JP3271250B2 - Maximum likelihood demodulation method and receiver - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a maximum likelihood demodulation method suitable for, for example, a portable telephone and a receiver using the maximum likelihood demodulation method.

[0002]

2. Description of the Related Art Recently, the use of digital cellular portable telephone systems in business and the like has been expanding due to the high degree of freedom of use places.

[0003] Such a cellular phone is, for example, 890-900.
915MHz transmission frequency band, 935-960MHz
Is connected to the nearest base station via a UHF band radio line with the receiving frequency band of, and a relatively wide range of movement is possible. In mobile phones, audio signals are digitized and TDMA (Time-Division Multiple Acces
s) The time axis and data amount are compressed and transmitted by the processing.

That is, as shown in FIG. 6, an audio signal from a microphone 201 is converted into a digital signal by an A / D converter 202 and converted into an efficient digital signal by an algorithm of an audio codec circuit 203. After that, the digital audio signal is subjected to error correction encoding processing in the encoding circuit 204. This is called a convolution code (convolution code) or a Viterbi code. After the error correction encoding circuit 204, the signal is modulated by a modulation circuit 205, supplied to an antenna 207 via an amplifier 206, and transmitted.

The transmission data is received by the antenna 207, converted into a baseband signal in the RF circuit 211, supplied to the demodulation circuit 212, and demodulated with respect to the modulation on the transmission side. Then, the demodulated data is decoded by a decoding circuit 213, further processed by a voice codec circuit 214 on a transmitting side, returned to an analog voice signal by a D / A converter 215, and transmitted to an amplifier 216. Speaker 2 through
17 and the sound is reproduced.

[0006] The demodulation circuit 212, the decoding circuit 213, and the audio codec circuit 214 are composed of a DSP (digital signal signal).
(Processor) 210. Demodulation is Viterbi equalization processing, and decoding is Viterbi decoding processing.

[0007] Viterbi decoding is a process for restoring a convolutionally coded signal. Viterbi demodulation, on the other hand, considers signal degradation due to transmission paths, etc., as being convolutionally coded by the transmission path, and uses an algorithm similar to Viterbi decoding to determine what signal was most likely to have been transmitted. Estimate using.

[0008] The Viterbi demodulation method is widely used as a demodulation method of a convolutional code in which information is sequentially convoluted into a code. Maximum Likelyho
od) Performs demodulation effectively.

[0009] As shown in FIG. 7, a reception signal of the above-mentioned portable telephone (for example, in the case of GSM) is transmitted to TB (Tsi
1 Bit), a data bit and a synchronization bit, and a data bit following the synchronization bit. In the case of GSM, the synchronization bit has eight patterns.

In Viterbi demodulation, a correlation between received data and a known synchronization bit pattern is obtained, and the power of the cross-correlation function (square of the absolute value) shown in FIG.
As shown in B, a section where the power of the cross-correlation function is most concentrated is cut out with a predetermined number of bits and used as channel estimation (impulse response).

The predetermined number of bits (constraint length), for example, 5 of the cutout is usually set according to the assumed maximum propagation delay of the transmission path. In FIG. 8, the time T is a bit interval.

[0012] constraint length C is, for example, in the case of C = 5, the Viterbi demodulator, as shown in FIG. 9A, there is a ^{2 (C-1) = 2} 4 states, the likelihood are referred to as state metrics Has a corresponding value. In the Viterbi demodulation method,
Each time 1-bit processing is performed, the state changes according to a certain rule, and the state metric is updated.

The 32 arrows in FIG. 9A are candidates for state transition, and as shown in FIG. 9B, when a new bit is added to the MSB side of the state at time n, LS
This indicates that one bit on the B side is missing and the state shifts to the state at time n + 1. As shown in FIG. 9A, when one bit newly added to the state at time n is “0”, the state transits as indicated by the solid line arrow pointing to the upper right,
In the case of "1", the transition is made as indicated by the arrow indicated by the broken line pointing to the lower right.

As is apparent from FIG. 9A, all of the 5-bit states in which one bit is newly added to the MSB side of the old state are different. In other words, the 32 arrows correspond to all the bit patterns that can be represented by the constraint length of 5 bits.

An estimated value of a received signal when each bit pattern is transmitted is made by associating a channel transition of 5 symbols with all the 5-bit bit patterns as described above,
The "distance" between this estimate and the actual received signal is then added to the state metric of the original state. The "distance" between the estimate and the actual received signal is called the branch metric of each bit pattern.

