the case of the tpic pulse - second, third, etc.) in the data sequence just now cm response curve 25 in FIG. B. would the vaccine has been caught. The value of f /, 'is' · 1 when impulse response determination means 50 output the to the (/ n) ^len time received signal (produce the digital signals corresponding to the value of A ₁ at the time 30 entscheidunjskreis 57) as a binary "1" designed t. (which have a value of -2 and have been, or I if it is interpreted as a binary "0" corresponds to the amplitude 31 in FIG. 4c) and w.-rde. The values of A represent the arrspliUidcn-A, with an amplitude * 1 at the time i, corresponding to the distribution curve 25 at the sampling times 46.
cfiend the amplitude 32 in FIG. 4 c. From FIG. 4d can be reached. that the equation The signal corrector 70 then actually uses 35 (1) the height of the curve 47 at the i "" bit time this series of signals (representing the values of It, an-. For example, at time f, this was shown) to generate a suitable correction / - 4 "data bit (corresponding to bit 42 in the signal that has just been received when combined with the received signal «, Then d, - _{rf 4} - \ 1. That gives, which corresponds exactly to the input data, the 40 previous (/ 1) ^le received bit (corresponding to the transmission channel 12 (Fig. 2) was fed to the Rit41 in the Fig. 4a) was a binary "0" and the are, so rf,, - rf, ^ - - 1. Likewise, rf _{( t} - </., - M According to Fig. 1, the received signal is 45 (since bit 40 is a binary "1"), and <if, _s is first passed to sampling circuit 51, in which the Am- - «', - - ¹ I (since bit 39 is also a The amplitude of the received signal for each data 45 is "1"). If this is inserted into equation (1), sampling time is determined and an output voltage X, it is evident
is generated whose amplitude is constant for a bit time and equal to the amplitude of the at the beginning of X _t = ("H) (4) ^J (-1) (- 2) + (+ I) (I) -I- (+1 ) (0) the relevant bit time of the received signal. In the = 4424 1 - 7, Fig. 4d, X ₁ is therefore shown as a dashed curve47- 50 n)
which corresponds to the received signal 45.

In this example, the signal 45 becomes the Zei- where of course Zi _n = 4, / i, = - 2, / 1, = ^: I and

ten ί _ρ , f ,, t _t ... sampled at what times the /?., - is 0. as from the impulse response curve 25 in FIG

Since impulses 46 occur. 4c can be seen. The equations (1) and (2)

5 shows an embodiment of the S'tinnlabias · - 55 describe the amplitude of the curve 45

circle 51. The signal 45 is then received (Fig. 4d).

via a Schaller 68 7U a capacitor 58 j: e- From FIG. 1 it can also be seen that the ms

directs. The switch 68 is normally open when the signal correction device 70 receives data

however, upon occurrence of the data sampling pulse via the decision circle 57 into the transmission

46 Decided to be entered by a suitable clock 6o register 52.

(not shown) generated once in each bit line. The decision circuit 57 leads the

becomes, as is known per se. As shown in Fig. 4d to register an input signal in the form of a binary "1"

is seen. is the duration of the Dalcnahtastimpulsc to. if the signal from the correction device »70

46 is much shorter than a bit time and can be in is positive. and one of a binary *> () <-.

a facility in which data is processed at a rate of 65 when the signal is negative. With a before / i-eten

4800 bits per second can be handled, a guide in the form of the lint / errer can ti. Love

Then show 50 microseconds. register 52 at least that number of bits

If, according to FIG. 5, the sampling pulse 46, which during the decay pulse of the im-

