Four antennas OFDM receiver. How sum "in phase"

Hi guys,


I have an OFDM signal (with DBPSK) received using four differents antennas
(four unknown delays then). 

how i can sum the four OFDM signals "in phase"?  in others words,  how i
can estimate the phase compensation in each "bin" for each signal
delayed?

many thanks
---------------------------------------
Posted through http://www.DSPRelated.com

JAlbertoDJ <12823@DSPRelated> wrote:

>I have an OFDM signal (with DBPSK) received using four differents antennas
>(four unknown delays then). 

>how i can sum the four OFDM signals "in phase"?  in others words,  how i
>can estimate the phase compensation in each "bin" for each signal
>delayed?

Not enough information.  Useful will be the following:

Do you have any control over the signal design (preambles, for
example) or is the signal format pre-defined?

Where in the receiver signal chain do you hope to combine these
signals, i.e. RF, IF, baseband, after demodulation, etc.?
Do you have control over this as well?

In addition to phase differences, are there also magnitude differences
among the four signals?  If the signals are of wildly different
magnitudes, a simple approach is to select the strongest signal
of the four and just use that.  This is called "selection diversity".
If they are of uneven magnitude, but still all four usable,
you want to use Maximal Ratio Combining.  If so, this is often
best done right before FEC decoding.


Steve

Yes, the OFDM signal is a short block predefinied of 70 OFDM symbols,
using DBPSK modulation, and with some "differencial pilot phases"
inserted. I recieve these four blocks in baseband.

Maximal Radio Combining is not a problem. What i need first, is how to get
the 4 signals "in phase". After i can calculate the S/N to do a MRC.

thanks

---------------------------------------
Posted through http://www.DSPRelated.com

JAlbertoDJ <12823@DSPRelated> wrote:

>Yes, the OFDM signal is a short block predefinied of 70 OFDM symbols,
>using DBPSK modulation, and with some "differencial pilot phases"
>inserted. I recieve these four blocks in baseband.
>
>Maximal Radio Combining is not a problem. What i need first, is how to get
>the 4 signals "in phase". After i can calculate the S/N to do a MRC.
>
>thanks

Well my first thought is that since this is DBPSK, maybe you
don't need to determine the abolute phase of each signal,
since after differential combining of consecutive bauds,
they will then be back in phase.

(But that might be sub-optimal.)

Steve

>JAlbertoDJ <12823@DSPRelated> wrote:
>
>>Yes, the OFDM signal is a short block predefinied of 70 OFDM symbols,
>>using DBPSK modulation, and with some "differencial pilot phases"
>>inserted. I recieve these four blocks in baseband.
>>
>>Maximal Radio Combining is not a problem. What i need first, is how to
get
>>the 4 signals "in phase". After i can calculate the S/N to do a MRC.
>>
>>thanks
>
>Well my first thought is that since this is DBPSK, maybe you
>don't need to determine the abolute phase of each signal,
>since after differential combining of consecutive bauds,
>they will then be back in phase.
>
>(But that might be sub-optimal.)
>
>Steve


Yes, I tried with the mean phase difference for the four blocks:
     
           Diff = (Diff1 + Diff2 + Diff3 + Diff4)/4

  I should to get 6dB gain, but only i get +4.5dB with this solution

 So, i think first it's neccesary to sum the complex numbers in phase,
before  calculating the phase difference.


---------------------------------------
Posted through http://www.DSPRelated.com

JAlbertoDJ <12823@DSPRelated> replies to my post

>>Well my first thought is that since this is DBPSK, maybe you
>>don't need to determine the abolute phase of each signal,
>>since after differential combining of consecutive bauds,
>>they will then be back in phase.

>>(But that might be sub-optimal.)

>Yes, I tried with the mean phase difference for the four blocks:
>     
>           Diff = (Diff1 + Diff2 + Diff3 + Diff4)/4
>
>  I should to get 6dB gain, but only i get +4.5dB with this solution

>So, i think first it's neccesary to sum the complex numbers in phase,
>before  calculating the phase difference.

This jibes with what I would intuitively expect.

I would think you could pick up most of that 1.5 dB loss with
a relatively crude phase-estimator.

Steve

On Tue, 15 Sep 2015 17:11:49 -0500, "JAlbertoDJ" <12823@DSPRelated>
wrote:

>>JAlbertoDJ <12823@DSPRelated> wrote:
>>
>>>Yes, the OFDM signal is a short block predefinied of 70 OFDM symbols,
>>>using DBPSK modulation, and with some "differencial pilot phases"
>>>inserted. I recieve these four blocks in baseband.
>>>
>>>Maximal Radio Combining is not a problem. What i need first, is how to
>get
>>>the 4 signals "in phase". After i can calculate the S/N to do a MRC.
>>>
>>>thanks
>>
>>Well my first thought is that since this is DBPSK, maybe you
>>don't need to determine the abolute phase of each signal,
>>since after differential combining of consecutive bauds,
>>they will then be back in phase.
>>
>>(But that might be sub-optimal.)
>>
>>Steve
>
>
>Yes, I tried with the mean phase difference for the four blocks:
>     
>           Diff = (Diff1 + Diff2 + Diff3 + Diff4)/4

>  I should to get 6dB gain, but only i get +4.5dB with this solution
>
> So, i think first it's neccesary to sum the complex numbers in phase,
>before  calculating the phase difference.

If you want to do some sort of traditional diversity combining, like
MRC or Equal Gain Combining, you need to fully synchronize and channel
equalize each stream separately before combining.   In an OFDM system
this adds a lot of complexity.


Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com