Why does SER calculated for higher SNR becomes zero?

Hi, I am trying to study and understand this code (link below) for a deep learning NOMA system, but I'm really stuggling with getting the results out properly. The SER points calculated for user one goes to zero after a certain point of SNR, why does it not have non-zero calculated until SNR = 28 like for user two. In this research paper, the graph goes to 28 SNR for both users. Please help me in understanding this.

This is the link where I got the MATLAB code from:

https://www.mathworks.com/matlabcentral/fileexchange/75478-deep-learning-for-signal-detection-in-noma-systems

Answer 1

[vidyesh]

Hi Jay,

Based on your observations in the NOMA system simulation, the SER for user one (U1) dropping to zero before user two (U2) can be understood within the context of NOMA's design principles. In NOMA, power domain multiplexing is used to serve multiple users over the same frequency band, with different power levels allocated to each user.

Typically, U1, often referred to as the 'near user', receives a higher power allocation, resulting in better channel conditions compared to U2, the 'far user', who is allocated less power. As a result, at the same SNR, U1 experiences a lower SER due to the power advantage.

With increasing SNR, the ability of the system to decode U1's symbols correctly improves more rapidly than that for U2, leading to a quicker reduction in SER for U1. Hence, it is expected for U1's SER to approach zero at a lower SNR threshold compared to U2, reflecting the disparity in power allocation and channel conditions inherent to the NOMA protocol.

In high SNR scenarios, the calculated SER may approach zero due to the stochastic nature of noise, which can lead to a very low occurrence of errors. To address this, running the simulation with more samples or averaging the results over multiple iterations can provide a more accurate estimate of the SER

Hope this answer helps.