DE2734865C2 - Methods and circuit arrangements for reducing interference - Google Patents

Description

The invention is based on a λ'ε ^ ηΓβη according to the preamble of claim 1. Circuit arrangements for carrying out the method according to claim 1 are in claims 2, 4, 7 and 8 described. The invention has the purpose of interference that occurs in the transmission of the useful signal via a Message channel occur to diminish.

According to the prior art, methods for correcting disturbed signals have so far only been used in the case of digital ones Signals applied. The correction refers to the elimination of errors (error correction) that caused by disturbed signals, the originally transmitted signals themselves being uncorrected stay. With this method, the correction can therefore only be carried out after the disturbed signal has passed through one or more ED decision processes and was determined, that there are errors. The way this procedure works is based on the »logical« derivation of a redundant Part of the signal from the useful signal to be transmitted.

Correction of analog signals was previously only possible after an analog-digital conversion, an error correction of the digital signal and a subsequent digital-to-analog conversion is possible.

From DE-OS 18 OO 545 a circuit arrangement for suppressing interference interspersed in the transmission path during message transmission is known. The transmission path is two-channel. At the sending end, means are provided to split the message into two mutually complementary parts that are assigned to the channels of the transmission path. On the receiving side, both channels are connected to an evaluating summing amplifier with the weighting factor 1/2, with one of the channels and the output of the summing amplifier being connected to the inputs of a differential amplifier which sends the message at its output. In this circuit arrangement it is assumed that the interference affects both channels in the same way.

The invention is based on the object of specifying a universal method by which the signal closes Interference power ratio at the output of a transmission system directly for both digital and for Can significantly improve analog signals.

This object is achieved according to the invention by the features specified in claim 1. To the realization the circuit arrangements described in claims 2, 4, 7 and 8 are used for such a method.

One advantage of the method according to the invention is that with a given transmission power and given Signal to interference power ratio at the receiver output the bridgeable transmitter-receiver distance can also be enlarged when transmitting analog signals.

In the following, the invention will be described in greater detail using an exemplary embodiment shown in the drawing explained. It shows

1 shows the block diagram of a transmission system with a device for signal correction,

2 shows the block diagram of a transmission system with a device for signal correction and a non-linear characteristic curve with a positive slope, Fig. 3 shows an example of the characteristic curve of a non-linear coding network,

I- ig. 4 an example of a non-linear characteristic which is pieced together from straight lines,

F i g. 5 shows the block diagram of a transmission system with a device for signal correction, a sawtooth characteristic and a syndrome jump correction device,

6 shows an example of a sawtooth characteristic and the associated signal quantities,

F i g. 7 shows an example of the signal curves in a device for signal correction with a sawtooth characteristic and a syndrome jump correction device,

8 shows the block diagram of a transmission system with a device for signal correction, a sawtooth characteristic and a syndrome jump marker,

F i g. 9 shows an example of the signal curves in a device for signal correction with a sawtooth characteristic and a syndrome jump marker,

10 shows the block diagram of a transmission system with an adaptive device for signal correction,

F i g. 11 shows the block diagram of a device for correcting burst interference with linear coding networks and a non-linear decoding network according to the syndrome mean comparison method,

12 shows an example of the signal profiles in a transmission system with a syndrome mean value comparison decoder,

13 shows the block diagram of a decoding network for correcting burst interference according to the test equation method,

14 shows the block diagram of a decoding network for correcting burst interference in the information system Control channel according to the test equation method,

15 shows the block diagram of a transmission system with a device for correcting burst interference according to the test equation procedure with weighted taps,

16 shows the block diagram of a transmission system with A: control channels for correcting burst interference,

I- ig. 17 the block diagram of a transmission system with two information channels and one control channel,

18 shows an example of the signal waveforms in a transmission system with two information channels and a control channel when a single bundle error occurs,

19 shows an example of the signal waveforms in a transmission system with two information channels and j

a control channel when two bundle errors occur, 65 |

F i g. 20 shows the block diagram of a transmission system with / information channels and a control channel, and j

Fig. 21 shows the block diagram of a transmission system with / information channels and A: control channels.

1. Basic mode of operation of the signal correction

According to the invention, the task of continuous signal correction is achieved with a transmission system solved according to Fig. 1.

The information signal x (t) to be transmitted is transmitted to the receiver in a known manner via an information channel IK. An additive interference signal n (j) is superimposed on it, so that the following applies to the disturbed channel output signal:

X ₁₁ U) = X (O + n U). (!)

For the spectral amplitude density X ₁₁ W) belonging to x ^ U) :

+ ~ +09

XAf) = X (f) + Nif) = [ Xd) Q- ^j2 - ^r 'dt + f ηU) e " ^{2r /} ' d / (2)

X (f) is the spectral amplitude density of the transmitted information signal ^ r (0 and N (J) is the spectral amplitude density of the interference signal n (t) - To simplify the illustration, it was assumed that the transfer function Ü, (f) of the information channel IK and the transfer function Ü _K (f) of the control channel KK assume the value 1 in the frequency range under consideration.

With the help of a coding network CW ₀ , which generally contains linear and non-linear network parts, is derived from the transmitted information signal x (0 a control signal kU) and transmitted in a known manner via a control channel KK to the receiver, with an interference signal mU) additively superimposed so that the following applies to the disturbed control signal k _s (t) in the receiver:

Ut) = k (t) + with). (3)

The following applies to the spectral amplitude density belonging to kjt):

4-ao +00

³⁰ Γ Γ

K, (f) = K (f) + Mf /) = J k (0 e ~ ' ^2r "dt + J m (t) e"' ^{2j7 /} 'dr

(4)

K (f) is the spectral amplitude density of the transmitted control signal k (i) and M (J) is the spectral amplitude density of the interference signal m (t) -

The receiver also contains a coding network CN _h , which is assumed to be identical to the one at the transmitting end if U ₁ (J) = Ü _K (/) = 1 in all properties. (If Ü, (/) and Ü _K (J) are different from 1, then the coding network CN _h must be selected so that, if the information and control channels are free from interference, UiU \ = with) = O) UD = A- _f ( /) is applicable.)

With the help of this network, a receive-side control signal k / U is derived from the disturbed information signal x _s (0 ) which, with interference-free transmission (n (r) = O) and neglecting the transit times, matches the transmission-side control signal k (0. If the Interference component n (0 however different from zero, then an interference component n _K U occurs at the output of the receiving-side coding network), so that the following applies:

k _f (t) = k (t) + It ₁ U). (5)

If the coding network also contains non-linear parts, then the interference component n _E (l) is not only dependent on the interference η U), but also on the information signal x (t). The following applies to the spectral amplitude density K _E (J) of the control signal k _E (t) at the receiving end:

K ₁ (J) <= K (J) + N _h (f) (6)

N _E (J) is the spectral amplitude density of the signal n _E (t).

In a comparator V ₀ , the control signal k _E (t) at the receiving end is subtracted from the disturbed control signal k _K (t). The difference is named syndrome signal sU) and it applies to them:

sU) = k, U) -k, U). Π)

60 For the spectral amplitude density of the syndrome signal s (t) one obtains:

S (J) = K _K (f) - K _E (J). (8th)

With the help of equations 3 and 5 one finds for the syndrome signal:

s (t) = mU) -n _r U) (9)

and with the gin. 4 and 6 can be found for the corresponding spectral amplitude density:

S (Z) = M (Z) - N _e (f). (10)

The decoding network DN has the task of transforming the syndrome signal s (t) in such a way that the component m (l) is suppressed as far as possible and the component -n _E (t) - apart from the transit time T ₁ - as possible in -n (t) transforms.

In addition point A , correction signal z (f), which occurs at the output of the decoding network, is added to the disturbed information signal x _x (t -T ₁ ) delayed by the transit time T \ of the coding and decoding network, which makes it possible to compensate for the disturbance will. If the spectral amplitude density Z (Z) of the correction signal z (t) contains signal components which are above the cutoff frequency of the information signal Helgen, these can be suppressed with the aid of the filter G (Z).

2. Correction with linear coding and decoding networks

If the coding networks CN _a and CN _{b are} linear networks with the transfer function C ₁₁ (Z) = C _b (f) = C (f), then the following applies to the spectral amplitude density of the control signal on the transmit side:

K (f) = C (J) X (f). (11)

Correspondingly, the control signal at the receiving end applies to the spectral amplitude density K _E (f):

KhW) = C (Z) X _t (f).
With Eq. 2 and 11 result from this:

KkW) = CW) [XW) + NW)] = C (J) X (J) + C (f) N (f) = K (J) + C (J) NW). . (12)

A comparison with Eq. 6 shows that:

NdI) = CW) NW). (13)

According to Eq. 10 thus applies to the spectral amplitude density of the syndrome signal:

SW) - MW) - CW) N (J). (14)

If the decoding network DN is also a linear network with the transfer function D (J), then the following applies to the spectral amplitude density of the correction signal z (t):

Z (J) = D (J) S (J) (15)

or with Eq. 14:

Z (J) = DW) MW) - DW) CW) NW) - (16) g

With the help of the Fourier transformation, this results for the correction signal:

z (t) = J D (Z) M (J) S ¹² ^ dJ- J DW) C (Z) NW) e ' ² "''d /. (17)

In this regard, D (J) and C (/) are to be chosen in such a way that z (t) becomes the best possible estimate of -n (t) and the influence of MW) or rn (i) remains as small as possible. If the mean square error is used as the evaluation criterion for the quality of the estimate, this leads to the requirement:

55 + T

lim -1- Γ I η (t) + z (t) pd / = Min |. (18)

r - IT J.

Is z. If, for example, C (J), N (J) and Af (/) are taken as given, the optimal (linear) decoding filter can be determined with the help of the Wiener optimal filter from the above condition (Eq. 18):

DW) - 4 * £ & (1.9)

^W)

/( â € œ.,.(/) is the cross-spectral power density from the noise signal n (t) and the syndrome signal

Λ (/) = m (t) -n (t) * c (t). (20)

(c (t) is the impulse response of the coding filter C (f); the symbol * indicates the convolution operation). According to Wiener - Khintchine's theorem, the cross-spectral power density L _11x (J) is the Fourier transform of the associated cross-correlation function / "" for which applies to ergodic signals:

+7

/ _{/ n} (r) = lim -L f n (t) s (t + τ) dt.

