DESCRIPTION

Channel estimation method and device for long echo in especially TDS-OFDM systems

Technical Field

The invention regards to a channel estimation method for long echo in especially TDS-OFDM systems according to pre- characterizing part of claim 1, and to a device for channel estimation for long echo in especially TDS-OFDM systems.

Time Domain Synchronous-OFDM (TDS-OFDM, OFDM: Orthogonal Frequency Division Multiplex) system is a variation of general OFDM system by combining both time-domain and frequency-domain processing. It is adopted as one operation mode in the compulsory national standard of Digital Multimedia Terrestrial Broadcasting (DMTB) in China. The TDS-OFDM offers a pseudo random (PN) sequence as guardian interval between every two frames of OFDM symbols. This special frame structure brings the challenge of novel channel estimation method.

Being the core part of general OFDM system, the quality of channel estimates has a direct impact on the overall system performance. The special frame structure of TDS-OFDM system requires novel channel estimation methods at the receiver side, instead of the ones used in general OFDM system. In addition, terrestrial environments usually have long-echo delays. This requires the channel estimation for terrestrial system be able to estimate long echoes up to 40 μs around.

Background Art

Channel estimation provides an estimate of channel impulse re- sponse to a frequency domain equalizer in a TDS-OFDM system. It

has a direct impact on the system performance. There are several requirements for the channel estimation.

At first, estimation should be as accurate as possible even un- der low SNR situation. At second, estimation should be able to cope with channels with long echoes up to especially 40 μs . However, longer echo coverage is desired. At third, estimation should be able to handle dynamic channels.

Fig. 5 shows a channel estimation block in the general OFDM system diagram. It recovers the channel response from received data and then feeds the estimate to a frequency domain equalizer 4. There is shown an arrangement of general channel estimation in the OFDM system diagram. A transmitter TX sends out sent data via an antenna to an antenna of a receiver RX. Path V between these antennas provides error data and multiple effects effecting sent data to become initially received data received by antenna of receiver RX. Initially received data are provided -to an- automatic gain control 1. Outputted data of automatic gain control 1 are forwarded to an IQ-demodulation block 2 (IQ: Inphase/Quadraturphase) . Data outputted of IQ demodulation block 2 are inputted as the received data y(i) into a block especially comprising the channel estimation block 3. Data parallel with channel estimation results outputted by channel esti- mation block 3 are forwarded to a frequency domain equalizer 4. Data outputted by the frequency domain equalizer 4 are forwarded to a forward error control 5, FEC.

The general channel estimation method in OFDM system can not be applied directly due to the special frame structure of TDS-

OFDM. Fig. 6 shows the TDS-OFDM frame structure for two different modes having 4200 and 4725 symbols per frame, respectively.

First part of symbols is cyclic shifted symbols and PN data be- tween such cyclic shifted symbols. These PN data between cyclic

shifted symbols have the length of 255 in case of 4200 frame structure or the length of 511 in case of 4725 frame structure. In other words, the frame is divided into 2 parts. The beginning part, e.g. 420 symbols in the 4200 frame structure, is called guardian interval, which is a cyclic PN sequence in time domain. And the later part, e.g. 3780 in the 4200 frame structure, is a OFDM data part.

Assuming the transmitted signal is s[n) and additive noise is n(n) , which is white Gaussian noise, then the received signal y(n) is as following

L-1 y{n) = s{n)®h(n)+n(n) = σ∑a(τ)s(n-τ)+n{n) , (1) τ=0

where h{n) is an impulse response of a multipath channel, and a{τ)s are wide-sense stationary, narrow band complex Gaussian random processes with longest echo at L-symbol delay. It is assumed that they are constant during the transmission of one OFDM symbol, which is a general assumption in OFDM systems. Here "®" means linear convolution.

The channel estimation problem is how to make close estimates of these α(r) from the received signal y(n) .

Existing state of the art solutions using clean part of PN in the received signal can only provide channel estimation for short echoes. For the PN length of 420, echoes longer than 255 symbols will become indistinguishable as shown in Fig. 7.

This is due to the cyclic nature of the guardian interval. As can be seen in Fig. 7, the clean parts of PN in the two received signals for training shown as shadowed parts are the same. The long echo and the short echo are indistinguishable if

the estimation method is based on clean part of PN in the received signal .

For short echo channel, i.e. within the length of 255-symbol, the situation in Fig. 7 will not happen. Time-domain correlation method or other method like LMS (Least Mean Square) could be used upon the clean part of PN in the received signal for channel estimation. For long echo channel, these methods can not distinguish the long echo. For example, the channel esti- mates for a long echo channel like channel 1 in Fig. 7 will look similar as channel 2. The estimate of long echo is "folded" into the range within 255 symbols if the method is based on clean part of PN in the received signal.

