JPH09270829A - Viterbi detection method for differential phase modulated wave - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain the same characteristic as conventional delay detection Viterbi decoding with a small arithmetic operation amount. SOLUTION: An input difference phase modulation signal is sampled by its symbol period (13), its sampled output zs is delay-detected (14), its detection output Δϕn ' is fed to a D symbol delay circuit 18 and to a differential coder 15, in which the signal is coded. The differential coded output ϕn ' is used to apply inverse modulation to the sampled output zs (16), and an obtained phase difference error with respect to the delay detection signal is decoded by a 3Q<-1> state Viterbi decoder (17). In the case of a high Eb /No , a discrimination error of delay detection is scarcely caused, and even when the discrimination error takes place, the error is either of ±2π/M radians, it is not required to conduct MQ<-1> state Viterbi decoding. A Viterbi decoding output ΔΨn- D' is added to an output of a delay circuit (18) for correction, to obtain a detection output Δϕn- D.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a detection method for differentially detecting a phase-modulated wave and estimating a transmission phase sequence by a Viterbi algorithm.

[0002]

2. Description of the Related Art M phase (M is an integer of 4 or more, usually M = 2)
ⁿ, N is an integer of 2 or more) Differential phase modulation
In the conventional differential detection of the signal (DPSK), the reception at the time point n
Signal z _nZ delayed by one symbol time_n-1Detect
As a reference signal for waves, Re [z_nz^* _n-1exp-j
Δφ_n] Is the maximum Δφ_nThe transmitted symbol and format
Set. Where z_nIs a complex number representation of the received signal,
Re [. ] Is a real part, and * is a complex conjugate. Differential detection
Since the carrier wave recovery circuit is not required, the configuration of the detection circuit is simple.
However, the error rate characteristic is worse than that of the coherent detection.

Therefore, the documents “1” and “D.
s and K.S. Feher, "Optimal no
n coherent detection of P
SKsignals, "IEEE Electronic
s Letters, vol. 26, pp. 398-4
00, March 1990 ”discloses a method for improving the error rate characteristic by Viterbi decoding of a combination ^MQ-1 state of consecutive Q-1 phase differences.

[0004]

The conventional method of estimating the transmission phase sequence by the Viterbi algorithm has a drawback that the calculation scale of the Viterbi decoder increases as M and Q increase, which is not practical. Therefore, an object of the present invention is to provide a Viterbi detection method for a differential phase modulation wave in which the number of states of the Viterbi decoder is reduced and the operation scale is reduced with a slight deterioration in characteristics.

[0005]

According to the present invention, reception
Of the transmitted M-phase differential phase-modulated signal to the transmission symbol period
And two consecutive received signal samples z_n-1And z_n
And are used for delay detection, and the detected phase difference is
Differential encoding by adding to the differentially encoded output signal before the time
Update the output phase and receive it using the differentially encoded output phase.
Inversely modulate the signal sample and inversely modulate the signal sample μ_nAt time n
State m^Q-1Number of states S (m is an integer smaller than M)_nTo
Q-1 phase difference error series {ΔΨ_nq: Q = 0, ..., Q
-2} (ΔΨ is the difference between the phase difference of the inverse modulation signal and the transmission phase difference
Difference S) and state S at time (n-1)
_n-1Reference signal η stored in_n-1(S_n-1) Is ΔΨ
_nAnd the inverse modulation signal sample z_n′ With
Find the inner product, and use this inner product as state S_n-1'From state S_nTo
A branch metric λ (S
_n-1→ S_n) And set this branch metric to state S
_n-1Path metric Λ (S_n-1)
State S_n-1Path metric Λ (S_n
| S_n-1), The branch metric and path
The operation for finding the trick is the state S_nThere is a path to enter
All states S_n-1Repeated about
State S ^ that compares the size of the trick and gives the maximum value_n-1
Select only this and discard the others, this state S ^_n-1State S_n
The most probable path point (n-1)
Stored in the path memory and the path metric Λ
(S_n| S ^_n-1) To state S_nPath metrics in
Ku Λ (S_n) Is stored in the path metric memory as
Reference signal η_n(S_n) Is Q consecutive inverse modulation signal markers
Using the book, η_n(S_n) = Μ_n+ Σ_{q = 1} ^Q-1μ_nqexp (jΔ
Ψ_n+ ΔΨ_n-1+… + ΔΨ_{n + 1-q}), Store it in the reference signal memory, and perform the above operations.
State S_nM for all^Q-1Pieces
State S that compares the size of tricks and gives the maximum value_n＾
And state S_nThe path memory is
State S reached by tracing back only at fixed time point D_nD＾
Phase difference error ΔΨ of_nD^ Is found, and the detected phase difference is D symbol
Phase difference error ΔΨ_nDAdd ^
Calculate and output as detection output.