9 Λ 10

Pulse response curve of the transmission catalysis used has the same switches 63 and 64, which can be transmitted by köinplernern. With the present tary inputs rf ,. "And W ~" can be controlled. As in the example in which the impulse response of the overrun can be seen from the table in FIG. 6a, the switch 63 shown by curve 25 (Pig.4c) closes when η des is set ^{in the / t "n} cell 54 , the shift register 52 should have a capacity S shift register 52 a binary "1" is stored, of at least two bits, since in the three at the same time the complete * bits comprehensive time span of the run-out of the curve becomes mental input for the switch 64 "false", so that 25 has a negligibly small amplitude. The switch 64 opens. The input signal X ₁ from The first cell 54a of the shift register 52 contains the sampling circuit 51 therefore appears at the output · then the binary bit which is in the (/ I) ¹ " ¹ data time point 65 of the digital multiplier 55 (via which the dot has been received. Likewise, the closed switch 63 contains) without a change in sign, second cell 546 contains the bit that contains the other ^{in the (/ 2) th data time} Cell 54η is a binary "0". point has been received, and the n ^u cell 54η so the input for the switch 63 becomes "false". then contains the bit. which was received at the (/ / i) ^len data point in time during the input for switch 64 "true". 15 will. In this case, the input signal from FIG. 1, each cell 54a, 546. Correlating the same correlators 53a, 536 ... output signal of the digital multiplier 55 therefore assigns 53 η to an input signal. Each correlator 53 has the same amplitude, but the opposite polarity, as well as the polarity of the output signal X ₁ from the sampling circuit 51 as the input signal X ₁ . fed. Each of the correlators 53 a, 53 6 ... 53 η As can be seen from the curves shown in FIG. 4, there is a digital multiplier 55 and from FIG. ₁ , the sampled received signal has an integrator 56, of which X 1 (curve 47) in FIG The value of preferred versions are shown schematically. -t 7 on. The data received at the time I ₂ - The following describes the operation of a typical 2 $ pulse consisting of a binary "0" and is now described for the correlator 53n, which is related to the n " ⁿ cell 54 η more in stage 54 a of the shift register 52 belongs to the shift register 52. The digital multipli- stores, which is why rf,, represents a binary "0." The kator55n feeds ckm integrator56η an output digital multiplier 55a interprets d ,, in a signal that has the same amplitude as X, Meaning which is a reversal of the sign de? One and the same or the opposite sign, 30 gan ^ ssignals X ₁ causes the output signal depending on whether the bit of the digital multiplier 55 a stored in cell 54 has the amplitude 7. from a binary “1” or from a binary “0 *. That is, the digital multiplexer 55 η holds the captured signal a binary "1", and this value maintains the polarity of X ₁ if d, "= f 1 (now in stage 546 of the shift register 52 a binary" 1 ") is, or reverses the polarity of X ₁ , 35 stored. It therefore represents d, "a binary" 1 ", if d," = 1 (binary "0"). those from digital multiplier 556 in correlator 536 in FIG. 6 is a preferred embodiment of one that is interpreted in such a way that no digital multiplier 55 is used as a block diagram showing the sign of X ₁ . At the exhibition. The digital multiplier 55 has the gain of the digitafmultiplier 556 is therefore a core 60 with an input resistance 61 of the value R 40 signal with the amplitude f 7 and a feedback resistance 62 likewise from the above equation (1) for the output with the value R . As is known per se, an output signal X _{1 of} the sampling circuit 51 can be seen, such amplifiers have the gain "1", but the output signal from the ^nth digital multi reverses the polarity of the input signal. The multiplier 55 π by the following equation - digital multiplier 55 also has two mutually 45 can be pressed:

d; _ "X, = (d, _") (d,) A _e + (rf, ") (rf,.,) *, + (<*, _") (rf, .YES, + · · · (Rf, ..) (OV _n ) A _n + · · -, (3)

where the digital multiplier 55n is the value of X ₁ 50 or -1, the product is always multiplied by rf, ". The symbols in the

(3) have the same meaning as in the above- ("in) (« / J ⁼ 1.

standing equation (1), and in particular _{SQ asked that m the equation (3) the expression} (d, ") a value of either +1 or -1, each ^6V '

after, cb the η bits earlier received signal 55 (d, _ ") (di") h "

interpreted as a binary "1" or a binary "0"

has been. is always equal to A _n . The equation (3) can therefore

It does not matter whether the value of (d _f ") is simplified to -f-1 as follows:

4 "X: = (rf, _e ) (rf,) A ₀ + (rf,.) (Rf,,) A ₁ + (rf,") (rf,,) A ₂ -r- - - - A _n + · - -, (4)

where equation (4) directs the output from the re- ren 56 α, 56 6 ... 56n. The integrators 56 priiscnlalivcn digilalmultiplikator55 /; (Fig. I) DAR 65 to determine a ver-Dic outputs the digital multipliers 55a 'nüllnismäßig long period of time the average value of the output voltage de ^r delivers, corresponding digital multipliers 55th Although 55 /) ... 55; i will / u the supervising intcgrato- this period of time is by no means critical, it is

