MLSE Equalizer Traceback depth and BER

I've modeled a simple communication channel with Rayleigh fading on 3 different patches and AWGN Noise on Simulink

I'm using the MLSE Equalizer with channel coefficients given by the Rayleigh SISO block and a Traceback depth of 20. I generate 100kb/s in frame of 100bit each via a Bernoulli block and send it to the channel with a QPSK modulatione.

At the reciever side, with the parameters written below, the BER is around 0.03. If I change the Traceback depth of any value different from 20, the BER explode always at around 0.5 so, I can't understand why specifically the MLSE equalizer works with Traceback depth 20 and doesn't work with another value

Answer 1

[vidyesh]

Hello Nicola,

The effectiveness of Traceback depth 20 in your MLSE Equalizer setup for a Rayleigh fading channel with AWGN and QPSK modulation can be attributed to several factors.

• The Traceback depth of 20 likely aligns well with the channel's memory, enabling effective error correction by considering a suitable history of received symbols.
• Lower Traceback depths may not be enough for accurate decoding, leading to underfitting, while higher depths could make the algorithm too sensitive to noise both resulting in higher BER.
• Other Traceback depths might introduce a delay that, if not properly accounted for, could adversely affect BER performance. Adjusting for this delay might allow effective BER performance with Traceback depths other than 20.

The choice of Traceback depth may require experimentation to find the optimal value that matches your specific channel conditions and modulation scheme. The significant increase in BER with Traceback depths other than 20 suggests that 20 is particularly well-suited to your system, but adjusting for delays may offer additional flexibility.

Refer to the below page for information on effect of traceback depth on the Viterbi algorithm.

https://www.mathworks.com/help/comm/ref/vitdec.html

Hope this helps.