(21)

The following applies to the cross-power density:

+ OO

LM) - J I ₁ Jt) e- ^j2 * "dt. (22)

Similarly, for the autocorrelation function / _v (r) and the power density L _S {J) of the signal / · (/) (see Eq. 20):

respectively.

+ Hm - ^ - Γ

τ s (t) sU - τ) dt (23)

LAf) = J IAt) C '"" dr. (24)

For the simple special case

DV) CV) = 1 (Γ, = 0) (25)

one obtains from equation 16:!

Z (J) = D (J) MiH- N (f). (26) j

i The following applies to the spectral amplitude density YV) of the corrected signal y (t): I.

YV) = -W) + Z (J) (27)

or with Gi. 2 and Eq. 26:

YV) = XV) + D (f) M (J). (28)

For the signal to interference power density ratio of the corrected signal> · (. ') One obtains: |

IW)

This equation shows that the signal to interference power density ratio can be specified for each frequency range of the useful signal by a suitable choice of the decoding filter D (J). A good correction, ie low interference at a frequency To require a Weines D Vo) and thus according to Eq. 25 a capital C (Jo) -, which equates to a large transmission power in the control channel KK . The advantage of using linear coding and decoding networks is therefore the possibility of selective choice of the signal

Interference power density ratio. In relation to the required transmission power, compared to a direct |

Transfer overall no profit achievable. |

The device according to FIG. 1 can also be used for fault detection (error detection) (dashed line

lated connections in Fig. 1), if the correction signal r (/) is fed to a threshold value decision maker, at whose output E there is always a signal! is applied when - according to the setting of the threshold value decision maker - the correction signal is significantly different from zero.

3. Correction with linear coding networks and
a downstream non-linear characteristic curve with a positive slope

Continuous signal correction with linear coding networks has the disadvantage that effective interference suppression requires considerable transmission power in the control channel KK . If the coding networks OV ₀ and CN _h according to FIG. 2 are converted into a linear coding filter CGO and a non-linear coding network NCN

broken down with the characteristic A- = / (</), you have the option of limiting the transmission power. The slope

, (30)

you you

the non-linear characteristic curve should always be greater than or equal to zero. The decoding filter D (J) must then be set so that the interference amplitudes n (t), which are in the range of the maximum slope s _{max of} the characteristic A- = f (u), are completely eliminated when the interference m (r) = 0 is. If, _{on the other hand, the interference amplitude} η (ι) lies in areas in which the slope j <i mev, the influence of the interference is also reduced when a linear decoding filter D (J) is used, but a complete correction no longer takes place. If a decoding network DN is used in which the linear decoding filter is followed by a non-linear network whose characteristic curve is inverse to the characteristic curves k = f {u) in the coding networks, a complete correction of the disturbance η (/) is possible, the influence however, the disturbance m (t) in the amplitude ranges with a small slope becomes very large.

By suitable selection of the non-linear characteristic curve k = f (u) and the associated non-linear characteristic curve in the decoding network DN , the signal amplitude ranges in which correction is preferred can be selected. When transmitting speech is z. B. the influence of a disturbance at low signal levels (pauses in conversation) more disturbing than at high signal levels. If, accordingly, the maximum slope j _{mov is placed} in the range u «0, the correction is most effective in this range.

F i g. 3 shows an example of the form of a non-linear characteristic curve with a positive slope in which the signal values at two different times i and I _{2 are} entered (where Ü _K (J) ~ 1 was assumed). It becomes clear that the interference amplitude η (/,), which is exactly as large as the interference amplitude n (t ₂ ), causes a significantly larger syndrome signal j (r,) than the interference amplitude n (t ₂ ). Since the interference amplitude n (t \) at time r is in the range of the maximum slope of the characteristic, it can be corrected almost completely with a suitable choice of the decoding filter D (J) , whereas the interference amplitude n (t ₂ ) - corresponding to the ratio of the Syndrome values s (t ₂ ) to s (Z ₁ ) - only a fraction is corrected. The slope of the characteristic is therefore a measure of the system's ability to correct in this amplitude range. As a non-linear characteristic curve, it is also possible to use characteristic curves that are piecemeal composed of straight lines, as is the case, for example, in FIG. B. Fig. 4 illustrates.

4. Signal correction with linear coding networks, a sawtooth characteristic
and a "syndrome jump correction device"

In order to avoid the disadvantage of the high transmission power in the control channel KK , according to FIG. 5 the linear coding networks are supplemented by a downstream sawtooth characteristic. The advantage of this method is, according to the invention, that the transmission power required for the transmission of the control signal is limited by the peak value of the sawtooth characteristic, whereas the signal to interference power density ratio depends on the slope of the characteristic. Since the same sawtooth characteristics are used on the sending and receiving sides, the transmitted control signal and the control signal derived on the receiving side always coincide with interference-free transmission, so that here, too, Eq. 9 remains valid. In contrast to coding with exclusively linear networks, however, the interference component n _£ (/) is also dependent on the amplitude distribution of the signal u, U). In the time segments in which the two signals u (t) and u _E (i) are in the same segment of the sawtooth characteristic (see example for the signal values at time t ₀ in FIG. 6), all equations of the segment are retained 2 "Correction with linear coding and decoding networks" is valid if C (J) is replaced by 5 · C (J) , where i is the slope of the sawtooth curve, for which the following applies with the values of Fig. 6:

In particular, Eq. 29 applies unchanged, while Eq. 25 the shape

It follows from this that a large interference power density ratio Q (J) connected by a small D (J) with a sufficiently large slope 5 of the sawtooth characteristic according to Eq. 32 can be achieved.

The limiter B at the output of the control channel KK suppresses all interference amplitudes that exceed the amplitude range U _{1 of} the transmitted control signal k (t) , with the result that the influence of the interference m (t) alone can never cause the transmitted and received correction signal come to rest on different branches of the sawtooth characteristic.

FIG. 7 shows, as an example, the signal profiles in the block diagram according to FIG. 5, it being assumed that the disturbance is m (t) ~ 0 (in addition, it is assumed that C (J) = const). Caused by the disturbance η (t) (Fig. 7a), the transmit and receive control signals arrive at different branches of the sawtooth characteristic _{in the time segments Z 1} to t ₂ , / 3 to r ₄ and /, to / _ft that the output signals u (/) and u _t (t) of the coding filter C (J) exceed the linear branches of the sawtooth characteristic at different times (see FIG. 7b). The jump points of the control signals A- (z) and k _E (t) (see FIG. 7c) therefore occur at different times. The syndrome signal s (t) = k _g (i) - k _E (t) is thus in the time segments in which u (t) and u ,, (t) lie on different branches of the sawtooth curve, or too large by U _c too small.

(In general, if the disturbance is sufficiently large, / i (r), s (t) can deviate from the correct value by vU _c , ν = ± 1, ± 2, ± 3, ...).

However, if the upper limit frequency of the control channel KK is selected to be higher than the upper limit frequency of the signal u (t) or u _E U), the jumps in the syndrome signal, which indicate the beginning and the end of an incorrect correction, are steeper than all signal edges and can therefore be used for their detection. The task of the syndrome jump correction device (see FIG. 5) is to separate the syndrome signal sU) from the superimposed square-wave signal ν U _c. For this purpose, the syndrome signal is fed to a differentiator whose output signal s' (t) has needle-shaped pulses, which in the subsequent pulse normalization device (e.g. by a threshold value decider with a downstream monoflop) with a corresponding pulse height in a short unit pulse of constant duration with the amplitude ν ί /, can be converted. Fig. 7e shows the output signal of the differentiating circuit, the pulse peaks proportional to the size (vUJ the signal jump which introduces a 'fake' syndrome signal is the normalized pulses (Fig. 7f) set with the first edge a multi-flop, the output signal s _b U) assumes the amplitude νU, and maintains it until the arrival of the next normalized pulse. 7g shows the output signal λ · _λ (/) of the multiflop. To adapt the edge steepness of this signal to the edge steepness of the syndrome signal, s _b (.t ) is fed to a pulse shaper with the transfer function E (J). (In the example of FIG. 7, E (/) = const, is assumed). The output signal s _c (t) of this filter is shaped in such a way that it compensates for the rectangles superimposed on the syndrome signal. The delay time T _{2 of} the delay element VG ₂ is to be selected so that it corresponds to the signal transit time through the syndrome jump correction device. A comparison of the disturbance n (t) and the syndrome signal s _d U) according to FIG. 7h shows that they are oppositely equal at each point in time, which enables a complete correction of the intruding disturbance η (/) in the addition point A when the Running time Γι of the delay element VG ^ the signal running time through the filters C (J), Ü _K (f) and D (J) and the running time 7 ^ is made the same. The following filter G (J) generally removes residual interference, in particular from the control channel, which exceeds the cutoff frequency de? Information signal x (t) lie.

5. Signal correction with linear coding networks, a sawtooth characteristic
and a "syndrome jump marker"

The basic function of this procedure corresponds to the procedure in point 4. Instead of the »syndrome jump correction device«, however, a »syndrome jump marking« is used here. So that such a marking is possible. a pulse-shaped signal / (/) is superimposed on the output signal v (t) of the sawtooth characteristic, which indicates that the signal «(/) has exceeded a linear branch of the sawtooth characteristic. The sign of the pulse marks the direction in which the signal u U) left the branch; while the size of the pulse is selected so that the control signal k (t) exceeds the range limits + U, / 2 of the sawtooth characteristic curve during the pulse duration, which enables the jump to be recognized at the receiving end.

F i g. 8 shows the block diagram of a transmission system which contains a device for continuous signal correction with a sawtooth characteristic and a syndrome jump marking. To simplify the representation, the transfer functions U ₁ (J) and Ü _K (J) of the information channel IK and the control channel KK were assumed to be equal to one.