11 * 12

in a preferred embodiment of the equalizing transmission channel is a delia function which For the period in which the integration is carried out, there is a cross-correlation between the input and output, approximately 10,000 bit times, but the impulse is still output signals of the transmission channel pointed out that the speed with the answer A, corresponds to the channel, the correction facility adapts itself to changes The value of the pulse response of the transmission channel is constant within the entire frequency spectrum, adapted from the time constant of the integrator 56, and this ideal function is of course physically dependent. The integrator 56 has a comparatively unrealizable. However, it does not require a short time constant, e.g. B. less than borrowed to use a real delta function to get a 100 bit times, the correction device adapts itself to the measure for the impulse response of a transmission channel • e? Ir (fish to the impulse behavior of the channel} 2. Instead, an input signal With such short time constants, however, the use of the disintegrator that contains a finite power can introduce a noise into the system. Its autocorrelation is approximately equal to that of the delta function and its power spectrum within

The F i g. 6 shows a possible embodiment of a 15 approximately equal to the bandwidth of the channel used. Integrators 56. In this case the integrator remains. Fine binary random sequence represents one such

tor % m simply from a capacitor 66, which represents a signal.

the resistor 67 is charged. The integration from the F i g. I can be seen that the slide

period of the integrator 56η is transferred from the values register 52 stored binary data to the

of the capacitor 66 and the resistor 67 input signal for the transmission channel 12 is the same

• is correct, as is known per se. Those on capacitor 66 are, moreover, that is passed to sampling circuit 51

Occurring output voltage represents the time integral received signal just that output signal. Yes

the voltage which is obtained at the output 65 of the digital from the transmission channel 12. This

Multiplier 55 η is present. both signals are determined by the impulse response If the incoming signals are sufficiently cross-correlated as the calculation device 50. around the impulse

if distributed, the transmission channel 12 will receive within a long time response A

period roughly the same number of binary (assuming that the binary input sequence

Receive ones and binary zeros. It is still sufficiently randomly distributed).

becomes the order in which the binary ones and after the F i g. 1 the output signals are off

Zeros appear to be essentially random. 30 of the impulse response determining device SO Under these conditions the output (corresponding to the measured values of h,) of

signals from the correlators 53a, 53 /> ... 53n which the signal correction device 70 uses to ".in ce-

relevant values of A ₁ . A ₄ . .. h " to derive both the appropriate correction signal that is used in a

Height as well as the sign. The cor- union rrit the received signal a return

relators 53 therefore generate an output signal that generates the original input data.

the impulse response of the transmission channel used is possible. An analysis of Fig. 4c leaves the essence

12 represents. In the example according to FIG. 4, the necessary correction signals would be recognized. As

which the output signals of the correlators 53β and can be seen. represents e.g. the amplitude of the curve

53 b the amplitude and the sign according to the 28 at time f., That part of the received

Values of A ₁ and h, respectively, represent those with 31 and 32 represent 40 signals that are directly a consequence of the transmission

are designated (see. Curve 25 in Fig.4c). a binary "1" is 42 (see Fig. 4a or 4b). in the

The received signal 45 contains that the outputs of the correlators 53 have the pulse time /

display response of the transmission channel, also (Fig. 4d) also energy pickup, to which the

from an analysis of equation (4). Are curves 26 and 27 which contribute to the impi- 'mtword-

received data is randomly distributed sufficiently so that 45 curves are obtained from the transmission of the data bits

is the mean value of (d, „) (Cl ₁ ) within a time · 40 and 41, respectively. The curve 26 therefore performs

period is zero. Likewise, the mean value of a contribution in the form of an energy component is + / ».,

(</, · ") (</, _,), Of (d, _ _n ) (d, _.,) Equals zero etc. This (point 320, whose amplitude and sign derri

is a consequence of the fact that point 32 correspond on average. Curve 27 does the same

Autocorrelation of the successive data bits 50 make a contribution in the form of a component - A (point

equals zero. That is, </, _, and d, · take un- 31 ^, the height of which is equal to that of point 31, ie-

dangerous same number of times the values +1 and - 1 at purely but the opposite sign ^aufwe: st (da

random occurrence. If the mean value via curve 27 is derived from the transmission of a binary »0 -

If the time is determined, these factors fall from the originates). The required correction signal (in

Equation (4), and all that remains is the point in time f.) Is evidently equal to the sum of the im-

pressure A _n left. Or, in other words, the pulse response amplitudes at points 3Γ and 32 '.

output signals from the correlators 53 a, 53 b ... 53 π The correction signal must of course from the received

can be represented by the equation catch signal 45 can be subtracted.

