DE3736125C2 - Equalizer for time invariant or slow time variant channels - Google Patents

Description

The invention relates to an equalizer in the preamble of claim 1 Art.

Equalizer to equalize time invariant or slow time-variant channels with disruptive multipath propagation work e.g. B. according to the best known Viterbi algorithm. However, this requires a very high computing effort and is therefore too slow for many applications.

The invention is based on the object of an equalizer the type mentioned in the preamble of claim 1  specify the without significant loss of equalizer power requires significantly less computing effort and is therefore faster and less expensive to implement is.

The task is in the equalizer mentioned by the in characterizing part of claim 1 Features resolved.

Advantageous refinements and developments of Invention are specified in the subclaims.

The invention will now be advantageous with reference to drawings Embodiment explained in more detail. Show it in detail:

Fig. 1 block circuit of the receiving device;

Fig. 2 block circuit of the equalizer;

Fig. 3 block circuit of a survivor extension.

In Fig. 1 is a block diagram of a receiving device for the equalization of messages that are transmitted over slowly time-variant channels in the double sideband method without carriers, is shown.

The received signals s1 pass from the input 1 via a bandpass-low-pass transformation device 2 , which generates the complex envelopes s2 of the received signal, to a channel correlation filter (channel-matched filter KMF) 4 and to a channel impulse response h (t) and the autocorrelation function r (t) determining and adapting device 8 .

The output signals s4 of the filter 4 are sampled by a sample / hold circuit 9 and the sample values s5 are fed to a first input of the equalizer 5 according to the invention. A second input of the equalizer 5 is connected to the device 8 for the transmission of the likewise sampled autocorrelation function r (i).

The equalized signals s6 are fed from the signal output of the equalizer 5 to a decoder and decision maker 6 , at whose output the recognized bit sequence s7 can be taken for further evaluation.

A device 8 for determining the channel impulse response h (t) is e.g. B. known from German published patent application DE-OS 35 40 716. The channel impulse response h (t) is used for setting the filter 4 and the autocorrelation function r (i) also generated in the device 8 for setting the equalizer 5 .

In the case of time-invariant channels, the channel impulse response h (t) and the autocorrelation function r (i) can be stored as a constant in the filter 4 or in the equalizer 5 , so that the devices 8 and thus also the device 7 described later can be omitted.

The equalizer forms a receive sequence s6 from the sample value sequence s5, which is created by sampling the output signal s4 of the filter 4 at the symbol clock k * Ts, which should correspond as closely as possible to the transmission sequence and from which a suitable decoder and decision maker 6 determine the recognized bit sequence s7 can.

In FIG. 2 is a block diagram of the equalizer 5 is shown. SE1 to SENs denote a set of Ns survivor extension units, AS a selection and sorting unit and ME a feature extraction unit. The survivor expansion units SE1 to SENs are supplied with the sample values s5 (cf. FIG. 1) and the auto-correlation function r (i). At the output of the Survivor expansion units is in each case via a plurality of output lines for parallel output of the possible survivors mS 1 determined by the Survivor expansion units. . ., mS _Ms the selection and sorting unit AS connected. This sorts the possible survivors mS₁ to mS _Ms determined by the survivor expansion units SE1 to SENs and outputs valid survivors gS₁ to gS _Ns . Of the valid survivors, a special, selected survivor, the so-called. Minimum Survivor gS₁ supplied to the feature extraction unit ME, which delivers the recognized symbols s6 to the output A of the equalizer. Furthermore, the valid survivors gS₁ to gS _Ns become their respective survivor expansion units SE1,. . ., SENs returned.

The equalizer 5 will now be described in more detail below. It works according to the principles of the Viterbi algorithm, which is known as optimal, but has some significant changes, so that the very high computing effort of the Viterbi algorithm is avoided without an excessive loss of equalization performance occurring.

The resulting impulse response from channel and modulation low pass 2 is the impulse response h (t) determined by device 8 . The following applies to a sample value s5 = z (K) at the output of the filter 4 , which is sampled at a time interval of a symbol duration Ts:

Here xs (K-i) denotes that sent at the time (K-i) · Ts Symbol and r (i) the (complex) autocorrelation function the channel impulse response h (t) for the shift i · Ts. For the autocorrelation function applies:

where h _* (t) is the conjugate complex shock response.

