DE3011828A1 - Demodulation circuit for unknown channel property wave - has adaptive phase corrector in digital hilbert filter pair - Google Patents

Abstract

Already in Patent Appl. P3006801 a synchronous demodulation receiver has been described in which analogue signals derived from a band pass channel are regenerated and transformed into analytical signals by means of a Hilbert filter pair. In present this filter pair is made also adaptive, that is, self-adjusting to the parameters of the received signal. This is obtained by a phase subtractor member producing the error vectors and quadrant error derived from the vectorial output signals (zp, zq) derived from the Hilbert filter pair summing circuits. The identified error vectors are utilised for modifying the coefficient input signal (h/p-n, h/q-n) to the multipliers whose other inputs derive from the amplitude-sampled wave form segments in the delay stages (T/1). The continually varying multiplication coefficient derives its sign from an algorithm connecting the sampling clock interval (T), the index (i) of the clock pulse sequence, the adaptor factor for the sample step width, the input signal sequence (u) the difference vector (z-q) the vectors at the input and output of the so called decision circuit (Es), and the carrier frequency (f).

Description

recipient

The invention relates to a receiver for synchronous demodulation of real bandpass signals that have been suppressed or equalized by means of filter filters, which are produced by means of a Hilbert filter pair are converted into analytical signals, which are rotated by means of a phase shift element, be scanned step by step and queried in a decision maker.

Such receivers are known, for example from the article Microprocessor Implementation of high-speed data modems "by Gerwen and others from IEEE transactions on Communications, Vol. COM-25, No. 2, Febr. 77, p. 238 and ff.

or from "A contribution to digital equalization and pulse shaping in data transmission over linear channels "by Heinrich Schenk, dissertation at the University of Erlangen.

In the case of data receivers of this type, the individual functions are in realized separate filter arrangements, for example the synchronous demodulation with phase separation in a digital Hilbert filter pair, the optimal noise suppression in a matched filter and the signal equalization in a special equalizer filter. With fast working data transmission devices are to achieve a small Bit error rate in addition to the Hilbert filter pair, the other filters mentioned are required, so that a relatively high circuit complexity occurs.

If the channel properties are not known at the receiver, the Equalization of the transmission channel to be done adaptively, which is an additional effort evokes.

The invention has therefore set itself the task of a receiver of the above type, which manages with significantly less circuit complexity. The receiver should achieve the best possible equalization of impulse crosstalk, even if the channel properties are unknown. Furthermore, a recipient should can be specified, which also enables optimal noise suppression.

These tasks are given with the in the claims Funds resolved.

The receiver according to the invention achieves that the entire Circuitry for realizing the filter functions is smaller, with the Receiver an optimal equalization of the impulse crosstalk even with unknown channel distortions supplies. In one embodiment of the invention, the receiver also enables an optimal noise reduction. With discrete-time or digital execution of the Filter functions is particularly advantageous that all filter assemblies with the low step act ar- can work and not for multiplication need the high sampling rate.

The invention is described with reference to the figures.

FIG. 1 shows the block diagram of a receiver according to the invention which works with quadrature amplitude modulation. The input signal supplied by the channel is sampled with kT / l, where T is the step cycle length, 1 is a natural number> 1, which is determined by the highest frequency fg in the received signal according to the sampling theorem l / T-2fg, and k is the sampling cycle index. The sampling samples are fed to a transversal filter, the two outputs of which supply the quantities v and v, which together form the -p -q vector received signal result. A prerequisite for good demodulation is that the phase responses of the transverse filter pair differ by n / 2 in the frequency range of interest as precisely as possible and that the attenuation responses have an all-pass character. The instantaneous values of the vectorial received signal at the times of the step clock iT are then rotated in a phase rotation element Ph by the angle AiTf, where f is the carrier frequency, and result with correct equalization, synchronous step clock and a phase rotation. the correct angle d desired vectorial output signal which approximately coincides with the transmission signal.

In a decision maker, it is checked whether the reception signal matches the setpoint signal.

coincides with the received signal z in non-reversible unambiguous Way, an estimated value q is assigned, which is the same with interference-free transmission the value sent, and where a target / actual comparison is made by forming the difference the vector components is performed.

