SK132994A3 - Method of transfer of digital signals and receiver for realization of this method - Google Patents

Abstract

Method in which signals are sent by modulation of a carrier frequency and, on reception, in the frequency domain, the received signals are equalized (11) before making them pass again into the time domain in order to carry out recognition thereon (20), and in which the signals are demodulated after equalization, the phase errors of the signals ( epsilon k) are measured according to prior equalized signals, and complex phase rotation coefficients (e<j theta i>) are determined, serving for the equalization (18) of subsequent signals. Application to digital television. <IMAGE>

Description

A method for transmitting numerical signals and a receiver for performing the method

Technical field

In particular, the invention relates to a method of transmitting numerical signals, in which the signals are transmitted by modulating the carrier frequency and receiving the received signals in the frequency domain before they are returned to the transition region and thus surveyed. The present invention is used primarily for hertz transmissions, but can also be used for any other type of transmission, for example by cables.

BACKGROUND OF THE INVENTION

At present there is an attempt to realize the transmission of signals in a hertz way in numerical form and no longer in analogous form. Thus, the transmission of radio or television programs in numerical form allows theoretically, upon reception, to restore analogue signals of sounds and images without defects caused by disturbances in the Hertz channel. In fact, the state of the received numerical signal is correctly detected, even if it is loaded with noise, and only a serious fault, which is unlikely, could cause an error by detecting a state other than the one being transmitted. In fact, Hertz channel transmission can accurately cause quasi-disappearance caused, for example, by the subtraction of homologous signal streams which, although containing identical signals, are differently delayed and thus differently phased and differently attenuated as they followed different paths. It is therefore necessary to compensate for propagation distortions. Numerization allows to attach a code word for error detection or correction to the signals, which, if necessary, further increases the transmission size due to disturbances by correcting detection errors without excessively increasing the number of signals being transmitted.

From IEE Proceedimgs E. COMPUTERS + DIGITAL TECHNIQUES, Vol. 128 č. 6, November 1981, STEVENAGE GB, p. 239 244, CORSINI et al. We already know the equalizer in which the modulated signal received passes into the frequency domain to be aligned there before it goes back to the transition region for recognition. The demodulation required to receive the signals does not take place there.

The invention is directed to the reception of numerical signals, which means even lower sensitivity to noise and signal distortions, without thereby reducing the spectral efficiency of modulation.

SUMMARY OF THE INVENTION

Thus, the invention relates to a method of the above type, which consists in demodulating the signals after receiving, measuring the phase errors of the signals according to the previously aligned signals and determining the complex phase rotation coefficients used to equalize the later signals.

In this way, an equalization performed continuously based on the phase errors that persist after it is used, which is a feedback loop, which tends to reduce errors caused by later signals. In particular, defects in the transmission channel and their development are taken into account and compensated.

The advantage is that after correcting the phases of already-matched transient signals with complex coefficients, the residual errors of the signals that have returned to the transition region are determined for the block of signals, which are errors that compose to calculate the quadratic error that area, to apply a set of modulation line corrections to carrier wave compensation, and use a set of modulation line corrections by multiplying each line by a specific coefficient obtained by subtracting the previous coefficient of a complex number proportional to the error in the frequency domain. relative to a given line and in relation to the signal associated with the signal representing the line immediately after Fourier transformation.

In this way, the energy of the lasting noise is reduced.

It is also advantageous that after subtracting the frequency (5) near the frequency of the carrier line and independent of their changes, the carrier wave received near the zero frequency is transposed before equalization.

In a first embodiment, before the emission, the signals to be transmitted are grouped into blocks of different size, which are individually supplemented in a circular manner, and on reception, the frequency domain alignment is applied in each block after the circular element has been removed. On this topic, reference may be made to Theory and Application of Digital Signal Processing by L.R.Rabinet and B.Gold, Prentice Halí, 1975.

Thus, the reception effects compensate for the marginal effects that create distortion in certain Fourier transform calculations for certain signals, since the discrete Fourier transform concerns only a limited number of points, creating distortion.

In the second embodiment, the received signals of blocks of a certain size are grouped upon reception, which are then added individually after the conversion to numeric data at the end of a sequence of a certain zero-size numeric data that transitions temporarily to an adjacent block. they shall incrementally add the offset data that is in the overlap zones and the offset signals of the adjacent block temporarily belonging there.

The above compensation is in this case carried out after transmission, without the need to take action on the issue or without recognizing the status of the received units.

