A study on time-domain approach for channel estimation Student: 王治傑 Advisor: Prof. 林大衛
Outline Introduction General LS-based estimation Frequency pilot time average method – FPTA-1 – FPTA-2 Most significant taps method Simulation
Introduction In OFDM systems, if complexity is affordable, time interpolation can be used to improve the performance. Most of the channel estimation approaches can be viewed as DFT-based approaches, where LS channel estimates are fed to IFFT block to get time domain channel estimates and transformed back to frequency domain by FFT.
General time domain and frequency domain channel estimation In usual time domain and frequency domain channel estimation, the pilot signal is P[k] with pilot ratio 1/K where K, m and i are integers A is pilot amplitude and M=N/K is an integer
General time domain and frequency domain channel estimation Suppose the channel has a discrete-time impulse response Hence, the received frequency domain pilot signal will be
General time domain and frequency domain channel estimation As for the received noise In this method, P[k] is PN-sequence, p[n] and P[k] are both noise-like. Hence the pilot-to-noise ratio in the time-domain and in the frequency are about the same.
Frequency pilot time average (FPTA) method The pilot signal is shown as: Hence
FPTA If the FPTA method is used for channel estimation with the pilot signal in the time domain, the pilot- to-noise ratio is Since there are K identical parts of time-domain pilot samples, the corresponding parts of received samples are averaged over K parts.
FPTA-1 Consider only the pilot affected by multipath channel and Gaussian noise
FPTA-1 Averaging the received signal over K parts Repeat the r avg K times to reconstruct the received samples y
FPTA-1 Hence, the channel frequency response at pilot tones can be estimated
FPTA-2 Use FFT{r avg } together with FFT{p o }, where p o is one of the identical parts of time-domain pilot samples.
Proposed time-domain method In practical situation, there are not so many channel paths with significant energy. Hence, many samples will have little or no energy at all except noise perturbation. Neglect those nonsignificant channel taps may introduce some performance degradation, but at the same time it will eliminate the noise perturbation.
Proposed time-domain method The time-domain received samples After averaging The raw LS channel estimation
Proposed time-domain method Just choose the most significant J taps Thus the frequency response estimate is
Proposed time-domain method Where Ignore the residual term and mse is
Simulation results
Reference H. Minn and V.K. Bhargava, “An investigation into time-domain approach for OFDM channel estimation,” IEEE Trans. on Broadcasting, vol.46 no.4, pp , Dec C.S. Yeh and Y. Lin, “Channel estimation using pilot tones in OFDM systems,” IEEE Trans. on Broadcasting, vol.45, no.4, pp , Dec 1999.