In general, the reciever upto the demodulator uses uniform quantization
and step size in the powers of 2 while the Low Density Parity Check
(LDPC) decoder uses uniform quantiztion and arbitrary step size. LDPC
decoder using min-sum algorithm needs different quantization schemes
based on the offset used. Offset varies based on code parameters.
Say the soft output of demodulator is 16 bits. The output is in the
format :2's complement,uniformly quantized and step size 2^-15. Let it
be called y_q15.
Say now I want to quantize the output of the demodulator before feeding
it to a decoder for channel code. The number of bits I needed for the
decoder input is 5 bits and step size 'beta'. The value of beta is
optimized for each code the decoder supports.
In matlab, I can achieve the above with the following.
stepSize_inverse = 1/beta; (%pre computed)
y_requantized = round(y_q15*stepSize_inverse); % has to be
computed for every value
However doing the above multiplication in hardware implementation is
costly. I can implement a Look up table approach to avoid
multiplication- but this is also cumbersome if I needed to parallel
processing on a vector. Could you please let me know if there is any
efficient way to tackle the re-quantization problem?
Other alternative approach I came upon is filtering approach which is
too expensive for what I need to solve.
"On the re-quantization of data to implement high-order narrow-band
filters using reconfigurable logic " Dick, C.H.; Harris, F.;
Signals, Systems and Computers, 1995. 1995 Conference Record of the
Twenty-Ninth Asilomar Conference on Volume 2, 30 Oct.-2 Nov. 1995
Page(s):1347 - 1351 vol.2
Thanks for your time.
This is the reference about the min-sum and the quantization issues.
"Near optimum universal belief propagation based decoding of
low-density parity check codes" Jinghu Chen; Fossorier, M.P.C.;
Communications, IEEE Transactions on
Volume 50, Issue 3, March 2002 Page(s):406 - 414