In the case of reception, it is always advantageous that the signals can still be regrouped into blocks of a certain size, which partially overlap, before the equalization on the blocks and after the equalization, eliminating those of the equalized signals that come from the received signals,

- IbAs located in the overlap zone of two consecutive blocks.

The procedure applies only to the transmitted signals themselves.

The invention also relates to a receiver of numerical signals transmitted by modulation of a carrier wave comprising means of deciding whether or not a decision to recognize transmitted signals is made with respect to the received signals, comprising:

computing means for performing direct and discrete

Fourier transform on the received signal, providing a spectrum of modulated carrier wave frequency line signals, means for aligning the line signals relative to the propagation distortion estimated by the reference (reference) signal,

- computational means equipped to apply a discrete inverse Fourier transform to the spectrum of the line offset signals and to provide a transient result for the decision means, which is based on the fact that the decision means are equipped to measure phase errors of the offset signals relative to the reference signals; the phase rotation of the received additional signals was controlled.

The invention will be better understood by the following description of a preferred embodiment of a transmitter and receiver for carrying out the method of the invention with reference to the accompanying drawing, in which:

FIG. 1 illustrates a first embodiment of a frequency equalizer of the receiver; FIGS. 3 to 5 illustrate three variants of a transmission method according to the invention; FIG. 6 shows a diagram of a numerical receiver demodulator; 7 illustrates a method of performing an adjustment of the correction correction coefficients.

The principle of the method of transmitting numerical signals is illustrated in the diagram of Figure 1. Numeric signals for transmission, here by Hertz path, are applied to a Hertz transmitter 1 for modulating the carrier frequency, the modulation being situated in the frequency band that surrounds the carrier wave. Filtration is envisaged here to limit the necessary bandwidth. The carrier wave passes the hertz channel of transmission 2 to reach receiver 3 where it passes through the filter to eliminate out-of-band signals. The modulated carrier wave is then used for the mixer 4 obtaining the output of a free oscillator 5 oscillating at the nominal carrier wave frequency, but without feedback and thus does not include variations or possible derivatives thereof.

Thus, the carrier frequency at the output of the mixer 40 is brought back to the zero frequency and the baseband modulation thus obtained is applied to the frequency equalizer 11 supplying the numerical demodulator 20 making the decision to dissolve the downstream signals. The output of the demodulator 20 provides in the adapter circuit 18 of the equalizer 11 numerical signals representing errors measured between the state of the offset subsequent signals, loaded with noise, and the state that was predicted and acknowledged.

Figure 2 illustrates a first embodiment of a frequency equalizer 11. The sampling circuit 12 receives a modulated carrier wave and samples it with sample numbering at a frequency of 1 / T, where T is the modulation time or symbolic duration during which the signal is transmitted. Subsequent sampled signals are represented in parallel by blocks of N signals (N is quite positive) in a series / parallel verifier 13 coupled at the output to a calculation block 14 performing a Fourier transform discrete and direct on the set of N signals or sampling points. The calculation block 14 thus produces a spectrum of frequency lines. The amplitude and phase of each line represented by the numerical data are applied to a line-specific numeric multiplier belonging to the set of 15 multipliers and whose multiplication coefficient, which is complex, is adjusted to compensate for the attenuation and phasing / phasing that occurs in channels for transmission by the 2 frequency component of the modulation at the frequency of the line (s) considered.

The modulation of the carrier wave by each signal, distributed throughout the transmitted band and represented by a set of lines, is, as will be explained below, free of distortions induced by channel propagation

Second

The data of the block showing the amplitudes and phases of the lines so aligned are then exposed in a 16 discrete inverse Fourier transform calculation block that brings the modulated carrier wave back to the transition region, that is, each transition signal is obtained based on the data representing the amplitudes and phases of the aligned lines . The numeral parallel / series converter 17 then transmits the series transient signals coming out in the calculation block 16 to the numeric modulator 20. Thus, the few considerably weakened lines in channel 2 whose increased amplitude amplification during alignment is inaccurate because of noise load have only limited effect In the intermediate signal output at the converter output 17, it is even predictable, as will be explained below, to limit the amplification of very weak lines so as not to reduce the signal / noise ratio.

The adaptation circuit 18 receives, inter alia, phase error information from the demodulator 20 to adjust the phase of the complex multiplier coefficients 15.

A first variant of the transmission method is illustrated in Figure 3. In order to compensate for the distortion caused by the discrete-Fourier transformation relating to a finite number N, while the theory predicts an infinite length signal, the emission process prior to this distortion is performed.