In the Viterbi demodulation method, a state transition indicated by a thick solid arrow in FIG. 9B is selected so that the "sum" of the branch metric and the state metric is reduced. The sum of the branch metric of the selected arrow (bit pattern) and the original state metric is the state of the new state.
Metric.

When the constraint length is 5 bits, for example, FIG.
As shown in FIG. 9A, at time t00, “0” in FIG.
From L07 corresponding to the state of “111”, time t01 to t01
Data "10010101100010" over 13
Are sequentially input, the state transition indicated by the thick solid line arrow is sequentially performed, and at time t14, the state L04 of "0100" in FIG. 9A is reached.

When data different from the above-described data is input at times t03 and t08 to t13, state transitions indicated by thin solid lines, long and short broken lines and dotted lines are performed, respectively. At time t14, L
The states reach from 00 to L15. That is, even if the arrow is traced back from any of the states L00 to L15 at the time t14, it joins some flows.

The above-mentioned branch metric corresponds to the conditional probability that a certain bit pattern was transmitted under the condition that a certain signal is being received. In other words, assuming that a certain bit pattern is transmitted and the “distance” between the estimated received signal and the actual received signal is considered to be noise that follows a Gaussian distribution, the conditional probability that the bit pattern was transmitted is the magnitude of the noise. Can be represented by the probability that is “distance”.

This probability is expressed as exponential in a Gaussian distribution.
It can be expressed in the form of {square of distance}, where the square of distance is proportional to the logarithm of this probability.

Furthermore, the state metric, which can be said to be the sum of branch metrics, can also be said to be the logarithm of the product of the conditional probabilities as described above. It can be said that it seems to be the most transmitted bit sequence.

[0022]

By the way, conventionally, when a Viterbi demodulation method is adopted in a portable telephone, as described above,
The constraint length has been set according to the assumed maximum propagation delay of the transmission path.

For example, if the bit interval T (see FIG. 8) is set to 3.
Assuming that the propagation delay is 7 microseconds and the maximum propagation delay τ of the transmission path is 20 microseconds, this propagation delay τ is equivalent to 5 bits, and the constraint length is set to 5 bits.

Also in demodulation by a decision feedback adaptive equalizer that updates a tap gain based on a comparison between a filter output and a training reference signal (synchronization bit), a shift register is used in accordance with an assumed maximum propagation delay of a transmission path. Number of taps was set.

However, if the constraint length and the number of taps are fixed to values corresponding to the maximum propagation delay, demodulation with the number of bits corresponding to the maximum propagation delay even when the actual delay of the transmission path is small. Calculations and data processing are performed,
As compared to the amount of calculation required when the constraint length and the number of taps are reduced, about twice as much per bit or tap,
There has been a problem that the amount of calculation processing increases exponentially, so that useless calculations are performed by the increased amount, and that power is wasted in the arithmetic unit.

In particular, in the case of a battery-powered device such as a portable telephone, there is a problem that the operation time is shortened.

In view of the above, an object of the present invention is to reduce the amount of calculation for the maximum likelihood demodulation processing when the actual delay of the transmission path is small, thereby avoiding wasteful power consumption in the arithmetic unit. To provide a maximum likelihood demodulation method and a receiver.

[0028]

In order to solve the above-mentioned problems, a maximum likelihood demodulation method according to the present invention transmits a digital modulated signal received via a transmission path based on a cross-correlation function between a synchronous signal having a predetermined pattern and a synchronization signal. In the maximum likelihood demodulation method for obtaining the obtained bit sequence, the absolute value of the cross-correlation function is 2
A means for determining the degree of concentration of the power is provided, and the square of the absolute value of the cross-correlation function is more than a predetermined value and concentrated in a section having a smaller number of bits than the predetermined number of bits corresponding to the maximum propagation delay in the transmission path. When it is determined that the channel is estimated to be smaller than the predetermined number of bits, a section where the square of the absolute value of the cross-correlation function is maximized is cut out to perform channel estimation of the transmission path. It is a feature.

[0029]

According to this configuration, when the actual delay of the transmission path is small, the square of the absolute value of the cross-correlation function is concentrated at a rate equal to or more than the predetermined value in the section of the bit number smaller than the predetermined bit number. As a result, the section where the square of the absolute value of the cross-correlation function is maximized can be cut out with a smaller number of bits than the predetermined number of bits, and the calculation amount for the maximum likelihood demodulation processing is significantly reduced. Thus, wasteful power consumption in the arithmetic unit is avoided.

[0030]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to FIGS.

A portable telephone as an example of a receiver to which the maximum likelihood demodulation method according to the present invention is applied includes, for example, a radio reception system 10, a baseband system 20, and a transmission system 30, as shown in FIG. Is done.