Generally the desired correction signal is

- f (d-) Xdt = h (5) ^{6o m emem} special sampling time by the subsequent

J ^v '"" given equation: In the device correction signal which determines the impulse response

tion 50 a Krcuzkorrela- = (rf _f ,) A, f- (d, _a ) A ₂ I ··· (</, ") //" I ·, (6) tion is used to calculate the total impulse response A. , the 65th

To determine transmission channel 12. In particular, where the various symbols correspond to the anr.c-

the principle is applied in the establishment, led symbols correspond,

that when the autocorrelation of the input signal Um this correction signal / u are obtained in

i 762

the signal correction device 70 (Fig. 1) uses a plurality of digital multipliers 71a, 716 ... 71 η , of which each multiplier receives the output signal from the corresponding correctors 53a, 536 ... 53n of the device 50 determining the impulse response as an input signal. As a second input signal, each digital multiplier 71 receives a signal which indicates the binary value stored in the corresponding stage 54 of the shift register 52. The digital multiplier 71a therefore receives w as a first input signal a signal (which indicates the impulse response value Λ) from the correlator 53a . The digital multiplier 71α also receives an input signal d, ,, which indicates whether a binary “1” or a binary “θα, which was received one bit time earlier, is now stored in the cell 54a of the shift register 52.

The function of the digital multipliers 71 is the same as that of the dipital multipliers 55 and can be derived from the function shown in FIG. 6, there is a signal at the output of a typical digital multiplier 71 η , the amplitude of which is equal to that of the impulse response signal / i "(from the correlator 53 α), and which has the same polarity as h" (if d, _n - - ^l is 1, whereby a binary »appears η in step 54 1 ') or the opposite polarity of / i" (if d "= - 1, indicating that in the stage of shift register 54η 52 a binary »0« is saved). It can be seen from this that the output signal of the / i ^lcn digital multiplier 71 ;; is given by (d, ") h" . The desired correction signal is then obtained by adding up the outputs of the digital multiples? Orcn71 α. 71 b ... 71 η in the tuning amplifier 72. The correction course 23 obtained from the amplifier 72 is represented exactly by the line (6) given above and can have the general appearance shown in FIG. 4c.

To restore the original input signal the correction signal from the summing unit is amplified 72 combined with the received signal in the subtraction circuit 73, whereby the corrected received signal results. As can be seen from the example of I i 7. 4. is made by subtniklion the correction signal 23 (FIG. 4e) from the received signal 45 that is shown in FIG. 4 f shown corrected groove / signal cr / cugt (24). It will be on it pointed out that in the idealized example the corrected useful signal 24 exactly to the data input signal of the transmission channel corresponds, as in So F i 1 ·. 4; i or 4 b shown. Furthermore, the amplitudes of all received data pulses equal to one another (FIG. 4 c). d. H . those with the received Sicnal 45 previously associated ambiguity has been completely eliminated. SS

During the sampling of the received pulse, however, the signal 23 cannot be exactly the same as the signal 4rt or Ab . For this reason, the digital decision circuit 57 in FIG. 1 is used to generate a binary "1" when the signal 23 is positive and a binary "0" when the signal 23 is negative. The digital output signal of the digital computer circuit 57 then always has the suitable form for the transfer to the shift register 52 of FIG. it

From the above description it emerges that for the pcniiuuslc measurement of the impulse response values /;,. Il ₁ ... // "and thus for the most favorable effect of the correction device it is desirable that the data are transmitted as randomly as possible. The randomness of the data can be ensured by combining the digital information to be transmitted with the output signal of a generator which generates a pseudo-random sequence. The combination takes place in a modulo-2 adder. The modulo-2 addition of a pseudo-random sequence with the input information results in a data sequence which, in turn, is pseudo-randomly distributed. To recover the original information, the corrected useful signal can be combined with the output of the same generator generating a random sequence in a further modulo-2 adding device. If the input data introduced into the transmission channel 12 have been combined with a pseudo-random sequence at the transmitter end (FIG. 1), the data output signal from the subtraction circuit 73, for. B. be combined in the modulo-2 adder 76 with the pseudo-random sequence from the generator 75. If the generator 75 generates the same random sequence as is used in the transmitter, the output signal of the modulo-2 adder 76 precisely reproduces the originally transmitted data signal.

