DE2133037B2 - Automatic setting circuit for adaptive equalisers in PAM TDM systems - has additional periodic signal as frame synch. signal - Google Patents

Abstract

The automatic setting circuit which is used for an adaptive equaliser (i.e. a transversal filter equaliser in a TDM system for PAM), has an additional periodic signal transmitted besides the communications signals and acting as the frame or group synchronising signal. At the receiver, samples of this regenerated additional signal amplitude are taken from samples of the amplitudes of all incoming signals and setting criteria for the equaliser are derived from the difference signal and from the regenerated additional signal. This additional signal or synchronising signal alternates in polarity so as to give equal positive and negative amplitudes.

Description

T 4

Method only inadequate equalization, since the difficulties outlined above are as possible the course of the cable attenuation only roughly approxi- mately completely avoiding method for automatic Is beer. Adjustment of an adaptive equalizer in a time-next is from the publication »An adaptive multiplex system for pulse amplitude modulation to receiver for Digital Signaling Through Channels 5.

with intersymbol interference "by JG P roa kis Starting from a method for automatic _e nd JH. Miller in "IEEE Transactions on In- Setting an adaptive, the reversal formation theory", Vol. IT-5, No. 4, July 1969, linear distortions on a transmission link pp. 484 to 497, in particular p. 488, a method for Serving equalizer, in particular known transversal filter automatic setting of an equalizer, io equalizer, in a time division multiplex system for PuIs- <Jas to form the setting criterion of an amplitude modulation, in which, in addition to the known test signal sequence, makes use of which in directional signals a periodically recurring additional regular intervals in the signal to be transmitted is transmitted, in which information flow is inserted at the receiving end. These test samples of the amplitudes of the regenerated signal sequence can be reproduced error-free on the receiver side 15 additional signal from samples of the amplitudes and is used to minimize the incoming signals and the mean square error contained in the signal. This setting criterion is not useful if the additional signal setting criteria for setting elements of the high crosstalk attenuation between neighboring equalizer are derived, this task is required for barren channels. This requires zero resolved according to the invention in that the frame synchronization signal serves as additional passages of the impulse response of each channel to the signal.
Points in time at which the information is scanned in the other channels. results when the frame synchronization signal in "Computer Simulation of an Automatic Adaptive" alternately positive and negative, equal in amplitude Equalizer for Real Telephone Channels and Free 25 and, measured on the transmitted signal, Data Format "by F. E gg ima η η, »International pulses with maximum amplitude are used. Seminar on Digital Processing of Analog Signals «, it is advantageous if for the formation of the setting Zurich, Switzerland, 11th to 13th March 1970, the method will criteria the regenerated additional signal compared to the insertion of a known test pulse sequence in the difference signal step by step by integer multiples The information to be transmitted is extended to the effect that the time interval between two successive samples is also taken into account, if products of the zero crossings of the impulse response of each sample of the step-wise shifted regene channel to the left and right of the maximum value ^ atzsignals and the difference signal formed this impulse response forces lying points in time, if a sum is formed from the products of each step in which the information in the adjacent channels 35 and if these sums are queried as. However, a procedure will serve setting criteria here.

used, in which to set the variable Next, it is advantageous if, to form the parameters, not the deviation of the distorted setting criteria, the regenerated additional signal against the ideal impulse response, but the distorted over the difference signal gradually determined by integer impulse response at the sampling times 4 ° multiples of the temporal distance between two each other. As in the publication by RW Lucky, the following samples are added when the products "Techniques for Adaptive Equalization of Digital from the samples of the gradually shifted Communication System", in "The Bell System regenerated additional signal and the sign of the Technical Journal", February 1966, pp. 255 to '286, difference signals are formed, if from the is shown, this method is unfavorable because a sum of 45 products of each step converges only slowly. In addition, the following maximum value, referred to as the main value when these sums are used as setting criteria, came about. Finally, it is advantageous if the setting of the impulse response is not regulated to the setpoint value of the setting elements when interference signals occur. For the correlation the delayed equalizer is slowed down or blocked.
Output signals with the reference signal at the output 50. In a preferred embodiment according to the reference signal generator or the decision of the invention, analog multipliers are necessary adjustable circuits are provided as setting g'.ieder step-wise circuit, which borrowed with the help of what makes the instrumentation difficult . For de-digital-to-analog converters or for magnetic distortion of the following oscillators, a feedback memory, in particular transfluxors, is used who implements transversal filters. This structure can, in particular 55 den.

