INTERPOLATION METHOD AND APPARATUS USING TRACKING FILTER

IN MULTI-CARRIER RECEIVER

FIELD OF THE INVENTION

The present invention relates generally to communication technology, and particularly to an interpolation method and apparatus using tracking filter in multi-carrier receiver.

DESCRIPTION OF RELATED ART

In wireless communication environment, signals suffer time and frequency domain fading. To solve this problem, OFDM technology is one solution for high rate wireless communication. OFDM partitions the overall system bandwidth into a number of orthogonal subbands. Each subband may be viewed as an independent transmission channel that may be used to transmit data. With OFDM, each subband is associated with a respective subcarrier upon which data may be modulated.

In a wireless communication system, an RF modulated signal from a transmitter may reach a receiver via a number of different propagation paths. For an OFDM system, the subbands may experience different effective channels due to different effects of fading and multipath and may consequently be associated with different complex channel gains.

An accurate estimation of the response of the wireless channel between the transmitter and the receiver is normally needed in order to transmit data on the available subbands. Channel estimation is one key technology in OFDM system, which can estimate the channel information and compensate the multi-path fading on the data. In pilot-aided OFDM-based systems, such as DVB-T/H, channel estimation is usually carried out by using interpolating technology, such as Linear, Spline, Gaussian, wiener interpolator and so on, to obtain the channel state information (CSI) of all subcarriers. Some pilot subcarriers are sent from the transmitter and the pilot subcarriers are measured at the receiver. Since the pilot is made up of symbols that are known a prior by the receiver, the channel response can be estimated as the ratio of

the received pilot symbol over the transmitted pilot symbol for each subband used for pilot subcarrier transmission.

Fig.l shows a conventional channel estimation system structure 100 in the DVB-T/H system. And Fig.2 shows the status of original pilots, pilots after time domain interpolation, and pilots after time domain interpolation and frequency domain interpolation during the channel estimation process.

As shown in Fig.1 , the received data from an OFDM transmitter is received by a pre-process module 10, in which the received data at a scattered pilot (SP) location is divided by the corresponding values of local SPs, and then the CSI of SP is obtained in pre-process module. Secondly, the CSI of every three subcarriers can be obtained by the interpolation in the time domain at a time interpolator module 11. Finally, the interpolation in frequency domain is processed with the CSI of the 3 interval subcarriers at a Wiener interpolator 12. Thus the CSI of all subcarriers in one OFDM symbol are obtained. The conventional Wiener filter used for interpolation in frequency domain has the maximum bandwidth of Tu/6 (where Tu indicates useful duration of an OFDM symbol). On the consequence, the maximum delay time of the echo that the conventional interpolating method can deal with is the Tu/6.

However, in the NorDig Test Specification, the delay time that a demodulator should deal with must exceed Tu/6 and the maximum delay time of the echo is Tu/ (24/7) which is about Tu/3. Furthermore, there are the system performance requirements for the long echo issue. On these cases, the conventional interpolating methods are not robust, so the new interpolating method is needed to meet NorDig Test Specifications requirement.

SUMMARY OF THE INVENTION

In an aspect, a method for performing channel estimation in a receiver is provided. The method comprises steps of receiving orthogonal frequency division multiplexed (OFDM) signals including pilot subcarriers and data subcarriers; and performing frequency domain interpolation by using a tracking filter whose bandwidth and center position can change with the maximum difference among the

transmission time of multi-path channel from the transmitter to the receiver.

In another method, a receiver for receiving OFDM (Orthogonal Frequency Division Multiplexed) signals is described, wherein the receiver includes a frequency interpolation module, which includes a tracking filter whose bandwidth and center position can change with the maximum difference among the transmission time of multi-path channel.

In view of the above, and as will be apparent from reading the detailed description, other embodiments and features are also possible and fall within the principles of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG.l shows a conventional channel estimation system structure in the DVB-T/H system;

FIG.2 shows the status of original pilots, pilots after time domain interpolation, and pilots after time domain interpolation and frequency domain interpolation during the traditional channel estimation process;

FIG.3 shows a channel estimation module of a present embodiment;

FIG.4 shows the channel status information correction module in the channel estimation module of the present embodiment in FIG.3;

FIG.5 shows a process performed by the channel estimation module of the present embodiment in FIG.3; and

FIG.6 shows a MMSE frequency interpolation method used in the tracking filter of the channel estimation module in FIG.3.

DETAILED DESCRIPTION

To meet NorDig requirements, in the embodiments hereafter, a Minimum Mean Squared Error (MMSE) interpolating method using tracking filter in frequency domain is proposed. The position and the bandwidth of the tracking filter are changeable with the maximum delay time of the echo T _max , so it is called "tracking filter". With this tracking filter and the MMSE interpolation, the maximum delay time

of Tu/3 and the system performance requirements in NorDig Test specifications can be achieved.

Now look at Fig.3 which shows a channel estimation module 300 of an embodiment. The channel estimation module 300 includes a pre-processing module 30 in which the received data at a scattered pilot (SP) location is divided by the corresponding values of local SPs, and then the CSI of SP is obtained in pre-process module, a time interpolator module 31 which is used to achieve the time domain interpolation. After the time domain interpolation, the virtual pilots at the positions of every three subcarriers (which is the interval between neighboring pilots) are obtained. In the embodiment, a tracking filter module 32 is used to achieve the frequency domain interpolation. After the time interpolation and the frequency interpolation, the channel state information (CSI) of all useful data subcarriers is obtained. Meanwhile a CSI correction module 33 is used to correct the CSI obtained by the tracking filter module 32 with the virtual pilots output from the time interpolator module 31. And a data FIFO module 34 is used to buffer or delay the arrival of a data when performing the channel estimation using the previous pilot.