In the radio receiving system 10, a signal received from the transmitting / receiving antenna 11 is transmitted to the selector 12 and the bandpass filter 1.
3 and supplied to an RF amplifier 14 for reception, and the output of the amplifier 14 is supplied to a pair of frequency mixers (mixers) 1.
5 and 16 respectively. The output of the oscillator 17 is
While being directly supplied to one of the mixers 15 and supplied to the other mixer 16 through a 90 ° phase shifter 18, the respective mixers 15 and 16 obtain orthogonal baseband signals (I, Q). .

The baseband system 20 includes memories 21 and 22
Including path memory 20PM and arithmetic circuit 23
P20D is provided. As described above, the DSP 20D performs Viterbi demodulation, Viterbi decoding, and audio codec processing.

Then, the quadrature signals (I, Q) from the mixers 15, 16 are converted into digital signals by A / D converters 24, 25 and stored in one memory 21. The other memory 22 is used for Viterbi demodulation,
A predetermined synchronization bit pattern is written in advance. The path memory 20PM is a memory that stores, for example, the arrow illustrated in FIG. That is, by adding the pushed bit to the memory, the transition from state to state is stored.

Then, the arithmetic circuit 23 calculates a cross-correlation function between the received data in one memory 21 and the synchronization bit pattern in the other memory 22, and performs a Viterbi demodulation process using the cross-correlation function. . DSP20D
Then, after Viterbi demodulation, Viterbi decoding is performed, and the algorithm of the audio codec is executed.

The audio data from the arithmetic circuit 23 is D / A
The signal is converted into an analog audio signal by the converter 26,
The signal is supplied to a receiver (speaker) 28 through an amplifier 27.

On the other hand, the transmission system 30 is exactly the same as the transmission system of FIG. 6 described above, and converts a voice signal from the microphone 31 into a transmission signal and transmits it from the antenna 11.

Next, the maximum likelihood demodulation processing according to one embodiment of the present invention will be described with reference to FIGS.

FIG. 1 shows a maximum likelihood demodulation processing routine according to an embodiment of the present invention, and FIG. These routines are executed by the arithmetic circuit 23 shown in FIG.

In the maximum likelihood demodulation processing routine 100 shown in FIG. 1, first, at step 101, the correlation between the known synchronization bit pattern and the received signal is obtained as described above, and the initial value of the constraint length, for example, 5 A section where the power of the cross-correlation function is most concentrated is cut out using bits, and the transmission path is estimated (channel estimation).

Then, the process proceeds to step 102 to evaluate how concentrated the power of each symbol of the correlation function is.

In the next subroutine 110, as shown in FIG. 2, the length (constraint length) C used for channel estimation is determined from the correlation function according to the power concentration of each symbol of the correlation function. Can be

That is, in step 111 of the subroutine 110 of FIG. 2, when i = 0, the initial maximum constraint length C is assumed, and the power of each symbol of the correlation function may be shortened. Is set. In this embodiment, the initial maximum constraint length C is set to 5 bits, and the power concentration α capable of shortening the constraint length is 85.
%.

In the next step 112, the sum of squares (power) P0 of the C-bit channel estimation is calculated, and the routine proceeds to step 113, where i = i + 1 is substituted.

In the next step 114, the power Pi for channel estimation of Ci bits is calculated.

Then, in step 115, it is determined whether or not the power ratio Pi / P0 between the C-bit channel estimation and the C-bit channel estimation is equal to or higher than the concentration α.
If the power ratio Pi / P0 is smaller than α, step 1
Returning to step 13, the processing up to step 114 is repeated.

If the power ratio Pi / P0 is equal to or greater than the concentration α, the process proceeds to step 116, where the constraint length to be used is C
−i + 1, and the constraint length variable processing subroutine 11
Exit 0.

The maximum likelihood demodulation processing routine 10 shown in FIG.
Returning to 0, in step 121, the channel estimation and the convolution are calculated for the 2 C-th power bit pattern in which the 2 (C-1) th power state shifts to the next stage, and the received signal estimation value is calculated. Desired.

In the next step 122, for each of the two state transitions that reach each state, the magnitude of the received signal estimated value and the sum of the distances of the received signals are compared, and one of the state transitions is selected.

Proceeding to step 123, the bits pushed by the state transition are added to the path memory.

In the next step 124, the path memory is processed from the state having the lowest metric, and the demodulation result is output.

The routine proceeds to step 125, where it is determined whether or not the determined bit has been output. If the determined bit has not been output, the process returns to step 121 and the processing up to step 125 is repeated. . When the determined bits are output, the maximum likelihood demodulation processing routine 100
Ends.