when faults occur, instabilities and as adjusting elements, bridge self-excitation can also be advantageous to lead. A circuit is used as the test pulse train, which is a voltage binary bipolar pseudo-random signal, from which negative feedback linearized field effect transistor as at regular intervals an impulse in the sequence you contain controllable element.

information-carrying PAM-Sipnale is inserted. 60 It is also advantageous to provide this reference signal as controllable elements on the receiving side via setting elements, photo resistors,
by means of a pseudo-random generator that is defective in a suitable manner for the practical implementation of the invention, it is advantageous to generate fixed pre-equalizers without a learner. This does not have to be neglected.

borrowed effort, in particular because of the relative 6 ₅ In the following, the invention is based on

moderately long period of the binary pseudo-exemplary embodiments explained in more detail.

Random sequence. It shows

The invention is based on the object of the F 1 g. 1 is a block diagram and

opposite g ^ ^ ^ _{yg] icacTn r r>} 3 ^ _{4 and} 19

w response to the Leitung.13 is _{nu and Ejner Anzah,} _ _{"to + M Einste} ll _g incorporate 6 \ T,

Line 11 ™ J \ S- ¹ _add ^ "« disappears the 8 ', 9' and 20. The outputs of these setting elements see below

Äli S / b di Li IV 1213 14 'd 21 with the

_add ^ «the 8, 9 and 20 disappear. The outputs of these properties correspond to the remaining 6 ₅ via the lines IV, 12,13, 14 ' and 21 with the Vr and SSwinger, and as inputs of the ausgangsse.t .gen summator 16 ver-vor and ^N ^ f ^scn ^ '"Y output 17 that is linked. The input of the transversal filter is 5,

^ ϊΑΑίί ^ ¹ -The distortions of the output w.ed.rum denoted by 17. The GE-

d £ ze k;

F i _g . 2 PuMane for the purification of the principle of the ^ "^ t ^^ " tigt. _%

^ TeSoÄlddesadaptiveEqualifier - ^ - ^ SC £

^ge ^. ^ - Sa ^ er at the entrance and S ^ limbs. With each ™ Α ^ ο

at the exit of the V «^ c ^ ^ .Jfden fignd 1 n- ^ kleme ^ übe g ^ _der ^ JJ

As equalization filter is a zweckmab.g '' _{is d} ^at _{k] the zeiU} i _c e D.spers.on the ■

Salfilter (echo equalizer) uses this Hlter ™haben f _impu ls response ζ but always greater than _B

can zero durations of the impulse response, n department ^ esu.n «^ P ^ _that ^ ^

Times left and ^right y ° ™ ^ * ^ma ££ _{2, the schwinger} almost completely be eliminated, _d,

Force impulse response and g ^^ J ^ J ™ ^ _{0n d} f _e ^ transversal filter contained delay «. _d

mutual influence of the neighboring Jg ^ is chosen to be sufficiently long if

and with it the crosstalk ^ b-. G. l show there are enough taps and ₃ ,

Block diagram of a transmitter and hg ^ ^ _{Wmam in input} 5 at

Pulse planes for a more detailed explanation Hegenden Signal are not extremely large. _s ,

Function. _{d It} goes without saying that the described transversal is _f ;

The transversal filter _in , F ^ g. J exist a ^ ^ _{he equalizer structure} . _L.

As already mentioned, the number of ^ν "™ | 8 in a _{chain can also be r}

divided 1,2, 3 and 4. Each V ^{e '0} ^ ™ S ° ^f _{nd the "ruckg} e _ko apply ppelte filter structures, further _A

and 4 has a ^Ve ™ ^. ^u ° ^ Vde ° SSS Zei- corresponding combinations and branching network- _{f (}

Distance between two consecutive Gesenaeur j _{as it is for example} i _swe i _{se in} , »Expensive

chen on. Both the S input and any AD ™ ^ j adaptive equalizer for fast

grabbed by the Ver _Z <3gerun _g sgheckr 2 3 and 4 ^ ■ s.rung ^ me ^ p _{R> M δ hf ffl a "n>1> NTZ <1>} J

The delay line formed is described in a banste booklet 1, pp. 36 to 42. D: r

member 6 to 10 ^ \ ^ l '^ \\\ ^ l ^m lLn "For the sake of simplicity, the invention in the following _E

Inputs of a summation η 16 verbu ^ ^ ₁ . _{can only be} described using the transverse driver.