The signals to be emitted are a block-to-block circuit 31 grouped by blocks of N points (N already defined) and a circuit 32 is connected to each block by a prefix (or suffix) of a circular shape. Then, if we have a block of N signals for emission (and _Q , a ^, ..., ^and N_ _x ) / we complete it with a circular prefix of symbols M to create a block with the length M + N we transmit. Motif, prefix or suffix, so to speak circular, because we have subsequent influences during calculations of convolution with position shift each time. The signal coming from the mixer 4 passes through the down-pass filter 33 and then the sampler 34, which operates at 1 / T cadence, thus selling the received block (r _o , r, ··, to the sync circuit block 5) which recognizes the position of each block and removes the last samples M and sells the block (r,, ..., ^r w) obtained and thus pre-aligned to the equalizer 11, to pass to the frequency domain This method of using a circular prefix (or suffix) causes the linear convolution channel 2 is equal to circular convolution, which is the kind used for discrete Fourier transform.

A second variant of the transmission method using the overlapping and summation technique is shown in Figure 4. The carrier wave in the base band passes through the equalizer-numberer 42 with 1 / T cadence and reaches the series / parallel 43 converter with L points feeding the block to calculate equalizer 41.

The calculation block 44 also receives numeric additional data N-L with a value of zero for each block of numerical ‘samples L.

A completed block at one end (χ _θ , χ _χ , ..., ··· » ^θ ) of length N passes through a computation block 44 where it is subjected to a direct discrete Fourier transform that transforms it into a block ( _θ , X) , ..., x _N _ _x) · This block is sequentially multiplied in a multiplier 4.5, controlled by the adaptation circuit 48, the block (C, C, ..., ^C n / "which is the frequency response of the equalizer 41st

This gives a block (Y, Y, ..., Y), where Y = C .X, where i = a cleft from 0 to N-1, which is subjected to a discrete inverse

Fourier transform in the second block for calculating 46 of the equalizer 41, providing a block of transient signals (y,

Since the calculation block 44 operates with a 1 / LT cadence, overlap or overflow occurs at the NL points between two consecutive blocks, after an inverse discrete Fourier transform. The output of the equalizer 41 is the sum calculated in the correction circuit 47, with blocks of length N representing an overlap at the NL points at the ends of the blocks. For a block at the output of the transitional order n, represented by y ' _o (n), y'Jn), ..., y \ (n), ..., y' _N _ ₁ (n), we have the following relations depending on 2 blocks of respective orders nl and nt-1 within:

y'o ( ⁿ ) = yo (n) ⁺ yLM) y'l (n) = yi (n) + y _{L +1} (n-1) y'NLi ( ⁿ ) = - 'yN-Li ( ⁿ ) ⁺ yN-i ( ⁿ - ¹ ) y'NL (n) = yN-L (n) y'L1 ( ⁿ ) = yL-1 (n) y'L < ⁿ ) = yt_ ( ⁿ ) ⁺ yo y ' L + 1 (n) = y | _ + 1 (n) ⁺ yi (n + 1) y'Ni ( ⁿ ) = yN-i ( ⁿ ) ⁺ yN-Li ( ^{n + 1} )

Thus, the alignment can exist without a prefix or suffix and thus without loss of the spectral efficiency of the modulated carrier wave.

Figure 5 shows a third variant of the transfer procedure, employing an overlapping and selection technique. The elements of Figure 5 are homologous to the elements of Figure 4 and the number of units of their reference is the same, preceded by a decade 5. In this variant, the circular convolution of the equalizer is converted to a linear convolution. For this reason, all periods of modulation L undergo a discrete Fourier transform, thus forming a block (X, X, ..., X) in the frequency

O 1 N - · X areas starting with the intermediate block (x, χ _χ , ···,) of the received signals. The block (X _q , X> _· , Χ _Ν _ _χ ) is successively multiplied by the block of complex xh coefficients (C _q , C, ..., C) representing the equalizer conversion function, thus providing the frequency block (Υθ, Y,. .

). The inverse discrete Fourier transform transforms it into a transient block of balanced signals (y, y, ..., Y _N _ _X )>, which is based on the computation block 56, from where the part of the block corresponding to the transient zone is eliminated in the correction circuit 57. overlaps between blocks whose convolution products are misleading.