The structure and mode of operation of a generator 75 generating a pseudo-random sequence are known per se (cf. for example "Digital Communications With Space Applications" by S. M. Golomb et al., Prentice-Hall, New Jersey [1964]). The modulo-2 adder 76 consists of a circuit that acts according to the following table:

Output signal
of pseudo-random
gcncraiors to
transferring
Data Ran out of
mis dem
Modiilu-2 · adder C - C - CC - - - cc -

On the other hand, by using linear coding switches together with the data to be transmitted, a sequence can be generated which is almost randomly distributed. In order to recover the original data, the corrected digital signal from the digital decision circuit 57 can be passed through a reverse-coding switching mechanism. Linear coding switching mechanisms, which are useful for the correction device described, are described in the article "The Synthesis of Linear Sequential Coding Networks" by D *. A. Huffman, reprinted in the book "Information Theory". Colin Cherry (Ed) Academic Press, New York (1956). When using linear coding switchgear, the Pscudozufatfsgcnercn on the transmitter and receiver do not need to be synchronized with each other.

In order to be able to utilize the transmission channel as effectively as possible, two orthogonal lower edges with a multi-component module can be inserted can be used with any alcohol. With this arrangement, more old one can be used in each NyquiM interval tniormalionsbil are transmitted (a Nyquist * lnicr · viiiI is the line period in which one another * following impulses from a can! can be transmitted without the peaks of the received pulses stttrcn each other, with the corresponding

(0 \ 762 361

15 16

Nyquist speed, measured in hits per second, shows that there is a fairly long swing ahead.

sen which is numerically equal to twice that, i.e. H. the end of the pulse curve 100 points

available channel band width in Hertz) · still has considerable amplitudes. By pre-distorting

An example of such a modulation system for the input data by means of the shaping filter 81 without multi-level amplification modulation is shown as a block. 5 shows the effective pulse behavior of the same channel circuit diagram in Fig. 7 (transmitter) and in Fig. 8 (curve 101 in Fig. 10a) a substantially kiir- (receiver) shown. The following description of the decay time than with unfiltered embedding with the associated IMg. 7.8 is only used for output data. This improvement in the impulse response the explanation of the use of the described Ent- the transmission channel itself supports the equalizer, ίο distortion of the transmitted data.

This system sends out a four-gate signal. According to FIG. 7 the phascnglcich modudas receives by combining two modulators 82 phase-inverted by 90 °, a carrier from the oscillating plate AM wave is generated. The working gate 84, the frequency of which is within the pass band mode suitable modulators and DC modulators of the transmission path 15 is located. Using for such a Vicr-Vcktor modulation device 15 of commercial telephone lines the free on pages 202 and 203 of the book “Data Transquenz. des Os / illalors 84 are approximately 1800 Hz, mission «by William R. B cn nett and James R. The output signal of the modulator 82 consists of Davcy, McGraw-Hill Book Company, New York. a carrier that was in phase with the starting 1965, described. The signal of the oscillator 84 shown in FIGS. 7, 8 is when out of the forest Setup will create a pseudo-random sequence vcr- ao mungsfillcr 81 a binary "1" is received, and helix to ensure 711 that the iilicr reverses the radar with respect to the signal from oscillator 84 data transmitted through the transmission channel are random. 180 '^ is out of phase when the datcnhil is off Furthermore, in this system, a second version of the standardization filter 81 is a binary “0”. Form of the correction circuit used in the case of the phase shift of 90 modunicln only a compensation of the impulse response the as lierendc modulator 83 receives a signal from the In-phase oscillator 84, phase shifted by 90 ', whose phase is controlled by phase shifter 85 Transmission canal successful, but also a grain of 90 ° in relation to the os / illalorsinn il pc «sation of the distortion as a result of the mutual shift. Is the shaping filter 81 ' Influencing the channels. receive a binary "1", the output

