WO2009124568A1

WO2009124568A1 - Method and circuit device for correlation and reshaping based channel estimation in a dtmb receiver

Info

Publication number: WO2009124568A1
Application number: PCT/EP2008/002886
Authority: WO
Inventors: Yuanli Wang; Bowei Song; Xiaoxiang Li
Original assignee: Trident Microsystems (Far East) Ltd.
Priority date: 2008-04-11
Filing date: 2008-04-11
Publication date: 2009-10-15

Abstract

Method and circuit device for correlation and reshaping based channel estimation in a DTMB receiver The invention relates to especially a method for correlation and reshaping based channel estimation in a DTMB receiver by processing received data (r(n) ) being transmitted via a multipath channel (CH) within data frames (f), the frames (f) having varying phase pseudo-random sequence (PN) guard interval. The received data (r(n)) are manipulated and correlated with a local pseudo-random sequence (S) of the receiver. Correlated data (C) are reshaped for reducing effects of a non-ideal auto-correlation (A) of the local pseudo-random sequence. Noise reduction filtering is applied to the channel reshaping outputs. Especially, data of the pseudo-random sequence (PN) are extracted from different frame headers (fh) through a fixed window (t) and are combined as an uncontaminated pseudo-random sequence (PNz). The correlation is formed of the combined pseudo-random sequence and of the local pseudo-random sequence (S). Reshaping is done by applying inverse matrix of the local pseudo-random sequence (S).

Description

Method and circuit device for correlation and reshaping based channel estimation in a DTMB receiver

Technical Field

The invention regards to a method and circuit device for correlation and reshaping based channel estimation in a DTMB receiver.

Background Art

Pseudo-random sequences are widely used as framing sequence in frame structures of many communication systems, such as ATSC (Advanced Television Systems Committee) and DTMB (Digital TV (Television) Broadcasting Standard) as Chinese terrestrial digital television standard. In these systems, a pseudo-random sequence is inserted into a frame header of each data frame. Especially, the pseudo-random sequence is inserted into the data frame in front of a data section of the data frame. Generally, such pseudo-random sequence can be used to estimate a channel response on receiver side. Careful selection of the structure of such pseudo-random sequence can simplify the design of channel estimation.

During transmission, the pseudo-random sequence being inserted into the header section is deteriorated by additive white Gaussian noise and Inter-Symbol-Interference (ISI) caused by echoes. In order to recover the data from the interfered receive signal, channel information must be extracted before Inter-Symbol-Interference is compensated. Time-domain correlation and fast Fourier transformation are the usual ways for pseudo-random sequence based channel estimation. To simplify the design of channel estimation, the pseudo-random sequence can be constructed in a special way, e.g. pseudorandom sequences with varying phases between data frames. In the Chinese terrestrial digital TV broadcasting standard each data frame consists of pseudo-random sequences segment as the frame header and of a data segment or data section comprising load data. For some modes of it, the phases of pseudo-random sequences between data frames are different.

Pseudo-random sequences are quite commonly adapted in digital communication systems for facilitating the signal recovery on receiver side. The pseudo-random sequence can be inserted before or after data segment d and the phase of the pseudorandom sequence PN of each frame can be different as shown Fig. 5a and 5b. Given a pseudo-random sequence segment of length L, to kinds of method can be used to generate a varying phase pseudo-random sequence as shown in Fig. 5a and 5b. The frame structures shown above in Fig. 5a can be used for the proposed method. Each data frame f consists of a frame header fh of the length L, comprising the pseudo-random sequence PN as a training sequence, and of a data segment d following the frame header fh and comprising the load data, which are to be transmitted.

Pseudo-random sequence with varying phase means that the pseudo-random sequences between the different data frames f are just cyclic shifted to each other. When the pseudo-random sequence has the structure shown in Fig. 5 (a) the maximum estimated length of echo approximates to the length of pseudo- random sequence. When the pseudo-random sequence has the other structure shown in Fig. 5 (b) , the maximum estimated length of echo approximates to the length of pseudo-random sequence without the cyclic part. In Chinese terrestrial digital TV broadcasting standard, the structure of pseudo-random sequence for some modes is the same as the one shown in Fig. 5 (b) . Fig. 6 shows components of a receiver, which executes channel estimation for such frame structure according to the prior art.