Output 17 at the same time the output aes ^, _o ll however expressly pointed out _E

filters _{shows Puiso} lane in FIG. 2 it is clear that the following considerations apply to _v

to the respective sampling points in α multiplex system, which both signals _Ξ

r sampled "J ^ he" ^^^ Ϊ dner capacity "pure in the form of pulse amplitude modulation as well? fvSS2rSS?" bSder next sample multi-stage (quantized) Pulsampl.tudenmodula- _f

did, kept constant, allowed on, uc tion to equalize. c

The transversal filter can, because on the reception _γ

The ^el " ^stel ^ ^h ^ f _e ^r ! _N ^b the ^ eternal taps with side only individual, pbkation of the signals at individual sampling times: at the J ^ g = numerical value. ₄ o discrete amplitude values are queried before \

emem real P ~ X "o ^d _k ^e ^ S principle from one that contain the transmitted information, as an analog-Such an output can be developed in M ^ shift register. An analog value-

Potentiometer and a "™ ^ ^al" _{practice sch} i _e beregister provides a derailleur of Abtastverstarker consist NAU ^ ch ^ erimg _{hoe. The individual} samples are circled

the realization with ™ ^ ^^ _{45 the} value of sample and hold circuit to sample and hold Kre.s ver assumption, _therefore, the b.nsteugne ^ ^^^ ^ _{α dic Anordnun replaces a}

x ° ^{= +} Sd eb'enso on line 13 the delay line. In the case of a 32-channel system, the transversal filter must be used and, _{for example, the} required sampling rate ^{on the B b demens}

unchanged emgang sgL ts ^ e.stgg _second ^. _{Pulsa litude modulation}

Sisnal at the entrance 5, only one zeiuicn B ^ ^ ^ _{samples überlra} g _s, ie 256,000 values.

De ™ ^a function of the transversal filter is at hand, the delay time of a delay element is the pulse diagrams in F i "r. 2 π ^ «· There is one therefore 3.91 ^ ^ ^^^^ ^ _{adapt | ven}

Γ '1, 0' twnrtTdarlsteit This Imputan * word equalizer according to the invention. This equalizer In rf ^ i VorschSefund two post-echoes. 55 consists of several modules. A first Bauenthalt two wrapping. C ^ vl (n.? T e p r n U vorschwinße rs the group again represents a transversal filter. To eliminate the ^ men Vo ^ chw.ngers g ^ ^^ ^ ^ ^^ ^ ^^ ^

^VerZe ? HerTirerungsSun _g tapped, and the named branching network replaced. Ek-? ^ng . 11 i ^ 7ä S to such a value _C _ ₂ m: nte, which corresponds to the in F i g. 1 contained elements Emste Ig ed 6 will _{have the same effect on a α g hanlv / 0Ti} 6o, are in F i g. 3 same as beze.chnet.

set that ^de ". ^% ^e _o ^l _rs ° _hwinger ^ _er impuls- The transversal filter in turn consists of an opposite gl, ch the _{vor hw ^ g P ^ Ve Iiedern r 2} - _{3;> 4} - _and 19

(ο

Dashed lines 22 and 23 indicate this. other frame synchronization signals are also possible Other transversal filters, not borrowed here. The description should be in the following of the drawn delay and adjustment elements contain, for the sake of simplicity, known from the example mentioned, restrict.

The input 5 of the transversal filter is simultaneously 5 Under the following circumstances, the use of this is connected to the input of a circuit 24, which signal can be used to derive a setting criterion: clock recovery from the time-division multiplexed frame synchronization signal contained in the signal. The output of the transmission channel must be so low, for example, at output 25 of the syn- chronization clock appears at the frequency of 8 kHz, at output 26 io impulse responses, which as a result of the 32nd channel send the sampling clock for the analog value shift register with deten pulses occur, do not overlap and that the frequency of 256 kHz and the active equalization of these impulse responses at the output 27 with the help of a generated frame synchronization signal, which is possible with a transversal, without the temporal 30/32 channel system from the signal in the 32 . channel dispersion of the equalized impulse responses was found to be large, which for the transmission of the frame takes ₁₅ values. If a 30/32 channel sync signal is used. This signal is designated in the system shall in any case, the post-oscillations of the following with a'k . The signal on the impulse response in the 32nd channel will have decayed before line 27 is subtracted from the output signal j », - at the output significant pre-oscillator or the main value of 17 with the aid of a differential amplifier 28. chronologically next following impulse response in this At output 29 therefore appears a signal which occurs in the a ₀ channel.

