KR101103165B1 - Delayed Decision Feedback Equalizer Channel Tracking Algorithm for EDGE Systems - Google Patents

Abstract

The present invention relates to a channel tracking algorithm used in an 8 PSK equalizer in a base-band modem of a GSM / EDGE system. In particular, the present invention relates to a channel tracking algorithm that is suitable for a large number of calculations required for channel tracking. The present invention relates to a channel tracking algorithm.

A channel tracking method of a delay determination feedback equalizer in an EDGE system according to the present invention includes extracting a delayed DFE symbol for a predetermined symbol period after sequence estimation is performed, and extracting the extracted symbol for channel update. And performing a channel update with the LMS using the symbol vector.

According to the configuration of the present invention as described above it is possible to prevent the performance degradation due to error propagation compared to the case of using the best conventional survivor (survivor), and compared to the PSP method according to the prior art has the advantage that can greatly reduce the calculation to provide.

Channel Tracking, Error Propagation, Delay Decision Feedback Equalizer

Description

Delayed Decision Feedback Equalizer Channel Tracking Algorithm for EDGE Systems

1 is a view for explaining the burst format of the GSM / EDGE system

2A and 2B are error distribution diagrams according to the positions of the bursts in the normal channel and the RA250 channel, respectively.

3 shows a schematic structure of a PSP-MLSE receiver;

4 illustrates a structure of a PSP-MLSE for removing tail ISI from a received signal using a DFE.

5 is a diagram illustrating a delayed DFE symbol used in a symbol vector in a channel tracking method of a delay determination feedback equalizer in an EDGE system according to the present invention.

FIG. 6 is a diagram illustrating performance improvement when a channel is updated according to a channel tracking method of a delay determination feedback equalizer in an EDGE system according to the present invention in a rapidly changing RA250 channel.

<Explanation of symbols for the main parts of the drawings>

410: PSP-MLSE 420: initial channel estimator (Initial CE)

430: Decision buffer 440: Symbol vector determination unit

The present invention relates to a channel tracking method of a delay decision feedback equalizer in an EDGE (Enhanced Data Rates for GSM Evolution) system.

8 In the global system for mobile (GSM) / EDGE system using phase shift keying (PSK), accurate estimation of the channel is very important.

1 is a diagram illustrating a burst format of a GSM / EDGE system, FIGS. 2A and 2B are error distributions according to positions of bursts in a static channel and a RA250 channel, respectively, and FIG. 3 is a PSP-MLSE receiver. Figure is a schematic structure of the.

As shown in FIG. 1, the burst format of the GSM / EDGE system includes two training segments (TSs) in a midamble section and two data segments consisting of 58 data symbols. , 2nd data segment). In addition, there are three tail symbols TB before and after the burst, and 8.25 guard symbols GP at the end of the burst.

For accurate estimation of the channel, cross-correlation (LCC) or least square (LS) algorithms are used in the mid-amble section to obtain inter-symbol interference (ISI) in the equalizer. Channel Impulse Response (CIR) is estimated.

FIG. 2A is an error distribution diagram in the static channel, which shows that it is constant regardless of the position of the burst. In this case, since the CIR does not change within a burst in the normal channel, the CIR estimated in the mid amble section does not need to be updated in the data section.

On the other hand, Figure 2b shows that the error distribution according to the channel change as the error distribution in the RA250 channel from the midamble to the tail direction increases. As described above, in a channel environment such as TU50, HT100, and RA250, the CIR estimated in the midamble section and the CIR in the data section are different, and thus, when the channel update algorithm is not used, the channel is estimated toward the tail bit section. The error gets worse. As a result, the ISI estimation error increases in the equalizer, thereby degrading the performance of the equalizer. In particular, when the channel changes rapidly, such as the RA250, the performance degradation is even greater.

Therefore, in order to keep up with the channel change, the CIR is updated with a Least Mean Square (LMS) algorithm in the data section in order to improve the problem. A representative receiver using the method is Per-Survivor Processing (PSP) -Maximum Likelyhood Sequence Estimation (MLSE).