The channel estimation methods based on clean part of PN is accurate if based on existing methods like time-domain correlation, but it could only estimate short echoes less than 255- symbol delay. In the terrestrial broadcasting environments, long echoes are common, especially in single frequency network. So it is desirable to have a large coverage range of echo channel estimation in the receiver.

Technical Problem

It is an object of the invention to provide another channel estimation method and device for long echo in TDS-OFDM systems making possible estimation of long echoes, especially, estimation of long echoes up to 40 μs around should be possible. In other words, especially estimation of long echoes more than 254-symbol delay shall be possible.

Technical Solution

This object is solved by a channel estimation method for long echo in TDS-OFDM systems having features according to claim 1,

and by a device for channel estimation for long echo in TDS- OFDM systems having features according to claim 9. Preferred aspects and embodiments are subject-matter of dependent claims.

Especially, there is provided a channel estimation method for echo in communication systems, wherein there is estimated a short echo, wherein a long echo being longer than the first short echo is estimated by executing a long echo range channel estimation, and wherein the estimation of short echo and the estimation of long echo are combined to get a combined channel estimation.

Advantageous Effects

Further, there is provided a method, wherein the short echo estimation is based on clean pseudo random sequence part of received signal and long echo range channel estimation is based on mixed PN and data part in the received signal. Especially, in the long echo range channel estimation the mixed PN and data part in the received signal is selected and passes through a series of especially programmable low-pass filters to suppress a data interference. Preferably, the low-pass filter outputs of the low-pass filters are used as decisions to generate errors for training coefficients of an adaptive FIR filter. The trained coefficients are the estimates of long echoes. These long echo estimates can be used to remove an ambiguity in the short echo estimates introduced by cyclic guardian interval by the step of combining the estimation of the short echo and the estimation of the long echo to get the combined channel estima- tion.

Especially, a selection of mixed pseudo random sequence and data part for long echo estimation is started after an end of a guardian interval and having duration of symbols less than the non-cyclic pseudo random sequence length.

A method passing mixed PN and data part to a series of low-pass filters to remove data interference can be used for guardian intervals with either variable or fixed PN phase. For variant phase guardian interval a shift according to the phase of current frame' s pseudo random sequence guardian interval is applied to a selected pseudo random sequence and data mixed part, and the shifted inputs are then passed to low-pass filters ensuring that each low-pass filter is processing the same pseudo random sequence symbol with data interference over frames.

Further, there is preferred a channel estimation device for echo estimation in communication systems, especially channel estimation device for long echo in communication systems exe- cuting such a method, comprising a short echo estimation block adapted for estimating a short echo, a long echo estimation block adapted for estimating a long echo by executing a long echo range channel estimation, the long echo being longer than the first short echo, and a combining block adapted for combin- ing the estimation of short echo and the estimation of long echo to get a combined channel estimation.

Especially, the short echo estimation block is adapted to execute estimation based on clean pseudo random sequence part of received signal, the echo estimation block is adapted to execute long echo range channel estimation based on mixed PN and data part in the received signal, wherein in the long echo range channel estimation a selector unit is adapted to select the mixed PN and data part in the received signal and to pass it through a series of especially programmable low-pass filters adapted to suppress a data interference.

Especially such device comprises an adaptive FIR filter and an error generator, adapted in such a way that the low-pass filter outputs of the low-pass filters are inputted and used as deci-

sions to generate errors for training coefficients of the adaptive FIR filter, wherein the trained coefficients are the estimates of long echoes.

Such method and device can be executed or arranged in a system that uses tome-domain cyclic pseudo random sequence as guardian interval between frames. Especially, such method and device can be executed or arranged in a TDS-OFDM system as the communication system.

Method and device makes possible a long echo channel estimation basing on time domain and utilizing the special guardian interval. The time domain methods using clean part of received PN are only able to estimate short echoes, especially short echoes around 30 μs for PN length of 420. This is due to the cyclic nature for guardian interval. In contrast, present long echo channel estimation method and device uses filters to process the contaminated PN part before estimation training. Thus it can provide long echo channel estimation, especially long echo channel estimation up to e.g. 48 μs for PN length of 420.

Description of Drawings

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

Fig. 1 a diagram of preferred aspects for channel estimation,

Fig. 2 data for training for short and long echo estimation,

Fig. 3 a diagram of preferred components for long echo estimation,

Fig. 4 combination of short and long echo estimates,

Fig. 5 shows a channel estimation block in the general OFDM system diagram,

Fig. 6 a TDS-OFDM frame structure for two different modes, and

Fig. 7 signals received through two different channels.

