JPH0281543A - Demodulation system for tdma burst signal - Google Patents

Abstract

PURPOSE:To improve the efficiency of a burst signal and to demodulate the signal without a preamble of a data series for carrier recovery and clock recovery by storing the received burst signal at once and demodulating it. CONSTITUTION:Function approximation is attained by using, e.g., the least square error method from a frequency series estimated for each sample interval through the use of the sliding DFT method. Then the frequency of an original signal stored in memories 9, 10 is corrected by using the approximated function to correct the initial phase of the corrected signal and a clock recovered from the corrected data series is used, discriminated for each bit to obtain a demodulation data. The burst signal is stored entirely at once in terms of digital value and demodulation is applied by the digital arithmetic processing using the sliding DFT method. Thus, even a received burst mode PSK signal wave with a bad line state is demodulated with high accuracy.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a method for demodulating a TDMA burst signal by establishing synchronization and demodulating the TDMA burst signal by synchronous detection.

(Prior Art) In the conventional TDMA communication system, the received signal consists of a plurality of mutually asynchronous burst signals, so in order to demodulate this signal, it is necessary to perform the demodulation operation while creating a reference signal for each burst. Therefore, the carrier regeneration circuit had to be able to establish synchronization within an extremely short period of time. In addition, the transmitting side cannot transmit information data until synchronization between the received signal and the reference carrier reproduced by the demodulator is established.
In a normal burst mode signal, a preamble, which is a synchronization data sequence, is added to the information data 111.

As shown in FIG. 7, the preamble consists of a data sequence for carrier wave recovery, a data sequence for clock recovery, and a unique word sequence for frame synchronization or phase uncertainty.

Here, the length of the preamble is determined by the line conditions normally used. That is, when the reception C/N is small and the line condition is poor, it takes a long time to establish phase synchronization between the carrier of the received signal and the carrier of the reference signal, so a long preamble length is required. . On the other hand, in a satellite link, the carrier frequency of the received signal is usually different from the transmitted carrier frequency due to the stability of the local frequency of the earth station transceiver or frequency conversion within the satellite. Especially in low-speed digital communications,
The signal bandwidth is narrow, and the frequency deviation due to the above causes may be 4 to 5 times the modulation signal bandwidth. In order to demodulate such a burst signal with a frequency deviation of 4 to 5 times the signal band, a very long phase synchronization time is required, and therefore the preamble length must also be very long.

On the other hand, in maritime and aeronautical satellite communications, in addition to the above-mentioned fixed frequency deviations on the line, short-term Doppler due to mobile station fluctuations is a major problem. In short-time Doppler, the carrier frequency changes over time, and the speed is 1
It may reach about 1/10 of the modulation signal bandwidth per second.

Conventionally, a PLL method or a frequency-multiplying method has been used as a carrier wave regeneration method, but in order to compensate for frequency fluctuations due to short-term Doppler, it is necessary to increase the loop bandwidth of the PLL or the narrow-band filter of the frequency-multiplying method. Bandwidth as P
It is necessary to take 1/10 or more of the SK modulation bandwidth. However, in this case, when the received C/N is high, carrier wave regeneration is possible with good efficiency, but when the received C/N is low,
The reproduced carrier wave contained a large noise component, resulting in a large deterioration of error rate characteristics.

(Problems to be Solved by the Invention) As mentioned above, under a line with a low C/N and a very large steady frequency deviation, when transmitting a burst mode signal, it takes a very long time. A preamble length must be prepared, and the length may be equal to or longer than the length of the information data following the preamble. In this way, essentially, taking a long preamble that is not information data greatly reduces the communication efficiency of the entire TDMA system, which is particularly a problem in low-speed TDMA communication.

On the other hand, when short-time Doppler occurs on a line, even if a long preamble is added, carrier wave recovery itself becomes difficult, and information data cannot even be demodulated.

