Modulator_2d::demodulate_soft_bits does not work well
for high SNR. For high SNR value, it produces inf and
-inf, which generates error when channel decoding is
operate on the demodulated bits. I have change to
demodulate_soft_bits_approx and it works well then.
Regards,
Arthur.
Logged In: YES
user_id=1284695
Arthur, Could you send us an example of code where you
reproduce this error? On my system, I was able to set N0 to
well below 1e-300 before the demodulator output was inifite
(this infinity output seems to be caused by the 1/N0
calculation) - note that such a low noise level is zero for
all practical purposes. Possibly system dependent.
Updated documentation to suggest Modulator_ND class be used
instead -- faster and better numerical stability.
Erik
Logged In: YES
user_id=1004597
Finally, we managed to solve the problem in Modulator_2d
demodulate_soft_bits() function. The fix is already in the
repository (rev. 599). See ChangeLog for more information.
Closing this bug report.
/ediap