In the constraint length variable processing subroutine of FIG.
As shown in FIG. 3A, the power of the channel estimation in the case of the maximum constraint length of 5 bits is used as the denominator. The ratio of the channel estimation power when the constraint length is 1 bit, as indicated by the halftone dot portion in FIG. 3C, is obtained, and it is possible to select the minimum constraint length at which this ratio is equal to or greater than the predetermined concentration α. it can.

If the propagation delay of the transmission path is shorter than the maximum assumed value and the power of channel estimation is concentrated on a small number of symbols within the constraint length, the power is concentrated even within the constraint length. The missing symbols have little effect on the estimation of the received signal. Therefore, in such a case, even if the constraint length is shortened, the error rate after the maximum likelihood demodulation processing does not deteriorate.

When the channel can be sufficiently estimated with three symbols, and the constraint length is reduced to 3 bits by the constraint length variable processing subroutine 110 as described above,
The state transition is represented by eight arrows as shown in FIG.

In FIG. 4, the number of arrows is reduced to 1/4 and the amount of calculation for the maximum likelihood demodulation is also reduced to 1/4 compared to FIG. 8 in which the constraint length is 5 bits. Have been.

Therefore, the power consumption of the arithmetic unit is also significantly reduced, and the operation time of a battery-powered device such as a mobile phone is prolonged accordingly.

Although the above embodiment has been described with reference to Viterbi demodulation, the present invention can be similarly applied to demodulation by a decision feedback adaptive equalizer by making the shift register variable length. it can.

[0059]

As described above, according to the present invention, there is provided a maximum likelihood demodulation method for obtaining a transmitted bit sequence based on a cross-correlation function between a received digital modulation signal and a synchronization signal having a predetermined pattern. When it is determined that the square of the absolute value of the cross-correlation function is greater than or equal to a predetermined value and is concentrated in a section having a smaller number of bits than the predetermined number of bits corresponding to the maximum propagation delay in the transmission path, Since the section where the square of the absolute value of the cross-correlation function is maximum is cut out with a smaller number of bits than the predetermined number of bits and the channel is estimated for the transmission path, the amount of calculation for the maximum likelihood demodulation processing is reduced. It is possible to remarkably reduce the power consumption, and it is possible to avoid unnecessary power consumption in the calculation unit corresponding to the amount of calculation.

[Brief description of the drawings]

FIG. 1 is a flowchart for explaining the operation of an embodiment of a maximum likelihood demodulation method according to the present invention.

FIG. 2 is a flowchart for explaining an operation of a main part of FIG. 1;

FIG. 3 is a conceptual diagram for explaining the operation of one embodiment of the present invention.

FIG. 4 is a conceptual diagram for explaining the operation of one embodiment of the present invention.

FIG. 5 is a block diagram illustrating a configuration example of a receiver according to an embodiment of the present invention;

FIG. 6 is a block diagram illustrating a configuration example of a mobile phone to which the present invention is applied.

FIG. 7 is a conceptual diagram for explaining a conventional maximum likelihood demodulation method.

FIG. 8 is a conceptual diagram for explaining a conventional maximum likelihood demodulation method.

FIG. 9 is a conceptual diagram for explaining a conventional maximum likelihood demodulation method.

FIG. 10 is a conceptual diagram for explaining a conventional maximum likelihood demodulation method.

[Explanation of symbols]

 Reference Signs List 10 wireless reception system 20 baseband system 21, 22 memory 23 arithmetic circuit 30 transmission system 100 maximum likelihood demodulation processing routine 110 constraint length variable processing subroutine

Claims (2)

(57) [Claims] 1. A maximum likelihood demodulation method for obtaining a transmitted bit sequence based on a cross-correlation function between a digital modulation signal received through a transmission path and a synchronization signal of a predetermined pattern, comprising the steps of: The degree of concentration of the square is determined, and the square of the absolute value of the cross-correlation function is more than a predetermined value and concentrated in a section of a bit number smaller than the predetermined bit number corresponding to the maximum propagation delay in the transmission path. When it is determined that the channel number is smaller than the predetermined number of bits, a section where the square of the absolute value of the cross-correlation function is maximum is cut out, and the channel is estimated for the transmission path. Maximum likelihood demodulation method characterized in that: 2. A receiver comprising means for demodulating a digitally modulated signal received through a transmission path by demodulation means using a maximum likelihood demodulation method, comprising: Determining means for determining the degree of concentration of the square of the absolute value of the cross-correlation function; and means for controlling the number of bits for channel estimation of the transmission path based on the determination result of the determining means. Features receiver.