The present invention has been made in order to solve the problems of the prior art described above, and has no preamble even under a line with a low reception C/N, a large steady frequency deviation, and a short-time Doppler. TDMA enables demodulation with
The purpose of this invention is to provide a method for demodulating burst signals.

(Means for Solving the Problems) A feature of the present invention is that in a TDMA burst signal demodulation method in which a TDMA burst signal, which is a received signal, is demodulated by synchronous detection, a first Quasi-synchronous detection is performed using a reference signal and a second reference signal that is orthogonal to the first reference signal, and two output signals that are orthogonal to each other obtained by the quasi-synchronous detection are sampled at a predetermined period. A
A/D conversion is performed, and the digital received signal, which has become a digital value through the A/D conversion, is stored in a memory in burst units,
After removing the modulation component of the digital reception signal read from the memory to make it an unmodulated signal, convert the unmodulated signal on the time axis to an unmodulated signal on the frequency axis, and convert the unmodulated signal on the frequency axis to an unmodulated signal. The temporally varying frequency of the modulated signal is estimated on the frequency axis, an approximation of a continuously changing frequency function is obtained from the estimated frequency sequence, and the obtained approximation function is used to calculate the frequency of the received signal. The goal is to compensate for frequency fluctuations and demodulate.

(Principle of the Invention) The present invention performs quasi-synchronous detection of a TDMA burst mode signal on the receiving side using a reference carrier wave having a value close to the carrier frequency of the received signal, and the received carrier frequency and reference carrier frequency obtained by the detection are A burst mode modulated wave having a frequency different from the above frequency is converted from analog to digital, and all sample values of the digital values obtained by the conversion are stored in a memory for each burst. The digital signal stored in the memory is, for example, an unmodulated signal with modulation components removed by a multiplication method or the like. The signal made into an unmodulated signal is
For example, a signal on the time axis is converted to a signal on the frequency axis by a slide, an ing DFT method, or a frequency estimation method using an adaptive digital filter, and then the electric scum vector is obtained by squaring the absolute value. The electric charge vector obtained here has a very large power value at a frequency that has a frequency component that is the difference between the carrier frequency of the received signal and the reference carrier frequency. Therefore, by finding the maximum value of the electric charge vector value, it is possible to detect the frequency difference between the two carrier wave frequencies mentioned above and the reference carrier wave frequency signal.

Now, assuming that the frequency of the reference signal wave is constant, after all,
This means that the carrier frequency of the received signal has been successively estimated.

Here, since the carrier frequency is calculated on the frequency axis, even if the received signal has a large steady frequency deviation, it can be estimated as a product. Further, even if Doppler exists for a short period of time, it is possible to estimate the frequency value for each sample, that is, the frequency value that fluctuates over time, so it is also possible to estimate the frequency that follows the Doppler frequency.

In addition, although the sliding DF method will be explained below as a method for converting a signal on the time axis into a signal on the frequency axis, the sliding DFT method is described in the literature (L., R. Rabiner et al., “Theory
and Application of De
Digital Signal Processing”
Englewood C11ffs, NJ: Pr.
entice-Hall 1975. ), in which DFT operations are performed sequentially on digital values on the time axis for each manually inputted sample value. this is,
Like conventional DFT operations, block by block,
It does not convert a signal on the time axis to a signal on the frequency axis.

FIG. 6 is a signal state diagram on the frequency axis according to the present invention,
By using the sliding DFT method, the accumulated burst signal can be converted sample by sample into N (N=L) frequency axis signals using a certain number of samples on the time axis. By using the obtained frequency axis signal, it is possible to obtain the frequency value of the difference between the carrier frequency of the received signal and the reference carrier frequency for each sample.

Here, assuming that the detected frequency difference is based on the constant frequency of the reference signal wave, this means that the carrier frequency of the received signal can be successively estimated. Therefore, since the carrier frequency is calculated on the frequency axis, it can be estimated with high accuracy even if the received signal has a constant large frequency deviation. Further, even if Doppler exists for a short period of time, it is possible to estimate the frequency value for each sample, that is, the frequency value that fluctuates over time, so it is also possible to estimate the frequency that follows the Doppler frequency.

