JP3221173B2 - Frequency offset correction method for digital communication equalizer. - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a frequency offset correction method for an equalizer of a digital mobile communication device.

[0002]

2. Description of the Related Art FIG. 4 is an operation flow diagram of an equalizer having a frequency offset correction function in this digital mobile communication device, and FIG. 5 is a format diagram of an example of a received signal for one burst. In the case of GSM / PCN, which is a European digital cellular system, one burst is composed of 142 symbols, the central 26 symbols are a fixed value called training sequence B, and the other A,
C is an encrypted bit indicating information. Hereinafter, description will be made assuming the GSM / PCN system.

[0003] The equalizer processing will be described with reference to FIG. The processing of this equalizer depends on the nature of the processing.
It is often realized by an SP (Digital Signal Processor). In FIG. 4, a block indicated as a RAM 10 indicates a built-in RAM of the DSP, and other blocks are processed by the DSP by a program. Or indicate functions that are handled by special hardware.

First, in step 11, a signal for one burst is received and stored in the RAM 10. In step 12, the communication path is estimated from the training sequence of the received signal for one burst stored in the RAM 10. The data obtained in the communication channel estimation (12) is the impulse response of the estimated communication channel, and this impulse response is stored in the RAM 10 in step 13.

Next, the impulse response estimated in step 13 is used as a filter coefficient,
In step 6, a matched filter is calculated for the received signal. The carrier component of the received signal can always be extracted by the matched filter operation in step 16, and the result is also stored in the RAM 10 in step 17. In the next step 18, the frequency offset is estimated from the output signal of the matched filter stored in the RAM 17, and in step 19, the frequency offset is corrected for the output signal of the matched filter using the estimated frequency offset value. The result is also stored in the RAM in step 20. A term dependent on the output characteristic of the matched filter of the Planch metric (metric A)
And This frequency offset estimation and correction will be described later.

On the other hand, based on the estimated impulse response stored in the RAM in step 13,
Calculates a term (metric B) depending on the channel characteristic of the branch metric, and this result is also calculated in step 15 by RA
Stored in M. When the metric A and the metric B from the RAM are added in Steps 20 and 15, a branch metric at each reception time is obtained in Step 21. The branch metric obtained in this way is
In step 22, a decoded burst can be obtained by performing an ACS (add-compare-select) operation or inputting it to an ACS circuit constituted by hardware. This series of equalizer processing is a known technique.

Next, conventional frequency offset estimation will be described. Mean square value E of matched filter output y (k) ′ over NT time from time 0 to time (N−1) T
The ratio between the real part Re and the imaginary part Im of (y ′ ² ) is represented by the following equation (1), where θ is the amount of phase rotation change per symbol time due to frequency offset and φ is the reference phase.
It is determined only by φ and N.

[0008]

If the output of the matched filter at time 0 is the input of the center of the training sequence used to determine the filter coefficients of the matched filter,
Since the reference phase φ can be regarded as 0, θ can be estimated if N is determined in advance. It is known to obtain the amount of phase rotation change from the root mean square value of the output signal of the matched filter.

The θ _A in the NT time from the time 0 to the time (N−1) T and the θ _B in the NT time from the time − (N−1) T to the time 0 are expressed by the following equation (2).

[0011]

Further, the time-(N-1) T to the time (N-
1) θ at 2NT time of T = (θ _A −θ _B ) / 2
Is expressed by the following equation (3).

[0013]

Here, these expressions are replaced by the following symbols.

[0015]

These are shown in the Re, Im coordinate system as shown in FIG. That is, A is from time 0 to time (N-1) T
Of the phase rotation over time NT, B is the time-(N-
1) The amount of phase rotation change over the time NT from T to time 0, and the real parts R _A and R _B of the root mean square value of the matched filter output
And its imaginary part I _A, indicated by I _B.

As described above, the time-(N-1) T ~
Phase rotation cos of phase rotation change amount (AB) due to frequency offset over 2NT time at time (N-1) T
The component and the sin component can be obtained using the following equations (4) and (5).