According to such a configuration, the inversely modulated signal becomes an M-phase DPSK signal modulated with a decision error sequence when the conventional differential detection is used. Most errors are plus or minus 2π / M radians when the error rate is small enough. Therefore, if the error rate is small, the inverse modulation signal takes almost three values even if it is the M-phase DPSK signal. That is, 3 ^Q-1 states (m =
3) The Viterbi decoder can estimate the error sequence.
It will be easily understood that not only m = 3 but also m <M can be similarly decoded, and the amount of calculation is smaller than that in the case of m = M.

[0007]

BEST MODE FOR CARRYING OUT THE INVENTION
The complex display in the band is as follows. z_n= √ (2E_s/ T) expj (φ_n+ Θ) + w_n (1) where φ_n= {2mπ / M; m = 0, 1, ..., M-
1} is the modulation phase, E_s= (Log_TwoM) E_bIs 1 thin
Received energy per bolt (E_bIs per bit
Received energy), T is the symbol length, θ is the unknown phase,
w_nIs thermal noise and its variance is 2N₀/ T (N₀Is sloppy
Sound power spectral density). Phase difference Δφ_n= Φ _n−φ_n-1
Is the transmitted symbol (log_TwoM bits) and {2mπ
/ M; m = 0, 1, ..., M-1}.

FIG. 1 shows an embodiment of the present invention (configuration of demodulator).
Is shown. The differential phase modulation signal from the input terminal 11 is converted to baseband, and unnecessary noise is generated by the reception filter 1.
It is removed in step 2, and is sampled in the sampling circuit 13 for each symbol period and supplied to the differential detector 14. Delay detector 1
The determination in 4 is performed as follows.

[0009]

## EQU1 ## where * is the complex conjugate and Re [z] is the real part of z. That is, the transmission symbol is estimated from the phase difference between two consecutive reception samples. Although z _n-1 is a reference signal for determination, it is disturbed by thermal noise, and therefore its characteristics are deteriorated as compared with the ideal synchronous detection / differential decoding. Now, such differential detection determination output is differentially encoded by the differential encoding circuit 15 as follows.

Φ_n′ ＝ φ_n-1′ + Δφ_n′ ＝ Σ_{k = 1} ⁿΔφ_k′ (3) The received signal is received by the inverse modulator 16 using this differentially encoded output.
Inversely modulate. Ie μ_n= Z_nexp-jφ_n′ (4) If equation (4) is rewritten using equation (1), μ_n= √ (2E_s/ T) expj (Ψ_n+ Θ) + w_nIt becomes like ′ (5). Where w ′ is new noise but variance
Is immutable. Now, Ψ _nIs the next ΔΨ_nDifferentially encode
It has become a thing.