A transmitter TX sends data frames via a channel CH to a receiver T. Data frames are modified by a channel response h during transmission via the channel CH. Accordingly the receiver T receives a radio frequency signal from its antenna and converts it to a desired intermediate frequency signal. This signal is applied to a in-phase- and quadrature-modulation module IQM, which comprises components for shifting a frequency spectrum and some digital filters. This module moves the intermediate frequency signal to the base band. The signal outputted from the in-phase- and quadrature-modulation module IQM is applied to a carrier recovery module CR. The carrier recovery module CR is used for estimation of the carrier frequency offset. The carrier recovery module CR is part of a phase locked loop to the in-phase- and quadrature-modulation module IQM.

Furthermore, the output signal or output data of the in-phase- and quadrature-modulation module IQM is applied to a clock recovery module TR, which serves for the frame header of the data frame. The output signal of the clock recovery module TR is applied to a channel estimation module CE, which estimates channel information and channel response h, respectively, and outputs its data to an equalizer EQ. The equalizer compensates the channel's impact on the data. Data outputted out of the equalizer EQ are applied to a forward-error correction module FEC, which is used for decoding.

Some existing methods of channel estimation correlate received data with the local pseudo-random sequence s directly. However, its performance would be affected by the adjacent data segment and by additive white Gaussian noise. Especially, when frequency selective fading is serious or if the signal to noise ratio is low, the estimated result is not sufficiently accurate for signal recovering. By a simple mathematical deduction, the correlation between the received signal or data r(n) and a local pseudo-random sequence s(n) can be got, as shown in Eq. (D

L-I Af-I

= ^f∑_ls(i)[∑h(τ)x(n+i-τ)+w(n+i)] ι=0 r=0 :i)

L-I L-I

= ∑h(τ)∑s(i)x(n+i-τ)+ Σ∑s(i)w(n+i) r=0 1=0 /=0

H = O,...,Z-I

10 If the transmitted pseudo-random sequence has cyclic property then x(n + i-τ) = s((n+i-τ)_L) n,i = \,...,L-\; r = 0,...,M-l. (2)

Then Eq. (1) can be rewritten as

R_sr=∑h(r)K(n-τ)+RJn)

J-J r=0 \3 I

= h(n)+RJn).

In Eq. (3) R_ss is the auto-correlation function of the pseudorandom sequence, which is assumed to be ideal delta function and R_sw is the cross-correlation function of the pseudo-random 20 sequence and noise, which has a zero mean.

So the estimated channel response h(n) is:

h(n) = RJn) = h(n)+RJn). (4) 5

However, the method has some drawbacks. Firstly, the auto-correlation function of a pseudo-random sequence may not be an ideal delta function as shown below for the case of Chinese terrestrial digital TV broadcasting standard:

So Eq. (3) does not hold strictly. This will bring estimation error.

, secondly, Eq. (2) does not hold because the correlation result will be interferred by adjacent data section of the data frame and that will cause serious performance loss.

Thirdly, this method will introduce extra estimation error when the signal-to-noise-ratio is low, because the later part of Eq. (3) can not be ignored when the noise power is large.

Technical Problem

It is an object of the invention to provide in alternative way a method and circuit device for correlation and reshaping based channel estimation in a DTMB receiver to reduce interference effects from data out of data frames data segment.

Technical Solution

This object is solved by a method having features according to claim 1, and by a circuit device or DTMB receiver having features according to claim 10. Preferred aspects and embodiments are subject-matter of dependent claims. Advantageous Effects

Especially, there is provided a method for correlation and reshaping based channel estimation in a DTMB receiver by processing received data, the received data being transmitted via a multipath channel within data frames, the data frames having varying phase pseudo-random sequence guard interval, wherein the received data are manipulated and are correlated with a local pseudo-random sequence of the receiver and wherein correlated data outputted out of correlation are reshaped for reducing, especially for removing of effects of a non-ideal auto-correlation of the local pseudo-random sequence.

