Re: QDMC progress, updated

> > And if you compare the psd with the one in the pdf the similarity is 
> > striking.

Did you make make the same comparisons with alaw or other codecs?
Just to make sure.

> I'm not sure of the reduction this gives, I think its 2:1 (16->8 bits). 

Me too. Look at ffmpeg/libavcodec/pcm.c

> Hmm...so you propose that the decoder converts linear PCM -> log
> PCM and then does the compression? I wish I knew enough about DSP to judge
> whether that made any sense.

I think lin->log conversion does pratically so the same as a quantizer:
Decrease the number of bits by dereasing the accuracy, only less
sophisticated. The same could also done in the frequency domain (i.e. after the
fft when encoding), but I don't think that the frequency sweep you made would still
look the same then.

> However, a small problem with your hypothetical flow: I know that
> at the end of the QDM* decoder, there is a standard "clamper" function,
> the one that converts an array of single-precision floats to signed 16-bit
> numbers. I don't have my notes right now, but it's an easy function to
> spot.

Maybe they use an alaw -> float table and convert to int16. The table should be
256*4 bytes then.

Just my $0.02

Burkhard
______________________________________________________________________________
Horoskop, Comics, VIPs, Wetter, Sport und Lotto im WEB.DE Screensaver1.2
Kostenlos downloaden: http://screensaver.web.de/?mc=021110



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click