FIG. 9a shows the output signal u (t) of the coding filter C (/), which is fed to an amplitude comparator according to FIG. 8, and the output signal v (f) of the sawtooth characteristic. The thresholds of the amplitude comparator, the output signal h (t) of which FIG. 9b shows, are set so that h {t) has an amplitude jump at the times at which the signal u U) exceeds a linear branch of the sawtooth characteristic. The downstream pulse generator generates a standard pulse of the corresponding polarity from every edge of the signal h (t)

(see Fig. 9c), which adds the control signal k (t) in the addition point A \ to the signal v (i) (see Fig. 9d). The disturbed control signal k _g U) is also fed to an amplitude comparator whose thresholds are set so that its output signal always has a signal jump when its input signal significantly exceeds the range limits ± UJl of the sawtooth characteristic. The following pulse generator generates a standard pulse of the corresponding polarity from every jump. F i g. 9e shows the from

50 control signal k _B (t) derived standard pulse sequence / _A - (r) provided that the control signal is transmitted without interference;

The disturbed signal u _£ (/) and the associated output signal of the sawtooth characteristic v _E (t) are shown in FIG. 9f. The receive-side and transmit-side amplitude comparators and the downstream pulse generators are constructed identically, so that the signal i _E (t) has a standard pulse of corresponding polarity at the times at which the signal u _E U) crosses a linear branch of the sawtooth characteristic. F i g. 9g shows the course of this signal for the example on which it is based. The following counting decoder has a positive input for the signal i _f: (t) and an inverting input for the signal i _K (t), in which all pulses are negated before they are counted. The counter counts the pulses of both inputs taking their polarity into account and, according to its counter reading, emits a signal whose amplitude corresponds in polarity and size to the respective counter reading (see FIG. 9h). A comparison of the signals u U) and u, (t) shows that the output signal s, U) of the counting decoder is always different from zero when uU) and u _E U) lie on different branches of the sawtooth characteristic, where: :

s _c U) = -vU _c v = 0, ± 1, ± 2 ... (33)

v denotes the count of the counting decoder. It indicates how many branches of the sawtooth characteristic lie between u U) and u _E U). Since the syndrome signal sU) specifies the deviation of the received signal from the correct value within a branch of the sawtooth characteristic and s _c (t) takes into account the number of branches between

see u (t) and u _E (t) lie, is the sum

s _: (t) = S ₁ XO + s (ή (34)

to the fault η u) that has occurred, the opposite is true if, as in FIG. 9 assumed the disturbance m U) = 0. 6. Accuracy and automatic adjustment during signal correction

The signal to signal-to-noise ratio improvement that can be achieved with the signal correction devices described in points 1 to 5 is mainly dependent on the accuracy with which the coding networks are compared. With an ideal adjustment, the following must apply for n (t) = m (t) = 0:

• s- (0 = 0. (35)

If this requirement is only approximately fulfilled due to imperfect implementation or temporal fluctuations in the control channel and the components, the effectiveness of the correction process is reduced as a result. The use of an adaptive coding network CN _{6 is} required to achieve high signal-to-interference power ratios or to use signal correction in control channels whose properties are subject to fluctuations over time. 10 shows the block diagram of a transmission system with an adaptive device for continuous signal correction. Assuming that the mean value of the interference signals η U) and m (t) is zero, the integral must also be in the case of an ideal adjustment

s U) dt (36)

result in zero via the syndrome signal. If one further assumes that this requirement is still met if only a certain frequency band of the syndrome signal is considered, then a criterion for setting the adaptive coding network CN _{h can be} obtained therefrom. With the aid of a filter set F] to F _r , partial signals can thus be derived from the syndrome signal sU) which result in an integrated zero with an ideal adjustment, ie with the correct setting of the adaptive filter. If, on the other hand, one of the integrator signals is different from zero, it can be used to set the adaptive filter in this frequency range.

7. Signal correction of bundle interference *)

7.1 Decoding according to the syndrome mean comparison

F i g. 11 shows the block diagram of a transmission system with a device for correcting burst interference, the coding network of which is made up of linear delay elements of the delay time T. The delay time Γ of the individual delay elements determines the duration of the reliably correctable bundle error length, while the number of taps is a measure of how many such bundle errors can be corrected simultaneously, ie within the total delay time μΤ of the coding network. The number of delay elements connected in series is denoted by μ. The taps must be chosen so that the number of delay elements that lie between any two taps are all different from one another. In the block diagram of FIG. B. between the first (X ₁ (O) and the third (x _} (t)) tap 3 delay elements. Between any two other taps of the coding network there must therefore not be another 3 delay elements. After between the taps X ₁ U) and X ₁ U) one and there are two delay elements between X ₂ U) and x ₃ (r), the tap x _A (t) _{must be separated from the tap .v 3} (/) by at least 4 delay elements. In general, the following relationship applies to the number of delay elements between any two taps with the consecutive numbers./? and y:

α = γ - ß {\ < Co, rl <μ + 1, β Ψ γ). (37)

55

For a certain a> 0, Eq. 37 can be fulfilled by at most one Paarjö, γ. The function of the correction device according to FIG. 11 will be explained with the aid of an example, it being assumed that the coding networks each contain seven delay elements (μ = 7) and four taps (p = 4). Fig. 12a shows the time course of a source signal x (t) = XiU) as it leaves the source and enters the coding network on the transmission side and the information channel (solid curve). When transmitting .v (/) via the integration channel, a bundle error b (t) should occur, the duration of which is less than or equal to the duration T (bundle disturbances whose duration is greater than T are subsequently converted into two or more bundle errors b (l) decomposed), so that:

[«/) <> </ <Γ / any 0 <r < 7th

| 0 otherwise [Q otherwise ^v;

♦) The following applies to the syndrome signal in the following sections: sU) = k ^ U) ~ kg (t)

This interference is additively superimposed on the information signal, so that the signal X _1E (dashed curve of FIG. 12a) enters the coding network CN _{h on the receiving end.} At the second tap, the signal x ₂ (t) or X2 _E U) appears, which the Fi g. 12b shows and that the input signal delayed by the transit time T is X ₁ U) or XitU) . Delayed by a further 2 T or a further 4 T , this signal also appears at the third and fourth taps of the delay elements (FIGS. 12c and d). The control signals k (t) and k _E U) shown in FIG. 12e result in each case as the sum of the signals which are present at the four taps of the coding networks CN ₁₁ and CN _h . Assuming initially that k (t) is transmitted without interference via the control channel KK (w (?) = 0), then the syndrome signal s (?) (See FIG. 12f) in the time segment 0 to T is identical to that in the interference signal η (?) penetrated the information channel IK. Depending on the selection of the taps of the coding network CN _b , this syndrome signal form and thus the bundle disturbance b (t) is repeated in the sections T -IT, 3 j - AT unaTT - Sris see FIG lie. The signals S _x U) to S ₁₁ (I) appear at these taps, which are shown in FIG. 12f to 12i are shown. It can be seen that in the time segment 7 Γ to 8 Γ for the first time the occurring interference signal b (t) is applied simultaneously to all taps, while previously it occurred only at one or none of the taps of the syndrome delay network. The task of the devices connected downstream of the syndrome delay network is to determine the degree of correspondence between the syndrome signals s _x (t) to s _p (t) at each point in time t . If the agreement is very good, as shown in the example of FIG. 12 is the case in the time segment 7 T- 8 T , it is assumed that a fault has occurred and a corresponding signal is enabled for correction. Is the

The degree of agreement among the individual syndrome values, however, is very low, e. B. in the time segment 0-7 T, it must be assumed that various uncorrectable disturbances are present; no correction will be made in this case. To carry out this task, all syndrome signals s _x (t) to S ₁₁ U) (ρ = 4 in the example) are added up in a syndrome summing element SU, and in an amplitude divider AT by the factor /? divided, resulting in a mean syndrome signal s _m (t) . In the comparative

Ehern K, to V ₁ , it is then determined how large the deviation e,. (f) ν = 1 ... ρ of the individual syndrome signals j, U) to s _l; U) from the mean syndrome signal s _m (t) . Since the evaluation of these deviations also depends on the absolute size of the syndrome signals, the subsequent formation of the relative error is useful, but not absolutely necessary. In the block diagram of FIG. 11, the formation of the relative error £,. = e, (/) / s, "(?) with the help of the division device DE, which is the reciprocal of the mean syndrome value s", (t)

s, “U) and the multipliers M _x to M _p performed. If the mean syndrome value sjt) becomes zero at a point in time t = ? ", The outputs of the multipliers M _x to M _{1 are} based on the relative error fi (/ ₍ i) = ... = ε ,, (ι ₀ ) = 1. The following devices S, ... B ₁ , form from the relative errors E _x U) ■ ■ ■ C ₁₁ U) the amounts which in the summing element SU _F to a mean error signal £ ", (/) are added up (see Fig. 12k). The threshold value decider £>, to which this signal is fed, is set in such a way that it is set in the event of a sufficiently large mean error and thus syndrome values that differ greatly from one another

Si (?) ... i "(i) the correction switch S ' _Ä opens and thus prevents a correction, while it is very small

relative errors and thus good agreement of the syndrome values s _t U) ■ ■ ■ S ₁₁ U) closes.

The simplest way to obtain the correction signal is to use the mean syndrome

signals s, “U) (dot-dash connection between points A and B in FIG. 11). If malfunctions occur ί

the control channel KK or several bundle errors at the same time, the formation of a special correction |

however, it is preferable to use a door signal. For this purpose, the amounts of the relative errors ε _χ (ι) to c “U) are each one

Threshold value decision maker E _x ... E ₁ , supplied. These determine which syndrome signals S ₁ U) ■. ■ S ₁₁ U) deviate particularly strongly from the medium syndrome signal s n, {t) and which match well. The outputs of the threshold value decision maker E _x ... E ₁ then open the switches S _x ... S ₁ , those syndrome signals that have a strong

Deviation from the mean syndrome signal s n, (t) . In contrast, the “well matched” syndrome values are added up in the summing element SU ₁ and sent to a variable amplitude divider AT ₁ . The number j of the syndrome values switched through at any given point in time, by which the divider AT ₁ has to divide the signal S ₁ U) , is calculated with the aid of the summing element SU _E , at whose input the output signals of the decision maker £ ", ... E ₁ , applied, is determined. The thus determined correction signal z (t) still allows

Such an error-free correction if within the time μ7 "less than p / 2 bundle errors, the duration of which must not exceed T , have occurred (p = number of taps in the coding networks, μ = total number of delay elements in the coding networks). The phase reversal stage Ph inverts this Signal, so that it runs out of phase with the fault that has occurred. In addition point A , the correction signal z (?) Is added to the delayed signal x _E (t - aT) . So that the delay element VG _x (see FIG. 1) can be saved, is in the present case x the delayed also by T signal _e (t) of the last tap as signal x _x U ~ T _x) herf attracted. the the correction signal z {t) outgoing dashed connection has the function of the signals

have already led to a correction in the syndrome delay. to be deleted so that this is free for the correction of new errors. This connection can also be omitted. The filter G (J) removes remaining very high-frequency residual interference.