[0018]

The value obtained by dividing these values by 2N gives the phase rotation co per 1T time, ie, per symbol time.
s component (= cos θ) and sin component (sin θ). In the program, sin (AB) or cos
Square the value of (AB), remove the square root from the formula,
It is convenient to obtain cos θ and sin θ from these values by referring to a table. Here, cos θ and sin θ are called frequency offset estimation values.

Next, the frequency offset correction based on the frequency offset estimated values cos θ and sin θ obtained here will be described. The frequency offset correction is to correct the phase of the received signal so as to be the reference phase φ. In the following description, RA _X is c of the output signal of the matched filter
os component, the IA _X is a sin component of the output signal of the matched filter. However, it is assumed that the signal is received from time 0 to time (N-1) T. RA _X ′ and IA _X ′ indicate signals after frequency offset correction. The suffix _X indicates the sample number.

RB _X is the cos of the output signal of the matched filter.
The component, IB _X, is the sin component of the output signal of the matched filter. However, it is assumed that the signal is received from time − (N−1) T to time 0. RB _X ′ and IB _X ′ indicate signals after frequency offset correction. This correction of the frequency offset is performed by dividing the signal received from time 0 to time (N-1) T and the signal received from time-(N-1) T to time 0. .

In the correction of the frequency offset with respect to the signal received at time 0 to time (N-1) T, RA ₀ , I
Frequency offset cos0α, sin0α for A ₀
And frequency offset correction RA ₀ ', IA _0' is described as follows, The frequency offset cos1α for RA _1, IA _1, sin1α and frequency offset correction RA ₁ ', IA _1' and, ... RA _N-1, IA frequency offset cos (N-1) for the _{N-1 α,} sin _(N
-1) α and frequency offset correction RAN _-1 ', IA
_N-1 'is also shown as follows.

[0023]

Similarly, from time- (N-1) T to time 0
In the correction of the frequency offset for the received signal, the frequency offset cos0 for RB ₀ and IB ₀
β, sin0β and frequency offset correction RB ₀ ′,
IB ₀ 'and, also frequency offset cos1β for RB _1, IB _1, sin1β and frequency offset correction RB _1', and _{IB 1 ', ... RB N-} 1, IB N-1 for a frequency offset cos (N-1) β , Sin (N−
1) β and frequency offset correction RB _N-1 ', IB
_N-1 'is shown as follows.

[0025]

As described above, the phase rotation si per symbol determined from the average value of the phase rotation over 2NT time
When frequency offset correction is performed using nθ and cosθ as frequency offset estimation values, this situation is shown in FIG.

[0027]

According to this conventional method, the frequency offset is corrected by using the phase rotation sin θ and cos θ per symbol obtained from the average value of the phase rotation change over 2NT time as the frequency offset estimation value. Therefore, there is no problem when the phase rotation occurs uniformly around time 0, but as shown in FIG. 3, time-(N-1) T to time 0 and time 0 to time (N-1) If the amount of phase rotation change is significantly different between NT and T, there is a problem that the phase after frequency offset correction does not become the reference phase φ.

FIG. 8 shows a case where frequency offset correction is performed by a conventional method for a received signal having a significantly different phase rotation change amount from time- (N-1) T to time 0 and time 0 to time (N-1) T. It is a coordinate diagram which shows a situation. That is, 2
If the frequency offset correction of the conventional method is performed using the phase rotation sin θ and cos θ per symbol obtained from the average value of the phase rotation over NT time as the frequency offset estimation value, the corrected phase does not become the reference φ. In the case where the carrier frequency for transmitting information is a relatively low frequency of several hundred MHz, such a problem hardly occurs because the fluctuation of the phase rotation is small through one burst, but the phase is relatively low in the case of a relatively high GHz frequency. Because it is easy to rotate, it tends to be a problem.
Further, as the number of symbols in one burst increases, the fluctuation of the phase rotation increases, which causes a problem.

An object of the present invention is to solve such a problem.
It is an object of the present invention to provide a frequency offset correction method for a digital communication equalizer that can correct the phase of each received signal to a reference phase φ even when the amount of change in phase rotation is significantly different throughout one burst.

