SK8798A3 - Digital demodulation method - Google Patents

Abstract

A method of numerical demodulation in which a receiving device (120) performs N different demodulations (N> 2) while delivering to each of the relevant estimates consecutive binary symbols resulting from different coding sequence of bits transmitted by the transmitting device, where the different coding is in the form if = ck Φ af (k) or if and ck means a binary symbol series k and bit series k, f (k) means an integer that is at most equal to k-1 and znamená means a one-sided operation OU, with each estimate of the binary symbol k is a real number with (i) k (1 <i N), whose character is the most likely value for this symbol whose module estimates the probability of this values as the most likely. Value of one the bit-to-sequence sequence is estimated based on the number.

Description

Numerical demodulation method

Technical field

The invention relates to a method of numerical demodulation.

It is mainly used in receiving devices carrying out selective reception methods.

BACKGROUND OF THE INVENTION

Selective reception methods are known in the art of numerical transmission. These methods include: spatial, selective reception useful especially in radio transmission if several reception sensors are located at different locations; frequency diversity reception if the same transmitted information is transmitted simultaneously at different frequencies; time sampling if this information is repeated.

These different methods of selective income can also be combined with each other. The advantage of these methods is that they make it possible to reduce the binary error rate in estimates produced by the receiving device. On the other hand, they generally have the disadvantage that they require additional means at the edges of the passing band and / or that the transmission and reception devices are complicated.

In order to combine the multiple estimates obtained by the selection receiver, there are a number of ways among which a selection method can be mentioned, which consists simply of selecting according to the best signal / noise ratio; a method which is called combining equal gain, where it is decided by the sum of the signals after phasing; a method which is called the highest combination ratio, in which the decision is made according to the sum of the squares of the phased signals and divided by the estimated noise value with which the signals are affected. This latter method provides a maximum signal / noise ratio after recombination.

It is possible that the source signals are not subjected to interference (noise, channel) completely decorated (especially with respect to channels). In this case, classical recombination methods do not yield acceptable results. Finally, the numerical transmissions themselves calculate the probability of observation and the value of the resulting decision in terms of maximum probability; this aspect does not explicitly appear in classical ways.

It is an object of the invention to solve an alternative based on maximum probability from classical methods of recombination of estimates affected by various disorders.

SUMMARY OF THE INVENTION

The invention thus provides a method of numerical demodulation, in which the receiving apparatus performs N different demodulations (N> 2), supplying for each of them corresponding estimates of subsequent binary symbols and _k resulting from different coding of the bit sequence c _{to be} transmitted by the transmitting apparatus. where a _k = c _k Φ and _{f (k)} where a _k and _{k k k} represents a binary symbol of the order k and k of the k series, f (k) means an integer at most k-1 and Φ represents a unilateral OU operation, _{The k-} series k is in the form of a real number whose character represents the most probable value of this symbol and whose measuring module measures this value as the most probable. According to the invention, the receiving device comprises a value of one bit c _k of the sequence k on the basis of the shape number X _k -Y _k because:

X _k = max {| 4 ° + <|} lai <N

Yk = max {| 4 ° - $) |} l ^ <<N

We demonstrate that since the useful signal level is sufficiently large relative to the broadcast noise level, the above estimation X _k -Y _k is proportional to the probability of bit c _k ie the logarithm of the probability densities of the conditionally received signals from bit c _k and conditional to the logical complement. bit c _k Total demodulation, to act according to the rules highs following (a posteriori) probability even in the presence of correlated errors of various estimates and symbols _to.

The invention is used not only for recombination of multiple estimates obtained by the diversity reception technique, but also in the case where at least two of the N symbol estimation sets a and _k are obtained by demodulating this segment of the signal received by the diversity reception method (or such case where estimation errors will often be correlated).