Referring to FIG. 3, the PSP-MLSE receiver includes an MLSE 310, a mid amble detector 320, and a channel estimator 330.

In order to estimate the channel in the PSP-MLSE receiver configured as described above, first, the channel estimation unit 330 of the channel estimator 330 is used by using a training sequence symbol of the midamble section detected by the mid-amble 320. Each channel estimator (CE) 331, 332, 333 obtains an initial CIR of a corresponding channel by a CC or LS algorithm and outputs the result to the MLSE 310. The LS algorithm is more complex than CC, but the LS algorithm is mainly used because it shows better performance.

Thereafter, the initial CIR value estimated through the training sequence symbol of the midamble is updated through the LMS algorithm in the data interval. In this case, the trellis of the equalizer proceeds from the mid amble section to the tail beat section.

M-ary PSK modulated signals S _k , such as Binary Phase Shift Keying (BPSK) and Quadrate Phase Shift Keying (QPSK), pass through a multipath fading channel and receive additional white Gaussian noise (AWGN) When the received signal r _k is expressed as Equation 1 below.

(Equation 1)

In Equation 1, {h _{k, l} } is a set of channel tap coefficients and h _k is a Gaussian noise sequence.

If the trellis state of the equalizer is expressed as μ, the transmission metric when the trellis state transitions from m to m at time index k is defined as Equation 2 below. .

(Equation 2)

In Equation 2, e is an error value representing a difference between the received signal r _k and the signal regenerated by the CIR and the trellis sequence in the equalizer. h is a CIR vector according to a time index and a trellis state, and X is a symbol vector when the trellis state transitions from? to m.

The method of updating a channel using the LMS algorithm is shown in Equation 3 below.

(Equation 3)

The method is very effective for following fast fading channels but has limited use since it requires a lot of computation to update the channel parameters for each survivor.

The PSP-MLSE has too many trajectory states at high modulation orders or long channel lengths, such as 8 PSK, making it impossible to implement with today's digital signal processing technology. The method used to solve this problem is Decision Feedback Estimation (DFE). With this approach, decision feedback can effectively reduce the memory length of the channel. If the CIR length of the channel is L, it is divided into two parts, N and L-N. The first part, N, is treated as PSP-MLSE, and the second part, L-N, is processed by Decision Feedback Equalizer (DFE).

4 is a diagram illustrating a structure of a PSP-MLSE for removing tail ISI from a received signal using a DFE.

The PSP-MLSE, which removes the tail ISI from the received signal using the DFE, includes a PSP-MLSE 410, an initial channel estimator 420, a decision buffer 430, and a symbol. The vector determination unit 440 is configured.

In the channel estimation of the PSP-MLSE receiver which removes the tail ISI from the received signal using the DFE configured as described above, N, the first part of the channel CIR using the output of the initial channel estimator 420, is PSP. Dealing with the MLSE 410, the LN, which is the second part of the channel CIR, is processed by the decision buffer 430 by using the output of the initial channel estimator 420.

There are two types of DFEs, a global DFE and a local DFE. The global DFE uses a DFE value determined in the best survival path as one decision vector affects all survival paths in the calculation of transmission metrics of all survival paths. This method causes performance degradation due to error propagation if an error occurs in the best survival path.

Equation 4 below represents a method of using the global DFE as a formula, and illustrates a method of previously removing ISI using the DFE.

(Equation 4)

In this case, when there is no DFE error, the ISI is removed as shown in Equation 4, but when the DFE error occurs, the ISI may increase.

Local DFE, another method of DFE, changes the DFE value for ISI removal in each survival path to compensate for performance degradation due to error propagation, which is a disadvantage of global DFE. have.

(5)

On the other hand, when the DFE is applied not only in the PSP-MLSE method but also in the general MLSE method that does not update the channel, that is, in the case of Decision Feedback Sequence Estimation (DFSE), the local DFE is used rather than the global DFE.