Figure 6 shows a schematic diagram of the numerical demodulator 20, which is located downstream of the equalizer 11, while receiving sequentially aligned transient signals γ. The demodulator 20 comprises a complex multiplex numeric circuit 21 that receives balanced transient numeric signals y _R (and all this represents a transient signal line (s)) and converts them into corrected numeric signals z _r applied to the decision circuit 22. This decision circuit 22 compares the signal z _r and in particular its phase with several numerical signals representing possible states of modulation and selects a signal w which it considers to be the least different from the corrected signal of _r . The decision circuit 22 calculates the error signal e _R = z-3, which it applies to the numerical comparator of phase 23 when receiving the signal and _R as a reference. The comparator 23 delivers a signal resulting from a phase error ktorý that undergoes a down-pass filter in the numeric filter 24 and may be transverse or backward before it is applied as a signal u to the perfect accumulator 25.

The accumulator 25 acts as an adder and subtractor and provides an angle value that serves as an address for determining the sine, cosine table in the dead memory 26. The quadrature, sin, and cos defining angle correction angle _r are supplied when returning to the phase correction multiplier 21. later equalized signals in the form of:

eJ®k = cos θΐς + j.sin 9k

These same values of sin 60 with cos are also supplied by the upward correction multiplier 15 of the demodulator 20. Thus, the demodulation, performed downward at the equalization, intervenes after applying the corrections due to the early measured errors and thus with reduced distortion and without causing significant delays in due to the Fourier transform in the receiver loop.

The adaptation circuit 18 of the equalizer 11 operates starting from a block of predetermined signals ( ^α , α and _i ), transmitted periodically over the K1 blocks of the signals (K is all positive), K being the higher the channel for transmission 2 more stable. The predetermined block (Αθ, A, ..., corresponds to the discrete Fourier transform of the predetermined block (a _Q , a, ..., ^and N_ ₁ ) ·

The block (χθ, x,. .., _ _Χ Ν _χ) received at the input buffers 11 corresponding to the emission of the block (a _Q, a _{I (.} .., ^a t4_ ₁₎ ^and converted to j ^much fast discrete Fourier transform on the block (X , X, ..., X) After returning to the transit area, we have to find the posted one again

N-1 block ( ^and oi ^and xt '' -t ^and N) at the output of demodulator 20 and thus (a _o -e00 ₍ and _ie -j91, ···>^{e'j QN_1} ) ^{at the} output of the equalizer li, including the downward phase correction e- ^oi (i = 0 to N1).

If we call A '= (A ¹ , A''o' by block transformation Y = (Υ _θ , Υ _χ , ..., Υ _Ν _ _χ equal to block A '.

discrete Fouriero ..... and _N _ ₁ e-j6N-1) ₎ equalizers multipliers 15 must be

If Y _± ^{= c} -X, the coefficients ^c i ^{= A} ' ₁ / X ₁ are derived. · As mentioned above, the coefficients can be limited to a maximum of C = A' / ^s when the signal X ^ has a module lower than the threshold

WITH.

Another embodiment of the equalizer adaptation circuit is shown in Figure 7. The received signals x _± (n) pass through the calculation block 74 of the equalizer 71 to go through a discrete Fourier transform and provide signals X (n) that are aligned in the multiplier circuit 75 earlier, before undergoing an inverse discrete Fourier transform in the calculation block 76 while providing the signals y _± (n). The signals γ _± (n) are converted to the signals z (n) in the demodulator 81, which supplies the numerical decision circuit 82, which then supplies the acknowledged numerical signals α _± (n). The instantaneous error numerical signals β _χ (η) - ζ _χ (η) - and ((n) are converted to the rebalanced output signals of the compensator 71 by a phase rotation defined by θ ίθ ^ Π) and made in circuit 83. whose output e ' _± (n) = y _A (n) - e' J '' (n) .a (n) is applied to subtractor 84., receiving signals y (n) coming from equalizer 71. Numeric signals of difference e '(n) = y (n) _e -jSi (n), and _± (n) pass through a discrete Fourier transform in the calculation block 85, and thus can pass to the frequency domain to determine the error E' (n) for each order of the order i.

Thus, the quadratic error accumulated in the frequency domain is defined:

N-1

J (n) = ΣΣ] E ', (n) l 2 i = 0 corresponding to the quadratic error in the transition region:

N J_

J (n) = ^ L · - | e'j (n) | 2 i = 0

Thus, the optimization of the coefficients C of the equalizer 71 is performed for each line i by determining a transient gradient of the function J (n) evolving according to the reception of the blocks of the transient order n which are consecutive. We determine the new coefficients:

Cj (n + 1) = Cj (n) - J (n) let

Cj (n)

Ci (n + 1) = Cj (n) - | E'i (n) | 2 Cj (n) what can still be written

C (n + 1) - C (n) - © (.X * j (n). E j (n)

X * is an associated complex X. _L (n) resulting from a discrete Fourier transform.