In a particular embodiment of the invention, at least two of the N different demodulations are converted on the same signal segment corresponding to one symbol frame of the numeric signal modulated by the transmitting device, the signal segment being received by the receiving device after transmitting the numeric signal modulated by the transmission channel. demodulation includes the following steps:

- estimating the first demodulation parameters at the first end of the segment; and

- calculating the first estimates of the frame symbols based on the first estimated demodulation parameters and the signal segments passed from the first edge to the second edge, and the second of the two demodulations comprises the following stages:

- estimating the second modulation parameters at the second edge; and

calculating second estimates of the frame symbols based on the second estimated demodulation parameters and signal segments that passed from the second edge to the first edge.

This method leads to evaluable gains on the binary error value, starting with the signal capture itself.

Other features and advantages of the invention will become apparent from the following description and non-limiting examples of the invention, and from the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

Fig. 1 is a block diagram of a transmission apparatus according to the invention; Fig. 2 is a diagram showing the structure of a signal frame in the case of an embodiment of the invention; FIG. 3 and 4 are block diagrams of a demodulation method performed by a receiving device in two directions of demodulation; Fig. 5 is a graph showing examples of probabilities obtained in each direction of demodulation; Fig. 6 is an organization chart showing a method of combining forward and backward estimates according to the invention; and FIG. 7 is another example of an embodiment of a receiving device according to the invention.

DETAILED DESCRIPTION OF THE INVENTION

The invention is further described in its use in numerical radiocommunications between a transmitting device 10 and a receiving device 20. The transmitting device 10 comprises a code source 12 (audio digitizer in the case of a telephone system) that transmits a numeric data stream x _k organized in the following frames. In the embodiment shown in FIG. 2, the signal x _{k is} organized in 126 bit frames at a rate of 1 / T = 8 k bit / s).

The channel coder 14 processes the bits transmitted by the code source to improve transmission fault tolerance. In the example of FIG. The 2-channel coder 14 utilizes a convolutional code CC (2, 1, 3) with an efficiency of 1/2 in the first 26 bits of the frame x _k .

The resulting 52 + 100 = 152 bits e _k is then subjected to a binding designed to cancel a burst of errors that may introduce a Rayleigh escape event. The 8-bit synchronization word is inserted after each 152-bit frame bound to produce the signal c _k that the encoder 14 sends to the modulator 16. This last shape of the radio signal s (t) is amplified, then transmitted to the antenna 18 of the transmitting device 10. In a preferred embodiment, the symbols c _{k are} binary (c _k = 0 or 1).

The modulation used is, for example, GMSK modulation with BT parameters of 0.25 (see K. MUROTA and IEEE Trans. On Communications, Volume COM-29, No. 7, July 1981, pages 1044-1050).

The receiving device 20 comprises a demodulator 24 receiving the signal received by the antenna 22 and amplified. The demodulator 24 issues estimates of the transmitted symbols c _k . These estimates are denoted by S _k for fine decisions and _k for rough decisions. If the symbols c _k are from the region M and are between 0 and M1, the possible representations for the fine estimation S _k are:

Sk = p k-exp (Ijndk / M) it means that in this case his argument IjndklM represents the most probable value d _{k of the} symbol c _k , where its module p _k is the probability measure of this value d _k . In the case of binary symbols (M = 2), the number S _k is a real and named softbit and its designation 2d _k -l gives directly the most probable value of the symbol labeled 2c _k -l.

The receiving device 20 comprises a channel decoder 26 double to the channel decoder 14 of the transmitter. In an exemplary preferred embodiment, the channel decoder 26 performs a frame after permutation bits inverse to the permutation corresponding to the binding used by the transmitter and decodes 52 redundant bits using a Viterbi lattice corresponding to the convolution code used. As it is customary in numerical transmissions Viterbi decoding can be rough decisions when demodulator 24 delivers only _the d, or soft decisions when the demodulator 24 supplying S _to.

The channel decoder 26 returns the estimates y _{k of the} bits x _k and transmits them to the source decoder 28, which again puts them in the form of broadcast information.

As shown in FIG. 1, the demodulator 24 includes a radio stage 30 to convert the received signal into a baseband.