In the conventional technology as described above, as shown in Equation (3), it is essential to follow the change of channels by using LMS for each survival path. In the process, to reduce the complexity of the MLSE equalizer, global / local DFE is used to reduce channel memory and eliminate tail ISI to reduce overall computation. As a result, in order to keep track of channel changes with LMS for all survivors in the conventional manner, channel updates must be made by the number of states in the MLSE equalizer for each time index. This requires a great deal of computation and is difficult to implement in practice.

For example, if the state number is 8 and the channel memory length L is 7, then 56 channel tap coefficients for one time index must be updated using the LMS. In addition, since the performance gain of the receiver that can be obtained in comparison with the complexity is not large, there are many limitations to use.

On the other hand, if the channel update is made by selecting only the best survivor to reduce the LMS calculation amount, performance degradation due to error propagation occurs as in the case of using the global DFE to reduce the state of the equalizer.

The present invention provides a method for channel tracking of a delayed decision feedback equalizer in an EDGE system suitable for estimating a symbol vector used for channel update, in which case the amount of states of the equalizer is large and a large amount of calculation is required for channel tracking. The purpose is to provide.

In the symbol vector estimation used for channel update, the present invention uses a delayed DFE as a symbol vector for channel update instead of using MLSE or DFE symbols for a current received signal. It is an object of the present invention to provide a channel tracking method of a delay determination feedback equalizer in an EDGE system that can be reduced.

A channel tracking method of a delay determination feedback equalizer in an EDGE system according to the present invention for achieving the above object comprises: extracting a delayed DFE symbol for a predetermined symbol period after sequence estimation is performed; And performing a channel update with the LMS using the symbol as a symbol vector for channel update.

The step of extracting the delayed DFE symbol may include setting a predetermined depth D of a predetermined depth for determining a symbol used for channel update in an equalizer, and setting the next symbol following the symbol to be set. If the trellis has proceeded by the predetermined width (D) of the predetermined length, characterized in that it comprises a step of determining the corresponding symbol for each symbol as a symbol for the channel update.

In the step of performing a channel update to the LMS, a channel vector h having a memory length of L for the LMS for the channel update,

The error signal e and the symbol vector X are each represented by

It is characterized by the expression of.

According to the channel tracking method of the delay determination feedback equalizer in the EDGE system according to the present invention configured as described above, the performance reduction due to error propagation is prevented compared to the case of using the best survivor according to the conventional method. Not only can this be done, but the computation is greatly reduced compared to the PSP method.

Hereinafter, a channel tracking method of a delay determination feedback equalizer in an EDGE system according to the present invention will be described in detail with reference to the accompanying drawings.

In the present invention, the channel update is used in the data section to prevent the CIR estimated by the training sequence symbol of the midamble from being different from the CIR of the data section in a fast fading environment. The LS algorithm is used for initial channel estimation and the LMS algorithm is used for channel update.

In the present invention, in order to reduce the enormous calculation required to update the channel, the present invention uses the channel changing characteristic.

Looking at the change characteristics of the channel, in a situation where the channel changes very rapidly, such as the RA250 channel, the channel shows a great difference in the midamble and the tail section even if it is within one burst.

However, even if there are many differences at the beginning and end of the burst, it can be assumed that they have almost the same value between neighboring symbols. This is because changing the channel a few symbols away does not change the receiver's performance significantly.

Based on the above assumption, the channel update is not applied immediately as the equalizer traverses as in the conventional schemes, and the channel update is performed by the LMS with respect to the symbol for which the sequence estimation is performed by a constant delay.

5 is a diagram illustrating a delayed DFE symbol used in a symbol vector in a channel tracking method of a delay determination feedback equalizer in an EDGE system according to the present invention.

According to the conventional method, the MLSE state of the equalizer and the global / local DFE symbol are used to obtain the symbol vector X. However, in the present invention, the symbol vector X is represented as a delayed DFE symbol (as shown in FIG. 5). X _KD ).

In order to reduce the memory size of the metric in the Viterbi channel decoder, when the trellis proceeds by a predetermined decision depth D before the trellis ends, the decoded bits are output one bit at a time.

