The first figure on the left (in this paper) provides the channel NMSE performance and the the second figure from left provides detector's BER performance.

I am curious, when the channel MSE overlapped in the first figure, why there is a difference between the BER performance.

I think, that there should be no difference, Moreover, even if there is a gap between MSE the use of quantizer of quantizer can bridge the gap in BER performance.


  • $\begingroup$ its a bit tough to answer this without access to the paper. Maybe you can share the figure and the system model? $\endgroup$ Jun 27 '18 at 18:53
  • $\begingroup$ @MaximilianMatthé please visit this link for detailed information. link. The question is about the figure 2b and 2c in this paper. THe model under consideration is a linear system given by eq. 6 $\endgroup$
    – azi
    Jun 28 '18 at 3:02
  • $\begingroup$ the link is not accessible, as it asks me for permission $\endgroup$ Jun 28 '18 at 6:30
  • $\begingroup$ @MaximilianMatthé please, use this link $\endgroup$
    – azi
    Jun 29 '18 at 3:54

From the simulation section of the paper:

The SER for each Monte Carlo simulation is calculated as follows: only when the activity state and the transmitted symbol are both detected correctly for a certain user, the corresponding user’s information is considered to be correct, and all symbols transmitted by the user whose activity state is mistakenly detected are considered to be incorrect. To provide a benchmark for performance comparison, we consider the oracle LS by assuming the actual active user support set known at the BS.

However, the SER performance gap of the proposed joint CE and MUD framework and the oracle LS always remains. This is mainly due to the CS-based sparse signal reconstruction algorithm with non-orthogonal training matrix suffers significantly, so that the data detection error is the main cost of user activity detection.

The non-oracle receiver has to figure out who is transmitting and hence can do additional errors. The oracle knows this information already. That's why you have the difference in the SER.

  • $\begingroup$ I think that the channel MSE also follows the same concept as BER. so if the MSE is converging, BER should converge too as MSE is also a function of the active users. $\endgroup$
    – azi
    Jul 2 '18 at 2:23

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.