The output, the error vector is fed to the transversal filter pair for setting its coefficients. The transversal filter contains a chain of delay elements Til, at the beginning of which the sampled values of the input signal are entered. The principle of the transversal filter now consists in the sampling values of the input signal u (k / lT), u ((k 1) / lT) ... u. ((Kn) / lT), delayed by whole multiples of Til, by coefficients hn 'h n + 1 ... hg valued, add up. This is carried out both for the normal component and for the quadrature component, so that the equalized signal components or vq arise at the two summing outputs. The evaluation is carried out by means of a setting multiplier, at the second input of which the individual coefficients h are present.

In addition to FIG. 1, FIG. 2 shows a circuit arrangement for adaptive setting of the filter coefficients h, with one of the coefficient branches is shown for the normal component. In the arrangement, they become the delay chain continuous real input signals with phase-corrected error size 6 bp ', that is the p in-phase component of the by means of a further phase rotation element around the Angle 2sitz rotated back error vector 6, multiplied in a further multiplier. The products are fed to another multiplier, at its second input the size -2a with a as the setting increment- factor pending. the Error output variables are stored in an accumulator, which consists of an adder feedback delay element T exists, accumulated. The at time (i-1) T The coefficient value stored in the accumulator is corrected and used for Time of the step cycle i.T as a new coefficient value hp (iT) on the setting multiplier given, whereby the evaluation of the input signal mentioned above takes place. With this Arrangement, an optimal equalization of the impulse crosstalk can be achieved.

In one embodiment of the invention, noise suppression is also used achieved. For this purpose, there is a further feedback of the accumulated error output variables via another multiplier, via which a weighting with a weighting factor g takes place, multiplied by the phase-corrected error size b Input signals, p where the sums of both by means of a further adder are formed, which are then subsequently, as described above, in the following multiplier be multiplied by -2a.

The weight factor g is a measure of the two transversal filters let through noise power, which is the dissertation of Schenk "A contribution to digital equalization and pulse shaping for data transmission via linear channels, Page 14 and ff. is more closely defined. Through the described feedback loop becomes an absolutely necessary noise suppression device in fast todems dispensable in front of the input of the transversal filter pair.

In a further embodiment of the invention is used to obtain the Filter coefficients only use the sign of the input signal sample. this is indicated by the dashed block SIGN, which precedes the input of the first Multiplizie rers, in which the error size multiplication takes place, tet is carried out. As a result, the error size multiplication is reduced by less effort Switching elements realized, and it can be a rough, but fast adaptation of the equalizer take place in the running-in phase.

In a further embodiment of the invention, a quantization takes place, i.e. a reduction in the word lengths of the Error output variables or the current coefficient values h. The latter is p by the dashed block QUANT, which is inserted here before the setting multiplier is presented, layed out.

Claims (7)

Claims »Receiver for synchronous demodulation of real, filter suppressed or equalized bandpass signals, which are filtered by means of a Hilbert filter pair are converted into analytical signals, which are rotated by means of a phase shift element, are scanned step by step and queried in a decision maker, with the Equalization and conversion functions combined in a single pair of filters are, according to P30 06 801 (registration according to BK 79/92), characterized that the filter pair is equipped adaptively. 2. Receiver according to claim 1, characterized in that the coefficients h of the filter pair by means of the iterative gradient method according to the algorithm # [(i + 1) T] = (# -2α | u / oo / u | # #) [iT] with # = | cos2 # iTf # sin2 # iTf | | # sin2 # iTf cos2 # iTf | , with the step cycle length T, the step cycle index i, the adjustment step width factor a, the input signal sequence u, the difference vector - = zq, the decision input or f output vectors z or q and the carrier frequency f. 3. Receiver according to claim 2, characterized in that the algorithm an additional summand -2ag.hLi-T] can be added, which is a measure of the Filter pair transmitted noise power and where g is a weighting factor. 4. Receiver according to claim 2 or 3, characterized in that of of the input signal sequence u in the algorithm only the sign (SIGN) is used. 5. Receiver according to one of the preceding claims, characterized in that that the filter pair contains the function of noise suppression. 6. Receiver according to one of the preceding claims, characterized in that that the pair of filters is time-discrete or digital. 7. Receiver according to claim 6, characterized in that the Generation of the coefficients h a reduction (QUANT) of their or the word length of the generating quantities takes place.