Michael Cheng wrote:
> On Sun, 25 Oct 1998, Rafael W. Luebbert wrote:
>
> > Actually, I am already here. If you would like some info feel free to ask, and
> > I'll
> > do my best (within the bounds of my NDA) to answer your questions.
>
> First question: NDA with whom? FhG?
It is not with FhG, but I can't say more than that, sorry to be vague.
>
> Second question: Are you using the dist10 ISO code, or are you working
> with different code purchased from FhG?
I started with the dist10 code, but very little of that code remains. So much of
the code was (intentionally?) wrong, and there were so many omissions from
both the code and the standard that I had to write many routines from scratch.
I would love to have the FhG code, as I know where all the speed savings are,
but FhG wants US$100000 for their code!
>
> Third Question: In your experience, what part of the code has the most
> effect on quality? Hypothetically, if we wanted to improve the quality of
> the ISO encoder, where should we start: joint stereo? psychoacoustic
> implementation? remove errors in the iso code (if we can find them)?
The psychoacoustic model is hugely bug ridden in the dist10 code. However,
rewriting it is not likely to give you a great improvement in quality.
Adding the joint stereo routines would probably yield the best return
on your "coding" investment, especially MS stereo. I have "played"
with intensity stereo, but really was not impressed with the results.
Looking at the FhG encoder output, it appears they agree, as their
encoder produces either stereo or MS stereo frames when joint stereo
is activated. (I have not seen any evidence of intensity stereo, although
honestly I did not check markedly low bitrates.)
>
>
> later
> mike
More later,
Rafael