Mode for Invention

Fig. 1 shows a preferred device or a preferred arrangement of components being arranged to receive data and to execute a channel estimation. Especially, instead of hardware components there can be used a software algorithm. There are inputted received signal including data y(i) into such a channel estimator device 3. A number i of received data y(i) corresponds to data positions out of a sequence of received data y(i) for training to be used for channel estimation.

A preferred aspect to long echo channel estimation is to keep short echo estimation method and reuse it for long echo estimation with a correction method as shown in the diagram in Fig. 1 for the combined channel estimation.

There is inputted received signal including data y(i) and a frame head fh into a first step or block 31. First step or block 31 executes an estimation for short echoes and outputs data y(i), a frame head fh, and a channel estimate ce into a second step or block 32. Second step or block 32 executes an estimation for long echoes and a correction and outputs received signal or data y(i), a frame head fh, and a channel estimate ce .

The short echo estimation in first block 31 uses time-domain

correlation or other LMS training method based on clean part of PN in the received signal y(i) . The correction in second block 32 is responsible for estimating long echoes, especially for estimating long echoes more than 255-symbol delay, and for cor- recting the "folded" long echo in the results from short echo estimation in first block 31. In this way, the result of short echo estimation in first block 31 is reused as pre-correction estimation for long echoes.

If the long echo estimation yields none estimates, zero correction will be added to the short echo estimation results. This is the case when the echoes are within 255 symbols, especially around 33 μs . Otherwise, the long echoes detected in the long echo estimation in second block 32 are used to remove their "folded" counterparts in the results of short echo estimation in first block 31. This combination of short and long echo estimation provides a unified estimation performance for echoes within especially 0 - 48 μs delays. This means the position of the echo - especially either short or long - will not affect estimation performance.

The short echo estimation method could use existing time-domain correlation method or LMS training method based on the clean part of PN shown in shadowed part of Fig. 2.

For long echo estimation, it is preferred to use programmable low-pass filters 41 to eliminate the data interference on PN as shown in Fig. 3. Thereafter, LMS training method is applied to the filter outputs for long echo estimation. The long echo es- timation method uses mix part of PN and data to estimate long echo echoes shown as dotted part in Fig. 2 instead of the clean part of PN shown as shadowed part in Fig. 2. The long echo's contribution can be observed in this dotted window without ambiguity caused by short echo, which is shown in Fig. 7.

These training data are then fed to low pass filters in order to soften the data's interference. The number of low pass filters 41 is equal to the length of non-cyclic PN e.g. 255 for cyclic PN length of 420. Over a certain number of frames, the data part can be treated as random with mean value of zero while the PN part remains the same. So the output of low pass filters 41 will contain strong PN part with comparatively low data interference. Here the phase of PN in each frame can be assumed to change in a pre-defined way according to the stan- dard for digital terrestrial TV broadcasting in China. A changing PN phase means that there can be got the scenario of the whole PN by observing the dotted window in Fig. 2 over several frames. Based on the filter outputs, LMS method is applied for estimation of the long echoes. Here the LMS method is preferred over time-domain correlation method because LMS method offers better interference suppression in case that there is still data interference residue after low-pass filters.

The diagram of second block 32 for long echo channel estimation is shown in Fig. 3 in more detail. Mixed PN and data part in received signal is inputted into an adder 42 of one or more low-pass filters 41. Signal or data a outputted out of the adder 42 is forwarded to a delay element 43 having a delay value D. Output of delay element 43 is inputted for adding into a further input of the adder 42.

Furthermore, Signal or data a outputted out of the adder 42 and being PN with suppressed data is forwarded to an error generator 45 within a LMS training block 44. Error generator 45 esti- mates and outputs an error e, the error e being forwarded to an adaptive FIR filter 46 (FIR: Finite Impulse Response) . In addition, a local PN is inputted into the adaptive FIR filter 46. The adaptive FIR filter 46 outputs a filtered value to the generator 45 to be used for generating a following value of error e. In addition, the adaptive FIR filter 46 outputs long echo

estimates .

The training coefficients of the adaptive FIR filter 46 are the long echo estimates correspondingly. Once the long echo esti- mates are ready, the whole long range of channel estimates can be achieved by combining the short and long echo estimates and removing the "folded" image of long echo in the short echo estimates. A correction procedure is shown in Fig. 4. The channel 1 is taken as example to illustrate the correction.

The advantage of this combined channel estimation method is that it can provide a unified estimation performance for arbitrary echoes within a large propagation delay range of for example especially 0 - 48 μs delays for PN length of 420. This means the position of the echo will not affect the estimation accuracy as long as the echo is within the detection range of this combined method. For example, the estimation performance for a 0 dB echo at 48 μs or a 0 dB echo at 5 μs will be the same-. In addition, the LMS method used in short and long echo estimation ensures fast tracking performance under dynamic channel and noise suppression under low Signal to Noise Ratio (SNR) situations.