However, since the frequency estimated here includes noise on the transmission path, correct values are not estimated at all sample points. Furthermore, if the estimated frequency at a certain sample value includes a large error due to the influence of noise, this corresponds to a case where a cycle skip occurs in the reproduced carrier, and the error rate characteristics will deteriorate significantly from then on. Therefore, the regenerated carrier wave is required to change continuously. Therefore, in the present invention, a sliding DFT method is used to perform function approximation from a frequency sequence estimated at each sample interval using, for example, the least squares error method.

The approximation function for the frequency fluctuation obtained here is a temporally continuous function. Next, the frequency value of the original signal stored in the memory is corrected using an approximation function, the initial phase value of the corrected signal is corrected, and the clock regenerated from the corrected data sequence is used to generate one bit. Determine each
This is to obtain demodulated data.

In this way, the present invention accumulates all the burst signals as digital values once and demodulates them through digital arithmetic processing using the sliding DFT method. It is also possible to accurately demodulate a received burst mode PSK signal wave in which the signal wave is very large and Doppler exists for a short period of time.

In addition, since the demodulation operation is performed on the accumulated digital values, unlike the conventional method that performs phase synchronization in real time, the preamble only needs to be a unique word sequence for frame synchronization. It is possible to increase the proportion of information data in the TDMA burst signal, thereby increasing communication efficiency.

(Example) Hereinafter, an example of the present invention will be described in detail using the drawings.

FIG. 1 is a diagram showing an embodiment of the present invention, and the TDMA burst mode modulated wave as an input signal is an M-phase PSK wave having a signal sequence as shown in FIG. 2.

In FIG. 1, +00 is an input terminal for the M-phase PSK modulated wave, 1.2 is a multiplier, 3 is a π/2 phase shifter, 4 is a fixed frequency oscillator (hereinafter referred to as "fixed oscillator"), 5, 6 is a low-pass filter, 7 is an analog/digital converter, 8 is a memory switch for memory A side 9 and memo 98 side 1o that stores one burst of manual data, 11 is a digital processing type demodulator, 200 is demodulation Indicates a data output terminal.

Here, the input terminal 100 has an M-phase PS expressed by the following equation.
The K modulated wave is generated manually.

5%4(t) = JTA cos (Wct+ψ(1
)+θ to 10o) +n(t) −(1)(
k-1) T≦L≦kT In equation (1), A indicates the amplitude level of the received signal, Wc is the carrier angular frequency, ψ(1) is the short-time Doppler frequency, θ
5 is phase information corresponding to the second information bit, and in two-phase PSK (M=2), θ5 is, for example, 0. Take π and get 4
In phase PSK (M=4), θ5 is 0. π/2. π, 3π
Take /2. Also, θ.

indicates the initial phase of the input burst signal, and T indicates the time interval per symbol. n (t) represents noise added on the transmission path.

The carrier wave frequency Wc of the received signal expressed by equation (1) is determined by the stability of the local frequency of the transmitter/receiver, the stability of the local frequency of the frequency converter included in the line used, etc.
This value is different from the carrier angular frequency Wc° transmitted by the transmitter. In other words, within the line, (wc-wc'
) will have a frequency deviation of

The received signal shown by equation (1) includes a reference carrier wave 401 having a frequency Wc'' of the fixed oscillator 4, and a reference carrier wave 401 having a frequency Wc'' of the fixed oscillator 4,
Quasi-synchronous detection is performed with reference carrier wave 301 having a phase difference of /2. However, since the frequency Wc'' of the fixed oscillator 4 has a value very close to the transmission frequency Wc', in the following description, it will be assumed that Wc''=Wc.

Rc(t), R,(t) of reference carrier waves 401 and 301
are respectively expressed as the following equations.