On the reception side it is assumed that the car correlation function r (i) over a period of N · Ts only changes insignificantly and that applies:

r (i) = 0 for | i | <N (3)

Since the autocorrelation function r (i) from the known measured or adapted resulting channel impulse response h (t) can be determined for any given one Symbol sequence x (K) with Eq. 4 a model value sequence calculate z ′ (K):

The task of the equalizer 5 is to generate a sequence s6 = x (K) in such a way that the model value sequence z '(K) matches the sample value sequence s5 = z (K) as closely as possible. Then the transmitted sequence xs (K) and the generated sequence x (K) match well. If such a sequence is found, the sequence s6 = x (K) is referred to as the reception sequence and its elements as reception symbols.

At any point in time K, the state of the equalizer 5 according to the invention is characterized by a set of Ns (K) valid survivors

gS _i (K) with i = 1,. . ., Ns

Typical values for the amount Ns (K) are 8 to 128. The values for calculating the initial survivors are stored in a memory. The required number the survivor depends on the length of the channel impulse response and the link length of a possibly existing one Encoding.

Each Survivor S _i (K) consists of at least:

• a) a sequence of (L + 1) <N
  Quadrature amplitude modulation symbols
  (QAM symbols) x _i (K, h)
  with h = 0,. . ., L
• b) a sequence of (N + G + 1) symbol errors y _i (K, h)
  with h = 0,. . ., (N + G) and
• c) a size designated with total costs U _i (K).

The sequence lengths depend on the length of the channel impulse response h (t).

A coding state C _i is advantageously carried at the beginning of each survivor. This makes it possible to equalize the transmission signal using the convolution coding (trellis coding).

The coding state C _i (K) describes the symbol alphabet X _i (K + 1), from which the next symbol of the sequence x _i (K, h) must be selected. In the case of uncoded transmission, the coding state C _i (K) is constant for all i and K. The symbol alphabet X _i (K + 1) consists of M _i (K + 1) symbols.

In each Survivor expansion unit SE1,. . ., SENs a transmission symbol generator is provided in each case, which per sampling cycle all to the QAM symbol sequence x _i (K, h) of the returned survivor gS₁,. . ., gS _Ns suitable QAM symbols x _j (K + 1) are generated. The appropriate symbols are taken from the symbol alphabet X _i (K + 1), which depends on the coding state C _i (K) of the returned survivor gS _i (K). A new coding state C _j (K + 1) is generated for each symbol x _j (K + 1) generated by the symbol generator, depending on the convolutional code used.

At time K, all sampled output values s5 = z (k) of filter 4 up to and including k = K are evaluated in the equalizer. The evaluation results are included in the survivors.

At the transition to time (K + 1), the equalizer evaluates the Sample value z (K + 1) as follows.

Per survivor S _i (K) with i = 1,. . ., Ns and per symbol x (K + 1) of the symbol alphabet X _i (K + 1), which consists of M _i (K + 1) symbols, a possible new survivor mS _j (K + 1) is calculated. Thus at the time (K + 1)

possible survivors. Of these, a predetermined number Ns (K + 1) selected by valid survivors.

The new possible survivor mS _j (K + 1) is calculated from the returned survivor gS _i (K) and the special symbol x _j (K + 1) of the symbol alphabet X _i (K + 1).

Every new possible survivor mS _j (K + 1) consists of:

• a) the coding state C _j (K + 1). This state is determined using the coding rule, e.g. B. a transition table or from the transmission symbol generator, from the old coding state C _i (K) and the symbol x _j (K + 1).
• b) the QAM sequence of (L + 1) <N symbols x _j (K + 1, h) with h = 0,. . ., L. It is: The QAM symbol sequence x _j (K + 1, h) of the new survivor is created by appending the generated symbol x _j (K + 1) to the symbol sequence x _i (K, h) of the returned survivor gS _i (K)
• c) the sequence of (N + G + 1) symbol errors: y _j (K + 1, h) with h = 0,. . ., (N + G) The new symbol error is determined by gradually compensating the values of the autocorrelation function r (i) as follows: The preliminary symbol error results from Eq. 5 taking into account Eq. 3 to: with h = 0,. . ., (N + G)
• d) and the cost U _j (K + 1). These result from correction of the total costs U _i (K) of the returned survivor gS _i (K) with the symbol error y _i (K + 1, N) according to the equation: U _j (K + 1) = U _i (K) + | y _j (K + 1, N) | ² (7)

The new possible survivor mS _j (K +1) is now fully calculated.