[0030]

According to the present invention, a frequency offset of a digital communication equalizer for performing a decoding process by reducing the intersymbol interference of a received signal due to multipath fading with a received signal of one burst as an input. In the correction method, the received signal for one burst is divided into a plurality of blocks, the frequency offset of the received signal belonging to each block is estimated from the received signal for each block, and the frequency offset estimated for each block is estimated. The frequency offset of the received signal belonging to each of the blocks is corrected using the estimated value to perform the entire offset correction.

[0031]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A frequency offset correction method according to the present invention estimates a frequency offset of a received signal belonging to a block from a received signal obtained by dividing one burst into a plurality of blocks,
The frequency offset of the received signal belonging to the block is corrected using the frequency offset estimation value estimated for each block.

For example, as one embodiment of the present invention, when one burst is divided into two blocks, time 0 to time (N-
1) A frequency offset cos component (= cos) per symbol time obtained from the received signal at the NT time T
θ _A ) and the sin component (= sin θ _A ) are used to correct the frequency offset of the signal received in this section. Similarly, a frequency offset cos component (cos θ _B ) and a sin component (= sin) per symbol time obtained from the received signal in the NT time from time − (N−1) T to time 0 are obtained.
The frequency offset of the signal received in this section is corrected using θ _B ).

The processing when one burst is divided into two blocks will be described below. First, time 0 time (N-1) T
Phase rotation c due to frequency offset over NT time
The os component and the sin component can be obtained by the following equations.

[0034]

The value obtained by dividing these values by N is calculated from time 0 to time 0.
A phase rotation cos component (= cos θ _A ) and a sin component (= sin θ _A ) per 1T time in NT time at time (N−1) T, that is, per symbol time.

On the program, as in the prior art, sin
It is convenient to obtain the phase rotation per symbol time from the value of ² A or cos ² A by referring to a table. Phase rotation cos [theta] _A per symbol time obtained by table reference, the sin [theta _A and frequency offset estimate.

Similarly, from time- (N-1) to time 0 and NT
The phase rotation cos component cosB and sin component sinB due to the frequency offset over time are represented by R in the above equation (8).
_A, a I _A R _B, can be obtained by replacing the I _B.

The value obtained by dividing these values by N is given by
The phase rotation cos component (= cos θ _B ) and sin component (= sin θ _B ) per 1 T time in NT time at time (N−1) T, that is, per symbol time. Phase rotation cos [theta] _B per 1 symbol time obtained from this result, the sin [theta _B and frequency offset estimate.

Next, the frequency offset estimated values cos θ _A , sin θ _A , cos θ _B , and sin θ obtained in this manner are obtained.
_The frequency offset correction by _B will be described. RA
_X is cos component of the output signal of the matched filter, the IA _X is a sin component of the output signal of the matched filter. However, it is assumed that the signal is received from time 0 to time (N-1) T. R
A _X ′ and IA _X ′ indicate signals after frequency offset correction. RB _X is the cos component of the output signal of the matched filter, I
BX is the sin component of the output signal of the matched filter. However, it is assumed that the signal is received from time − (N−1) T to time 0. RB _X ′ and IB _X ′ indicate signals after frequency offset correction.

In the correction of the frequency offset for the signal received at time 0 to time (N-1) T, RA ₀ , I
Frequency offset cos0α, sin0α for A ₀
And frequency offset cos1α, sin1α and frequency offset correction RA ₁ ′, IA ₁ ′,... RA
_N-1, IA _N-1 for a frequency offset cos (N-
1) α, sin (N−1) α and the frequency offset corrections RA _N−1 ′, IA _N−1 ′ are θ in the above equation (6).
Is replaced by θ _A.

In the correction of the frequency offset with respect to the signal received from the time − (N−1) T to the time 0, R
Frequency offset cos0β, s for B ₀ , IB ₀
in0β and frequency offset correction RB ₀ ′, I
B ₀ 'and, also frequency offset cos1β for RB _1, IB _1, sin1β and frequency offset correction RB _1', and _{IB 1 ', ... RB N-} 1, IB N-1 for a frequency offset cos (N-1) β , Sin (N−
1) β and frequency offset correction RB _N-1 ', IB
_N-1 ′ is represented by an equation in which θ in the above equation (7) is replaced with θ _B.