hereinafter referred to as the error signal et-. (Any- if these assumptions, possibly by

A decision circuit 30 is shown in dashed lines; fixed pre-equalization must be enforced, and those which meet the sign at their output can supply the regenerated frame of the applied input signal as the ideal signal. The line 31 used, connects the output of the circuit 30 to the _25. This also applies if, as in our example, the input of a delay element 32 of the delay, the frame synchronization signal is a periodic signal delay time N · τ. Whose output 33 is with the first one. The mentioned restrictions with regard to the inputs of multipliers 34, 35, 36, 37 and 38 dispersion of the impulse response do not need to be combined. The line 39 connects the output 27 to be held when the frame synchronization signal with the input of a further delay line, ₃₀ binary bipolar pseudo-random text with a sufficient number of delay elements 40, 41, 42, 43 and 44 of large period duration is used, with the delay time τ is formed. The taps The frame synchronization signal can be obtained from the

this delay line are connected to the second incoming signal sequence with the aid of a peak value input of the multipliers 34, 35, 36, 37 and 38 detectors in the in FIG. 3 designated 24 connected. Their outputs 45, 46, 47, 48 and 49 ₃₅ derive the assembly. In our example, which will be described in more detail below, the ideal signal sequence only has the value "+1" or, in the time intervals in which circuits 50, 5i, 52, 53 and 54. Lines 55, 56, 57, 58 and 59 lead to the setting "-1", otherwise the value "0" and is called a '{f). divide 6 ', 7', 8 ', 9' and 20 and provide the necessary If the signal y _k at the sampling time / ₀ + k ι

Information about their automatic setting. ₄ o at the output of the not yet correctly set

In the following, before going into more detail on the apparatus equalization filter, the structure of the structure shown in FIG. 1 applies. 3 shown block diagram + m

of an adaptive equalizer that Λ = ^> fy ' xUp + kt - Jt)

Setting criterion based on some theoretical J '"~ ^N (I)

Explanations derived. 45 _y— _c . _χ

Prerequisite for deriving a setting or ' * ~ ^J

Error criterion is that on the receiving side, ^ _{G] n} ~ _h ~ _{ng sinddie diedurchdie} setting _n

Estimated values for the sent correct S.gnale or _Heder ^ _n ^^^ dn, the x, _. are you "ideal signals" can be derived. Be. pure ^& _yerzög J _{th distorted} signals on; th Xbgriff de, pulse amplitude modulation is generally only ₅ o delay line in the transversal filter. At the same time it is known that the signals to be transmitted have the form -j _t ^ee ^ of certain pulses. Since any number of y _k = y a "■ h (t ₀ + kr - ητ)

Amplitude levels can occur, the derivation is ■ = - « ^)

an ideal signal difficult. But also the structure
a quantizer for multistage pulse amplitudes - 55
modulation becomes very expensive as soon as more than four Here h (t) is the impulse response of transmission

different amplitudes are transmitted. transmission channel and equalizer and / 1 * = h (r ₀ + Jt τ The transmitted message flow can therefore be The a " contain the information, so do not use the height to derive a criterion, especially since the amplitude levels of the transmitted signal pulse often only a part of the Channels is occupied with calls. 60 in the individual channels. Equation (2) therefore states that the ideal signal here is that the signal at the equalizer output is made up of a recurring frame synchronization signal. 32-channel system in the 32nd channel is transmitted with factors a _n- weighted, impulse responses zi and which is composed, for example, of a. A: and η as well as j are here continuously alternating positive and negative impulse with, 65 integer number variables.

measured on the transmitted signal, maximum ampli- The deviation of the distorted impulse response k (

tude exists. This synchronization signal sequence then points from the ideal pulse normalized to the value "1" the fundamental oscillation is 4 kHz. Of course, the answer to the sampling times is