For the selection of the valid survivors (survivor selection), sizes D _j (K + 1) designated with "provisional costs" for each new possible survivor are calculated in the survivor extension units SE1 to SENs or in the selection and sorting unit AS. This is done with the help of the symbol errors

Y _j (K + 1, h) with h = 0,. . ., N.

The new preliminary costs are:

When choosing the weighting function w (i) and the exponent b there is some freedom. However, it should w (N) = 1. A cheap choice follows z. B. from:

Another choice, especially with short shock responses is advantageous is

In both cases, b = 2 is a cheap choice.

The new possible survivors mS _j (K + 1) are sorted in the selection and sorting unit AS according to increasing preliminary costs. The _survivor mS _J0 (K + 1) with the smallest preliminary costs D _j (K + 1) with j = J0 is called the minimum survivor. It is selected first as a valid survivor gS₁ (K + 1). The selection of the other valid survivors gS _i (K + 1) with i = 2,. . . , Ns happens according to the provisional costs and possibly the symbol sequences of the survivors already selected. Here, for. For example, the selection systems proposed by AP Clark can be used:

A.P. Clark, S.M. Asghar. Detection of digital signals transmitted over a known time-varying channel. IEE PROC., Vol. 128, Pt. F, No. 3, June 1981.

The equalizer output is based on the minimum survivor, which has the number j = J0. From the equalizer output:

• a) the recognized symbol s6 = x _J0 (K + 1, L) delayed by L. The recognized symbol is fed to the decoder 6 .
• b) the provisionally recognized symbol s8 = x _J0 (K + 1, N + G), delayed by (N + G) symbols.
  This symbol is fed to the device 8 and possibly a phase and frequency control loop 7 . In device 8 it is used to adapt the channel impulse response to the transmission channel.
• c) if applicable, the symbol error s9 = y _J0 (K + 1, N + G) of the provisionally recognized symbol s8 = x _J0 (K + 1, N + G). The symbol error s8 can also be fed to the phase and frequency control loop 7 . Together with the provisionally recognized symbol s8, this can be used to determine a phase error which can advantageously be used to control the phase and the frequency in the demodulator.

The phase error can be determined using the following equation:

Δρ (K + 1) = arc (1 + y _J0 (K + 1, N + G) / x _J0 (K + 1, N + G)

In Fig. 3, the signal processing is shown in a Survivor- expansion unit must be performed per created symbol x _j (K).

The transmission symbol generator of the symbol selection unit SA generates a new symbol x _j (K) during the transition from the symbol clock K-1 to the symbol clock K, which is the previous one in a filter F1 and a stored symbol sequence X _i (K-1, h) in the sense of a predetermined coding is appropriate. In the convolutional coding used, which at
G. UNGERBOECK, Channel Coding with Multilevel / Phase Signals.
IEEE Trans. Inform. Theory, Vol. IT-28, No. 1, pp. 55-67, Jan 1982
is described in detail, the old coding state C _i (K-1) is used and a new coding state C _j (K) is generated.

The distortions contained in the input signal (s5) are in filters F1 and F2 using the autorrelation function r (i) compensated.

The "followers" of the autocorrelation function are in filter F1 and in filter F2 the "precursor" of the autocorrelation function compensated.

The intermediate results of the filter F2 result in the symbol errors y _j (K, 0) to y _j (K, N), which control the device VK for determining the provisional costs.

The symbol error y _j (K, N) is used to update the total costs U _j (K), the previous costs U _i (k-1) being used. The previous costs U _i (k-1) are also used to determine the preliminary costs D _j (k). The old symbol errors y _j (K, N + n), with n = 0, 1,. . ., G and the old QAM symbols x _j (K, N + n), with n = 0,. . ., L are stored in the delay lines V1 and V2, respectively. They are used by the characteristic extraction unit ME.

The sizes x _i (), y _i (), C _i and U _i are part of the returned survivor gS _i (K-1), while the sizes x _j (), y _j (), C _j and U generated from them _j then belong to the generated possible survivor mS _j . The Survivor also includes all filter content.

