When the DC component is missing in DFT

On Thu, 05 Mar 2009 09:20:29 -0600, "m26k9"
<maduranga.liyanage@gmail.com> wrote:

>This is regarding OFDM and is a transmission through air. The circular
>convolution is intentional because then the FFT cant be treated as a
>multiplication of the channel and the data.
>
>Thank you.

Most wireless OFDM systems (e.g., 802.11a/g/n) don't use the DC
subcarrier due to the potential for LO bleed-through in the direct
downconversion circuit.  What are you trying to recover from the DC
term?   As rb-j pointed out, you can reconstruct the time-domain
signal without the DC term, if that's acceptable for your purposes.

Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.ericjacobsen.org

Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php

Thank you very much for the replies.
Yes this is with cyclic prefix for circular convolution.
What I am actually trying to do is to estimate the impulse response of the
channel. My original idea was to measure the frequency-response and then
IFFT it back to get the channel response. At this moment I can obtain every
bin except for the DC which is infinity due to a division by zero. I am
using two identical transmission signals in the channel estimation but one
is made from re-arranging the samples of the first signal. So the DC
component of both these signals are the same. There is a (signal1-signal2)
term which gives a zero for the DC term, so I am unable to use that value
for the evaluation of the DC bin value.

But may be I can try to find the average  and use it directly.
If you guys know any FFT tricks please let me know. I'll post back if this
worked out.

Thank you very much.