In general, the term data means informational segments of a data frame, and in present method the term data means especially data or data segments extracted from a frame header of the data frames and not load data comprised in a data segment of the data frames.

Further, there is provided a method, wherein parts of data of the pseudo-random sequence are extracted from different frame headers of the received data through duration of a fixed window and are combined as an uncontaminated pseudo-random sequence. In other words by appropriate choice of the fixed position of a time window there are extracted information out of data frames header for extraction of a pseudo-random sequence or for extraction of at least segments of a pseudo-random sequence. The extraction is done in a way such that the extracted data are not interfered by load data of data segments of data frames received via parallel paths of the multi-path channels. Thus, it is possible to extract data of a pseudo-random sequence, which is uncontaminated by load data, to use such uncontaminated pseudo-random sequence data for further processing.

It is preferred, if a position of the fixed window starts in relation to the data frame after a maximum echo length of a data frame having no phase shifting and ends at an end of a segment of the pseudo-random sequence and/or ends at the end of the frame header of the data frame of the first of the received data. The first of the received data means said data of the data frame received at a first time. In other words such data are data of a data frame being without face shift in relation to other data frames running via other of the multi-path channels .

Especially, the uncontaminated pseudo-random sequence is extracted and combined from at least two data frames, which are received via the same path of the multi-path channel. It is also advantageous individual or in combination, to extract and combine the uncontaminated pseudo-random sequence from at least two data frames, which are received via different paths of the multi-path channel.

Especially, in step of extraction there are extracted only data from frame headers in such a manner, that these data from frame headers being not interfered or contaminated by data of a data segment following the frame header and being received via other paths or with other phases of the multi-path channel. Thus, it is possible to compose an uncontaminated pseudo-random sequence out of the received data and data frames, respectively.

Preferably, the correlation is formed by correlation of at least parts of such uncontaminated pseudo-random sequence or sequences and the local pseudo-random sequence to get correlation result for further processing. Especially, for reshaping an inverse matrix of a auto-correlation matrix of the local pseudo-random sequence is applied to the correlation results of the correlation. In other words it is assumed, that output values outputted out of this correlation are formed out of the channel or channel impulse response and out of the auto- correlation of the local pseudo-random sequence, the pseudorandom sequence being provided in such a manner as described above. It is possible to take advantage of the realization, that the inverse matrix of the auto-correlation of the known local pseudo-random sequence can be constructed in an easy manner. Thus, effects of a non-ideal auto-correlation of the pseudo-random sequence can be reshaped reducing or completely taking away effects of the non-ideal auto-correlation. The result is a useable channel estimation, which however, comprises additive white Gaussian noise.

Especially, a low pass noise decreasing filtering is applied to correlated and reshaped data. A noise reduction of channel estimates is achieved by executing this step.

Furthermore, there is provided a circuit device or DTMB receiver for correlation and reshaping based channel estimation in DTMB receiver by processing received data, having components for processing the received data being transmitted via a multipath channel within data frames, the data frames having varying phase pseudo-random sequence guard interval. The device or receiver being designed and/or programmed to manipulate the received data and to correlate the manipulated data with a local pseudo-random sequence of the receiver and to reshape correlated data outputted out of correlation for reducing, especially for removing of effects of a non-ideal autocorrelation of the local pseudo-random sequence.

Especially, the circuit device or the DTMB receiver is designed and/or programmed for executing such a method. Further, there is claimed use of such a method in a circuit device or in a

DTMB receiver.

The proposed method can be used in digital communication systems with a varying phase pseudo-random sequence structure between frames. For such frame structure, the channel information can be extracted with sufficient accuracy based on the proposed method. Thus, the preferred channel estimation method bases on the concept to eliminate interference of the data section by transforming the auto-correlation of the pseudo-random sequence to an ideal or nearly ideal delta function and by reducing the impact of noise by filtering the estimated and correlated data.

Description of Drawings

An embodiment will be disclosed in more details with respect to enclosed drawing. There are shown in:

Fig. 1 schematically components of a receiver for channel estimation,

Fig. 2 schematically several data frames for illustrating the extraction of not disturbed and uncontaminated, respectively, parts or segments of the pseudo-random sequence of the header section without influence of data section data,

Fig. 3 a method for composing or reassembling of data of a composed pseudo-random sequence basing on data of frame headers of different received data frames,

Fig. 4 components of a filter for noise reduction,.

