FR2898227A1 - Data transmission system, has transmission microprocessors transmitting binary element in different stable periods, and reception microprocessors evaluating data transmission quality at each period by operating on received data - Google Patents

Abstract

The system has a transmission part (10) with transmission microprocessors (60) forming correcting words with binary elements from data to be transmitted. A transmission frequency synthesizer (40) and a transmitter circuit (42) transmit the elements. A reception part (20) has reception microprocessors (62) providing the data received from the elements, to a user. The transmission microprocessors transmit one of the elements in different stable periods, and the reception microprocessors evaluate data transmission quality at each period by operating on the received data.

Description

DATA TRANSMISSION SYSTEM USING THE EVASION METHOD OF

  FREOUENCES. The present invention relates to a data transmission system using the frequency evasion method which consists of operating, during time steps Tp, each channel of a set of channels, a system comprising on the one hand a transmission part provided with coding to form, from data to be transmitted in the form of words with "k" bits, words cor-rectors with "n" (n> k) bits, means of transmission to emit N elements binary at each step and furthermore a receiving portion provided with receiving means for reproducing the transmitted bits and transcoding means for providing the user with data transmitted from the transmitted bits. Such systems are known. The article by MBPURSLEY and WESTARK entitled "Performance of Reed-Solomon Coded Frequency-Hop Spread-Spectrum Communications in Partial-Band Interference" published in the journal IEEE Transaction on Communications, Vol. 33, No. 8, August 1985 and WESTARK's article "Coding for Frequency-Hopped Spread-Spectrum Communication with Partial-Band Interference - Part II: Coded Performance" published in IEEE Transaction on Communications, Vol. .COM-33, No. 10, October 1985. There is an attempt to protect this kind of system from unwanted enemy interference. It is important to determine the scrambled channels in order to erase the information they transmit because, in general, an error correction code is able to correct twice as many erasures as errors. In the aforementioned systems, the detection of the interference is effected by a processing on signals - 2 -

  analogs at the receiver means, which requires additional circuits, and so is considered a disadvantage. The invention proposes a data transmission system using the frequency evasion method which does not have the aforementioned drawback and which therefore provides an indication of the quality of the bearings without using analog means. For this purpose, such a system is remarkable in that the transmission part is furthermore provided with diversity means for transmitting in L different levels the same binary element of the correction words and in that the reception part is further provided with means evaluating the transmission quality at each level by operating on the data received. Thus, the invention recommends a digital processing on the received data. This gives the additional advantage that the system can be formed of circuits using microprocessors, which thus facilitates the manufacture. The following description accompanied by the accompanying drawings, all given by way of non-limiting example, will make it clear how the invention can be realized. Figure 1 shows the diagram of a system according to the invention. Figure 2 illustrates the frequency evasion method. Figure 3 shows how the various bits are stored in memory. Figure 4 shows a flowchart explaining the operation of the transmission. Figure 5 shows a flowchart explaining the overall operation of the reception. Figures 6 and 7 show flow charts ex-35 complying with the data processing according to the invention. - 3 -

  Figure 1 schematically shows a transmission system according to the invention. This system comprises a transmission part 10 and a reception part 20 and allows the transmission of useful data between an input terminal 22 of the part 10 and an output terminal 25 of the part 20. This transmission is carried out by radio between a transmitting antenna 30 and a receiving antenna 35 respectively connected to the parts 10 and 20. The carrier frequency of the emission is fixed by a transmission frequency synthesizer 40 cooperating with a transmitter circuit 42. In the transmitted wave, a receiver circuit 50 is provided using the output signal of a reception synthesizer 52 whose frequency of the output signal is in agreement with the frequency of the received signal, taking into account a possible change in the frequency of the received signal. frequency. The different processes are performed both on the data to be transmitted and on the data to be received by means, respectively, of a transmission microprocessor assembly 60 and a reception microprocessor assembly 62. A first buffer register 64 is connected between the input terminal 22 and 20 a BUSE communication line of the set 60, a second register 66 makes it possible to carry out a parallel-serial conversion between the data supplied in parallel via the BUSE line and the circuit transmitter 42 so that transmission occurs binary element per bit. Analogously, two buffer registers 74 and 76 are also provided in the receiving part 20. The register 74 is used to convert the serial data from the receiver circuit 50 into parallel data compatible with a connected BUSR data communication line. at the set 62, the register 70 is connected between this line BUSR and the output terminal 25. An emission time base 80 determines the chronology of the transmission and in particular provides a signal EM and the input INT for interruption of the assembly 60 for controlling the loading of the register 66 by an interrupt program. - 4 -