If a single bundle error mU) = b (t) of the maximum duration T occurs in the control channel KK , only one of the syndrome signals s _x (t) ... s _fl (t) can be different from zero at any given point in time, da. ? (/) = m U) applies and m U) passes through the syndrome delay network unchanged. This has the consequence that ρ - 1 relative error assumes the magnitude l / p and one the magnitude (p - l) / p and the mean error signal c _m (l) can thus not be less than 2 (p - l) / p This ensures that a burst error in the control channel KK does not lead to a disruption of the information signal if the threshold of the threshold value decision maker E _{F is} lower than 2 (p - Mp) Ip (see Appendix 1). If, on the other hand, a bundle error «(/) or m (i) occurs both in the information channel IK and in the control channel KK , the syndrome signal in the time segment 0 to T assumes the value _w s (t) ⁼ " U) + rnU) , while at all other time segments it either coincides with n (t) or zero

is. The mean error signal can therefore not fall below the value / i - 1 in the time segment 0 ... uT, which prevents incorrect correction. In the time segment t = aT ... (Μ + \) T (correction phase of the bundle error ηU)), however, good for the mean error signal:

(39)

1 mU) 2 (p- 1) P.

The interference signal n (t) is only corrected here if ε _η does not exceed a certain limit value, 4- h. if the amplitude of the beam error m U) is not significantly greater than the interference amplitude. iö η U) of the bundle error in the information channel (see Appendix 2). Similar conditions are obtained for the cases in which the burst error mU) adopts a different timing for burst error ηU).

7.2 Decoding according to the test equation method

From the signal curves of Fi g. 12 shows that the occurred in the information channel IK burst error η U) = b (l) is applied after the delay time μΓ (7 T) at all taps of the syndrome register, such that for the period μΤ </ <(μ + I) T :

ί, (/) = bit -uT)

X ₂ U) = bit - vT) (40)

25th

s (t) = bU-T)

However, if further burst errors occur in the time segment 0 < ί <(μ + 1) Τ - either in the information channel / A or in the control channel KK - or if the delayed syndrome signals of a previous error have not yet disappeared from the syndrome delay network, everyone can Tapping an "interference term" c ,, (f), ν = 1,2 ... superimpose on the signal b (t). c ,, (?) stands uniformly and represents all the different causes that can lead to the occurrence of such a "fault cluster". If, for example, in addition to the bundle error bU) to be corrected, another bundle error b ₂ U) occurs in the control channel, a c _r (t) can be different from zero - and apart from a delay time shift of r ₀ mil I) ₂ U) match (c,. (t) ⁼ b ₂ (l - t ₀ )). In the worst case z. For example, if additional bundle errors occur in the time segment 0 </ <(μ + \) T , an "interference term" C ₁ U), v-1,2 ..., can be superimposed on each tap, see above that then applies to the output signals of the syndrome delay network:

S ₁ U) = bU - vT) + <, (/)

S ₁ U) = bU - μΤ) + C ₂ U) S ₃ U) = bU - vT) + ad)

■ xT </ <(μ + I) T (41)

45

■ s "U) = bU - vT) + C ₁₁ U)

This system of equations consists of ρ equations that contain ρ + 1 unknowns, which excludes a unique solution of the system. If, however, it is assumed that only a limited number of burst disturbances occur during a major part of the transmission time, ie only a limited number of unknowns differ from zero, then a solution of the system of equations may be possible. Since it is not known which of the unknowns (bundle disturbances) bU ~ μΤ "), C \ (t) to C ₁₁ U) are zero at any given point in time, a hypothesis must be drawn up which is subjected to a subsequent check. However, this check requires that at least one (or more) derp equations (test equations) are not used for the determination of the unknowns other than zero. This leaves the maximum number of sp - I equations (trunk equations) that can be used to solve ρ - 1 unknowns the ρ + I unknowns 2 are to be set as zero, while the remaining ρ - 1 unknowns can be determined from the trunk equations

J ^ MLiP

Hypotheses are set up and as many trunk equations are solved. Then, with the help of the test equations, it must be determined which hypothesis was correct. As a rule, of the {* ' t' j hypotheses, only those are of interest in which b (t - vT) is assumed to be different from zero, so that the solution

of ('^ different systems of trunk equations and test equations is sufficient. The formation of the hypotheses, the solution of the trunk equations as well as the control with the help of the test equation is described in the following

for ρ = 4 and α = 7 described in more detail. The entire system of equations in this case has the form:

j | (/) = b (t - 7 T) + C | (0

₅ ⁵ "- ^U) = bU-! T) + c ₂ U) _αΓ < _{/ <(μ + 1) Γ} . ₍₄₃₎

J ₃ (O - bO - 7 T) + C ₃ U)

S ₄ U) = b (t-lT) + C ₄ U)

According to Eq. 42 there is a total of (p + \) p / 2 = 10 hypotheses.

10

1st hypothesis b (t -IT) = 0; c, (0 = 0;

Solution of the trunk equations: C ₂ (O = ₂ U)

, 5 CU) =

Test equation: S \ U) ⁼

= J ₄ (O ⁽⁴⁴⁾

2nd hypothesis b (t - 7 T) = 0; C ₂ (O = 0; solution of the trunk equations: C | (0 =

^C3 ? L * " ^J ) '<(45)

C ₄ (O = J ₄ (0 ^l '

Test equation: J ₂ (O =

3rd hypothesis b ( 1-7 T) = 0; C ₃ (O = 0;

Solution of the trunk equations: C | (0 ⁼ Ji (0

(46) 30 test equation:

4th Hypothesis 6 (/ - 7D = O; C ₄ (O = 0;

Solution of the trunk equations:

Check equation:

5th hypothesis c, (0 = 0; C ₂ (O = 0;

40

Solution of the trunk equations:

* - ₄ \ t. / ° ₄ V * y JHV

Check equation:

45

6. Hypothesis c, (/) = 0; c ₃ (/) = 0;

Solution of the trunk equations:

r. (t \ = Γ.ίΛ - r. (F \

_. f. ι π = ei π - pi ii \ '^ /

Check equation:

. 7. Hypothesis c, (/) = 0; c ₄ (r) = 0;

Solution of the trunk equations:

(50) Test equation:

8. Hypothesis C ₂ (O = 0; C ₃ (O = 0;

Solution of the trunk equations:

n- (t \ = C. (A - C- (A

MW - ^J 4W - »2W

Check equation:

S ₃ U) = J ₄ (O
= 0 J ₃ (O - SiO)
- s, (0 C | (0
C ₃ (O
J ₄ (O = S ₂ O)
= 0 Sl (O
S ₂ (O
S ₄ O)
SlO) -SiO) bO-
C ₃ (O
C ₄ (O
Jl (O 7 7 -) - SlO)
S ₂ O)
S ₃ O)
SlO) - SiO)
- J, (0 Kt-
c-, 0)
C ₄ (O
S ₃ O) 7 7 ·) = S ₂ O)
S ₄ O) bO-
C ₂ (O
C ₃ (O
J ₄ (O 7 Γ) = Kt-
C ₁ O)
C ₄ O)
J ₂ (O 7T) =

9th hypothesis φ) = 0; <r ₄ (0 = 0;

Solution of the trunk equations: b (t - 7 T) = s ₂ (t)

-S ₂ (O

-S ₂ (I) (52)

Test equation: s ₄ (0 = S ₂ (O

10th hypothesis φ) = 0; C ₄ (O = 0;

Solution of the trunk calibrations: b (l) = s _} (0

C (O = J ₁ (Z) - SiU)

C ₂ (O - i ₂ (0 - i ₃ (0 ⁽⁵³⁾

Test equation: _Si (t) = s _A (t)

If only the correction of the bundle error of the information channel IK is of interest, it is sufficient if the parts of the solutions of the trunk equations are instrumented that are required to determine b _x (t - 1 T) . How to use the example for /? = 4 (hypothesis 5 to 10), there are ρ - 1 different solutions of the form

h (t - 1 T) = s _v U) ν = 1 ... ρ - 1. (54j

Whether one of these and, if so, which of these trunk equation solutions is the correct one is determined by the ('' J ') test equations determined.

13 shows the block diagram of a decoding network for correcting burst interference according to the test equation method described. It consists of a syndrome register with corresponding taps, at the output of which the syndrome signals J ₁ (O to s “U) are present, a device for solving the test equations and ρ - 1 switches S _x to S ₁ , - _x . The device for solving the test equations has the task of determining which of the hypotheses applies or which test equation is fulfilled, in order to then activate the corresponding switch S ¹ I ... S _1,. , to close, which releases one of the syndrome signals S _x U) ■ ■ -S ₁ ,, (/) for correction. If none of the test equations proves to be fulfilled, then several alsp-1 unknowns are different from zero ₍ and a correction is not possible (all switches remain open). In this case, a special output signal 11 _K U) from the device for Solution of the test equations «can be used to indicate to the receiver of the transmitted information signal that he is receiving a signal which is very likely to be disturbed.

7.3 Test equation procedure with tightened test conditions

The execution of a false correction in this method occurs when the fulfillment of a test equation is simulated, although the underlying prerequisite that at least two unknowns are zero is not fulfilled. With the first hypothesis according to Eq. 44 this would be e.g. B. the case if: 6i (/ -TT) = -C _x U) Φ 0. 7i To reduce the probability of an incorrect correction, the test condition can be extended to two (or more) test equations. Beip equations can then only be determined p - 2 unknowns (bundle error), so that 3 unknowns are required for the formation of a hypothesis

Must be set to zero, which leads to ("* ¹ J different hypotheses. For the above example m \ tp = 4 this also results in 10 different hypotheses, the first of which has the following form, for example:

Hypothesis c, (0 = 0; C ₂ U) = 0; c, (0 = 0;

b _x U)

= S ₄ U) -

Solution of the trunk calibrations: b _x U) = S _x U)

Test equations: _S] U) = S ₂ U); S ₁ U) = S ₃ (O ^'55 ^

In this case, the greater security in suppressing incorrect corrections is bought at the price of a reduced performance of the correction system (in the above example, a reliable correction is only possible if a maximum of 2 bundle errors have occurred).