Fig. 5 frame structures with pseudo-random sequences having varying phase according to the prior art, and

Fig. 6 schematically components for channel estimation in a receiver according to the prior art.

Mode for Invention

Fig. 1 shows a preferred receiver composed of a plurality of components. In the main there are described only components, which are advantageous for a preferred method for channel estimation. Instead of constructive components it is possible to use software modules or combined hardware/software modules, which are executable by means of adequate software programming for executing the respective method steps. Necessary software programs are stored within the device.

Especially, there are shown five steps illustrated as modules, which are preferred for processing of received data r(n) received in the receiver. In a first method step the received data r(n) are applied to an extraction module EXT. The extraction module EXT extracts the uncontaminated, i.e. not- influenced data of the pseudo-random sequence PN of the header section depending on the header section position of the data frame and of the maximum delay m of the channel. The maximum delay m of the channel is provided by a module for estimation of channel's maximum delay m and is applied to the extraction module EXT. The module for estimation of channel's maximum delay CMD gets the respectively necessary information from a channel length estimation block or from a predefined and/or preset value as maximum delay m depending on the modules construction.

The extraction result of the extraction module EXT is applied to a collector module COL. The collector module COL collects a plurality of data from different header segments from different data frames to compose a pseudo-random sequence PNz as reference pseudo-random sequence for the received data r(n) . The pseudo-random sequence PNz is applied to a correlator module COR. The correlator module COR gets applied a local pseudo-random sequence PNl as further data sequence. The local pseudo-random sequence PNl is applied to the correlator COR by a local training sequence module LTS. The correlator's COR correlation result is applied to a reshaping module RES for reshaping. The result of the reshaping module RES corresponds to a first estimated channel response, which however, is interfered by white Gaussian noise. Thus, this channel response is applied to a noise decreasing module NDE. The noise decreasing module NDE processes the first estimated channel response and outputs the estimated channel response h(n) . Fig. 2 illustrates schematically an extraction process for extracting the data of the pseudo-random sequence PN from a plurality of data frames f. In the uppermost line there is illustrated an exemplary data frame f, which is composed in its first section out of a data sequence providing a header segment fh in known manner. Thereafter the data frame f comprises a data segment d comprising load data as to be transmitted data information. In the line below there is illustrated a channel response h which is convoluted with the originally sent data during transmission via the radio channel. Thus, the receiver receives a data frame s, which comprises originally sent data folded with the channel response h.

In the lower section of Fig. 2 there are three received data frames vnPN, which has been transmitted via the radio channel from transmitter to receiver via different channels of the multi-path channels. These three data frames have offset phases and are received within the receiver at consecutive times.

Exemplarily, the third of these received data frames is offset by a delay longer than the channel response h. In other words, the third data frame is received past the maximum delay m of the channel at the receiver with respect to the receiving time of the first data frame. In times after this time of maximum delay m there are received only header segment data from each of the data frames before the duration or length L of the first received header segment fh ends. In other words, the length L of the header segment fh has to be longer than the maximum delay m of the channel to make possible present method for channel estimation. During the time between the maximum delay m of the channel and the length L of the header segment fh of the first received data frame it is guaranteed that there are received only data of header segments fh from all the received data frames without any interference of load data of the data segment d. As shown in Fig. 2, a multi-path channel will cause Inter- Symbol-Interference while transmitted signal or data passes the channel, and thus the pseudo-random sequence segment will be interfered by data of the adjacent data segment d. To get a good estimation of the channel response h it is very important to remove this interference from the pseudo-random sequence segment .

In a digital communication system, especially in a television broadcasting or in a mobile phone communication system, the area which a base station or broadcaster covers is restricted and the delay m of echo is also limited. The delay m corresponds to the echoes on the multi-path channel. For a system in which the length of the pseudo-random sequence in each data frame is longer than maximum delay of the echoes, the proposed method can extract the uncontaminated pseudo-random sequence segment without interference from data part. Once the channels maximum delay m is known, an uncontaminated pseudorandom sequence part can be located in the received signal. In other words, as soon as the maximum delay m of the channel is known, an uncontaminated pseudo-random sequence PNz can be extracted out of the received data frames. There is used the time window t, which starts at the end of the maximum delay m and which ends at the length L for the first received data frame. When starting with index 0 then it ends at the length L- 1.

