I am trying to perform practical diversity combining implemented in GNU Radio by using Ettus N210 USRPs, one as a transmitter and two as receivers.

To compare Equal Gain Combining (EGC) and Maximal Ratio Combining (MRC) techniques, I simply add the same received signal from two different receivers constructively for EGC and multiply each branch with a gain derived from square root of its signal-to-noise ratio (signal+noise/noise) before adding for MRC.

Contrary to the literature, for a few cases MRC is not performing better than EGC. I have noticed that when there is no signal, the noise level for both receivers is not same. In theory, it is assumed that all branches have equal mean power and noise level.

Now, I am confused. Should I attenuate/amplify the received signal (signal+noise) from each receiver to make the mean signal power same across both receivers or should I attenuate/amplify the received signal (signal+noise) in a way that noise level become same in the signal from both receivers before performing EGC.

Once I am done with EGC, should I simply multiply the weight at each branch to perform MRC?

  • $\begingroup$ You can't solve this by amplifying or attenuating the received signal, since that doesn't change the SNR. Try a simple test: get a BER with receiver A only, then put receiver B in the same physical location and find the BER with it. If they're not the same, you know your receivers are different. Your best bet at this time would be to modify the equations to account for different noise power at each receiver (or just fix your radio! :) $\endgroup$ – MBaz May 2 '16 at 19:04
  • $\begingroup$ Thanks for the answer. I have already tried the mentioned test and BER is not the same for both receivers. I am aware, amplifying or attenuating the received signal does not change the SNR but it does change the ratio of signal added for combining and there comes the asked problem. $\endgroup$ – Rok May 2 '16 at 20:03
  • 1
    $\begingroup$ I think that if you make the received signal power equal in both receivers, you'll get the theoretical diversity gain. However, the total noise power will be higher than it should be, so your BER plot will be shifted to the right. In other words, the BER plot should have the right slope, but the BER will be worse than it should be. $\endgroup$ – MBaz May 2 '16 at 20:25
  • $\begingroup$ I think you are right about the BER plot in such a case but then how would it be compared with other techniques of diversity in which the signal is not combined and the best one is selected e.g. selection diversity? $\endgroup$ – Rok May 3 '16 at 8:39
  • $\begingroup$ The BER difference between both receivers may be due to many reasons. I suggest to evaluate each receiver's performance in terms of noise figure. Maybe including that in your calculations will give you results closer to the theoretical ones. $\endgroup$ – vaz May 3 '16 at 9:09

Since, in theory, it is assumed that long term average SNR and noise remain same across all branches, first, I configured my USRPs to have same noise levels and then placed them in a way that average SNR across both branches also remain the same. Finally, simply combining them or first multiplying instantaneous SNR factor with each branch leads to EGC and MRC, respectively.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.