By means of two mixers 32, 34 of the received radio signal, it is mixed into two radio waves in the quadrature with the carrier frequency transmitted by the local oscillator 35, and the resulting signals pass through the lower band filters 38, 40 to obtain one component in phase and one component in the quadrature. The two components are sampled and quantified by analogue-numeric converters 42, 44 with a frequency at least equal to the frequency of the transmitted bits. R _n complex samples of the numeric baseband signal transmitted by the convectors 42, 44 are recorded.

In the embodiment shown in FIG. 1, the demodulator 24 operates according to a sequential algorithm for demodulating binary symbols. In the case of GMSK modulation, sequential demodulation can be performed using the following approximation for a base-modulated signal with (t):

s (t) = Σ j ^k and _k (h-kT) (4) k = ~ <x>

This expression corresponds to the approximation in the first order of degradation proposed by PA Laurent in his paper Precise and approximate construction of digital phase modulation in superposition of amplitude modulated pulses (AMP), IEEE Trans. on Communications, Vol. COM-34, no. 2, Feb. 2, 1986, pp. 150-160. This article also explains how to calculate the function h (t), which, in the case of GMSK modulation with BT = 0.25, corresponds to a stimulus of 2T of the surrounding, aimed at t = 0. In expression (4), binary symbols and _k values ± 1 corresponding to c b _k coded differently: α * = α * _ι. (2c * - 1).

The radio channel is affected by outages that correspond to the sum of opposite phase signals coming from multiple paths, which are caused by different reflections of the transmitted signal at near or distant obstacles. The time scattering of these paths is typically in the range of 12μ * ν, a short time before the one bit time (T = 125 µs in the considered numerical example), representing the variable complex A (t) propagation channel corresponding to Rayleigh leakage and single path phasing. The frequency of failure is 2f _d , where f _d is the Doppler frequency associated with the variation of the distance between the transmitter and the receiver: f _d = f ₀ .v / c, when f ₀ is the center frequency of the channel, is the relative speed of the transmitter and receiver ac is the speed of light . Thus, for a speed of 100 km / h, the Doppler frequency was found to be 41.67 Hz when f ₀ = 450 MHz, from where the failure (83.33 Hz) is a full 12 ms. This therefore allows for a larger frame outage and in particular a higher outage frequency than the synchronization word frequency (50 Hz).

The presence of these rapid outages and, in general, a rapid channel change before the frame time, necessitates frequent channel estimation and thus a serious risk of spreading errors caused by the feedback of the decisions. In fact, if errors occur on the decided binary symbols when demodulating, these errors lead to erroneous channel estimates, which themselves lead to new demodulation errors.

A _k = A (kT) (k = 0 to 167) are complex channel values propagated sampled at 8 kHz in baseband. The channel is inter alia affected by the Gaussian additional white Noise B (t) of the NO / 2 variation labeled B _k after sampling and adjusted filtration. The received signal after the signal has been filtered by the response filter h (t) 46 is thus in the form:

R _a -N- {kT) Σ j ⁿ "H (NK) T) + B (kT) = N == - <x>

= A _k {j ^k -'ak-, H (-Tb} + j ^k ifH (Q) + j ^{k + 1} and _{k + 1} H (+ T)] + B _k where H (t) is a known autocorrelation function h (t), in this expression an approximation is made of neglecting H (t) for | / |> 2T, which simplifies the calculations.

Samples of the output r _{to the} conditioned filter 46 are stored in memory 48 for processing by the radio 50 of the demodulator 24.

Controller 50 processes the filtered signal r _k using segments, each corresponding to one frame with 168 transmitted binary symbols and _k (0 < 168). As shown in FIG. 2, the appropriate framework for different coding includes bits c _k, with 152 bits of information one frame framed 8 bits passing sync word and the next 8 bits sync word.