The F i g. 7 shows as a block diagram the transmission part 30 .signal of the with a phase shift of 90 of a digitnldalcniihcrtriigungsaniagc. lvi the one modulating modulator 83 from a carrier, the four-Vtkloi -! * '«?:! ··' ·:! · !!!! Dt.!: cI'l ·; '!. ithe phase: uifwei-.t like the Air .g-jnjs · ' ·. ·,?. 'Vm-inij'.cn in the modulo-2 adiler device 78 of the phase shifter 85 (i.e. on the left the carrier rushes by 90 with the output signal of the pseudo-infill indicator. With reference to the output signal of the ovillary device 84 77, the data to be transmitted are first in a 35 before). If, on the other hand, the output consists of the for-zuiallsonlnimg. These circuit elements mungsfillcr 81 'from a binary "0" have an effect in the same way as a carrier with a phase difference with the fi μ already in connection with the modulator 83. 1 described. The random signal of the phase shifter 85 obtained from the Mo shift of 180 with respect to the output di '! 40 bczuj: on the output signal of the oscillator 84 Fig. 2 shown transmission channel 12 supplied to).

will. The forming filters 81 and 8i '(FIG. 7) form the output stage of the two modulators 82

speak the shaping filler 14 in the IMp. 2, wäh- and 83 become in the buzzing noise 86 to

rcnd the rest of the Fig. 7 the modulator 13 combined in a single Atispangssipnal that the

the IMg. 2 corresponds. 45 transmission path 15 (IMg. 2) is supplied. the

As from the IMp. 7 can be seen. the frequency liic-cs output signal corresponds to the f'n -

fnlNv ^ rlcillen input data of an Atifspaltcinrich- qucnz. of the oscillator 84 and has periodic l'ha-

limg 80 supplied, which the Einpangsdattnbits abwceh- scniinderunpcn on. In the IdcalfaMc the pure

We first applied a phase shift modulating in phase with this signal either

Modulator 82 and then to one with a Plinsen- 50 - ¹ 45. 45. 1 1.15 or 135.

Displacement of 90 modulating modulator 83 Are from the file Aufspalliinpscinriehlung 80

directs. That is, the first of the data ".pa! Lunps" received two consecutive binary "l" -cn.

cinriclttun |! HO received Dntcnbit becomes the Modit- so the output of the modulator 82 from one

lalor 82, the second Dntcnbit tin modulator 83, da * LF signal, which phase the AtwgfinpMigiinl des

drille Diilcnbit back to modulator 82, the fourth modulator "8.1 tim" ΛΓ nnclicill. Here (toamt phase

Difference in the modulator 89 and so on.

gilalsehallung, which rungsschaltunn to carry out this function. 8fi therefore cheats I 45 ". As well as that

is required, is required. The phase diagrams in FIG. 7n identify the modes

Intorcn 82 and 83 ztipeflllirtcn Dntnnbils are rare, a (Icsamlphascnverschicbunp of

passed through the shaping area 81 or SI '. The 60 135 "generated when from the Dnteniuifspnliunp-

Preforming of the input channels leads to a small device HO two ati-differentiating rings

fieren decay in the received data '»0« -cn cttipfiinpcn become. Solllcn ntifeinnder-

Impulses. For example, the graphic shows · the following bits consist of a »0« and an »I«,

development in Fig. 9a the impulse indicator of the phase so there is a phase shift by * '·) ·; ,, *

same channel "with a typical telephone line * s Ebcn" Mi a phase shift of \ ψ> takes place

de «US type 4B using a carrier - if two from binary» 0 «-en bcsiehemfe DflCenW»

frequency of 18110 Ha, however, occur without use. ..,

a »pulse shaping fiHcrs. From the P ig. 9n is to Die PI g. 8 / clfit nls ninckMhnlfMftr «fci» ranpwnjc ·

17 18

part of a data transmission system, the elements together have means for determining · ii dir ι.
with the transmitter part shown in FIG. 7, the impulse response of the in-phase channel rope · ■ '. ">> l»

can be turned. One from the transmission path via the digital multipliers 71 / (only a multi

incoming signal is shown at the same time to the phase multiplier) and via dew Sumnii'Min j; < · I-

equal to the demodulating product detector91 and the 5 stronger 72 / a correction symbol generates winl, il: u die 90 "phase-shifted demodulating product distortions compensated.