Rc(t)=cos (Wc' is 10I) −(
2) R, (t) = sin (Wc' is 10I)
-(3) The input signal 100 is multiplied by the reference carrier wave 401° 301 of equation (2) (3) in multipliers 1 and 2, and further passes through the low-pass filters 5 and 6 to remove harmonics. is removed, and finally the output signal of the low-pass filter 501゜6
01, signals shown in the following equations appear, respectively.

e +k(t) = S k(t)・Rc(t)=Ac
os((WC-Wc')L+ψ(L) 10, 10.-〇+ ) + nc(t) −(4)e 2b(
j) = S k (j) ” R-1:) = -Asi mouth ((wc-wc')t + ψ (L) + θ
k ten 〇. −〇, ) + n, (
t) (5) However, formulas (4), (5)
where nC(t), n, and (t) represent the noise after passing through the low-pass filters 5 and 6 having center frequencies (w, -wc'), which are orthogonal to each other.

Here, (wc-wc') indicates the steady frequency deviation amount between the received signal wave and the reference signal wave, and is hereinafter referred to as ΔW.

The signals represented by equations (4) and (5) are converted by the analog/digital converter 7 into digital values as shown by the following equations.

X, = e+k(t;) =Acos(△Wt, +ψ(ti)+01100-〇+
) +nc(t,) −(6)yt
”” e2b(ti) −Asin(△Wt, +ψ(ti)+θ is 10. −
〇 + ) + n5(ti)
(7) However, it is assumed that tl indicates the i-th sample time, and i ranges from 1 to N, which corresponds to one burst length.

X, , Y are manually input to the memory switching circuit 8.

The memory switching circuit 8 has a memory A side 9 and a memo 98 side 1.
Two memories 0 are connected, first memory A side 9
is accumulated in When N pieces of digital data for one burst are stored in the memory A side 9, the N pieces of digital data are transferred to the digital processing type demodulation circuit 11. The accumulated data on the memory A side 9 is transferred to the digital processing type demodulation circuit 1.
1, the next burst signal input manually is connected to the memo 98 surface 10 by the switching circuit 8 and stored therein. That is, the data accumulated in the memory A side 9 is processed during the reception time of one burst, and is outputted from the output terminal 200 as demodulated data. By alternately switching between the memory A side 9 and the memo 98 side 10 in this way, the burst signal can be demodulated without interruption.

FIG. 3 is a detailed configuration diagram of the digital processing type demodulation circuit 11 according to the present invention.

N sets of xl are output from the memory switching circuit 8 to the digital processing type demodulation circuit 11 shown in FIG.

Y, are input from input terminals 801 and 802, respectively. These N sets of X, , Y are operated by the multiplier 12 as shown in the following equation, and the modulation component θk is removed, resulting in a non-modulated signal.

Z, = (X+ +j Y+)' - (a
) In equation (8), Z represents the digital signal output from the multiplier 12, and M represents the number of phases of the received PSK signal.

21 in equation (8) ends up being an unmodulated signal having a carrier frequency shown in the following equation.

W(t,)=M(△W+□ ψ(ti))dti
- (9) In Equation (9), if there is no Doppler for a short time, it will be an unmodulated signal with a constant frequency M△W, and even if △W is large, for example, the present inventors have already applied for a patent. PSK signal demodulation method using FF7 method (Patent application No. 62-2653
74), it is possible to estimate with high accuracy, and demodulation using the estimated frequency value is also possible. However, in a line where short-time Doppler ψ(1) exists, as is clear from equation (9), the frequency of the unmodulated signal becomes a function of time.
, ) could not be determined with high accuracy.

In the present invention, the unmodulated signal of equation (9) is converted into a sliding D
The signal is input to an FT circuit 13 (hereinafter referred to as "rS-DFT circuit"), and a signal on the frequency axis is obtained for each sample point. Here, the 5-DFT circuit 13 obtains a signal on the frequency axis -F corresponding to a signal on the time axis that is inputted sequentially for each sample. The output of the 5-DFT circuit 13 is:
L frequency components corresponding to L signals on the time axis from the first sample to the second sample are output. That is, frequency components that change over time can be determined at each point in time. Thereafter, it is possible to obtain frequency components corresponding to L signals on the time axis that have been manually input up to that point in time for each time series that is manually input one sample at a time.