The controller 50 performs demodulation according to a sequential algorithm, the first phase of which is shown in the block diagram of FIG. 3. In this first phase, the complex channel response estimate begins at the beginning of the segment, then the segment is demodulated from beginning to end by inserting in the frame at each time bit the complex channel response estimate.

At the initialization 60 of this first phase, the bits b ^A and u are respectively read in accordance with the known binary symbols a0 a, and the index k is initialized to 2. In step 62, the index k is compared to 8, i.e. the length of the sync word. If k <8, bit b ^A is taken as well as the known bit k of the sync word in step 64. then proceeds to step 66, instantly estimating the propagation channel and performing a complex division:

y ^A -_í_J_ / 5-) ^k - ^l ) ^k - ² b ^{A -} 2H {-T ^ + j ^k ^ b ^A _} H (Q) + j ^k b ^A H (+ T) ^V '

Filtering the instantaneous estimates V * allows to smooth out the effects of Gaussian noise to obtain an estimation A _k _ _x for bit demodulation. In the example illustrated in FIG. 3, this filtration is simply performed by calculating the arithmetic mean of six instantaneous estimates F 1. Other types of filtration can also be used. After step 66, the index k is compared to 167 (frame length) in step 68. If kel67, the index k is increased by the unit increment in step 70 before returning to step 62.

The channel estimation at the beginning of the frame is complete when k = 8 in test 62. Thus, an estimate A ^A obtained by knowing the synchronized word is prepared. For each k> 8, the softbit is estimated in step 72 according to:

= Re (r .A ^and _to the _ ^. J ^k>) (6) and the estimate of a bit b ^and gave _the name of softbitu%. When the received bit is a ^A, the controller 50 re-estimates the channel at step 66 as previously exposed. Forward demodulation is complete when k = 167 during test 68 duration.

In FIG. 3, it can be seen that the caused error at bit b ^A in step 72, for example, due to a channel failure or inciting noise, causes distortion in the instantaneous estimates f f, F ^ and V _{k + 1} performed in the next three steps 66, leading to errors channel estimates that propagate over a period of time for the smoothing filter. These errors in A _k may cause further bit estimation errors on their path.

Fig. 5 shows a case where the received signal has energy evolving according to the curve E drawn by the dashed line (with a signal droplet occurring at the time k ₀ ), whose probability Js * J estimates (the curve drawn by the dashed line) is good before the failure it takes some time to recover the values in proportion to the energy E of the received signal.

To improve performance in the period following the outage, controller 50 performs further demodulation of the signal segment corresponding to the 168 bit frame from the end of the segment to its start. This makes it possible to obtain a probability | s ^ as expressed by the solid line in FIG. 5. It may be noted that demodulator performance will be improved if softbit is preferred to outage and softbit with% post-outage.

The back demodulation is performed in a second phase, similar to the first, whose block diagram is shown in FIG. 4th

In this second phase, the complex channel response estimate is started at the end of the segment, then the segment is demodulated from its end to its beginning by placing a grid on each time bit of the complex channel response estimate.

At the build-up 160 of this second phase, the respective bits ff ₆₇ and ff _{66 are taken} equal to the known binary symbols α ₇ and a ₆ and the index k is set to 165. In step 162, the index k is compared to 159. b * is taken to be equal to the known bit and _{k -160 of the} sync word in step 164. then, in step 166, an immediate estimate of the propagation channel V% _{+ 1} is continued and a complex distribution is made:

p * _Sfetl_ (ί \ ^{A + 1} / ⁺² 6L ^ ( ⁺⁸ ) + / ⁺¹ ^ _{+ i} w (0) + / ^ // (- O ^{V 7}

Instantaneous filtering lets you smooth out the effect of Gaussian noise to obtain an estimate of A% ₊₁ for bit demodulation. In the example illustrated in FIG. 4, this filtration is done simply by calculating the arithmetic mean of the six most recent instantaneous estimates. After step 166 ie the index k is compared to 0 at step 168. If k> 0, the index to k at 170 is reduced by one unit before returning to .

