[FFmpeg-soc] [soc]: r3046 - in aac: aac.c aac.h

[Reimar Doeffinger]

On Thu, Aug 07, 2008 at 10:50:35AM +0100, Robert Swain wrote:
> > In that case though I think it would be nice if there was some additional information
> > on the particular numbers used, and also mentioning that the last bit is always alternating
> > for this kind of generator (so if possible, only the upper bits should be used).
> > Believing some random guy on the internet: http://random.mat.sbg.ac.at/~charly/server/node3.html
> > 2147001325*a+715136305
> > at least has some statistics, I could not find any for the current numbers on a quick search.
> 
> A quick search for 'LCG 1013904223' throws a number of results.

Yes, what concerns me a bit that in difference to other LCGs none of those
results seem to have analyzed this one, though maybe that is done in the
book they are from.

> I should probably >>16 the
> values returned by lcg_random().

At least getting rid of the lowest bit probably is a good idea. I doubt
it will be possible to hear, but it might cause weird effects when
doing a spectrum analysis of the audio (though how much depends on how the noise
is added exactly - probably it does not _really_ matter either way).