ΔΨ_n= Ψ_n−Ψ_n-1= Δφ_n-Δφ_n′ (6) ΔΨ_nAs can be seen from the above equation, the transmission phase difference Δφ_nAnd conventional
Judgment value of differential detection Δφ _n′, That is, the judgment error
It has become. If E_b/ N₀When the value of is not small
Most of the errors are
S-minus 2π / M radian. Therefore,
In most cases, ΔΨ_n= 0, and sometimes ΔΨ_n=
It becomes ± 2π / M. This means that the error sequence estimation
Set the number of states of the Viterbi decoder of "1" to 3^Q-1Can be in a state
It is shown that. Therefore, the error sequence from the inverse modulator 16
ΔΨ_n3^Q-1Number of states, most accurate in the Viterbi decoder 17
A new error sequence is detected, that is, 3 at time n^Q-1Of
Find the path with the maximum metric among the surviving paths, and
Go up the path by D steps and configure the state where you have reached
Phase determination error ΔΨ_nD′ Is output. This judgment error Δ
Ψ_nD′ Is the detection output Δφ of the delay detector 14._n’Delay
Delayed by DT in circuit 18 Δφ_nD′ To adder
Final judgment output Δφ by adding in 19_nDOutput as ￣
Output to the terminal 21.

Δφ_nD￣ = (ΔΨ_nD′ + Δφ_nD′) Mod2π (7) Where mod2π is modulo 2π operation
It is. The processing in the Viterbi decoder 17 is the same as at time n.
State S_nBe Q−1 phase difference error series {ΔΨ_nq: Q =
0, ..., Q-2}, and when in the reference signal memory 21
State S of point (n-1)_n-1Reference signal η stored in
_n-1(S_n-1) Is ΔΨ_nOpposite to the one that only phase is rotated
Modulation signal sample μ_nAnd the inner product of
_n-1To state S_nBlanc showing the certainty of the transition to
Chimetric λ (S_{n- 1}→ S_n) And this branch
Trick the state S_n-1Path metric Λ (S
_n-1) And state S_n-1Of candidate series passing through
Metric Λ (S _n| S_n-1). This branch
State S is used to calculate the metric and path metric_n
All states S where there is a path to enter_n-1Repeat about
And compare the magnitude of the obtained path metrics to obtain the maximum value.
State to give S ^_n-1Select only and discard the others, this state
S ^_n-1State S_nThe most probable time to reach
When the path memory 23 stores the state of (n-1),
And its path metric Λ (S_n| S ^_n-1) To state S
_nPath metric Λ (S_n) As path metric
The reference signal η._n(S_n)
It is calculated by the following equation using Q consecutive inverse modulation signal samples.

Η_n(S_n) = Μ_n+ Σ_{q = 1} ^Q-1μ_nqexp (jΔΨ_n+ ΔΨ_n-1+… + ΔΨ_{n + 1-q}) (8) This reference signal η_n(S_n) Is stored in the reference signal memory 22.
I do. The above calculation 3^Q-1Individual state S_nFor all of
Done and obtained 3^Q-1Compare the size of the path metric of
The state S that gives the maximum value_nFind ^ and state S_n＾
Using the path memory 23 as a starting point for the trace point at time D
State reached by clicking S _nD^ Phase difference error ΔΨ_nD＾
Is output as a Viterbi decoding result.

In the Viterbi decoding of the document “1”, the Viterbi decoder in the M ^Q-1 state is required, whereas in this embodiment, 3 is used.
^Since the Viterbi decoder in the ^Q-1 state is sufficient, the processing amount can be significantly reduced. Not limited to m = 3, if m is large, ± 2
There is a possibility that a decision error other than π / M radian can be correctly detected, and if m is smaller than M, a smaller processing amount can be obtained than when m = M.