Equalizers that use the Viterbi Algorithm (VA) are also able to track the performance and depth (D) of a sequence decision after tracery is complete and then trace back. There is little difference in performance between deciding one symbol after a rally. In addition, unlike the channel decoder, the decision width D for symbol determination shows little performance deterioration even when a small value of 16 or less is used.

Therefore, even if the channel changes rapidly, the channel is updated using the decoded symbols, assuming that the channel changes little during the D symbol. This reduces the calculation in proportion to the number of survivors compared to the conventional method of channel update for all survivors, and propagates errors compared to channel update only for the best survivors. There is almost no degradation in performance.

To this end, in the channel tracking method of the delay determination feedback equalizer in the EDGE system according to the present invention, Equation 3 regarding the LMS for channel update described above is changed to Equation 6 below. .

(6)

In Equation (6), h is a channel vector having a memory length of L, and an error signal e and a symbol vector X are defined by the following equation.

(7)

(Equation 8)

FIG. 6 is a diagram illustrating performance improvement when a channel is updated according to a channel tracking method of a delay determination feedback equalizer in an EDGE system according to the present invention in a rapidly changing RA250 channel.

In FIG. 6, circles indicate when channel tracking is turned off, and squares indicate when a channel tracking is performed, a bit error ratio (BER) / block error (BER) according to a signal-to-noise ratio ( _ES / N _O ). Indicates.

As can be seen in Figure 6, when the channel tracking method of the delay determination feedback equalizer in the EDGE system according to the present invention is used, the same signal-to-noise ratio (E _S ) as the signal-to-noise ratio (E _S / N _O ) increases. / N _O ), the lower the BER / BLER.

As described above, when the channel tracking method of the delay determination feedback equalizer in the EDGE system according to the present invention is used, performance degradation due to error propagation is reduced compared to the case of using the best survivor according to the conventional method. Not only can this be avoided, but the computation is much reduced compared to the PSP method.

For example, if the equalizer has 8 states and the channel has 7 memory lengths, the PSP method must calculate 8 x 7 x (58 + 3) = 3416 tap weights per data segment. . On the other hand, when using the crystal width D as 16 in the channel tracking method of the delay decision feedback equalizer in the EDGE system according to the present invention, 1 x 7 x (58 + 3-D) = 315 tap weights. weights) only needs to be calculated. In other words, the calculation is reduced by more than 10 times.

In the above, the channel tracking method of the delay determination feedback equalizer in the EDGE system according to the present invention has been described in detail. However, the present invention is not limited thereto, and the present invention is included in the present invention even if the present invention is carried out without change and modification without departing from the spirit of the present invention, and the fact will be apparent to those skilled in the art.

According to the channel tracking method of the delay determination feedback equalizer in the EDGE system according to the present invention, the performance degradation due to the error propagation can be prevented and the PSP can be prevented compared to the case of using the best survivor according to the conventional method. It offers the advantage of greatly reducing the computation over the scheme.

Claims (3)

Extracting a delayed DFE symbol for a predetermined symbol period after sequence estimation is performed; And A channel tracking method of a delay determination feedback equalizer in an EDGE system, comprising: performing a channel update by an LMS using the extracted DFE symbol as a symbol vector for channel update. Extracting the delayed DFE symbol, Setting a predetermined length of decision depth (D) for determining a symbol to use for channel updates in the equalizer; And The next symbol following the corresponding symbol is configured to include the step of determining the symbol as a symbol for the channel update for each symbol when the trellis has proceeded by the predetermined predetermined width (D) of the EDGE system, characterized in that configured Channel Tracking Method of Delay Decision Feedback Equalizer in. delete The method of claim 1, wherein performing a channel update with the LMS: A channel vector h having a memory length of L for the LMS for channel update is

Is expressed as The error signal e _kD and symbol vector X are each,

A channel tracking method of a delay determination feedback equalizer in an EDGE system, characterized by the following equation.

Images