Ah, =

ίο

Using circuits 50, 51, 52, 53 and 54. Is

j + 0 / \ i, _{f is} greater than zero, then Cj becomes Ac] in the direction

(3) changed, resulting in a decrease in Ah]. Is

¹ ' ^J ° ■ Ah ₁ less than zero, there is a change in c

_TI . ,. . ^{,, 5} in the other direction. That attitude came

Here, A _{0 is} the main value, ie the maximum simultaneously and in parallel for all in the transversal

value of the impulse response, assigned whose setpoint filter contained adjustment elements

is normalized to the amplitude + 1. A further simplification is possible through

If one now compares the actually arriving application of the relationship

Signals shown in FIG. 4 are denoted by y (t) , for io

Time t = r ₀ + j τ with the ideal signal a ' (/), κ

that in Fig. 4 is shown, which alternately Je, > a _k -i- _N ■ sgn e _t _ _w (6)

made up of "+1" and "—1", followed by thirty-one * = 'each

Zeros, there is an "error signal" of the

^{Large 1} S for automatic adjustment of q.

^e j - yj ~ ^a 'i = 5 ^a n IA- + a - α' (4 \ Sgn <? * _ "means the sign in front of it

* '"''''e _k - _N , Aq the amount of change in q.

Tv "c» cc · ι · ♦ · c · λα ,, _ ■ · The in FIG. 3 adap shown as a block diagram

W .hn _P ? ^lg FV "/ ' ^? ' ί ^represent f" I "" ⁰ , ^{with e (t) tive equalizer ka} ™ ^ h according to the equation

Sit H Ih Γ ^{0!? 1} the signal can be set up in the method described in 32nd> o (6). YOU

T α α c Idealsigna fo ge α '(ή compared, instrumentation can largely be done digitally

while the signals, η the other channels, which with there only a pure VoSeShenkSation.

»5 new point created in which the scanning of the 32nd channel takes place.

Circuit 30 is used to determine the sign of et . The sign sgn e _k appears aul

k = 1

of the line 31 and can now be delayed with the aid of a digital τ- · ii j · ³ ° shift register 32

mentioned equal to zero, this delay rune is required if

The sum terms that contain this factor "ntton ^ Ä ^ ¹¹ ^ T ⁸ " ^! T ^

so usually do not provide a contribution. Only every 32 characters should be 35 Sde ™ n Corresponding to w.rd

results in a sum term, because here ^ gS VeltaZ moÄT " H ί ^V ^ g ^ g. '" is ^a

»+1« or -U The signals in the 1st to 31st channel G] ^ the «ίΐ Λ«, 52 ° ^ ¹⁶¹ * ™ « ¹ . ⁵¹ "- ^

which are here regarded as disturbances, are given by quantities i Ή M, e ^ewonne "· ^ be'den

the summation made in equation (1,5) un "» ^ ^ _u & « ^ t Ώ u'Ti" ³ ", 'h

integrated out. ΛΓ is a sufficiently large to be chosen 40 essentially corresponds to 2n μ multiplication

lende integer. The delay of both quantities to Z 1 hde G _r öß η ^ 1 ^ 'h ί n', 2

the right side of equation (5) to N- r is can n ^! L ν ί * '»- ¹ * ^and » 0 «occur

required to equalize pre- oscillators rTn d ^ ita ^ en μΙΖΓ ^^^ Τ ^ "T

can, ie to be able to determine Ah] for. / <0 wdse can dfe s £ Γ ^Sg ^ «ΐ \? ^{£ 15} Α

According to equation (5), the adjusting members 6 '45 iSt _s 1, 7 T, ³⁴ ' ³⁵ ' ³⁶ ' ? L ^

T, 8 ', 9' and 20 in FIG. 3 are continuously regulated in such a way that 'D sets f o-Zwe, -Add.erer to be executed,

that the distortions, that is, the Ah ₁ for the different n «FrS I d V ° ^{lie in the following way:}

nen y, go to zero. This method has the K "Ä a ^ ^Pr ° ^{domestic product} _w ^{d enblick Au} 8"

Advantage that di * a ' _t -, - _K only the states "0", "+ 1" speaking M _n , Τ · λ ^ ⁵ ° * ^ " ^Λ ' ^am χ

"-1" can assume what the instrumentation ₅ o ^ f ™ ^the M ^ Z ^ ei adder 37 about "ne gate

simplified. However, is a quick multiplication set fn to ScSS ⁸ S **? - ^HllfSta ^ ^S _fl "

an analog one with a sign size necessary, stores W, SS }? f * ^e " ^{thal e} f ^en I ^

fn ⁿ takeS ^ "*""«" ^Α -Ρ« —Α Ä ^^ ÄtSÄ

It was initially assumed that the painted digital ^de _Sc Y ^e ^ ^o S ^chan g ^sleit "ng through a simple