If other disturbances of very small amplitude (e.g. thermal noise) occur in addition to the bundle disturbances, this can be taken into account by allowing suitable tolerances in the instrumentation of the test conditions. ^ ₀

7.4 Test equation procedure with determination of the bundle error c ,, (0

If several signal coding and signal decoding networks are connected in cascade, then in some cases it is also advantageous to determine the bundle error C ₁₁ U) - which has occurred in the control channel.

14 shows the block diagram of a decoding network which, in addition to the device for solving the test equations, also has devices for solving the trunk equations for the bundle error C _{1 of} the control channel. Since the trunk equations for the determination of C ₁ coincide in some hypotheses, are

a total of only r (r <p) devices needed to solve the trunk equations. For ρ = 4, μ = 7, the following four types of trunk equations result from hypotheses 1 to 10 (see Eqs. 44 to 53):

1. Body equation type q (i) = 0

2., Body equation type Ci (O = J ₄ (Z) “.

3. Body equation type q (z) = s _A (t) - s ₂ (l) '

4. Body equation type C ₄ (O = s ^ U) - s \ (t)

Since the equation C ₄ U) = O does not have to be implemented (all switches S _cl .. .S _cr are oiled-net in this case ), the implementation of the above example r = ρ - 1 = 3 devices for solving the trunk equations necessary, which can be implemented in the last two cases with the aid of a differential amplifier.

7.5 Signal decoding according to the test equation method with weighted taps

An improvement of the method according to points 7.1 and 7.2 is possible in that the individual taps of the transmitting and receiving cable coding networks according to Fi g. 15 can be weighted with the weighting factors g _t ... #, _H. If a single cluster error of the form occurs in the information channel IK

" ⁽ⁱ⁾ JO otherwise ⁽⁵⁷⁾

while the control channel KK is free of interference, the following signals appear at the taps of the syndrome delay network:

S ₁ (I) = g, ■ bU - μΤ)

S ₂ (I) = g ₂ b (l - μΤ) + _gl ■ b (/ - (μ - 1) Γ)

(58)

S, ₊ , (0 = * .. +, · Kl - μΤ) + gjU - (μ - Dn + ... gl- Kl)

The considered bundle error bit) is used for decoding and signal correction in the time segment μΓ to (μ + I) T. The following applies to the signals at the taps of the syndrome delay network in this time segment:

StU) = g \ b {t - μΤ) + c, (ί)

= g ₂ b0 - μΤ) + C ₂ U)

(59) j

S, * iU) = g, _{+ l} bU- μΤ) + C _{1 + 1} U)

The additive terms c _v (i), ν = 1,2. In contrast to Eq. 41, however, a single additional bundle error in the information channel can lead to several φ) differing from zero. The solution and instrumentation of this system of equations can be analogous to Eq. 41 take place. In addition, all statements made there can be transferred accordingly. From the system of equations, Eq. 59 in total

Λ-Μ-Μ, - (μ + 2) (μ + 1)

various hypotheses and test equations of the following form: 55

Hypothesis 1 to (μ + l): ö (z-μΤ) = 0; C ₁ XO = O ν = 1.2, ... μ + 1. (60)

Test equation 1 to (μ + 1): j _v . (0 = 0 ν = 1.2 ... μ + 1. (61)

λ -L ο ·) ζ- j- 11

Hypothesis (u + 2) to ζ φ) = 0; c _o (0 = 0; ν, σ = 1.2 ... μ + 1. (62)

Test equation (α + 2) to. _Φ) _ ₌ £ α (0_ υ, σ = 1,2 .-. Μ + 1, σ> ν: (63)

2 g _r g _a

The means for solving the test equations in FIG. 15 must therefore be capable of satisfying the two to solve the types of test equations described. In principle, with continuous-time signals, all must

Equations are solved in parallel. If, however, a short "loss time" is allowed in which no signal correction is carried out, a serial solution of the two types of test equations is also possible.
For the trunk equations belonging to the above hypotheses, the following applies:

The selection of the currently required trunk equation solution is carried out depending on the result of the test equation analysis with the aid of switches S ₁ to S _a, which are controlled by the device for solving the test equations. Before a syndrome signal s, (t) is switched through to the signal correction line (signal z (D) , the weighting of the bundle error with the factor g, / g ,, occurs (see FIG. 15). If a coefficient g ,, = 0, the corresponding tap must be omitted in the decoding network can be formed from equations, which enables a more reliable signal correction. This advantage usually has to be bought by a significantly larger number of test equations and a greater amount of instrumentation Control channel have occurred lead to an incorrect correction 0 < t <uT several bundle errors in the information channel IK , however, each additional error causes that during the signal correction of the bundle under consideration lchler bit - μΤ) several c "ν = 1 ... μ + 1 differ from zero, which makes an optimal evaluation of the test equations difficult. The weighting factors g " ν = 1 ... μ + 1 are therefore expediently chosen so that the largest possible number of test equations that can be easily instrumented is possible.

8. Signal correction of burst interference with k control channels

In the case of a very close chronological sequence of the burst disturbances, several control channels K] to K _{k can} also be used in parallel to secure the information signal. With the help of A different coding networks, which can also contain non-linearities and delay elements with weighted taps, see Sections 2 to 7, control signals A, (r) to k _K 0) are derived from the information signal x (t) k and via the channels KK ₁ to KK _k transmitted to the recipient. In comparators F ₁ to V _k , these transmitted signals are compared with the control signals k _E , (/) to k _Ek U) derived at the receiving end, and the results are passed to the decoding networks ZW ₁ to DN _k (see FIG. 16). The structure and mode of operation of these are matched to the coding networks and thus enable the formation of the A: correction signals z, (/) to z _k {f). If one considers a period of time in which _{interferences which can be corrected by the decoding networks ZW 1} to DN _k have occurred in all channels, then the following applies:

Z ₁ U) - _z, U) ... - z _k U) = nU -; ± T), (65)

ie all correction signals ζ, It), ν = 1 ... A "are equal to the (delayed) interference nil - μΤ) that occurred in the information channel IK . Contain the interference signals n (t), m _x (t) to m _k U) on the other hand, so many bundle errors that uncorrectable error patterns or incorrect correction signals occur in the decoding networks DN] to DN _k that in this case applies:

= nit - μΤ) + Z ₁ U) = nU - '[JT) + C ₁ U)

^z k0) ~ ⁿ (t - yT) + CtO)

This system of equations can be calculated using the test equation method described in Sections 7.2 to 7.5 to solve.

Since it includes A + 1 unknowns, the A-1 unknown interference signals assumed to be different from zero can be determined with the aid of (* ^ 'j hypotheses (see Fig. 16) of delay networks with weighted taps (see Section 7.3). The k correction signals z, (/) to z _k 0) are sent to a "device for solving the test equations",

in which for each point in time the ( ^Α 'ΐΜ tests according to the equations

= 0, v = 1.2, ... A-

and

(67)

Z ₁ (Z) = zj.1), ν, σ = 1,2 ... A- v < σ

are carried out (see ζ. E.g. Eqs. 60 to 63). If one of the test equations proves to be fulfilled, one of the switches S ₁ to S _{k is} closed and the resulting correction signal z, (/) in the addition point A from the received one - if the associated hypothesis does not require π (ζ— -xT) = 0 and delayed signal x _K (t) subtracted In addition, the digital signal u _K " ν = 1 ... k, which indicates whether the correction signal z _v (t) of a decoding network is error-free according to the associated test equations or not, can be used to generate additional test conditions or restricted hypotheses can be used (dashed connections in Fig. 16). Signals z. For example, if the signal u _K,. (/) Indicates that the associated correction signal z _v (i) is "wrong," then the hypothesis c,. ( T) = 0 is very likely wrong. In this case it is therefore useful if the tests belonging to the hypothesis (or partial hypothesis) c £ t) = 0 are not carried out, since they can lead to incorrect signal corrections. (In principle, only AI switches 5, are necessary to determine a resulting correction signal z, (/). However, since individual signals z _v (i) are to be regarded as "unreliable" in certain cases when the error detection signals ιΐχΜ are taken into account, the use of the signal is to be considered z _k (t) for determining the resulting correction signal zjit) is very useful.)

If none of the test equations (Eq. 67) is met, the receiver can be informed by means of the signal u _Kr (t) that the signal present at the output at this point in time is very likely to be faulty.

If the interference signals n (t), m \ (t) ... and m ^ U) also contain weak noise signals in addition to the bundle ^{interference, then the syndrome mean value comparison decod T} emetwork, which is also used here analogously to FIG. 11, can be used can be advantageous if the signals s \ (t) to s _p (t) are replaced by the correction signals z, (?) to z _k U).

9. Signal correction of bundle interference in / information channels and a control channel

Is the probability of the occurrence of a bundle error not very high or is it not very high? close together, several information channels can be transmitted through the transmission of a single control signal to be protected. First, the basic procedure will be explained using an example.

9.1 Signal correction of bundle interference with 2 information channels and one control channel

F i g. 17 shows the block diagram of a transmission system with 2 information channels which are protected against trunk errors by a control channel KK. To simplify the illustration, the coding networks C) (/) and C ₂ (/), which consist of 13 delay elements with 4 taps and have the same structure on the receiving and transmitting sides, are only shown in detail on the receiving side. The output signals A | (0 and A ₂ (O of the two coding networks C \ {f) and Ci (J) are combined in the adder A \ to form a common control signal A- (i), transmitted and combined with the control signal k _{E formed at the receiving end} (t) are compared. the resulting syndrome signal s (t) ^{w 'r} d in the two decoding networks D \ and D ₂ after Prüfgleichungsverfahren analyzed and z in the two correction signals (0 and z ₂ reacted (/). the location and the number of taps of the coding networks C | (Z) and C ₂ {f) must be chosen so that a mutual influencing of the bundle errors from the channels IK \ and IK _{2 is} prevented as possible. This is the case when all delay times, which can occur between any two taps of the coding network Q and C _{2 selected at will} (see Eq. 37).