A second step is to store the uncontaminated pseudo-random sequence segment of each of the frames to make up a complete pseudo-random sequence PN. Because the phase of the pseudorandom sequence changes in each frame, the uncontaminated pseudo-random sequence part or segment is different for each frame within the fixed window t. Therefore, it is possible to store uncontaminated pseudo-random sequence parts in continuous multiple frames, which are not interfered by load data contents of the data frame's data segment d and a hole uncontaminated pseudo-random sequence PNz can be manipulated and used for channel estimation.

Also, in order to track the change of the channel, this uncontaminated hole pseudo-random sequence PNz will be updated partly frame by frame. A more detailed operation for composing the composed, uncontaminated pseudo-random sequence PNz is illustrated in Fig. 3. By this step, the pseudo-random sequence PNz is manipulated such that it satisfies Eq. (2) .

The third step is to cyclic correlate the collected pseudorandom sequence PNz being not contaminated with the local pseudo-random sequence PNl by correlator COR. The local trainings sequence provided by the local training sequence module LTS preferably corresponds to the last of the composed pseudo-random sequences PNz.

The data outputted out of the correlator module COR corresponds to the data of the channel response h which are convoluted with the respective cyclic auto-correlation values R_ss of the pseudo-random sequence, shown in the first part of Eq. (3) .

For the pseudo-random sequence PN need as local pseudo-random sequence S, its cyclic auto-correlation is not the ideal delta function shown in Eq. (5) . In order to get the sufficiently accurate channel estimates, the reshaping module RES is used to eliminate the effect of the pseudo-random sequences un-ideal auto-correlation in a following fourth step.

Given that A= [do, ai, ..., αi-j] is the cyclic auto-correlation of the known pseudo-random sequence S, H= [ho, hi, ..., h_L_i]^T represents the channel impulse, L is the length of the pseudo-random sequence, and C= [co, Ci, ..., c_L-i]^T corresponds to the cyclic correlation result between the local pseudo-random sequence PNl and the collected and composed uncontaminated pseudo-random sequence PNz, we have C = S*H*S

Here "*" means the cyclic convolution,

From Eq. (6) the results of the cyclic correlation can be taken as estimation of H only if R is a unit matrix. But the non- ideal property of cyclic auto-correlation of the pseudo-random sequence makes R being not such a matrix. Fortunately, the auto-correlation matrix R is a cyclic matrix and has an inverse matrix. The inverse matrix can be got easy in usual manner. So the estimation of H is

For the m sequence used in the Chinese terrestrial digital TV broadcasting standard, the auto-correlation matrix R in Eq. (6) has the following format:

L -1 1

-1 L ^■•• -1

R = :8)

-1 -1 ^■•• L Then, according to Eq. (7), the estimated channel response is

So in the fourth step in Fig. 1 the reshaping module RES processes the correlator modules COR output in the way described in Eq. (9) to remove the effect of the pseudo-random sequence's un-ideal auto-correlation.

After the reshaping in the fourth step, we get the channel estimates without considering the effect of additive white Gaussian noise, whose effect will be removed in a fifth step. In the fifth step, an anti-noise filter is used to improve the performance more, as shown in Fig. 4.

The input values of the noise decreasing module NDE are divided by (N+l) , which can be done by an amplifier module. The divided values are applied to an adder, the output values of which corresponds to the noise reduced output values and being the noise-reduced estimated channel response h(n) . Further, the output values of the addition process are fed to a delay module z^"1 and are fed after the delay to a further amplifier module, especially provided by a multiplication module. In this module delayed data are multiplicated by N/ (N+l) . The multiplication result is applied to the adder to be added with further values outputted at later time of the first amplifier of the noise decreasing module NDE. Reference signs:

Components :