Circuit 30 is non-existent, so that the line W Sie cherfe PrTIf -T ^ ™ · ^ t ^S °

with the line 31 is directly connected. The quantity grSöder W ^, ^{in turn, an int} 'e * - * can be derived from e _k with the help of the delay "! 6o S ^ ΓΐΚΜ ^ ₄ " * ¹ * ⁰ ί ΪΓ

win line 32. The individual variables * '* _, ν Ss ₃ controls U ^} η τ ° " ^A f ^{gang dcr SC} '! I"

from a ' _k with the help of the elements 40 41 enlaiffr π ^ ^{6 Le} " ^{Ung 58 turn the}

42, 43 and 44 generated delay line S _er u? _c Lr T ^? ⁹ '" ^Inf °' ^{ge der 2} ^ ¹ *"'

This may be a suitably modified digital shift Ta ion SL% ^orzeic hen _p roducts may be the Interegister. The multipliers 34, 35, 36, 37 and 38 6 _{5 received} Tchaln! ^ ¹¹⁰¹ ; , "^ ^U " ^{d gCn} f ™ *

form the pro-new required according to equation (5)! ^ ProduS ^ ³ ' ¹ "" ^ " ^success S ^en -da only every 125 μ ₅

products from e _k - _N and α '^ - χ. The necessary sum- The circuit α. β ». ·,

tion via K products of sampled values takes place with not ι «Γ, κ, your control logic is here in detail

ε ment as it is not used to clarify the

contributes to the method according to the invention. The control pulses can from the in F i g. 3 labeled 24 Circuit to be delivered.

In practice it is often beneficial not to use the relationships given by equations (5) and (6), since here the summation or integration according to K products has to start again and since the setting elements 6 ', 7', 8 ' , 9 'and 20 must be able to retain the respective setting reached, ie to save it until a new value is available to improve the setting.

Instead, relationships can

Cj = - q lim

K - Jf k

instead of equation (5) or

c, = - q ■ lim

ü-jw ■ sgn e _k . _N (8)

instead of equation (6) for setting the C) . As the equalization improves, the Ahj approach zero, the controlled variables described by equations (7) and (8) approach constant values. The starting point in time of the summation or integration can be chosen arbitrarily, q is a sufficiently small constant proportionality factor to be chosen, which determines the speed at which the q enters.

Due to the influence of the signals in the 1st to 31st channel in the formation of the controlled variables given by equations (5), (6), (7) and (8), an additional "interfering element" occurs in these equations temporary "emigration" of the equalizer, ie to a deviation from the desired behavior. In order to keep the influence of disturbances according to this partial sum as low as possible, the integration or summation time constant is chosen to be sufficiently large, ie the equalizer readjusts only slowly. In equations (5) and (6), K is chosen to be sufficiently large, whereas in equations (7) and (8) q is chosen to be sufficiently small.

The influence of disturbances that occur on the transmission paths and which can also lead to a non-optimal setting of the equalizer, as well as the influence of the signals in the other channels can be reduced or eliminated by the fact that the controlled variables are repeatedly checked for effects of in the interfering signals contained in the quantities to be correlated are subjected. On the basis of the results obtained from the check, the setting of the setting elements is slowed down or blocked over the period in which the interfering signals occur. In equation (5) and (6), for example, K can be increased, in equation (7) or (8) q can be decreased.

This can be done, for example, in that chronologically successive values of the controlled variables are formed in adjacent time segments and examined with regard to their signs.
The technical implementation of the adjusting members 6, 7,

ao 8, 9, 6 ', 7', 8 ', 9' and 20 in Figs. 1 and 3 can be given by step-by-step adjustable circuits with the help of digital-to-analog converters or with magnetic storage, such as transfluxors are realized. Next are as adjusting members Bridge circuits are possible that use a field effect transistor linearized by negative voltage feedback included as a controllable element. In addition, photoresistors can also be used as controllable elements be used.

The circuit according to the invention described has the advantage that no special test signal sequence must be inserted into the information flow to be transmitted, since in time division multiplex pulse amplitude modulation systems a frame sync signal is already present. As a frame synchronization signal a pseudo-random text does not have to be used, it is sufficient in most Cases to use a periodic signal, which is particularly easy on the receiving side can regain. The procedures and criteria described for the adaptive setting of the Equalizer allow particularly simple instrumentation and precise convergence of the dei Circuit.

