From: Magnus Holmgren <lear_at_algonet.se>
Date: Sun, 10 Jul 2005 22:30:34 +0200 (MEST)

On Sat, 9 Jul 2005 sascha_at_meta-x.de wrote:

> "Andrew 'ashridah' Pilley" <ashridah_at_netspace.net.au> writes:
>
> > I've got some 32KHz ogg files that sounded horrible with the resampling.
> > they played at the right speed tho, so the timing is right. There was
> > basically fast static ticking, like there's an off-by-one somewhere and
> > some pcm values aren't being written to, and thus, are coming out as
> > noise, but that's just my uneducated ear's interpretation.
>
> I can confirm that problem. Playing oggs with 32kHz has the noises at
> least since daily 20050703 (and still has them with the bleeding
> 2005-07-09 08:12).

I've been working on the code responsible for the resampling (in dsp.c),
and I must say I'm not entirely surprised. When I looked at how Vorbis
samples were processed (as compared to MP3), I was worried that some
calculations might overflow. Maybe that's what you're hearing?

And no, I don't have any code ready for test now, though making a quick
test for this would be quite easy actually. Anyone inclined could change
dsp.c to shift left 15 rather than 16 bits (and adjust min/max/fracbits
values appropriately). :) Maybe I'll run some tests tomorrow...

Magnus Holmgren


_______________________________________________
http://cool.haxx.se/mailman/listinfo/rockbox
Received on 2005-07-10