Forums

Equalization: FDE or TDE? SC-FDE/OFDM?

Started by wong September 30, 2003
I have some questions....
Given the elegance of FDE, why has FDE lagged behind in use compared
to TDE? Also, why is OFDM so popular when SC-FDE is so much better?
(No ICI, no peak to average issues)
wongchichian@hotmail.com (wong) wrote in message news:<34a4fe15.0309300431.5bb9a020@posting.google.com>...
> I have some questions.... > Given the elegance of FDE, why has FDE lagged behind in use compared > to TDE? Also, why is OFDM so popular when SC-FDE is so much better? > (No ICI, no peak to average issues)
I assume TDE is time domain equalization,FDE is frequency domain .. SC-FDE single carrier ...etc. If I follow your terminology correctly then here are my starting comments and it can grow depending on your feedback. The comparision is not very clear. I think choice of FDE and TDE depends on your application as well. I would also like to know what "elegance" one finds in FDE before commenting here-so please clarify? Since we work in GSM/EDGE receiver design we prefer TDE - average number of channel taps are very small(<10) hence it does not have long impulse reponse tail - computation complexity is not high, channel is also estimated using received samples which are readily available in time domain from receiver hardware - so no extra cost to fourier transform. My understanding is that if a CIR has a very long tail and TDE costs high computationally FDE approach would reduce the complexity. SC-FDE and OFDM are no comparision!! SC-FDE has to do very robust processing for ISI,ICI if the channel is corrupted by frequency selective fading. OFDM subcarrier model is much better against frequency selective fading. Since sufficient guard interval is provided in each OFDM symbol it is also very robust against ISI. Simpler OFDM receiver(DAB/DVB) can manage with one tap FDE. Question naturally comes whether a SC-FDE is possible at all when it required to carry such a huge bandwidth(54Mbps for 802.11a) through a purturbed channel? At least, I can not think of such SC-FDE e.g MCS9(highest data rate channel in EDGE system - that too can provide 384kbps theoretically) performance is much poorer than any other low data rate channel for the same receiver. These receivers are normally designed very robust on industrial norm. More on your comments next time ... Santosh
Hi...

For tde, we need to do adaptation first, and for most of the algos
used, like LMS, RLS etc, we need to devote so and so much time for the
adaptive algo to converge. However, in FDE, with cyclic prefixes we
just need to do least squares estimation, which is really simple...
maybe you are right, for certain channels, tde may be good enough.

sc-fde and ofdm are pretty  similar... in both cases we use a one tap
equalizer to do equalization and both uses cyclic prefixes.
furthermore, in the presence of multipath, uncoded ofdm as it is is
not usable as certain carriers are subjected to frequency fades. This
does not happen in sc-fde, since the fades are 'smudged out'. Unless
we do coding then their performance is similar. Also sc-fde does not
suffer from ici as there is only 1 carrier... and hence is more
robust. also, there is no peak to average issue, which usually means
more expensive oscillators and power amplifiers...

thanks...