  Similarly, on the receiving side there is provided a reception time base 82 synchronized from the received data from the receiver circuit 50. This time base provides in particular a signal RC to the interrupt control INT of the set 62 for indicate that the contents of the register 74 can be transferred to the set 62. The transmission by radio is carried out according to the frequency evasion method which is recalled below with the aid of FIG. This method uses a plurality of frequency channels CH1, CH2, CH3, CH4, ... Each of these channels is operated for a fixed duration Tp or plateau; the successive choice of these channels is fixed in a pseudo-random manner as complicated as possible. Thus, in FIG. 2, during the lapse of time Tp which flows between the instants t = 0 and t = Tp, channel CH3 is used and for the following period Tp the CHI channel. These different levels will be used to transmit 64 bits (N = 64). Figure 3 shows what these different bits represent. In this example, we will describe an information processing, according to the invention, 16 bearings PLI, PL2, PL3, PL4, PL14, PL15 and PL16. These 16 steps will be used to transmit 8 error correcting words which are repeated eight times (L = 8) and which comprise 16 bits (k = 16). .These binary elements shown in Figure 3 by a, b, c, d, ..., n, o, p are provided with a subscript greater to indicate the number of the step in which they are transmitted and a subscript lower to indicate the word of which they are part. Thus the word "0" formed of the elements ao, bo, co, do, ..., no, oo, po will be transmitted in the following way: - in the bearing PLI we will transmit the elements a10, oIo, mlo, , clo; in the bearing PL2: b2o, p2o, n2o,

  ., dao; In the PL3 bearing: c3o, a3o, o30, ..., e3o ... etc; The part of each bearing which transmits the word "0" forms a group referred to in FIG. 3 by GRO, a group GR1 relates to the transmission of the word "1" formed of the elements al, b1, cl,. GR7 for binary elements a7r b7, c7, d7, 05 ..., n7r 07, P7. It will be noted that each binary element occupies regularly distributed positions on each level. The operation of the system of the invention is based on the following considerations: First step: Evaluation of the quality of a set of pa- pers. This quality is determined by an estimate p of the average probability p of error for all the bits belonging to all the stages. Since the codewords are repeated once and the offsets are known, each element of a codeword can be compared with (L-1) other corresponding offset elements of L-1 code words. It is possible to define the possibilities of obtaining these results as a function of the probability "p". Thus, the probability Po of obtaining the result Ro 20, that is, L identical identical bits, is: Po = (1-p) L + pL The probability Pk of obtaining the result Rk, ie (Lk) identical elements and therefore k elements of opposite value is: Pk = CL. [Pk (1-P) L-k + pL.k (1-p) k] 25 Remember that: Ck = L! L k! (L-k)! Note that when L is even: PL ~ 2 = CL /? [PL / 2. (1-P) L / 2] It should be noted that the case where we have Lk similar binary elements considered fair and hence k bi- .. DTD: 20 false elements is indistinguishable from the case where we have Lk similar binary elements but considered as false and thus k just bits. The probabilities Pk thus verify: 20 - 6 - L / 2 k = OPk = 1

  We now examine for all the binary elements received the number of times we obtain the results Rk; let b (k) be the number of times that satisfies the relation: 05 L [2 b (k) = nN (1) k = O We can define L / 2 + 1, random variables Xk; these variables can take the values b (k) examined, each value being taken with a probability Pk. We can then write the following probability by noting that we have to deal with a MULTINOMINAL LAW:

  Prob [vke [0, 2], Xk = b (k)] = (n L)! L42 pb (k) (p) the [b (k)! ] k = 0 k To obtain the maximum likelihood estimate of p, (ie p this estimator), we cancel the logarithmic derivative of 15 (2). We then obtain the following formula: L ~ 2 b (k). Pk (P) = 0 k = O Pk (p) To be able to easily determine the value p we establish a table (see Table I below) for various values of p (with 0 <p <0.5) giving the function : POP) then we apply several times the formula (3), with different p and the value retained p is that which gives the lowest value of the sum indicated in the formula (3). (2) (3) TABLE I For L = 8 P 0.050 0.150 0.250 0.350 0.450 in - 8.421 - 9.412 -10.660 -12.061 - 9.140 IQ 12.632 - 1.567 - 5.289 - 7.285 - 4.911 02 33.685 6.305 0.260 - 2.153 -1.463 Q3 54.853 14.591 6.400 2.855 0.816 p4 75.789 21.961 10.667 5.275 1.616 no - 1.000 - 1.000 - 1.000 - 1.013 - 1.325 ni - 1.000 - 1.000 -1. 004 - 1.047 - 1.579 02 - 1.000 - 1.006 - 1.038 - 1.185 - 2.211 03 -1.056 1.214 - 1.500 - 2.167 - 5.500 04 0.056 0.214 0.500 1.167 4.500 05 0.000 0.006 0.038 0.185 1.211 06 0.000 0.000 0.004 0.047 0.579 07 0.000 0.000 0.000 0.013 0.325 Second step: Determination of the quality of each of the "n" levels. This quality is reflected by "pi" which is the average probability of error per binary element for the tier of rank "i" (where 1ti4n). For each binary element of a given level, one observes its correspondence with its replicas situated in the other levels (L '= L-1). NL 'binary elements distributed on (n-1) levels are thus involved. To simplify the calculations, it is approximated that these NL 'bits have the average probability of error p already evaluated at the first step. It is possible, in a similar manner indicated in the first step, to define the probability of obtaining the different comparison results for each binary element of the bearings. The probability Q 0 of observing the chords and 0 disagreement is: Qo = (1-pi) (1-p) L '+ pi p L' The probability Qk of observing L'- k 'agreements and k' disagreements are worth:

  05 Qk. = (1 μl) .CL, .pk (1-p) Lk + Pi.CL .PL -k (1-p) k, 10 and finally the probability QL 'to observe 0 agreement and the disagreements is: QL '= (1-Pi) pLt + pi (1-p) L'

  The different probabilities Qk verify: L. k'EO Qk. = 1 We put: Ak, = CL, .pk '(1-p) The -k' independent quantity of pi, we have: Qk '(Pi) _ (1-pi) .Ak' + pi.AL By analyzing the received data, we count the number of times c (k ') where we obtain the probability event Qn which satisfies: E' (k ') = N k '= 0 One can, like the previous case, define L random variables Mc. (0 'L') and show that:

  Prob [vke (O, L ') .X'k' = c (k ')] (5) N! The c (k ') The k' = O [Q k '(Pi)] [c (k')!] K '= 0 The maximum likelihood estimator of pi is denoted by (4) _ b ( 1) 2nN L (1-2p) -9-pi; it is obtained when we know p, by canceling the logarithmic derivative of (5) with respect to pi: L 'c (k') Q'ki (Pi) = L 'c (k') = 0 k '= 0 Qk '(Pi) k' = 0 pi + At AL'_k'-Ak 'We calculate in advance (see Table I above): qk P) At AL._ki-Ak. with 04p <0.5. Then, for various values of pi (0 <pi <0.5) the sum indicated in formula (6) is calculated. The sum closest to 0 corresponds to the value Pi sought. In the following, the precision of the estimators is analyzed. The estimator of the average quality of the bearings has L / 2 - 1 degrees of freedom. Indeed it is necessary to deduct from the number L / 2 + 1 of possible events a unit since there exists the relation (1) between the b (k) and another unit since p is estimated from the observations. In particular for L = 2, the number of degrees of freedom is zero and there is a relation between p and b (1): 1-I 1_ 2b (1) n N / LP __ 2 The standard deviation on the line XI embodiment b (1) is: 20 b (1)! nL Pt (p) [1-P1 (P)] = J nL 2p (1-p) (1-2p + p2) and by differentiation of (7) one has the sought standard deviation: hence ap = fP (/ 1-p) (1-2p + p2) I2L (1-2p) Remark: The same calculation carried out on the random variable, XI, would have led to the same formula, since Po (p) + Pl (p ) = 1. (6) 05 (7) (8) 20 - 10 -