The following values result for the differences in the delay time between the taps of the coding network C \:

.V ₁₁ (O τ JfI ₂ (O V / Λ - V (t \

x _u U) τ .Vi ₄ (O X ₁₂ (O τ -V ₁₂ (O τ

87 "

12 Γ

47 "

37 "

(68)

The following results for the coding network C _2:

6Γ 11 7 "

13 7 "57" 77 "27"

x ₂ i (0 τ X ₂₁ (OTA ₂₁ (O X ₂₁ (O τ X ₂₄ (O

X ₂₂ (O τ X ₂₃ (O τ X ₂₄ (O

(69)

Since all the differences in the delay times of the two coding networks are different, the decoding networks 17 the mutual influence is reduced to a minimum. This is also reflected in the

The following examples clearly show the signal curves in the syndrome delay networks to be examined more closely. First of all, the effect of a single bundle error in channel IK is examined

0 otherwise

(70)

it is assumed that the channels IK ₂ and KK are free of interference in the time segment under consideration. In Fi g. 18a, the time course of the interference signal n \ {t) is shown schematically by a triangle. The syndrome signal j (0, which is caused by this burst error and enters the syndrome delay networks SVN 1 and SVN ₂ , is shown in FIG. 18b.

The result is that at the taps of the syndrome delay network SVN ₁ the in FIG. 18c to 1 Sf shown signals occur. These signals are used in the "device for solving test equations 1" to determine whether and, if so, which test equations are met. In the present example, it should be assumed that two of the four equations that can be formed in each case are used as test equations (see Section 7.3). Therewith are of the 5 unknowns of the equation system to be solved

Α, (/ - 13 Γ)
Ai (Z-13 r)
Jn (O = A ₁ (Z -13T)
J ₁₄ (O = A ₁ (Z- 13 T)

C ₄ (O

<ζ <(μ + 1) Γ

(71)

2 unknowns can be determined, while three must be assumed to be zero. Overall there are C'jl ') ⁼ υ) ^{= 1} ^ different hypotheses of the following form:

1st hypothesis A ₁ (Z -UT) = 0; C ₁ (O = 0; C ₂ O)

Trunk equations C ₃ (O = Jn (O; C ₄ (O = test equations J || W = 0; J] ₂ (Z) = 0

2nd hypothesis O ₁ (Z-13 T) = 0; Ci (O = 0; C ₃ (O

Trunk equations C ₂ O) = Ji ₂ (O; C ₄ (O = Jh (O test equations Jn (O = 0; Jj ₃ (Z) = 0

3rd hypothesis i, (z -UT) = 0;

0; C ₄ (O =

Trunk equations
Test equations

C ₂ (O = Jn (O; C ₃ (O = Jn (O Jn (O = O; J ₁₄ (O = O

4th hypothesis / », (/ - 13 T) = 0; c ₂ (0 = 0; C ₃ (O =

Trunk equations C | (0 = Jn (O; C ₄ (O = Jw (O test equations J ₁₂ (Z) = 0; j _u (z) = 0

5. Hypothesis 6, (z-13 T) = 0; C ₂ (O = 0; C ₄ O) =

Trunk equations
Test equations

= Jn (O; C ₃ (O = Ji ₂ (O = 0; J ₁₄ (O = 0

6. Hypothesis b _} (t - 13 T) = 0; C ₃ (O = 0; C ₄ (O =

Trunk equations
Test equations

Ci (O = JnW; C ₂ (O = Ji ₂ (O Ju (O = 0; J, ₄ (0 = 0

7. Hypothesis C ₁ (O = 0; C ₂ (O = 0; C ₃ (O = 0

Trunk equations
Test equations

O ₁ (Z - 13 T) = J ₁₁ (O; C ₄ (Z) = J ₁₄ (O - Jn (O J | i (0 = Ji ₂ (O = Jn (O

8. Hypothesis c, (0 = 0; C ₂ O) = 0; C ₄ O) = 0

Trunk equations A, (z - 13 T) = Jn (O; C ₃ (O test equations j |, (0 = j, ₂ (z) = j _M (z)

-Jn (O

(72)

(73)

(74)

(75)

(76)

(77)

(78)

(79)

17th

* 9th hypothesis C ₁ (O = 0; c ₃ (0 = 0; C ₄ (O = 0

Ruimpf equations b _x {t -UT) = J _n (O; C ₂ (O = J ₁₂ (O - J _n (O *

Test equations s _u (t) = S ₁₃ U) = S ₁₄ U) ₍ 80)

10. Hypothesis C ₁ U) = 0; C ₃ (O = 0; c, (0 = 0

Trunk equations O ₁ U - 13 T) = J ₁₂ (O; C _x U) = J ₁₁ (O - J ₁₁ (O - Ji ₂ (O

Test equations J ₁₂ (O = J ₁₃ (O = J ₁₄ (O (81)

As long as one of these hypotheses is fulfilled, a reliable signal correction of the errors that have occurred is possible whereby only the core equations of hypotheses 7 to 10 have to be implemented, ie only switches S _{1x and} S _{12 are} required for this test comparison system. An uncorrectable error has occurred if none of the 10 hypotheses is met. In this case, the signal u _k {t) can be used to indicate that the receiver is on

Receiving a disturbed signal. An uncorrectable error can also occur if one of the 10 Test leaching is faked, although the associated hypothesis that 3 of the unknowns are zero does not apply. A "wrong signal correction" is carried out here.

From the signal curves of FIGS. 18c to d it can be seen that the hypothesis l (j,, (0 = 0; J ₁₂ (O = 0)) is fulfilled during the time segments 0 <t < 5 T and 6 T <κ i3 Γ , which means that there is no “wrong signal cor-

correction «carried out w.rd. In the time segment 5 Γ to 6 Γ, among other things, the hypothesis 2 (j, At) = 0 j ,, (0 = 0) is fulfilled, so that no "false signal correction" is carried out here either. The hypothesis 7 (j _n (0 = s "lt) = s _u Uä ast in the time segment 13 T to 14 T is fulfilled, which causes the closure of the switch S _x , and a correct correction of the error that has occurred (z, (0 = J ₁₁ (O = b _x {t - T 13)) in the summing point A ₃ the result Codiernetzwerk C ₂ and the decoding network D ₂ has have the same structure as the networks C, or

D _x . That's why you put in the GIn. 71 to 81 J ₁ , through j _2. , Μ = 1, 2, 3, 4, the test equations and return equations of the decoding network 2 are obtained 0 < t < 15 T at least two test equations are fulfilled, so that here a "wrong signal correction" of the information signal x _g , (/) due to a disturbance in the information channel IK, is avoided. For example, in the time segment 3 T <t < 10 T, the test equations J ₂₁ (O = 0 and

J ₂₃ (O = 0 fulfills (hypothesis 2) and in the time segment 13 T <t < 15 T the equations J ₂₂ (O = 0 · and J ₂₄ (O = 0 (hypothesis), so that here too a "wrong correction" As can be seen from the signal loss, several test equations are fulfilled in some sections, so that hypotheses other than the handled hypotheses can also be considered to be correct; in all cases, however, the result with regard to the signal correction is the same leads to hypothesis 5 \ 35 in the time segment 13 T <t < 14 T. the solution

b _x U -MT) = O, C ₂ (O = 0, C ₄ (O = 0

C ₁ (J) = j ₂ i (0; C ₃ (O = J ₂₃ (O ⁼ 0, / g2)

obtained with the help of hypothesis 10, the test equations of which are also fulfilled.

FIG. U shows the signal curves in the syndrome delay networks SVN _x and SVN _{2 of} the block diagram according to FIG. 17 under the assumption that disturbances occur in both IK _x and IK _2:

"''" [0 otherwise (83)

and

", -., ₌ Ib ₂ U) + bjU) 0.5 < ι < 2 Γ
"< ^ul [Ο otherwise (84)

Since the duration of the interference signal n ₂ U) is 1.5 T , which corresponds to two bundle errors, a total of three bundle errors have occurred.

In FIG. 19a the time profile of the interference signal (/) is shown schematically as a triangle and in FIG. 19b the profile of n ₂ U) is shown as a rectangle. The syndrome signal j (0, which caused by these three burst error in the syndrome delay networks SVN _x and SVN ₂ enters, 19c in FIG. Played. It causes _x at the taps of the syndrome delay network SVN the g in F i. 19d-g illustrated Signals that are further analyzed in the "device for solving test equations 1" occur in the time interval 0 <ί <12.5Γ at any given point in time at least 2 of the four syndrome signals v, μ - 1.2, ■> 4 g | _ei ch zero, so that a test equation system belonging to hypotheses 1 to 6 is always fulfilled, which means that no "false signal correction" is carried out in this section in the range S 12.5 T <K \ 3T none of the hypotheses is fulfilled, which is why a »false signal correction« also takes place here

not held; the signal u _k (t) to indicate that a fault cannot be corrected, however, signals the recipient

r-iu ''"' Ί ■ Relationship j,, (0 = J ₁₂ (O = J ₁₃ (O, so that the requirements for hypothesis 7 as

are to be regarded as fulfilled and for z, (0 = J _n (O = «, (/ - 13 T) what a» correct signal correction «applies (see l · Kg 18h). In the range 13.5 T <t < 14 T and 14.5 T <t < 15 T none of the applied test equations are fulfilled, iiCh that no signal correction takes place

Test equation fulfilled, but this does not result in a signal correction.

Soft signal curves would occur if the interfering signal «] (/) = 0 and only / I ₂ (O in the time interval 0.5 T <t < 1.5 T differs from zero, which corresponds to a total of two bundle errors, are shown in FIG The signals shown in dashed lines in Fig. 19. In this case, the _{test equation belonging to the hypothesis I0 (j, 2} (0 = Ju (O = Jm (O)) would be fulfilled in the range 13.5Γ </ <14Γ, so that in this case too Area a "correct signal correction" of the trunk error that has occurred would be possible.