For this purpose 3 sheets of drawings

Claims (9)

characterized in that as controllable EIe patent claims: elements of the setting elements photo resistors are provided. 1. Method for the automatic setting of a 9th circuit arrangement for performing the . jidaptive, the reversal of linear verse 5 Method according to one of the claims! to 5, * ^ achievements serving on a transmission link characterized in that fixed ! Equalizer, especially transversal filter-equalizer-pre-equalizer are provided, rers, in a time division multiplex system for pulse amplitude modulation, in which, in addition to the message signals, a periodically recurring additional io iignal is transmitted at the receiving end Samples the amplitudes of the regenerated Additional signal from samples of the amplitudes »Incoming signals are subtracted »Nd with the one from the difference signal and the 15th igen-generated additional signal setting criteria for the The invention relates to a method and a finstellglieder of the equalizer are derived, circuit arrangement for automatic adjustment thereby g e k e η η ze i c h η e t that as an adaptive, the undoing is linear ■ atzsignal the frame synchronization signal is used. Distortions on a transmission link 2. The method according to claim 1, characterized in that 20 equalizer, in particular transverse filter equalizer, is characterized in that as a frame synchronization signal from- in a time division multiplex system for pulse amplitude alternately positive and negative, in their amplitude modulation, in which, in addition to the message signals, the same and, measured on the transmitted signal, a periodically recurring additional signal is used via pulses with maximum amplitude, in which samples are taken at the receiving end. 25 of the amplitudes of the regenerated additional signal from 3. The method according to claim 1 or 2, characterized in Abiastamins the amplitudes of all incoming characterized in that to form the setting signals are deducted and from the criteria the regenerated additional signal compared to the differential signal and the regenerated additional signal the difference signal step by step by integer setting criteria for setting elements of the equalizer Multiples of the time interval between two can be derived. consecutive samples is offset that The advantages of time division multiplex transmission of Products from the samples of the stepwise pulse amplitude modulated signals that are shifted in the host regenerated additional signal and the economic efficiency of the effort and the bandwidth requirement Difference signals are formed that are based on the can only be fully exploited, A sum is formed for the products of each step, if those that arise on the transmission paths and that these sums are used as setting criteria for linear distortion of the signals to be transmitted to serve. undone in each individual multiplex channel 4. The method according to claim 1 or 2, thereby making. Linear distortions that are too labeled that for the formation of the constant crosstalk between the individual criteria The regenerated additional signal can be routed to 40 channels with fixed the difference signal step by step by integer equalizers over a longer period of time only inadequately Mastering multiples of the time interval between two, for example because temperature fluctuations follow Samples are offset so that changes in the transmission properties can produce products from the samples of the step-by-step transmission paths, shifted regenerated additional signal and the 45 An automatic equalization with the help of Signs of the difference signal are formed, special test signals at the beginning of the transmission that from the products of each step a sum with subsequent determination of the equalizer setting is formed and that these sums as setting is not useful, since changes to the transfer criteria are used. properties no longer during operation 5. Method according to one of the preceding 50 can be regulated and because it is not possible Claims, characterized in that it is, for example, at regular time intervals Adjustment of the setting elements to shut down at the same time when thirty transmission channels occur, is slowed down or blocked by interfering signals. to transmit a test signal sequence. 6. Circuit arrangement for carrying out the signals for pulscodemodulierie, which only the Method according to one of the preceding 55 states "0", "+ 1" and "- 1" can assume is Sayings, characterized in that as setting from the publication »An Automatic Equalizer for Digital Repeatered Lines «in» Proceedings of the are seen with the help of digital-analog-IEEE ", March 1969, p. 363 and 364, a method for Converters or from magnetic memories, automatic equalization during transmission in particular transfluxors are realized. 60 known. There it is shown that under certain 7. Circuit arrangement for implementing the conditions the maximum amplitude of the equalized Method according to one of Claims 1 to 5, the PCM signal is a measure of the cable attenuation in the characterized in that the setting element is "Nyquist frequency". As the typical cable bridge training are provided that have a known attenuation curve, this dimension can be used to set an automatic equalizer linearized by voltage negative feedback, Effect transistor included as a controllable element. However, this method cannot be applied to pulse amplification 8. Circuit arrangement for carrying out the tudenmodulated signals with more than two possible Method according to one of Claims 1 to 5, transmitting amplitude values. Also results in this