  We see that, when p varies from 0 to 0.5, ap varies from 0 to For L> 2, it suffices to remark that the standard deviation on each va XI varies as 1 / IL, ie as IL once these go normalized (the number of events giving rise to realizations b (k) varies as 1 / L, so the estimate of b (k) is IL times less accurate). In return, the number of independent variables used for calculating p varies as L / 2 and the accuracy of the maximum likelihood estimator as 1 / IL / 2. As a result, ap is independent of L. When choosing the values of p, the extreme values p = 0 where pk becomes infinite and the value p = 0.5 where the precision is zero is to be avoided. The quality estimator for each of the steps has L-2 degrees of freedom. For L = 2, the number of degrees of freedom is zero. We have, p being then fixed (p = p) ac (1) = Qt (Pi) [1-QI (Pi)] = (1-2p)! N (-PP + Pi + (1_215) 2 and by differentiation of the expression: C (1) N p P = 1 - 2p 1: a: c (1) where: P N (1-2p) api = fp + p + P (1) -p) 1 1 (1-2p) IF (9) We see that api is maximal for pi = 0.5 and is greater when p is closer to 0.5. that the standard deviation on each one goes X'k 'is independent of L, while the number of independent variables used for the calculation of pi varies like IL-1.This results in a reduction of pi with respect to the formula (9) equal to 1 / L-1 An insufficient precision of the process must be noted beyond a certain value of pp, but, in a favorable circumstance, this limitation is more severe for L low, a choice corresponding precisely to the values Their weak points can now be explained by the flow charts of the operation of the system according to the invention. RAMs implanted in read-only memory in the microprocessor assemblies 60 and 62. The assembly 60 is concerned with the flowchart of FIG. 4. In a FIG. 4, the reading operation of FIGS. bits contained in the register 64 and in box K2 the coding of these bits in an error correction code word for example a biorthogonal code. Box K3 relates to the storage operation in memory locations of each binary element of this error correction code, so that the binary sequences to be transmitted via register 66 are easily loaded into register 66. The program is then looped so as to re-read new bits in the register 64. The loading of the register 66 is performed 20 according to an interrupt program triggered by the signal EM appearing before the start of each step. In box K10 of FIG. 4b, the address of the word to be loaded is determined (this address changes at each occurrence of the EM signal) and the load-ment is indicated in box K11. The receiving part 20 operates according to the program contained in the set 62. The program comprises two parts: a part whose flowchart is shown in FIG. 5 and which concerns the data entry, and a part whose The flowchart is shown in FIG. 6 and relates to the processing performed on these data. The data is entered by interrupt program triggered by the RC signal; the operation shown in box K20 (Fig. 5) is to take data from register 74 and the operation indicated in box K21 to arrange the matrix data as shown in Fig. 3. 12 -

  The second part of the operating program of the reception part is intended to carry out statistical and estimation operations the basis of which is given in the appendix to this memorandum. In the flowchart shown in FIG. 6a, the position of the bits is indicated by three indices: the index ig varying from 1 to nGR where nGR is the number of group nGR = N / L and locating the number of the correction word error. The index in varying from 1 to n marks on a horizontal line of the table of FIG. 3 the position of the bits. the index iL varying from 1 to L indicates the rank of the line in one of groups GRO, ..., GR7. A variable d (iL, in) giving the indication of the shift of the binary elements from one line to the other is also used; these values of d (iL, in) are pre-established and are stored in read-only memory, the values of iL, in then defining the address of the value d. The flowchart of FIG. 6a explaining the evaluation of the average quality of the bearings is first examined. For this evaluation, we determine the distribution b (k) which, as a function of the probability Pk, represents the number of binary elements equal or not to each other in the L repetitions of appendix error code words. In box K50 we initialize the array to con-hold the different values of b (..), the initialization of the index ig is indicated in box K52. Box K55 indicates a test on the value of this index ig; if this index has reached the maximum value (N / L), the treatment indicated by the flowchart of FIG. 6b is carried out. If this index ig is less than the maximum value, this index is incremented by one unit (box K58). Then, we initialize the index in (box K60). In box K62, this value of in is tested to see if it is less than n. - 13 -