The channel estimation at the end of the frame is complete when k = 159 in test 162. Thus, an estimate of 61f ₆₁ obtained by knowing the sync word occurs. For each value &<159, the softbit s% is estimated in step 172 according to the relation:

(8) and the bit estimate _k is obtained by marking the softbit, controller 50 again estimates the channel in step 166 as being pre-exposed. Reverse demodulation is complete when k = 0 in test 168.

In the above example, the re-evaluated demodulation parameters as the demodulated segment traverses in each direction to the complex response A _to the propagation channel is constrained. It will be appreciated that other parameters could be included, such as those representing noise observed on the transmission channels. It is also possible to calculate the quadratic mean of the deviations (step 66) or (step 166) for each demodulation direction, to estimate the instantaneous noise value ΝΟΝΟ, ΝΟΝΟ in each demodulation direction. It is also possible to normalize the softness value S ^. or 4 when dividing this quadratic center. Estimates of ΝΟ ^, ΝΟ ^ could be constant over the frame under consideration; these are, for example, Wednesdays and | 4 £ ₊₁ -P £ ₊₁ | calculated across the frame. If these centers are obtained by sliding windows or by filtration, the noise value estimates can be constant, this is dependent on the index k.

Fig. 6 illustrates a method of utilizing forward and backward symbol estimates to find the maximum probability behind, (a posteriori), values of transmitted bits.

Thus, the value of software ^{k k} obtained after different decoding is

S _k = X _k -Y _k or (9)

X _k = max {4 i + 4 Y _k = max {

4-i + 4} 4-i-4} (10) (11) as shown in steps 90 and 92 in FIG. 6. The hard estimate d _{k of} bit c _k is taken to be equal to [1 + sgn (S _k )] / 2 in step 88. These steps 90, 92, 98 are made for each value of k between 8 and 159 (to calculate to be taken = Re (r ₇ .A6 * j ~ ⁷ ')) ·

The simulations made it possible to observe that, in comparison to demodulation in one direction, demodulation forward and back combined with maximum probability results (Fig. 6) leads to an improvement of 1.5 dB to 2 dB over the binary error value with signals generated in an analogous manner to described in connection with FIG. 2 and for conventional Doppler / Bit Rate ratio values.

Note that the estimates S _k calculated from steps 90 and 92 are shown in FIG. 6, correspond to the maximum probability when it can be assumed that noise has the same value in both directions of demodulation, which practically provides a satisfactory approximation. If this approximation is not made, it is appropriate to normalize the softbits with%, s% in proportion to the noise value before calculating the maximum according to (10) and (11), as previously explained.

In the above example, the symbols if depend on the Ck bits by different coding of the form if = Ck @ afik) or f (k) = kl and Φ denotes a one-sided OU operation which, if a _k has values of ± 1 and q of 0 or 1 equal to if = (2c _k - l) .a ^ k) · Normally, it is sufficient if the whole function f confirms ftk) <k-1, where the amount Xk and Yk are:

Z _fe = max {| ^ + 4J, | 4 + ^ (i) |} (12)

A = max {| ^ - ^) |, | 4 - ^) |} (13)

When f (k) = k-1, relations (12) and (13) correspond to relations (10) and (11). An exemplary embodiment of various coding wherein f (k) is not always equal to k-1 can be found in European patent application no. O 774 840.

Fig. 7 shows another example of a receiving device suitable for carrying out the invention. This device 120 sends a call for diversity reception, which in the considered example is spatial selection, a device comprising n antennas 22,

...... 22 _n and connected demodulators 24, ..... 24 _n . Each demodulator 24 acts in a single direction on a respective segment of the signal supplied by its antenna 22, and delivered by the respective softbit 5 (normalized or non-normalized) for each symbol and before different decoding. Thus, the device 120 prepares N = n estimates for symbols arising from different signal segments, instead of N = 2 estimates extracted from the same signal segment in the exemplary embodiment of FIG. 1 to 6.