Here, among the L signals on the frequency axis, the frequency value with the maximum electric charge vector value is the frequency value of the L non-modulated signals on the time axis up to this point. Can be done. However, the frequency value obtained here is the average frequency value of the L digital values on the time axis. Therefore, if L is made large, it becomes impossible to follow sudden frequency fluctuations. On the other hand, if L is made small, the frequency increment width of the DFT becomes large, which makes it susceptible to noise, making it impossible to estimate frequencies with good granularity. Therefore,
The value of L will be determined depending on the line status.

FIG. 4 is an estimated frequency diagram for each sample of the output of the 5-DFT circuit 13 according to the present invention. Specifically, the steady frequency deviation Wc and the short reception burst shown in equation (9) are
This is a frequency value for each sample value estimated when the DFT circuit 13 is input manually. As shown in FIG. 4, the frequency value estimated by the 5-DFT circuit 13 is 12 j? This makes it a discontinuous function.

Therefore, when the frequency of the original signal is corrected using the estimated value, it may differ greatly from the actual frequency value, cycle skips are likely to occur, and the demodulated data is highly error-prone.

In the present invention, the frequency function estimator 14 shown in FIG. 3 approximates a continuously changing frequency function using the frequency value for each sample value obtained by the 5-DFT circuit 13.

The frequency function estimator 14 performs the following operations.

Time 1. N frequency values Fi (i=1 to N
) is approximated by the polynomial shown in the following equation.

P (tt) = ao + a + ti + a2tl
'+ −+ a−tt−where, m unknown coefficients a in equation (10). al-a can be easily determined by using the least squares error method, for example. The least squares error method is a method of determining aoal""a- so that the squared error E shown in the following equation is minimized.

Equation (11) can be expressed as the following equation, however,
All Σs indicate Σ.

Equation (12) can be expanded and transformed as shown in the following equation.

Margin below Ru.

-(1:11) shows.

Here, since all the terms in the inverse matrix are constant terms, it can be calculated in advance, and ao a.

... ao can be determined by the following equation.

However, Bjk is a value obtained when the inverse matrix of equation (13) is restored to a normal matrix, and is a constant.

Ultimately, by using equation (14), the 5-DFT circuit 1
Continuous function p (
t, ) can be obtained.

FIG. 5 is a block diagram of the frequency function estimator 14 used in the present invention. In the output multiplier 21 from the 5-DFT circuit 13, 1. are multiplied and further processed by the adder 25, all the terms on the right side of equation (12) are calculated, and input manually to the calculator 2.

The arithmetic unit 28 performs a calculation corresponding to equation (14), and the output terminals 281, 282.28:3 output a. a, "a" are output respectively. In the frequency function synthesizer 29, ao
The function shown in equation (10) is synthesized using a, . . . al. The frequency function obtained here is a function that approximates the actual frequency fluctuation shown in FIG.

Output terminal 141 of frequency function estimator output 14 in FIG.
, 142, the signals shown in the following equations are output, respectively.

The frequency compensator 15 uses equations (15) and (16) to compensate the frequency of the original signal from the memory switching circuit 8 according to the following equation.

S+'== (X* +jYi) (x+°+ J Y
+ ') Output terminal 151.152 of frequency compensator 15
The signals shown in the following equations are output respectively.

X,” =X,X,'-Y,Y,'-(18
)Y,” =X,’Y, +x,y,° −(19
) Here, the signals of equations (18) and (19) pass through the waveform shaping filter 17, and code amount interference is minimized. The signal after passing through the waveform shaping filter 17 is phase-adjusted by passing through the initial phase compensator 18 . Next, in order to determine information data from the output signal of the initial phase compensator 18,
Data clock information is required.