  If the test is negative, return to box K55, if it is positive we go to box K64 where the value of in is incremented by one. In box K66 we initialize a variable n1 and in box K68 we initialize the index iL. In box K70 we perform a test on this index iL to know if it is less than L. If the test is positive we increment this index (box K72) and we go to box K75 where is indicated a sub program which determines the address of the binary element located in an array organized according to FIG. 3; the location of the line y, where this bit is located, is determined by: y = L (ig-1) + iL while the location of the column x is given by: x = d (iL, in) + in (mod n) + 1 In box K80, the bit whose location is fixed by x and y is added to the variable ni. We go back to box K70 as long as the test of box 70 remains negative; that is, the repetition of a bit of the code has been examined L times and ni represents the number of times that the value of this binary element is equal to "1". Therefore, in box K85, the number of times no where we find the value "0" is determined. We then determine in box K86 which of the values, ni or no is the smallest; if it is no, this value is incremented by one unit and is assigned to a variable ic (box K90); if it is not, this value is incremented by one and is assigned to the variable ic (box K92). This value ic makes it possible to define the value of b (ic) and is increased by one unit (box K95); then, we go to the next bit by going back to box K62. b (..) 30 being determined when the test indicated in box K55 becomes negative, it is possible to proceed with the treatment whose flowchart is shown in FIG. 6b. This process starts with the calculation of a normalization factor "h" (box K100) and then we give mn the value "MAX" which represents the maximum value of the variable considered (K102). The values of ek are - 14 -

  stored in a ROM of the set 62 for the value of L chosen, for the various values of k possible and for a certain number np of values of the error rate p, each of these values being indicated by the index ip of 1 to np. We initialize ip to zero (box K104). We then test if the value ip is less than np. If the test is positive, increment the value ip (box K107) by one. Then, we set the variables S and ic to zero (box K109). We then go to the test indicated in box K110 where we test if the value ic 10 is equal to the largest integer less than or equal to L / 2 + 1; if this test is positive, go to box K112 where the variable ic is incremented by one unit. Then, in box K115 we increase the value S of the quantity p (ic, ip) • b (ic) whose terms are read in the ROM and the RAM. When the test indicated in box K110 is negative, one tests (box K120) the value S, thus cumulated for the different ic; if this value is greater than the mn value, go directly to box K105; otherwise, the JSI value that has just been determined at the output of the test 20 (K110) is assigned to the value mn (box K122) and in a memory ip intended to finally contain the index corresponding to the quality of the estimated bearings. Box K130 recalls that when the K105 test is negative, indicating that the processing is complete, the ip value is available for the next processing in the flowchart shown in Figures 7a and 7b. This processing starts with an initialization phase (boxes K150 and K152), that is to say that the index in which the error level is estimated is used to identify the level at which the error rate is estimated ( box K150) and to zero a table intended to contain the realizations C (...) (box K152). Box K155 indicates a test of the value in; if in is not less than n, it means that the processing is finished; if in is less than n, it can be incremented by one unit (box K158), the index ig (box K160) is reset. We test this index - 15 -

