about MRC techniques for OFDM

Hello all,

Please, I wanna know more about maximal ratio combining post DFT
techniques. I'm studing this receving method, also i found a chipset
improving this technique:

http://www.teamcast.com/en/maj-e/c2a2i15040/products/dibcom-products/dvb-t-diversity-demodulation-chipset.htm

but i do not understand the theory. The combining is frequency
domain ? Someone can help me ?

Best regards

On Apr 13, 4:30 am, odysse...@tin.it wrote:
> Hello all,
>
> Please, I wanna know more about maximal ratio combining post DFT
> techniques. I'm studing this receving method, also i found a chipset
> improving this technique:
>
> http://www.teamcast.com/en/maj-e/c2a2i15040/products/dibcom-products/...
>
> but i do not understand the theory. The combining is frequency
> domain ? Someone can help me ?
>
> Best regards

If you want to just understand maximal ratio combining, any of the
standard communications textbooks will explain it in good detail. See,
for example,
Proakis: Digital Communications.
That theory should carry over to frequency-domain combining too.

Otherwise, your question is too generic. Please ask a more specific
question
on whatever it is you do not understand.

Thanks,
Dilip.

> Otherwise, your question is too generic. Please ask a more specific
> question
> on whatever it is you do not understand.

sorry i was too generic.
Digital Communications is a good book for MRC theory but i wanna know
more about MRC for OFDM modulation. In particular I am interested in
understanding where the combining is done (pre or post DFT).
In other words it's better combing the OFDM symbol in time domain or
combing the symbol which modulate the single sub-carriers (i.e in
frequency domain) ?

odyssey25@tin.it wrote:

>>Otherwise, your question is too generic. Please ask a more specific
>>question
>>on whatever it is you do not understand.
> 
> 
> sorry i was too generic.

> Digital Communications is a good book for MRC theory but i wanna know
> more about MRC for OFDM modulation. In particular I am interested in
> understanding where the combining is done (pre or post DFT).

Of course post FFT is better, since you can compute MRC separately for 
each subcarririer.

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

> Of course post FFT is better, since you can compute MRC separately for
> each subcarririer.

Ok.
Now, let us consider a system with two antennas. For these two
antennas, i think, the chipest has two different " RF front ends +
COFMD demodulator" . Some more, for each sub-carrier the signals from
different antennas are linearly combined. Now, a weight factor (with
zero phase) is associated with the k-th subcarrier. Well, i do not
understand how can be possible calculate these weight factors, if they
are equal for all the OFDM subcarriers of the same received signal or
they are different for each sub-carrier (in the same receive signal).
I hope I made myself clear

odyssey25@tin.it wrote:

>>Of course post FFT is better, since you can compute MRC separately for
>>each subcarririer.
> 
> 
> Ok.

What OK? Do you think you did me a favor?

> Now, let us consider a system with two antennas. For these two
> antennas, i think, the chipest has two different " RF front ends +
> COFMD demodulator" . Some more, for each sub-carrier the signals from
> different antennas are linearly combined. Now, a weight factor (with
> zero phase) is associated with the k-th subcarrier. Well, i do not
> understand how can be possible calculate these weight factors, if they
> are equal for all the OFDM subcarriers of the same received signal or
> they are different for each sub-carrier (in the same receive signal).
> I hope I made myself clear

OMFG what a clueless retard. Go read your textbooks, idiot.

VLV

On Apr 13, 1:46�pm, odysse...@tin.it wrote:
> > Of course post FFT is better, since you can compute MRC separately for
> > each subcarririer.
>
> Ok.
> Now, let us consider a system with two antennas. For these two
> antennas, i think, the chipest has two different " RF front ends +
> COFMD demodulator" . Some more, for each sub-carrier the signals from
> different antennas are linearly combined. Now, a weight factor (with
> zero phase) is associated with the k-th subcarrier. Well, i do not
> understand how can be possible calculate these weight factors, if they
> are equal for all the OFDM subcarriers of the same received signal or
> they are different for each sub-carrier (in the same receive signal).
> I hope I made myself clear

It's not completely clear but the gist seems to be that you have a 2-
antenna receiver. The signals from the two antennas are combined pre-
FFT, then the FFT is taken and the weights are applied post-FFT to the
combined signal. In this case, you can calculate the weights that need
to be applied by a training method based on transmitting and receiving
a known sequence. Yes, the weights will be different for each
subcarrier. This is also sometimes referred to as a frequency-domain
equalizer (FEQ/FDQ) as opposed to a time-domain equalizer (TEQ/TDQ).

Thanks,
Dilip.

On Mon, 13 Apr 2009 10:46:43 -0700 (PDT), odyssey25@tin.it wrote:

>> Of course post FFT is better, since you can compute MRC separately for
>> each subcarririer.
>
>Ok.
>Now, let us consider a system with two antennas. For these two
>antennas, i think, the chipest has two different " RF front ends +
>COFMD demodulator" . Some more, for each sub-carrier the signals from
>different antennas are linearly combined. Now, a weight factor (with
>zero phase) is associated with the k-th subcarrier. Well, i do not
>understand how can be possible calculate these weight factors, if they
>are equal for all the OFDM subcarriers of the same received signal or
>they are different for each sub-carrier (in the same receive signal).
>I hope I made myself clear
>

Not clear.

You seemed to be saying that you combined them first and then applied
weighting.

Are you talking about the EQ weights or the MRC weights?

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

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

On Apr 14, 7:39&#4294967295;am, Dilip Warrier <dili...@yahoo.com> wrote:
> On Apr 13, 1:46&#4294967295;pm, odysse...@tin.it wrote:
>
> > > Of course post FFT is better, since you can compute MRC separately for
> > > each subcarririer.
>
> > Ok.
> > Now, let us consider a system with two antennas. For these two
> > antennas, i think, the chipest has two different " RF front ends +
> > COFMD demodulator" . Some more, for each sub-carrier the signals from
> > different antennas are linearly combined. Now, a weight factor (with
> > zero phase) is associated with the k-th subcarrier. Well, i do not
> > understand how can be possible calculate these weight factors, if they
> > are equal for all the OFDM subcarriers of the same received signal or
> > they are different for each sub-carrier (in the same receive signal).
> > I hope I made myself clear
>
> It's not completely clear but the gist seems to be that you have a 2-
> antenna receiver. The signals from the two antennas are combined pre-
> FFT, then the FFT is taken and the weights are applied post-FFT to the
> combined signal. In this case, you can calculate the weights that need
> to be applied by a training method based on transmitting and receiving
> a known sequence. Yes, the weights will be different for each
> subcarrier.

Once the signals are combined, you can not apply different weights to
them. You need to apply different weights to
different signals before combining. However, if you want to combine
them pre-fft, as the training signal is in frequency domain, you would
need
to estimate the frequency response first and then ifft it to get the
response in time domain. If the frequency of the two antennas are not
perfectly
matched, things can get very ugly. I would say it is not a good
approach to combine them pre-fft.

>This is also sometimes referred to as a frequency-domain
> equalizer (FEQ/FDQ) as opposed to a time-domain equalizer (TEQ/TDQ).
>

This has nothing to do with frequency or time domain equalization.

> Thanks,
> Dilip.