Here, as methods for reproducing clock information, an envelope detection method, a zero crossing method, etc. are known, and reproduction can be performed using these conventional techniques. The initial phase compensator 1 based on the clock information regenerated by the clock regenerator 19
The output signal of 8 is determined by a determiner 20 and demodulated data is outputted to an output terminal 200.

(Effects of the Invention) As described in detail above, according to the TDMA burst signal demodulation method according to the present invention, on a line where the steady carrier frequency deviation of the received signal is large and Doppler exists for a short time, Demodulation can be performed with high precision even under extremely poor quality lines where the C/N of the demodulator's manual power is low. In addition, since this demodulation method stores the received burst signal and performs the demodulation operation, demodulation can be performed without preamplifying the data sequence for carrier wave recovery and clock recovery, which was conventionally necessary for burst signals. It is also possible to significantly improve the efficiency of burst signals.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a TDMA burst signal demodulation method according to the present invention, FIG. 2 is a frame format diagram of a burst signal used in the present invention, and FIG. 3 is a block diagram of a digital processing type demodulator according to the present invention. FIG. 4 is a diagram showing the estimated frequency for each sample of the 5-DFT circuit output according to the present invention, FIG. 5 is a diagram showing the configuration of the frequency function estimator according to the present invention, and FIG. 6 is a sliding DFT method used in the present invention. FIG. 7 is a frame format diagram of a conventional burst signal. 1.2... Multiplier, 3... π/2 phase shifter, 4...
・Fixed frequency oscillator, 5.6...Low frequency communication filter, 7...Analog/digital converter, 8...Memory switch, 9...Memory A side IO...Memo 98 side, 11・ ・ 12 ・ ・ 14 ・ ・ 15 ・ ・ 17 ・ ・ 18 ・ ・ 19 ・, ・ 20 ・ ・ 21, 2 24, 2 28 ・ ・ 29 ・ ・ 100 ・ ・ 200 ・ ・ ・ Digital processing type demodulator, ・Multiplier, 13...5-DFT, - Frequency function estimator, - Frequency compensator, - Waveform shaping filter, - Initial phase compensator, - Clock regenerator, - Determiner, 2.23... Multiplier , 5.26.27... Adder, - Arithmetic unit, - Frequency function synthesizer, - Input terminal, - Demodulated data output terminal. 1° hour

Claims (5)

[Claims] (1) In a TDMA burst signal demodulation method in which a TDMA burst signal, which is a received signal, is demodulated by synchronous detection, a first reference signal controlled by a fixed frequency oscillator and a phase shifter has an orthogonal relationship with the first reference signal. The received signal is subjected to quasi-synchronous detection with a second reference signal located at The digital received signal, which has become a digital value through D conversion, is stored in a memory in burst units, and the modulated component of the digital received signal read from the memory is removed to make it an unmodulated signal, and then the digital received signal is converted to a non-modulated signal on the time axis. Convert the modulated signal to an unmodulated signal on the frequency axis, estimate the temporally varying frequency of the unmodulated signal on the frequency axis on the frequency axis, and continuously change from the estimated frequency series. A method for demodulating a TDMA burst signal, characterized in that: an approximation of a frequency function is obtained, and the obtained approximation function is used to compensate for frequency fluctuations of the received signal for demodulation. (2) The memory comprises a plurality of memories that sequentially store digital reception signals for one burst, and the digital reception signal is read out by sequentially switching between the plurality of memories. Demodulation method for TDMA burst signals. (3) The TDMA burst signal according to claim 1, wherein the unmodulated signal on the time axis is converted to the unmodulated signal on the frequency axis and the estimation on the frequency axis is performed by a sliding DFT method. Demodulation method. (4) Converting the unmodulated signal on the time axis to the unmodulated signal on the frequency axis and estimating the unmodulated signal on the frequency axis are performed by a frequency estimation method using an adaptive digital filter. Demodulation method of the TDMA burst signal described. (5) The TDMA burst signal demodulation method according to claim 1, wherein the approximation of the frequency function is performed by a least square error method.