The module 25 combines these various software for decoding the bits q into the decoder channel 26. These combinations are ^{Xt =} forces ^with ' ^{1 + j} ' ( ⁴ ) (15) made according to formula (9).

The demodulation performed by each demodulator 24 corresponds, for example, to the block diagram of FIG. 3, wherein the received signals differ from one demodulator to the other and are marked 4 ° after the adjusted filtration and the forward estimates can be replaced

... right PO) ^r __ i_ t 'j ^a ~ ² bk-2 H (-T) + j ^a ~ ^bk} ^ H (0) + b j ^of _the H (+ T) bk's estimates of like bits recombination. After the respective softkeys 4 ° in step 72, the demodulators 24 are obtained, they deliver the softbits to the combination module 25, which calculates the estimates Sk and dk, then bit bk by different decoding the hard estimates dk, ie bk = dk (bfik). The bit bk thus obtained is returned to the demodulators 24j _; thus, they can calculate estimates based on channel responses at step 72.

Of course, the invention is very well applicable to other selective reception methods or receivers combining the selective reception method with the multiple demodulation method as described above.

Claims (9)

PATENT CLAIMS A numerical demodulation method, in which the receiving device (20; 120) performs N different demodulations (N> 2), supplying to each of the respective estimates the following binary symbols (a *) resulting from different coding of the bit sequence (ck) transmitted by the transmitting device ( 10), wherein the different coding is in the form ak = Ck ©, where if a Ck represents a binary symbol of the order k and the bit of k, f (k) means an integer at most equal to k-1 and a denotes unilateral OU operation, The binary symbol (a *) of the series k is in the form of a real number -M, the character of which represents the most probable value of that symbol and whose module estimates the probability of this value as most probable, characterized in that the receiving devices (20; 120) estimate the value of bit (ct) of the sequence k based on the number of the form Xk ~ Yk or: Numerical demodulation method according to claim 1, characterized in that the receiving device (20) produces hard estimates of each bit (q) of the k-series based on the number Xk-Yk. Numerical demodulation method according to claim 1 or 2, characterized in that at least two of the N different demodulations are performed on the same signal segment corresponding to one symbol frame (if) of the numerical signal modulated by the transmitting device (10), wherein the signal segment (r (t)) is received by the receiving device (20) after transmitting the modulated numeric signal (s (t)) via the transmission channel, the first of these two demodulations comprising the following steps: estimating the first demodulation parameters at the first edge of the segment; and calculating first estimates of the frame symbols based on the first estimated demodulation parameters and the signal segment that passed from the first edge to the second edge, and that the second of the two demodulations comprises the following stages: estimating the second demodulation parameters (A #) at the other edge of the segment; and calculating second estimates of the frame symbols based on the second estimated demodulation parameters and signal segments passed from the second edge to the first edge. Numerical demodulation method according to claim 3, characterized in that the first demodulation parameters are re-estimated at least once every time the segment passes from the first edge and the second demodulation parameters are re-estimated at least once during the passage of the segment from the second edge. Numeric demodulation method according to claim 3 or 4, characterized in that the first and second demodulation parameters each comprise at least one parameter representing a transmission channel response. Numerical demodulation method according to claim 5, characterized in that the receiving device (20) estimates the parameters representing the transmission channel response at the edges of the segment based on the synchronization sequence included in the frame of the numerical signal. Numerical demodulation method according to any one of claims 3 to 6, characterized in that the first and second demodulation parameters each comprise at least one noise-dependent parameter observed on the transmission channel. Numeric demodulation method according to claim 7, characterized in that the first demodulation parameters include a noise value whose estimate (NOfy is used to normalize the first estimates of the frame symbols, and in that the second demodulation parameters include a noise value whose estimate (jVO *) is used to normalize second estimates of frame symbols. Numerical demodulation method according to any one of claims 1 to 8, characterized in that at least two of the N different demodulations are performed on two respective signal segments received by the receiving device (120) according to the selective reception method.