detector92 in the demodulator 16 'supplied. The product also has the corrckluriiiilrrfiiiüilung 26 duktdetektor91 also receives a subcarrier from (Fig. 8) to a scanning circuit 51c / vinen digital the oscillator 93, whose frequency is that of the oscillator decision code 57 </, a .SthiJ'iTegister52r / and gate 84 ( Fig. 7) is the same. Likewise, the product receives the correlators 53 r / (from iIlik.i in FIG. 8 only detector92 an auxiliary correlator generated by the oscillator93 is shown) iii further means are carriers, the phase of which the phase shifter 94 is provided by I 90 ″ The mode of operation of the demodulator 16 'quadrature channel itself. The digital multipliers are known per se and are, for example, in the book "Data 7Ir / (only one shown) and the summation converter" (loc. cit.) from p 203. 15 stronger 72 q then generate a suitable correction Arn output of the in-phase product signal for compiling the channel distortions.
detector 91 is a signal not dissimilar to that. The correction timer device 95 also has a second set of correlators 113 / (in which the alternating bits of the original input F i g. 8 only shown in the correlator) that contains the data. The distortion of this phase-matching stages of the quadrature channel shift register signal affects in part the pulse input 52 q are assigned. The correlators 113 / can word Λ of the transmission channel used. That consist of the same design as the Vcrbin impulse behavior of this in-phase channel can be shown in FIG. 6 has been described. The typical correlation curves in FIG. 9a correspond to gates 113 / correlate the received signal in the (no pre-forming filter) or the curves in the 25-phase channel digitally with the data bits shown in FIG . previously received via the quadrature channel and stored in the output signal of the in-phase product shift register 52 q. The delektors91 shows also Vcr / errungscffekl ·.,: \\ Ausgangssignalc the correctors 113 / therefore represent a consequence of Kanalverkepplr strength and plcich / Citi effective impulse response of the phascngleichen Kagen transmission of alternating each other 30 Nalles is that a series of simultaneous Transfer data bits are in the quadrature channel. This coupling of data in the quadrature channel is. This transverse distortion can be shown as an impulse distortion between the channel impulse response corresponding to curve 102 in the quadrature channel and the in-phase channel. 9b or the curve IQ? in Fig. 10b.
posed and can throw the appearance of the curve The exit of the Konelatoren 113 / becomes a

102 in FIG. 9b (if the associated digital multiplier 114 / is supplied in the transmission channel 35 (in no form filters are used) or the curve in FIG. 8 shows only a multiplier), the

103 in Figs. 10b (if no COS ² filtering is also used for the content of the relevant cell of the shift). Register 52ο is received on this. The Digitalmultiplikatorcn pointed out that a COS ² filter significantly reduces the distortions caused by the channel 114 / enable the multiplication of the cross coupling. Quadrature channel received and in the shift register

The output of the quadrature channel product detector 52r / have been stored. The outputs of the tors92 consists of a signal similar to the curve 45 digital multipliers 114 / are also passed to the Sumin of Fig. 4d, which contains the other data bits, the output amplifier 72 /,
which have not been recovered from the in-phase channel. This quadrature channel signal 72 / represents a combined correction signal which is itself distorted as a result of the impulse behavior and compensates for the distortion of the in-phase channel and the transmission channel and the channel coupling as well as that of the quadrature channel. This effects from the in-phase channel. combined correction signal is derived from the

In the case of the data 50 phase equality product detector 91 shown in FIG

receivers are subtracted signals in the subtraction circuit 73 / for the two channels mentioned,

separate sub-devices 95 and 96 for the The output of the subtraction circuit therefore provides

Correction of the received pulses used. represents the corrected signal of the in-phase channel

Furthermore, each correction unit is so set up and contains those data input bits that have

directed that they were transmitted to both the cross-channel impulsant 35 and the phase modulator 82

word as well as the impulse response of the in-phase (see Fig. 7).