The signal profiles that occur at the taps of the syndrome delay network SVN ₂ are shown in FIG. 19i to 1 shown In the "device for solving test equations 2" it is determined for each point in time whether and, if so, which of the test equations are fulfilled. In the time segments 0 <ί <8.5Γ and 15 T <t < 17 T at least two of the syndrome signals are zero, which means that one of the hypotheses 1 to 6 is assumed to be correct here. A signal correction is therefore not carried out in these time segments. In the range 8.5 T <t <9 T , none of the hypotheses apply, so that no signal correction takes place here either. However, the signal u _k (t) indicates here that an uncorrectable error may be present. In the time segment 13.5 T <t < 14 T , the test equations belonging to hypothesis 10 are fulfilled, so that here a “correct signal correction” of the information signal X ₂ (O takes place (see FIG. 19m). In the following area 14 T < t < 15 T is the test condition J ₂₁ belonging to Hypothesis 8 (O = S ₂₂ (O = -Sm (O fulfilled, whereby the "correct signal correction" bi (t - 13 T) = s _n (t) is continued) The determination of the superimposed error C ₃ (Z) = S ₂₃ (I) - J ₂ ] (O delivers the correct result, namely the triangular cluster disturbance that occurred in the information channel / λ ".

It can be seen from the examples that one control channel also monitors several information channels and can be protected against the intrusion of bundle errors. However, it is necessary to that the taps of the individual delay networks are chosen so that mutual interference ; | flow of the various information channel interference is largely prevented, as z. B. at the

the following choice of taps is the case:

Coding network C _x (J): 1, 3, 7, 25, 30, 41

Code network C ₂ (J) 2: 1, 4, 16, 29, 36, 37 (85)

or

Code network C _t (f): 1, 2, 28, 31, 62, 74, 82, 84

Code network C ₂ (J): 1, 19, 24, 38, 59, 63, 76, 87 (86)

or

Code network Ci (J): 1, 2, 7, 26, 33, 73, 101, 109, 121, 131

Code network C ₂ (J): 1,24,40,58,61,75,102,104,113,117 (87)

The numbers indicate how many taps must be used to generate the control signal. The number of correctable bundle errors increases with the length of the delay network. In the case of time-discrete signals, the delay networks can use clocked analog shift registers be replaced.

9.2 Signal correction of bundle interference with 3 information channels and one control channel

If the probability of a trunking error occurring in a channel is very low, then more than 2 information channels can be monitored by one control channel. F i g. 20 shows the block diagram of a signal correction system with / information channels. For;' = 3 provides the following choice of taps a minimal influence on the various channel errors:

Coding network Ci (J):
Code network C ₂ (J):
Code network C ₃ (J): 1.4 (88)

or

1.2 16, 20th 1.3 IS, 19th 1.4 12, 14th 1.4, 1.8, 1.7,

Coding network Ci (J):
Coding network C ₂ (J):
Code network C ₃ (Z): 1, 7, 12, 14 (89)

or

Coding network C _x (f): 1, 10, 34, 38, 39, 98, 123, 130

Coding network C ₂ (J): 1, 12, 13, 24, 63, 77, 80, 124

Coding network C ₃ (J): 1, 19, 36, 51, 72, 78, 118, 126 (90)

19th

For / = 4, the following examples provide a minimal mutual influence of the channel errors:

Code network C, (/): 1, 2

Code network C ₂ If): 1, 3

Code network C ₃ If): 1.4

Code network Ci (/): 1.5 (91)

Code network C, (/): 1, 17, 21, 22

Code network C ₂ (Z): 1, 3, 11, 25

Code network C _} {f): 1, 15, 18, 27

Code network Q (f): 1, 12, 19.25 (92)

The use of coding networks with weighted taps can also be used with more than one information channel can be applied accordingly.

10. Signal correction of disturbances in / information and Ar control channels

21 shows the block diagram of a transmission system in which Ar control channels are used for monitoring and correcting / information channels. The code networks CWj ₁ to CN, _k are any linear networks and can, for. B. consist of a component that determines the spectral correction sensitivity (see section 2) and a delay network with various taps (see section 7). The non-linear code networks ./VCZV ₁ to NCN _k have the task of limiting the required transmission power and can, for. B. in the form of a kink characteristic (see Fig. 5) or a sawtooth characteristic. SK ₁ to SK _k are also non-linear networks that have the task of eliminating the syndrome signal corruption _{caused by NCN 1} to NCN _k. Is z. If, for example, a sawtooth-shaped characteristic curve is used for a non-linear code network NCN on the transmit side, the associated network SK on the receive side can be implemented by a syndrome jump correction device (see Section 5). The decoding networks DN _S to DN _lk are matched in structure and function to the associated coding networks CW _n to CN _ik. Is z. B. used for a code network a tapped delay network, the associated decoding network z. B. be instrumented according to the syndrome mean comparison method or the test equation method. Analogously to the method according to point 9, each correction signal z \\ U) to z _lk (t) thus supplies an estimated value for the interference that has occurred in the information channels IK ₁ to IK _1. The decoding network

Works DN \ to DN ₁ have the task of determining the degree of correspondence in the associated signal groups and of determining the most likely course of the fault that has occurred. This can e.g. B. can be carried out with the help of the test equation procedure described in section 8 or with the syndrome mean comparison procedure. The delay elements VG ₁ to VG ₁ have the signals X ₁ (I) to. *, · (/) To delay the signal propagation time of the entire correction device. The breakdown of the source signals into the individual information signals X ₁ (O to x, (t) is in principle arbitrary. However, the performance of the method can be increased if it is carried out in such a way that the information channels / K ₁ to IK _{1 are} as statistically independent errors as possible For example, a broadband source signal can be broken down for the correction into a low-frequency and a high-frequency information signal with interfering signals that are independent of one another.

11. Signal correction and signal types

No restrictions are required with regard to the information signal x (t) used , ie the signal correction described can in principle be applied to all types of signals as well as to modulated and unmodulated signals. In the case of the analog signals used to describe the signal correction, all relationships are valid for every point in time and every arbitrary amplitude value. In contrast, it is z. B. is sufficient for time-discrete signals if the requirements and statements made are met in the time segments in which the signal parameters are different from zero. This means that the processes and circuits discussed need to be functional only in these discrete time segments or points in time, which as a rule enables the methods and circuits to be simplified. So z. B. in the case of time-discrete signals, the syndrome delay networks of sections 7 to 10 are replaced by clocked analog shift registers. Discrete amplitude or digital signals also allow signal correction to be used, although the circuits can usually be simplified here as well.
The way in which the various signals are transmitted via the information and control channels is in principle arbitrary, e.g. B. they can be transmitted together in frequency or time division multiplex.

Abbreviations and designations used in the circuit diagrams

Ä addition element (subtraction element)

AT amplitude divider

B Limiter / device for amount formation

CN signal coding network (abbreviated to coding network in the text)

CN ₁₁ 27 34 865 Bundle fault j I CN ₁ , Signal coding network on the transmitting side Transmission function of the coding network Il DE Signal coding network on the receiving side Impulse response of the coding filter if DN Division facility Cluster error-like interference term E. Signal decoding network (abbreviated as decoding network in the text) Transfer function of the decoding network ⁵ I. F. Decision maker Error signal ι I. filter frequency I. IK Integrator Transfer function of a filter ¹ ί ι KK Inspiration channel Weights I.
ι LCN Control channel Number of information channels ίο i M. Linear signal coding network (abbreviated coding network in the text) Impulse signal I. NCN Multiplier pulse signal on the receiving end I. PIi Non-linear signal coding network (abbreviated in the text as non-linear coding network) spectral amplitude density of the control signal ti S. Phase reversal stage Spectral amplitude density of the control signal at the receiving end SU counter Spectral amplitude density of the disturbed control signal 15 1 Summing up Number of control channels I. K, Syndrome delay network Control signa! ί VC, Comparator control signal on the receiving end T Delay element disturbed control signal Delay element of duration T cross-spectral power density of signals n (t) and r (t) j Formula symbols Cross-correlation function of the signals «(■ /) and s (t) 1 bd) Spectral amplitude density of the interference signal in the control channel P. C (J) Interference signal in the control channel ²⁵ I. c (t) Spectral amplitude density of the interference signal in the information channel C (I) Spectral amplitude density of the interference signal on the receive side I. D (J) Interference signal in the information channel I. C '(t) interference signal on the receiving end I. f Signal to interference power density ³⁰ I. G (J) mixed interfering signal 5 a Spectral amplitude density of the syndrome signal i Snydrome signal I. i (t) Syndrome jump correction signal t. i, (i) Delay duration of a delay element 35 I. K (D Time I. K ₁ (S) Period duration of the sawtooth characteristic I. K (S) Height of the sawtooth characteristic * A. Signal in the transmission-side code network 1 A (O Signal in the code network on the receiving end 40 I. A /. Transfer function of the information channel ί kJi) Transfer function of the control channel P. LJS) Signal in the coding network at the sending end § UO Signal in the code network on the receiving end ' E-
in M (J) spectral amplitude density of the information signal 45 I. m (t) ispectral amplitude density of the disturbed information signal 1 N (S) Information signal N ₁ (S) disturbed information signal
21 I. n (t) t n ,. (I) ■ ^5o I Q (S) r (D \. S (J) I. S (I) S ₁ (I) 55 \ T I. i
i U, ί u, ί U (I) 60 I. Ui (I) i U ₁ (S) I. O _K (S) I. T ₁ '/) 1 Vi- (I) 65 I. X (S) I. XJS) X (I) I. X ₁ XD

YU) Spectral amplitude density of the corrected information signal

y (t) corrected information signal

y _k (f) filtered and corrected information signal

Zif) spectral amplitude density of the correction signal

z (t) correction signal

c (t) relative error signal

μ Number of delay elements in a coding network

ρ Number of taps in a coding network

Used indices

E receiving-side signals

g disturbed signals

/ [Information channel

K control channel

ν running index

attachment

Determination of the mean relative error c _m according to the syndrome mean comparison method

when a single bundle error occurs in the control channel

If a burst error occurs in the control channel in the time segment 0 < t <T , the following applies to the signals of the decoding network according to Fig. 11:

If the influence of previous disturbances in the syndrome register had subsided, i.e. if the last disturbance occurred further-nis μΤ ago, the following applies in the time segment 0 < t <T for the signals at the taps of the syndrome delay network:

-... = s _p - , (0 = 0; *, = bit). The syndrome mean values are thus obtained:

sjt) = bit) or

IT

P I

ä..i, (0 = I //,.,) for bit) Φ This results in the absolute errors:

., e, iCi) = s _v it) - SJt) = 0 - - ^ - = 1 ... ρ - 1 and e _p it) = _Sp it) - SJt) = bit) - - ^

P P

and thus for the relative errors:

V PJ bit) '

In the time segment 0 < t <T one obtains for the mean error signal:

22nd

Since εJO <t <T)> εJT <ί <(μ + I) T) is good for the present disturbance, if the threshold setting of the threshold value decider is below the value 2 (1-1 // ») it cannot result in a wrong correction.

