WLAN Demodulation in MATLAB

Started by cogwsn 6 years ago5 replieslatest reply 6 years ago288 views

I am using the WLAN Tool Box of #Matlab V2107b. 

A term named CSI for each data subcarrier is defined as 

CSI(i) = H(i) X conj(H(i)) + N(i)

H(i) = channel estimate of ith subcarrier

N(i) = noise estimate of ith subcarrier 

After obtaining the softbits thru QAM demodulation, these softbits from every subcarrier are multiplied by their corresponding CSI.

I am not able to understand why there is a need for such multiplication. Neither I could find any reference. 


[ - ]
Reply by cogwsnJanuary 26, 2018

Well I found this reference where it says that soft Viterbi could be used with additional channel state information. 


I guess that's why the softbits are multiplied by respective CSI before feeding them to deinterleaver! 

[ - ]
Reply by fred_harrisJanuary 26, 2018

your multiplication is performing a circular convolution with the channel power spectrum.. the channel and its matched filter response to account for the spectral response of the channel while minimizing observation noise in the channel estimate.  This same trick was used in GSM, where the mid-amble probed the channel and then convolved channel response with matched filter to be matched filter for received signal and then the Viterbi trellis estimated the transmitted sequence using the state described channel determined by the probe.    

happy new year


[ - ]
Reply by cogwsnJanuary 26, 2018

New year Wishes to you too :) 

To be honest I couldn't understand your answer at all ... could you please share some link for details. 

[ - ]
Reply by friedmanJanuary 26, 2018
Scaling the LLRs with their corresponding SNR (or CSI - the absolute value matters only when taking into account LLR quantization and saturation in the FEC decoder) comes directly from the definition of the LLR applied to the demapped complex constellation under the AWGN assumption. You can check standard texts, e.g.


Chapter 8.3. (in particular equations 8.21, 8.27)

Intuitively, you should multiply the soft-bit information with the SNR to place more trust in the bits coming from clean symbols/subcarriers, compared to the bits coming from noisy symbols/subcarriers.

Hope this helps
[ - ]
Reply by cogwsnJanuary 26, 2018

Thanks much. Its very clear now.