[0015]

The computer simulation results (Q = 2, 3, 4) of the error rate characteristics when the embodiment of the present invention is applied to the 8-phase DPSK are shown in FIG. The horizontal axis is the signal energy-to-noise power density ratio E _b / N ₀ per bit. For comparison, the error rate simulation results by the conventional differential detection (1DD) and ideal synchronous detection / differential decoding (CD) are also shown. A traceback depth D = 8 in Viterbi decoding was used. By increasing Q, the error rate characteristic can be made closer to that of synchronous detection / differential decoding (CD). The required E _b / N ₀ required to secure BER = 0.1% is ₁ in the conventional differential detection (1DD).
Although it is 2.9 dB, the required E _b / N ₀ can be reduced by 1.3 dB by using Q = 2 in the present invention. When Q = 4 is used, the required E _b / N ₀ can be reduced to 0.5 dB compared with that of the synchronous detection / differential decoding.

The figure shows the characteristic (Q = 3) of the 64-state Viterbi decoding of the document "1", but it is well understood that almost the same characteristic can be realized by using the same Q in the present invention. Therefore, when Q = 3, the same characteristic can be realized with an amount of calculation which is about 1/7 that of the document “1”.

[Brief description of drawings]

FIG. 1 is a block diagram showing an example of a differential detection device to which the present invention is applied.

FIG. 2 is a diagram showing a computer simulation result for showing the effect of the present invention.

[Procedure amendment]

[Submission date] March 29, 1996

[Procedure amendment 1]

[Document name to be amended] Drawing

[Correction target item name] Fig. 1

[Correction method] Change

[Correction contents]

FIG.

Claims (1)

[Claims] 1. The received M-phase (M is an integer of 4 or more) differential phase modulation signal is sampled at the transmission symbol period to obtain a reception signal sample z _n at a time point n, and two consecutive reception signals are received. The phase difference is detected by differential detection using the signal samples z _n-1 and z _n, and the detected phase difference is added to the differential encoded output phase one symbol time before to obtain the differential encoded output. The phase is updated, and the received signal sample z _n is inversely modulated using the differentially encoded output phase to obtain the inverse modulated signal sample μ _n , and Q−1 phase difference error sequences {ΔΨ _nq : q = 0, ..., Q
-2} (ΔΨ is a value that can be the difference between the phase difference of the inverse modulation signal and the transmission phase difference), and forms m ^Q-1 (m is an integer smaller than M) states S _n at time n, The inner product of the reference signal η _n-1 (S _n-1 ) stored in the state S _n-1 at the time point (n-1) phase-shifted by ΔΨ _n and the inverse modulation signal sample z _n ′ is Then, this inner product is taken as a branch metric λ (S _n-1 → S _n ) which represents the probability of the transition from the state S _n-1 to the state S _n , and this branch metric is the path metric Λ (S _n-1) in the state S _n-1 . S _n-1 ) to obtain a path metric Λ (S _n | S _n-1 ) of the candidate sequence passing through the state S _n-1, and an operation for obtaining the branch metric and the path metric is given to the state S _n . All states where there is an incoming path S _n-1
Repeating with respect to, the state of the obtained path metric is compared, and only the state S ^ _n-1 that gives the maximum value is selected and the others are discarded.
The time of the most probable path from _n-1 to the state S _n (n-
The state 1) is stored in the path memory, and the path metric Λ (S _n | S ^ _n-1 ) is stored in the path metric memory as the path metric Λ (S _n ) in the state S _n . _n (S _n ) using Q consecutive inverse modulation signal samples, η _n (S _n ) = μ _n + Σ _{q = 1} ^Q-1 μ _nq exp (jΔ
Ψ _n + ΔΨ _n-1 + ... + ΔΨ _{n + 1-q} ) and stores it in the reference signal memory, and the above calculation is performed for all the states S _{n to} obtain m.
The size of ^Q-1 path metrics is compared to find the state S _n ^ that gives the maximum value, and the state S _n ^ is used as a starting point to set the path memory at a certain time point D.
Then, the phase difference error ΔΨ _nD ^ of the state S _nD ^ reached by tracing back only is obtained, and the phase difference error ΔΨ _nD ^ is added to what is obtained by delaying the detected phase difference by the D symbol time to obtain a detection output. Viterbi detection method for differential phase modulation waves.