For ρ - 4, εJt) assumes the following value:

Appendix 2

Determination of the mean relative error e _m according to the syndrome mean value comparison method when a bundle error occurs at the same time in the control and in the information channel

The worst case with regard to an incorrect correction occurs when the disturbances in the control and inspiration channel occur simultaneously, since in this case the greatest correspondence occurs in the output signals of the syndrome delay network. In the time segment 0 < t <T, the following applies to the signals of the decoding network according to Fig. 11:

/ any 0 <f < T

If the influence of previous disturbances in the syndrome register has subsided, i.e. if the last disturbance occurred earlier than μΤ , the following applies in the time segment 0 < t <T for the signals at the taps of the syndrome delay network:

s \ U) = S ₂ U) = ... J, -, (/) - 0; S ₁₁ (I) = 6, (0 + b ₂ (t).

For the syndrome mean values one obtains:

sjl) = b _x (t) + b ₂ (t)

s ", U) = (6, (0 + b ₂ U)) / p ₃₅

sjl) = \ / s ", U) = l / (6, (r) + O ₂ U)) for 6, (r) + b ₂ U) * 0.

From this we get for the absolute errors:

<? "(*) = S, {t) - s", U) = 0 - (£>, (/) + O ₁ U)) Ip = 1 ... ρ - 1 40

e "U) = s" U) - SJt) = 6, (t) + O ₂ U) - (ö, (0 + b ₂ U)) lp = [6, (0 + b ₂ U)] Π - -j ₄₅

and thus for the relative errors:

Φ) -e. U) -j. (O-

6, (0

MO - MO · ϊ-ω = [*, ω + «οι (ι -1) · - ^ - (ι -}) ■

In the time segment 0 <;< T one obtains for the mean error signal:

cjt) = Σ I.-'- ■ - '" ^l

Since εJO <1 <T) > εJT <t < (μ + 1) 7 ") applies to the disturbances under consideration, they cannot lead to a wrong value given a suitable threshold setting of the threshold value decider which is below the value 2 (1 - Up) Lead correction.

If the two bundle errors are shifted from one another in time, the worst case occurs when only two taps of the syndrome delay network provide signals other than zero. For 65 the signals of the decoding network according to Fig. 11 applies in this case:

= ... = Sp- ₂ U) = 0; Sp- iW ⁼ bjO) ', SpU) - b] U) ·

For the syndrome mean values one obtains:

sJO = 6, (r) + b; H) s _m {t) = (A ₁ W +

+ 6 -, (O) for 6, (i) + O ₂ (O * From this we get for the absolute errors:

- J _m (r) -

e = 6, (0 -

6, (0 (P - D + A ₂ (Q

The following results for the relative errors:

6 (0 +

; v = l ... p-2

_ A ₂ (Q (ρ -

P (A ₁ (O + _ 6, (0 (ρ - 1) +

"ρ (6, (0 +

In the time period under consideration, one thus obtains for the mean error signal:

μ
^v Ir ( Dl- ^ρ - 2 ₊ A ₂ (O (P - ) + Α, (0 4- 6, (0 (P - D + A ₂ (O 1 »I. P. ρ (6, (0 H ι- A ₂ (O) P (6, (0 + A ₂ (O)

For the special case sig [6, (0] = sig [6 ₂ (0]) the following applies:

A ₂ (O [(P - D + 1] + 6, (0 [(P - D + 1]

ρ (6, (0 + A ₂ (O)

45 The observed disturbance can therefore not lead to an incorrect correction in the time segment 0 <t <yT if the threshold setting of the threshold value decision maker is below the value 2 (1 - l / p).

In the time segment μΤ <t <(u + I) T (correction phase), the following signals are present on the syndrome delay network when a bundle error occurs simultaneously in the information and control channel:

s, (0 = 6i (0 + 6, (0 S ₂ (O = J ₃ (O ... = Jp (O = fri (0. For the syndrome mean values in this period of time) we get:

Sm (O = PA ₁ (O + A ₂ (O

s _m (t) = A ₁ (O + Aj (0 / p

¹ 6, (0 + A ₂ (O *

A ₁ (Op + A ₂ (O 60) This results in the absolute errors:

c _r (0 = J ₁ - (O - J, "(0 = Ai (O - A ₁ (O -

• -2 ... P

* i (0 = 6, (0 + A ₂ (O - A ₁ (O - A ₂ (0 / p = A ₂ (O 1 -

For the relative errors one obtains: ε. = Φ) -IJt) = c, - A ₂ (O U--

PP 1

ρ 6, (r) +

-Σ

+ ö ₂ (0

p-1

A ₂ (O

ρ 6, (0 +

+ 1 A ₁ (0 1 P. A ₂ (0

The bundle _{error A 1} (O can only be corrected here if the mean error ε, _η ( ή does not exceed the threshold value set).

Since the threshold must not exceed the value 2 (1 - - (, a correction is possible if:

2 1

-i)

A ₁ (O

l + p

A ₁ (O

A ₂ (O

or transformed: 6, (0

10

20th 25th 30th 35

1 + p

A ₂ (O

This requirement is surely fulfilled if A ₁ (O and b ₂ (i) have the same sign. If this is not the case, a correction of the bundle _{error 6 (} (/) according to the syndrome mean comparison method is only possible if the following applies:

ρ IA ₁ (O l> 21 A ₂ (O I-

In addition 24 sheets of drawings

45 50 55 60 65

25th

Claims (8)

Patent claims: 1. A method for reducing interference in one or more information signals Cv (O) transmitted from a transmitting station to a receiving station via one or more information signal channels (IK), characterized in that a) that in the transmitting station via one or more coding networks (CN) from the information signal or signals (x (t)) first control signals (k (t)) are derived, which each contain relevant signal components of the information signal (x (t)) , b) that the first control signals (k (t)) are transmitted to the receiving station via control signal channels (ΑΆΤ), cj that in the receiving station via one or more coding networks (CTV) from the received information signal (s) (X ₁₁ U)) second control signals (k _E (t)) are derived and that the transmitting and receiving coding networks (CN) are designed in this way that with trouble-free Transmission in an information signal channel (/ AT) and in a control signal channel (A7O the first and second control signals (A- (O), k _E (t) are equal to each other) d) that syndrome signals (s (t)) are formed by forming the difference between the received first and the derived second control signals (k (r)), (A /. (0), e) that these syndrome signals (s (t)) are converted into correction signals (z (/)) in one or more decoding networks (DN) so that the decoding networks (DN) have the inverse characteristics of the coding networks (CN) , and 0 that the correction signals (z (t)) are additively superimposed on the received information signals (X ₁₁ U) , so that the interference that has penetrated into the information signals Cx (O) via the information signal channels (/ AT) is reduced. 2. Circuit arrangement for performing the method according to claim 1, characterized by linear coding networks (CN) and linear decoding networks (DN). 3. Circuit arrangement according to claim 2, characterized in that the linear coding networks (CjV) non-linear coding networks (NCN) and the linear decoding networks (DN) are followed by non-linear coding networks and that by appropriate selection of the characteristic of the non-linear coding network (NCN) and the inverse thereto Characteristic curve of the non-linear decoding network, the signal amplitude ranges can be selected in which the interference can be reduced most. 4. Circuit arrangement for performing the method according to claim 1, characterized in that the coding networks (CA ') and the decoding networks (ZW) each have a linear filter and a downstream Tetes network with a sawtooth-shaped characteristic and that the receiving station each have a limiter connected downstream of the control signal channel (s) (KK) , through which all interference amplitudes that exceed the amplitude range of the first control signals (k (t)) are suppressed. 5. Circuit arrangement according to claim 4, characterized by a syndrome jump correction device at the receiving end, via which an amplitude-discrete signal (v, U)) of uniform step height is derived from the syndrome signal (s (t)) with jumps and the syndrome signal (s (t)) for compensation the syndrome jumps is superimposed, and the value of the amplitude-discrete signal Cs ₁ (0) is reduced or increased by a step height depending on whether the slope of the syndrome signal (s (t)) exceeds a maximum value or falls below a minimum value. 6. Circuit arrangement according to claim 4, characterized by a syndrome jump marking device assigned to the linear coding network (CN) , via which a transmit and a receive-side pulse signal (/ (0, i _E U)) is derived, depending on the direction in which the coded transmission-side information signal (.v (0) or the reception-side information signal (χ ,, (ή) crosses an infinitely steep branch of the sawtooth-shaped characteristic curve, emits a positive or negative pulse, and the transmission-side pulse signal (/ (0)) via the control channel ( A * AT) is transmitted to the receiving station and, together with the receiving-side pulse signal (i _E (t)), is fed to a counting decoder, which emits an amplitude-discrete signal (5, (0), the amplitude of which is proportional to the count of the counting decoder and that of the compensation the amplitude jump of the syndrome signal Cs (O) is used. 7. Circuit arrangement for carrying out the method according to claim 1, characterized in that the coding network (CN) has delay elements which are connected in a chain, the taps auP · and in which all delay times between any two taps are different, that the decoding network (DN) has a syndrome delay network (SVN) which also contains delay elements and taps connected in a chain, and the taps of the syndrome delay network (SVN) are arranged as a mirror image of those of the coding network (CN) , and that it is established in the decoding network (DN) the similarity of the signals to the individual tap fations, and depending on the degree of similarity from the variably weighted mean of the Signals the correction signal (r (0)) is formed. 8. Circuit arrangement for performing the method according to claim 1, characterized by an adaptive coding network (CN _h ) in the receiving station, and by at least one filter (F) and a downstream integrator (/), wherein the filter (F) from the syndrome signal (s (t)) sub-signals are derived which are used for the adaptive setting of the receiving-side coding network (CzV ₄ ), the The nominal value of the partial signals is zero.