Because of the ripple of the tracking filter, the amplitudes of the pilot signals through the tracking filter 32 are reduced. And if the amplitudes of the pilot signals are reduced, the power of the pilot signals will be reduced, and thus the signal to noise ratio will be degraded and the system performance will be worse. To solve this problem, the amplitudes of channel impulse response need to be corrected by CSI correction module 33 using the pre-tracking-filter values, i.e. the virtual pilots.

As shown in Fig.4, an embodiment of the CSI correction module 33 is introduced. The CSI correction module 33 includes a scattered pilot (SP) extraction module 40, average processing modules 41, 43, absolute (ABS) processing modules 42, 44, a divider module 45, and a multiplier 46. The CSI output from the filter tracking module 32 is entered into the SP extractor 40 first, in which the CSI on the positions of the SPs are extracted. Then, the CSI on the positions of the every three subcarriers are averaged in the average processing module 41 and processed for getting the absolute value in ABS processing module 42. In addition, the CSI of

virtual pilots output from the time interpolation module 31 are averaged in the average processing module 43 and processed for getting the absolute value in the ABS processing module 44 too. The result output from ABS processing modules 44 divides the result output from ABS processing modules 42 at divider 45. And the dividing results from the divider 45 are used as a correct factor to correct the data CSI outputs from the tracking filter 32 in a multiplier 46. After the CSI correction, the corrected-CSI is sent out.

Fig.5 shows the process performed by the channel estimation module 300. The process starts in step 50. In step 51, the received data from an OFDM transmitter is received and is divided by the corresponding values of local SPs, and then the CSI of SP is obtained. In step 52, a time domain interpolation is performed on the pre-processed signals to get the CSI of virtual pilots at the positions of SPs. The time domain interpolation result is used to perform a frequency domain interpolation in step 53 to estimate the channel state information of all useful subcarriers. And the estimated channel state information gotten in step 53 is used together with the CSI of virtual pilots gotten in step 52 to get the corrected channel state information. The process ends at step 55.

Hereafter, the detailed process of the MMSE interpolation method used by the tracking filter 32 is introduced.

In an embodiment, see equation (1), the shift distance P _shlfl of the tracking filter showing the center position of the filter is in direct proportion to the maximum delay time of echo T _max , i.e. the maximum transmission time difference among different paths from OFDM transmitter to the receiver; the interval of the neighboring scattered pilots D _j , and the frequency difference δ/ between data positions. The relation

between the bandwidth B _flller of the tracking filter 32 and the shift distance P _shlfl of the tracking filter is shown in equation (2).

Bmr = y(« ^■ T^ ^■ Dj ^■ Af) (2)

In equations (1) and (2), a is a conversion factor and is an experiential value. In an optimal embodiment a is 0.38. And the interval of the neighboring scattered pilots is 3. To satisfy the Nyquist-Shannon sampling theorem, the B _βller can be

deduced easily. To track the echo position, the P _shlfl of the tracking filter changes

with B _ftller .

When designing the tracking filter, a MMSE criterion is used. Assume in an OFDM system, there are N subcarriers with Np pilots. The space of the subcarriers L=NZNp is selected to be an integer. The pilot subcarrier indices are given by n=rL, where r=0, 1, 2... Np-I . The estimated channel response at the pilot subcarrier locations for a given OFDM block are given by p = [H[rL] \ 0 ≤ r ≤ Np-\] ^τ , (3)

where H[rL] are the estimated channel responses at pilot locations. The estimated channel responses at all subcarrier locations can be represented by h = [H[n] \ 0 ≤ n ≤ N-l] ^τ . (4)

When using the traditional MMSE method, the estimated channel responses h at all subcarrier locations can be obtained by the convolution results between p (results output from time interpolation) and coefficients R^R^ . It is given by

h = R _hp -R _rp - ^] p , (5) where R _hp is the cross-covariance matrix between h and p . R^ is the

autocovariance matrix of p .

R _m = E{pp ^H } = R _nn + -^—I (6)

where R _pp is the auto-covariance matrix of p, and / is a unit matrix.

If we consider the channel response as a rectangle window, the R _h ~ and R~

are SINC(I I B _flller - I) , I is the order of the filter. Of course, the channel response can be set as other window functions.

To solve the Tu/3 echo issue, the shift distance of the filter P _shlβ is added to the

original cross-covariance and autocovariance matrix. That is R _h ~ and R^ are

changed to R _hp ~ - P _shlfl and % • -% respectively.

Now turn to Fig.6 which describes the above MMSE frequency interpolation method. In step 61 , the process starts. Then in step 62, the bandwidth of the tracking filter B _JUler is calculated according to T _max , β, D _f and δ/ . In step 63, the

cross-covariance matrix R _h ~ and autocovariance matrix R _j ^ are calculated. And then the coefficients of the tracking filter are calculated.

And according to the performance requirements, we can set a proper order / for the filter, such as 12, 13, or 14, the / can be odd number or even number.

In an example design, a tracking filter with 34 order and 12 coefficients are used. When the T _max equals to 260us, the starting frequency is -0.0105pi and the cut-off frequency is 1.8545pi, so the 3dB bandwidth is 1.865pi (corresponding to about 278us) and the center position is 0.922pi.

For the long echo test modes complying with NorDig Test Specification, the results are as Table 1.

Table 1 Performance under NorDig Test Specifications