So how about adaptive loading and the like? 
santosh.nath@ntlworld.com (santosh nath) wrote in message news:<6afd943a.0309301325.6c16cd30@posting.google.com>...
> wongchichian@hotmail.com (wong) wrote in message news:<34a4fe15.0309300431.5bb9a020@posting.google.com>... > > I have some questions.... > > Given the elegance of FDE, why has FDE lagged behind in use compared > > to TDE? Also, why is OFDM so popular when SC-FDE is so much better? > > (No ICI, no peak to average issues) > > I assume TDE is time domain equalization,FDE is frequency domain .. > SC-FDE single carrier ...etc. If I follow your terminology correctly > then here are my starting comments and it can grow depending on your > feedback. > > The comparision is not very clear. I think choice of FDE and TDE > depends on your application as well. I would also like to know what > "elegance" one finds > in FDE before commenting here-so please clarify? Since we work in > GSM/EDGE receiver design we prefer TDE - average number of channel > taps are very > small(<10) hence it does not have long impulse reponse tail - > computation complexity is not high, channel is also estimated using > received samples which > are readily available in time domain from receiver hardware - so no > extra cost to fourier transform. My understanding is that if a CIR has > a very long tail > and TDE costs high computationally FDE approach would reduce the > complexity. > > SC-FDE and OFDM are no comparision!! SC-FDE has to do very robust > processing for > ISI,ICI if the channel is corrupted by frequency selective fading. > OFDM subcarrier model is much better against frequency selective > fading. Since sufficient guard interval is provided in each OFDM > symbol it is also very robust against ISI. Simpler OFDM > receiver(DAB/DVB) can manage with one tap FDE. Question naturally > comes whether a SC-FDE is possible at all when it required to carry > such a huge bandwidth(54Mbps for 802.11a) through a purturbed channel? > At least, I can not think of such SC-FDE e.g MCS9(highest data rate > channel in EDGE system - that too can provide 384kbps theoretically) > performance is much poorer than any other low data rate channel for > the same receiver. These receivers are normally designed very robust > on industrial norm. > > More on your comments next time ... > Santosh
wongchichian@hotmail.com (wong) wrote in message news:<34a4fe15.0309301823.38fa38b9@posting.google.com>...
> Hi... > > For tde, we need to do adaptation first, and for most of the algos > used, like LMS, RLS etc, we need to devote so and so much time for the > adaptive algo to converge. However, in FDE, with cyclic prefixes we > just need to do least squares estimation,
Hi .. It is not a must to do adaptation for TDE. I guess I have given example for GSM/EDGE where each burst is provided with training sequences hence channel estimation does not strickly require adaptation. For BTS side inclusion of LMS(Block) later on does not improve much(may be 0.2 0.3dB). For MS side inclusion of LMS(Block) can give some gain(<0.5dB) so it is an option for performance. Least Square estimate does not perform well at low SNR so people go for LMMSE estimate(one such example is SVD channel estimate based on LMMSE) which is really simple...
> maybe you are right, for certain channels, tde may be good enough. > > sc-fde and ofdm are pretty similar... in both cases we use a one tap > equalizer to do equalization and both uses cyclic prefixes. > furthermore, in the presence of multipath, uncoded ofdm as it is is > not usable as certain carriers are subjected to frequency fades. This > does not happen in sc-fde, since the fades are 'smudged out'.
I could not follow you here. If it is SC-FDE it is likely to purturbed by frequency selective fading more, simple equalization may not be enough to mitigate it. In case of OFDM, each subcarrier constitutes nearly a flat fading profile hence on the whole frequency selective fading partially affects the OFDM symbol/spectra but not the entire symbol/spectra. So OFDM is more robust against frequency selective fading. Unless
> we do coding then their performance is similar.
I doubt a coded ODFM and coded SC-FDE are similar in the following case: -Bandwidth is very high(54Mbps) -severe frequency selective fading Could you provide a reference to check it!! Regards, Santosh Also sc-fde does not
> suffer from ici as there is only 1 carrier... and hence is more > robust. also, there is no peak to average issue, which usually means > more expensive oscillators and power amplifiers... > > thanks... > > So how about adaptive loading and the like? > santosh.nath@ntlworld.com (santosh nath) wrote in message news:<6afd943a.0309301325.6c16cd30@posting.google.com>... > > wongchichian@hotmail.com (wong) wrote in message news:<34a4fe15.0309300431.5bb9a020@posting.google.com>... > > > I have some questions.... > > > Given the elegance of FDE, why has FDE lagged behind in use compared > > > to TDE? Also, why is OFDM so popular when SC-FDE is so much better? > > > (No ICI, no peak to average issues) > > > > I assume TDE is time domain equalization,FDE is frequency domain .. > > SC-FDE single carrier ...etc. If I follow your terminology correctly > > then here are my starting comments and it can grow depending on your > > feedback. > > > > The comparision is not very clear. I think choice of FDE and TDE > > depends on your application as well. I would also like to know what > > "elegance" one finds > > in FDE before commenting here-so please clarify? Since we work in > > GSM/EDGE receiver design we prefer TDE - average number of channel > > taps are very > > small(<10) hence it does not have long impulse reponse tail - > > computation complexity is not high, channel is also estimated using > > received samples which > > are readily available in time domain from receiver hardware - so no > > extra cost to fourier transform. My understanding is that if a CIR has > > a very long tail > > and TDE costs high computationally FDE approach would reduce the > > complexity. > > > > SC-FDE and OFDM are no comparision!! SC-FDE has to do very robust > > processing for > > ISI,ICI if the channel is corrupted by frequency selective fading. > > OFDM subcarrier model is much better against frequency selective > > fading. Since sufficient guard interval is provided in each OFDM > > symbol it is also very robust against ISI. Simpler OFDM > > receiver(DAB/DVB) can manage with one tap FDE. Question naturally > > comes whether a SC-FDE is possible at all when it required to carry > > such a huge bandwidth(54Mbps for 802.11a) through a purturbed channel? > > At least, I can not think of such SC-FDE e.g MCS9(highest data rate > > channel in EDGE system - that too can provide 384kbps theoretically) > > performance is much poorer than any other low data rate channel for > > the same receiver. These receivers are normally designed very robust > > on industrial norm. > > > > More on your comments next time ... > > Santosh
no trg needed? do u have any references where i can take a look?
"santosh nath" <santosh.nath@ntlworld.com> wrote in message
news:6afd943a.0310010236.6dde9080@posting.google.com...



> So OFDM is more robust against frequency selective fading.
Santosh, Thank you for your easy and kind explanation. I add more reason in optimality point of view. OFDM is more optimal approach than SC-FDE because both pre-(W_Tx) and post-processes (W_Rx) are necessary to achieve Shannon capacity. More specifically, transmitter IFFT and receiver FFT are fortunately such optimal pre- and post-processes in ISI channel. With only post-process, we can not avoid the noise enhancement in the receiver; thus, SC-FDE would offer obviously a performance degradation because transmitter in SC-FDE system never exploits any pre-process.
> Regards,
> Santosh
> wongchichian@hotmail.com (wong) wrote in message
> news:<34a4fe15.0309301823.38fa38b9@posting.google.com>...
>> So how about adaptive loading and the like?
Furthermore, we should keep in mind that this optimality is still true in case of considering the power allocation. You can be helped with an example for this at R1-031142 by ftp of www.3ggp.org (/WG1_RL1/TSGR1_34/Docs/Zips/) (http://list.etsi.org/scripts/wa.exe?A2=ind0310&L=3gpp_tsg_ran_wg1&T=0&F=&S= &P=8229) In the figure-1 of R1-031142, the given example case, similar to 4 equal power paths CIR, shows that the SNR loss of receiver only processing is 6dB from that of both processing. However, this document also shows another possibility concerning about this problem. -- Best regards, James K. (txdiversity@hotmail.com) - Private opinions: These are not the opinions from my affiliation.