CE channel estimation module

CH channel

CMD module for estimation of channel's maximum delay

COL collector module

COR correlator module

CR carrier recovery module

EQ equalizer

EXT extraction module

FEC forward-error correction module

IQM in-phase- and quadrature modulation module

LTS local training sequence module

NDE noise decreasing module

RES reshaping module

T receiver

TR clock recovery module

TX transmitter

Data / signals:

A= [ao, ... a_L-i] cyclic auto-correlation of known or local pseudo-random sequence

C= [co, ... c_L-i] ^τ cyclic correlation of known or local pseudorandom sequence and composed uncontaminated pseudorandom sequence f data frame fh frame header d frame' s data segment h, h(n),h(τ) channel response h(n) estimated channel response

H= [ho, ... h_L-i] ^T channel impulse response i index for integrations / summations

L length of s (n) m maximum echo length

M length of h(n) PN pseudo-random sequence

PNz composed uncontaminated pseudo-random sequence r, r(n) received data

R_sr(n) channel estimation received data

R_ss(n) auto-correlation of local pseudo-random sequence

R_sw(n) cross-correlation of local pseudo-random sequence and of noise s, s (n) , S data of local pseudo-random sequence t time window for extraction vpPN pseudo-random sequence having varying phase w(n) additives white Gaussian noise (AWGN) x(n) sent signal / sent data τ index for integrations / summations

Claims

1. Method for correlation and reshaping based channel estimation in a DTMB receiver by processing received data (r(n)), the received data (r(n)) being transmitted via a multipath channel (CH) within data frames (f ) , the data frames (f) having varying phase pseudo-random sequence (PN) guard interval,

- wherein the received data (r(n)) are manipulated and are correlated with a local pseudo-random sequence (S) of the receiver and

- wherein correlated data (C) outputted out of correlation are reshaped for reducing, especially for removing of effects of a non-ideal auto-correlation (A) of the local pseudo-random sequence.

2. Method according to claim 1, wherein parts of data of the pseudo-random sequence (PN) are extracted from different frame headers (fh) of the received data (r(n)) through duration of a fixed window (t) and are combined as an uncontaminated pseudorandom sequence (PNz) .

3. Method according to claim 2, wherein a position of the fixed window (t) in relation to the data frame (f) starts after a maximum echo length (m) of a data frame (f) and ends at an end of a segment of the pseudo-random sequence (PN) and/or ends at the end of the frame header (fh) of the data frame (f) of the first of the received data (r(n)) .

4. Method according to claim 2 or 3, wherein the uncontaminated pseudo-random sequence (PNz) is extracted and combined from at least two data frames (0...L-1, Li... Li-I, L_n... L_n-I),

5. Method according to any of claims 2 to 5, wherein in step of extraction there are extracted only pseudo-random sequence from frame headers (fh) , these data from frame headers (fh) being not interfered or contaminated by data (d) of a data segment following the frame header (fh) .

6. Method according to any of claims 2 to 6, wherein the correlation (C) is formed by correlation of at least parts of such uncontaminated pseudo-random sequence or sequences (PNz) and the local pseudo-random sequence (S) to get correlation result for further processing.

7. Method according to claim 7, wherein for reshaping an inverse matrix of a auto-correlation matrix of the local pseudo-random sequence (S) is applied to the correlation results of the correlation (C) .

8. Method according to any of claims 1 to 8, wherein a low pass noise decreasing filtering is applied to correlated and reshaped data.

9. Circuit device or DTMB receiver for correlation and reshaping based channel estimation in DTMB receiver by processing received data (r(n)), having components for processing the received data (r(n)) being transmitted via a multipath channel (CH) , the data frames (f ) having varying phase pseudo-random sequence (PN) guard interval, said device or receiver being designed and/or programmed

- to manipulate the received data (r(n)) and to correlate the manipulated data with a local pseudo-random sequence (S) of the receiver and

- to reshape correlated data (C) outputted out of correlation for reducing, especially for removing of effects of a non-ideal auto-correlation (A) of the local pseudo-random sequence.

10. Circuit device or DTMB receiver according to claim 10, being designed and/or programmed for executing a method according to any of claims 1 to 9.

11. Use of a method according to any of claims 1 to 9 in a circuit device or in a DTMB receiver.