  with respect to L, as shown in box K162; if the value ig is less than L, we increment it by one unit (box K164) and we initialize a variable iL (box K166). We test this variable to find out if it is less than L (box 05 K168) before incrementing it by one unit (box K170). The index i'L is used to identify the L '= L-1 lines where is repeated the bit of the bearing considered and the index i'n is used to identify the corresponding columns. We determine in box K172 the value of i'n. In box K174, we initialize the variables i'L and 10 n'i (number of "one" in the L-1 repetitions). The variable i'L is tested against iL (box K176) before being incremented by one unit (box K178); after this incrementation, one tests again (box K180) i'L to know, this time, if there is equality with iL. If this is the case, return to box K176 where the test will necessarily be negative. When the test indicated in box K180 is negative, it is determined (box K182), in a manner analogous to that indicated in box K75, FIG. 6a, the coordinates of the binary element to be considered. But here, instead of using the indices iL and in, we use the variables i'L and i'n defined above. The binary element thus defined is added to the variable n >> (box K184) and we return to the test of box K176. When the test indicated in this box K176 becomes negative, we go to box K190 where we determine the variable 25 giving the number of the line (see Figure 3) of the binary element, also defined by the column in, that we will use in the test of box K192. The test is to know if this element is equal to "0". If this is the case, we determine a variable n i with n + n = L-1) (box K194) and a variable 30 i'c (box K196). In box K198 we increment by one unit the variable C (i'c) and we return to box K162. If the test of box K192 indicates a bit other than "0", the value i'c is the value of n'1 increased by one and we go to box K198. The flow diagram of the continuation of the process is shown in FIG. 7b. This continuation of the treatment is triggered when the test indicated in box K162 is negative. The values of qk 'are stored in a ROM for a certain number of values of the error rate pi (p) each of these values being indicated by the index ip varying from 1 to

  05 np. We initialize ip to zero (box K210); we test (box K212) this variable to know if it is lower than the value np before incrementing it by one unit (box K214). Different variables S 'and i'c are initialized to zero (box K216). The variable i'c is tested (box K218) for

  10 If it is the case, it is incremented by one unit (box K220). In box K222 we increase the C (i'c) value S 'of the quantity whose terms P (ip) + rl (i'c, ip) are read in the ROM and the RAM and we return to the test from box K218. When this test becomes negative, it means that

  The sum whose value is indicated by S 'is calculated. The value of this sum is examined using the test of box K224. If the absolute value of S 'is greater than the value mn, already defined, we go to the test indicated in box K212; if S 'is less than mn, "mn" takes the value IS'J and the value

  Ip that gave this minimum value is stored in an ip (in) array at the given in by address and returns to the K212 test. When this test becomes negative, we have filled a location in the ip array (which is indicated by box K230) and we return to box K152 to start a

  25 new treatment for a different in, if the indicated test

  box K155 does not indicate the end of processing (box K300). 30

Claims (2)

  1. A data transmission system using the frequency evasion method which consists in operating, during time steps Tp, each channel of a set of 05 channels, a system comprising on the one hand a transmission part provided with coding means for forming, from data to be transmitted in the form of words with "k" bits, corrector words with "n" (n> k) bits, transmission means for transmitting N bits at each 10 bearing and on the other hand a receiving part provided with receiving means for reproducing the transmitted bits and transcoding means for providing the user with the data transmitted from the transmitted bits, characterized in that the transmitting part is provided, in addition, diversity means for transmitting in L different levels the same binary element of the correction words and in that the reception part is additionally provided with means for evaluating the quality of transmission at each level by operating on the received data. 20   2. System according to claim 1, characterized in that the evaluation means operate in the following manner: determination of the average quality p of "n" steps: by measuring on the same L bits the quantity b (k ) representing the number of times the same element is received identically k times. by looking for the value p of p such that the following sum becomes minimal: L / 2 E = O b (k) pk (p) where pk (p) is determined in advance by the relation: pk (P) = P p (p) k pk = CL [pk (1-p) Lk + pL-k (1-P) k] - Determination of the quality of each step Pi:. by measuring for each binary element of a step Ni "the quantity c (k ') equal to the number of times that a binary element is received identical to itself on the other pa- ramels by looking for the value pi of pi such that the following sum becomes minimal: L-1 c (k ') k' = 0 Pi + nk '(P) where nk, (p) is determined in advance by the relation: Ak. nk, ( P) = AL, -k -Ak, where At = CL-1'Pt (1-P) L-1-t