Of course, all or part of the receiver elements may be integrated into a common calculation block, such as a microprocessor. Likewise, the time / frequency and frequency / time transformations may be performed by any suitable algorithm and in particular by modified Fourier transforms.

Claims (12)

PATENT CLAIMS A method of transmitting numerical signals, wherein the signals are transmitted by carrier frequency modulation and upon reception, the received signals are equalized (11) in the frequency domain before being passed to the intermediate region to effect recognition thereof (20), characterized in that the signal phase errors ()) are measured according to the previous signals and the complex rotational phases (είθ ¹ ) used to equalize (18) the signals for the offset coefficients of the later signals are determined. Method according to claim 1, characterized in that after the phases of the transient, already aligned signals (y (n)) have been determined by means of complex coefficients (eJ® *), the residual errors (e * (n) are determined for the signal block ) signals that have just returned to the transition region (y (n)), which are the errors that compose the quadratic error (J (n)) that is passed to the frequency region (85) to be applied (86) a set of corrections for modulation lines after modulation of the modulated carrier wave. Method according to claim 2, characterized in that a set of corrections is applied to the modulation lines by multiplying by a coefficient (C (n + 1)) which is specific to each line and obtained by subtracting a complex error-proportional number (E * (n)). from the coefficient of the earlier (C (n)) in the frequency domain related to a given line and proportional to the conjugate signal (X ' ₁ (n)) to the signal (X _± (n)) representing the line (s) considered immediately after Fourier transformation. Method according to any one of claims 1 to 3, characterized in that, by subtracting a frequency (5) close to that of the carrier wave and independent of its variations, the carrier wave received in the neighborhood of the zero frequency is transformed before equalization. Method according to one of Claims 1 to 4, characterized in that, prior to the emission, the signals to be transmitted by blocks of a specified size are regrouped (31), which are supplemented (32) individually with a circle symbol and receive the frequency domain alignment per block after the circle symbol has been removed (35). Method according to one of Claims 1 to 4, characterized in that upon reception, the signals received by blocks of a predetermined size (L) are rearranged and, after conversion into numerical data, are individually supplemented at one end by a numeric-size sequence (NL). Zero values that temporarily extend into the adjacent block and after alignment and return to the transition area are added (47) sequentially aligned data that are in the overlap zones and offset signals of the adjacent block that temporarily correspond to it. Method according to claims 1 to 4, characterized in that upon reception the signals are regrouped with blocks of a predetermined size (N), which partially overlap (N-L) before the alignment (54-56) takes place. on the blocks and after alignment, (57) those (N-L) are eliminated from the aligned signals that originate from the signals * located in the overlap zone of the two subsequent blocks. <· A receiver of numerical signals transmitted by carrier modulation, comprising a decision means (20), deciding, depending on the signals received, to make a decision to recognize the transmitted signals, comprising: computing means (14, 16) for performing a direct discrete Fourier transform on the received signals providing a spectrum of frequency line signals of the modular carrier wave, means (15) for aligning the line signals relative to the propagation distortion estimated after the reference signal, computing means (14, 16) ) designed to apply an inverse discrete Fourier transform to a spectrum of aligned signals and to obtain a transient result for the decision means (20), characterized in that the decision means (20) are arranged to measure phase errors ()) aligned with reference signals (a); subsequently controlling the phase rotations of the received later signals. 9. The receiver of claim 8, characterized in that the decision means (20) comprise a battery circuit (25) arranged in the addressing according to the phase error (£, u _{r) of} the memory (26) containing a trigonometric table for adjusting the numerical multiplier circuit (21 ), which sells said later signals to the decision circuit (22) and thus handles said errors (,,)). I Receiver according to one of claims 8 and 9, characterized in that the computing means are arranged (35) for recognizing the positions of blocks of a given received signal size before a direct discrete Fourier transform, each comprising a circular numeric symbol and removing said symbol. Receiver according to one of Claims 8 and 9, characterized in that the computing means are arranged (43) to rearrange the received signals with blocks of a given size (L) and to add (43) each of them at the end with a sequence given size (NL) of zero-value numeric data that temporarily extend into the adjacent block before the discrete Fourier transform is applied to the complemented data and to gradually compute the offset data that is in the overlap zones and the offset signals from the adjacent block temporarily correspond. Receiver according to either of Claims 8 and 9, characterized in that the computing means are arranged (53) for rearranging before a discrete Fourier transform of the received signals by blocks of a given size (N), partially overlapping (NL) over time, and for eliminating ( 57), immediately after the inverse discrete Fourier transform, of the obtained signals (NL) originating from the received signals belonging to the two blocks.