Claims (6)

1. Equalizer for time-invariant or slowly time-variant channels, via which messages are transmitted in the two-sideband method without a carrier, in particular for transmission with quadrature amplitude modulation (QAM) for use in a quadrature demodulator with a channel correlation filter and a device for determining and adapting the channel impulse response and the autocorrelation function of the channel impulse response for shifts of integer multiples of the symbol duration,
characterized in that a number of survivor expansion units (SE1,..., SENs) are provided, to which the sample values (s5) emitted by the channel correlation filter ( 4 ) and the device ( 8 ) for determining and adapting the channel impulse response determined Autocorrelation function r (i) are supplied,
that at the output of the survivor expansion units (SE1,..., SENs) a selection and sorting unit (AS) is connected, which the survivor expansion units (SE1,..., SENs) determined by the survivor expansion units (mS₁,. ., SENs) sorted and valid survivors (gS₁,..., GS _Ns ) selected,
that a selected survivor (minimum survivor gS₁) is fed to a feature extraction unit (ME) which delivers the recognized symbols (s6) to the output (A) of the equalizer, and
that the valid survivors (gS₁,..., gS _Ns ) are returned to their respective survivor expansion units (SE1,..., SENs), with a survivor at time K consisting at least of a sequence of QAM symbols ( x _i (K, h)), a sequence of symbol errors (y _i (K, h)) and total costs (U _i (K)), and the sequence lengths depend on the length of the channel impulse response (h (t)). 2. Equalizer according to claim 1, characterized in that a coding state (C _i (K)) is carried in the survivor in order to enable the equalization of a transmission signal with convolutional coding (trellis coding). 3. Equalizer according to claim 1 or 2, characterized in that in each Survivor expansion unit (SE1,..., SENs) there is a transmission symbol generator (SA), which per symbol cycle (K) all to the QAM symbol sequence of the returned survivor ( gS₁..., gS _Ns ) generates suitable QAM symbols (x _j (K + 1)), taking into account the coding state (C _i (K)) and a new coding state (C _j (K + 1)) for each generated QAM symbol (x _j (K + 1)) is generated depending on the convolutional code used,
that for each generated QAM symbol (x _j (K + 1)) a possible survivor (mS _j (K + 1)) is generated, the coding state (C _j (K + 1)) of which is generated by the transmission symbol generator (SA), whose QAM symbol sequence (x _j (K + 1)) by appending the generated QAM symbol (x _j (K + 1)) to the QAM symbol sequence (x _i (K, h)) of the returned survivor (gS _i ) arises, whose symbol errors (y _j (K + 1, h)) are determined by step-by-step compensation of the values of the autocorrelation function r (i), and whose total costs (U _j (K + 1)) are determined by correcting the total costs (U _i ( K)) of the returned survivor (gS₁,..., GS _Ns ) with a symbol error (y _j (K + 1, N)),
that for survivor selection either in the survivor extension units (SE1,..., SENs) or in the selection and sorting unit (AS) for each possible survivor (mS _j (K + 1)) using all symbol errors (y _j ( K + 1, h)) preliminary costs (D _j (K + 1)) are determined,
that from the selection and sorting unit (AS) of the possible survivors (mS _j (K + 1)) the survivors with the lowest preliminary costs (D _j (K + 1) as valid survivors (S _i (K + 1)) can be selected that meet the specified conditions with regard to the QAM symbols and that are processed in the next symbol cycle (K + 1), whereby specified conditions with regard to the QAM symbols are taken into account. 4. equalizer according to claim 3, characterized in that the preliminary cost (D _j (K + 1)) according to the relationship can be determined, w (i) preferably according to the relationship or, especially with short shock responses, according to the relationship is determined and b is preferably selected equal to 2. 5. equalizer according to one of claims 1 to 4, characterized in
that the selection and sorting unit (AS) selects the survivor (gS₁ (K + 1)) with the smallest preliminary costs (D _j (K + 1)) as the minimum survivor,
that the feature extraction unit (ME) extracts a preliminary symbol (s8) with the associated symbol error (s9) from the QAM symbol sequence of the minimum survivor in addition to the recognized symbol (s6),
that the provisional symbols (s8) with associated symbol error (s9) are fed to a phase and frequency control loop ( 7 ) which emits a first output signal (s10) for phase control to the device ( 8 ) for determining and adapting the channel impulse response and a second output signal (s11) a frequency multiplier feeds a multiplier ( 3 ). 6. Equalizer according to one of claims 1 to 5, characterized records that all the signal processing necessary for equalization steps using a programmable digital processor are realized.