Channel. There is a combined correction. Correction subdevice 96 is also included

signal derived that the distortion from both a second set of correlators 113 <? Mistake,

Sources compensated and for correcting the received the signal received via the quadrature channel

fungcncn signal in the associated phase or βο (which was tapped from the scanning circuit 51 q ) with the

Qundrainrkanul is used. Corrected data bits that were previously in the same-phase channel The correction program 95 (Fig. R) has received and stored in shift register 52 /

on a sampling circuit 5t /, a digital decision. The outputs of the correlators 113 <?

circuit 57 /, a shift register 52 / and the correlate »· therefore order from signals that the coupling

IVIi 5.1 /. of which in FIG. K only one corrclutor corresponds to 6 $ pulse rates. These values are JurtfcMvlll Im. Which circuit elements then the Indians from the Dißitalmultiplier 114? (in the Fig. I directly numbered and correspondingly numbered Fig. 8, only a multiplier is shown) with the added SchoolingM-lcmtntcn same. This circuit · »- before from the shift register 52 / of the same phase

Kansils received data multiplizicrl. The outputs of the digital multipliers 1I4 </ are also routed to the summing amplifier 11q . The output signal of the summing amplifier 72i / therefore consists of a correction signal which compensates for the distortion which is a consequence of the impulse response of the quadrature transmission channel and also of the frequency coupling and arises from the simultaneous transmission of data over the in-phase channel.

The correction signal from the Summierungsverstärker72r / q is from the subtraction circuit 73 together with the received quadrature channel signals (from the 92 Quaclraturproduktdetektor). The output of the subtraction circuit 73 <? therefore consists of the corrected received quadrature channel signal, ie the signal contains those input data bits which were originally sent out via the quadrature module 83 (FIG. 7).

Finally, a data combiner 97 is used which the corrected received signals from the in-phase channel and the quadrature channel combined into a single data stream, which is equal to the randomly distributed input data that are present at the input of the data splitting device 80 (FIG. 7). The original dates can then recovered using the pscudo random generator 75 'and the modulo-2 addic device 76 '. The generator 75 should generate a random sequence similar to that from the pseudo-random generator 77 generated random sequence, the output of the modulo-2 adder 76 'becomes the same the data that the modulo-2 adder 78 in the transmission system according to FIG. 7 were originally fed.

The described pulse equalizer can of course also be used together with micro level modulation devices. In this case, the impulse response of the transmission channel can be obtained by correlating the sampled signal received only with the most significant bits previously received. A suitable correction signal can then be obtained by first multiplying the impulse response with the previously received most significant data bits and with the previously received less significant data bits, after which the products are summed. To a grt
To obtain greater accuracy, the products derived from the most significant bits can be valued higher in the summation than the products derived from the less significant data bits.

Claims (5)

Patent claims: 1. Adaptive equalizer for a digital transmission channel with a shift register, with correlators and with a subtraction circuit, characterized in that the last received, corrected, digital data signal (24) fed to the shift register (52) of the signal received from the transmission channel (45) is fed to the correlators (53) for the purpose of correlation with the corrected signal in order to to generate a corrected signal (23), which in the subtraction circuit (73) from the received signal is subtracted to derive the corrected received signal. 2. Adaptive t. '/ Tzerrer according to claim 1, characterized in that the shift register (52) has η stages (54) for storing the last received digital data bits and a correlator assigned to each corresponding stage of the shift register, each correlator having a multiplier ( 55) for duplicating the signal received from the transmission channel with the size contained in the corresponding stage and an integrator (56) for the integration of the output signals of the multiplier. 3. Adapüver equalizer according to claim 2, characterized in that a circle (51) is periodic samples the signal received from the transmission channel and provides an output signal that has an amplitude which is equal to that of the signal of the previous sampling instant, where the multiplier digitally matches the output signal with that contained in the corresponding stage Size multiplied. 4. Adaptive equalizer according to claims 2 and 3, characterized in that η is at least as large as the number of bits transmitted during the decay of the impulse response. 5. Adapüver equalizer according to claim 3 or 4, characterized in that the multiplier supplies an output signal, the amplitude of which is equal to that of the output signal of the circuit (51) and its sign either corresponds to that of the output signal of the circle (51), if in a binary "1" is stored in the corresponding stage of the shift register, or from the cnlgi ^ engeseizien Sign of the output signal of the circle (51) is if the corresponding level contains a binary "0". For this purpose 2 sheets of drawings