FIG. 1 shows N at time 0 to time (N-1) T.
Perform frequency offset correction of the signal received in this section by using a frequency offset cos [theta] _A and sin [theta _A per symbol time calculated from the received signal at time T, the time - (N-1) T~ time NT time 0 results using a frequency offset cos [theta] _B and sin [theta _B per symbol time calculated from the received signal subjected to frequency offset correction of the signal received in this interval in time -
This shows how the phase is corrected to the reference phase φ even if the amount of phase rotation change of the received signal is significantly different between (N-1) T to time 0 and time 0 to time (N-1) T. Thus, time 0
At time 0 to time (N−
1) T, time- (N-1) T to time 0, the burst is divided to obtain the respective frequency offset estimation values, and the frequency offset correction can be performed based on these values.

However, when the number of symbols in one burst is large, if the phase is greatly changed in each block only by dividing into two blocks based on time 0 as described above, the corrected phase becomes φ. No. Therefore, when the number of symbols in one burst is large, it is necessary to perform finer block division to obtain the respective frequency offset correction values (cos θ, sin θ) and perform correction.

Next, an embodiment in which a burst is divided into four blocks based on time 0 assuming that the number of symbols in one burst is large will be described with reference to FIG. Frequency offset estimated value cos θ _AM , si from time 0 to time (M−1) T and time − (M−1) to time 0
nθ _AM, cosθ _BM, Determination of sin [theta _BM, frequency offset correction is also as described above.

Thereafter, the frequency offset estimated values cos θ _AN , sin θ _AN , cos θ _BN , and sin over time MT to time (N−1) T and time − (N−1) T to time −MT
θ _BN is similarly obtained. At the time of frequency offset correction, cos θ is used as the first frequency offset value in this section.
_{For AN} and sin θ _AN , the phase rotation change AM over the MT time from time 0 to time (M−1) T is added, and cos
theta _BN, for sin [theta _BN is time - (M-1) over MT time T~ time 0 adds the amount of change in phase rotation BM.
The subsequent processing is as described above.

As described above, one burst is divided into a plurality of blocks, the frequency offset of the received signal belonging to the block is estimated, and the frequency offset of the received signal belonging to the block is estimated using the frequency offset estimation value estimated for each block. It is possible to correct.

[0047]

As described above, in the frequency offset correction method of the digital mobile communication device equalizer of the present invention, the phase of each received signal is referred to even if the amount of change in phase rotation is significantly different throughout one burst. There is an effect that the phase can be corrected.

[Brief description of the drawings]

FIG. 1 illustrates a time- (N-1) T-
FIG. 9 is a coordinate diagram schematically illustrating frequency offset correction for a received signal having a significantly different phase rotation between time 0 and time 0 to time (N−1) T.

FIG. 2 is a coordinate diagram schematically illustrating frequency offset estimation when one burst is divided into a plurality of blocks according to another embodiment of the present invention.

FIG. 3 is a coordinate diagram schematically showing a state in which the amount of phase rotation change is significantly different from time- (N-1) T to time 0 and time 0 to time (N-1) T according to a conventional example.

FIG. 4 is an operation flowchart of a conventional equalizer having a frequency offset correction function.

FIG. 5 is a format diagram of an example of a general received signal for one burst.

FIG. 6 is a coordinate diagram schematically illustrating a state of phase rotation of a received signal in a conventional example.

FIG. 7 is a coordinate diagram schematically illustrating a conventional example of frequency offset correction of a received signal.

FIG. 8 is a coordinate diagram schematically showing a frequency offset of a received signal having a significantly different phase rotation change amount from time- (N-1) T to time 0 and time 0 to time (N-1) T according to a conventional example. .

[Explanation of symbols]

 11-22 processing steps

Claims (1)

(57) [Claims] 1. A received signal for one burst is input.
In a frequency offset correction method of a digital communication equalizer for performing a decoding process by reducing inter-symbol interference of a received signal due to multipath fading, a received signal for each block received by dividing the received signal for one burst into a plurality of blocks. Estimate the frequency offset of the received signal belonging to each of these blocks from, and correct the frequency offset of the received signal belonging to each of the blocks using the respective frequency offset estimation value estimated for each of these blocks, to correct the overall offset A frequency offset correction method for an equalizer for digital communication, wherein the frequency offset is corrected.