Speex dev - Jan 2010 - Fwd: Re: Fixed Point on wideband-mode: Single Frame loss on 2000 Hz sine causes "freak off"

Hi Jean-Marc,

do you have any other ideas what to look for? Or do you currently debug
yourself? As I already wrote, I'm out of ideas...

best regards,
Frank



Frank Lorenz <Frank_wtal at web.de> hat am 18. Januar 2010 um 16:39
geschrieben:
> Yes, I did.
>
> As mentioned earlier, only the enhancer inside the docoder produces a lot
of overflow messages (it points to lines 68 and 69 inside ltp.c). When I turn
the enhancer of (via SPEEX_SET_ENH control), the codec runs without any messages
from the FIXED_DEBUG thing, regardless if the output signal is o.k. or not.
>
> best regards,
> Frank
>
> Jean-Marc Valin <jean-marc.valin at usherbrooke.ca> hat am 16. Januar
2010 um 01:41 geschrieben:
>
> > Did you run the encoder with the fixed-point debug option?
> >
> >         Jean-Marc
> >
> > On 2010-01-14 11:11, Frank Lorenz wrote:
> > > Hi again,
> > >
> > > I tested lower cumul_gain limit values. In fact, I reduced the
value to 240000, 230000, 220000, ..., 20000, 10000, 5000, 2000, 1000, 500. 
(quality 7, complexity 2)
> > >
> > > For input signal 2000 Hz, the values 190000, 130000, 60000, and
1000 look o.k., all others don't. For 500, the (zoomed out) waveform shows
some kind of amplitude ripple -- i think, this value is definetely too low.
> > > I tested the good values for 2200 Hz input, too. Only the value
1000 looks more or less o.k. (no huge instability, but some kind of visible
"artifact" inside the fade in after the frame loss), for the other
values the codec is not stable.
> > >
> > > So, this does not seem to be a solution. Any further ideas what
to test?
> > >
> > > best regards,
> > > Frank
> > >
> > >
> > > Jean-Marc Valin<jean-marc.valin at usherbrooke.ca>  hat am
14. Januar 2010 um 13:15 geschrieben:
> > >
> > >> What happens if you change that line:
> > >>
> > >> if (cumul_gain>  262144)
> > >>
> > >> to use a smaller value? What value works OK (if any)?
> > >>
> > >> One more thing, when things go wrong, do they eventually go
back to
> > >> normal or does the codec never recover? It's unavoidable
that the audio
> > >> goes bad for a short period of time because of the long-term
predictor.
> > >>
> > >>          Jean-Marc
> > >>
> > > ______________________________________________________
> > > GRATIS f?r alle WEB.DE-Nutzer: Die maxdome Movie-FLAT!
> > > Jetzt freischalten unter http://movieflat.web.de
> > >
> > > _______________________________________________
> > > Speex-dev mailing list
> > > Speex-dev at xiph.org
> > > http://lists.xiph.org/mailman/listinfo/speex-dev
> > >
> > >
> >
>
> ___________________________________________________________
> Preisknaller: WEB.DE DSL Flatrate f?r nur 16,99 Euro/mtl.!
> http://produkte.web.de/go/02/
>
> _______________________________________________
> Speex-dev mailing list
> Speex-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/speex-dev
____________________________________________________________________
Hilfe f?r Haiti! Spende per SMS: Sende UI HAITI an die Nummer 81190.
Von 5 Euro je SMS (zzgl. SMS-Geb?hr) gehen 4,83 Euro an UNICEF.

Quoting Frank Lorenz <Frank_wtal at web.de>:
> do you have any other ideas what to look for? Or do you currently debug
> yourself? As I already wrote, I'm out of ideas...
I've been busy with other things (like a CELT release), so I haven't had
time to
look at it. The only thing I can think of is to look at what's happening in
the
fixed-point that's different from what the floating point does.

Cheers,

   Jean-Marc
> best regards,
> Frank
>
>
>
> Frank Lorenz <Frank_wtal at web.de> hat am 18. Januar 2010 um 16:39
geschrieben:
>
> > Yes, I did.
> >
> > As mentioned earlier, only the enhancer inside the docoder produces a
lot
> of overflow messages (it points to lines 68 and 69 inside ltp.c). When I
turn
> the enhancer of (via SPEEX_SET_ENH control), the codec runs without any
> messages from the FIXED_DEBUG thing, regardless if the output signal is
o.k.
> or not.
> >
> > best regards,
> > Frank
> >
> > Jean-Marc Valin <jean-marc.valin at usherbrooke.ca> hat am 16.
Januar 2010 um
> 01:41 geschrieben:
> >
> > > Did you run the encoder with the fixed-point debug option?
> > >
> > >         Jean-Marc
> > >
> > > On 2010-01-14 11:11, Frank Lorenz wrote:
> > > > Hi again,
> > > >
> > > > I tested lower cumul_gain limit values. In fact, I reduced
the value to
> 240000, 230000, 220000, ..., 20000, 10000, 5000, 2000, 1000, 500.  (quality
> 7, complexity 2)
> > > >
> > > > For input signal 2000 Hz, the values 190000, 130000, 60000,
and 1000
> look o.k., all others don't. For 500, the (zoomed out) waveform shows
some
> kind of amplitude ripple -- i think, this value is definetely too low.
> > > > I tested the good values for 2200 Hz input, too. Only the
value 1000
> looks more or less o.k. (no huge instability, but some kind of visible
> "artifact" inside the fade in after the frame loss), for the
other values the
> codec is not stable.
> > > >
> > > > So, this does not seem to be a solution. Any further ideas
what to
> test?
> > > >
> > > > best regards,
> > > > Frank
> > > >
> > > >
> > > > Jean-Marc Valin<jean-marc.valin at usherbrooke.ca> 
hat am 14. Januar 2010
> um 13:15 geschrieben:
> > > >
> > > >> What happens if you change that line:
> > > >>
> > > >> if (cumul_gain>  262144)
> > > >>
> > > >> to use a smaller value? What value works OK (if any)?
> > > >>
> > > >> One more thing, when things go wrong, do they eventually
go back to
> > > >> normal or does the codec never recover? It's
unavoidable that the
> audio
> > > >> goes bad for a short period of time because of the
long-term
> predictor.
> > > >>
> > > >>          Jean-Marc
> > > >>
> > > > ______________________________________________________
> > > > GRATIS f?r alle WEB.DE-Nutzer: Die maxdome Movie-FLAT!
> > > > Jetzt freischalten unter http://movieflat.web.de
> > > >
> > > > _______________________________________________
> > > > Speex-dev mailing list
> > > > Speex-dev at xiph.org
> > > > http://lists.xiph.org/mailman/listinfo/speex-dev
> > > >
> > > >
> > >
> >
> > ___________________________________________________________
> > Preisknaller: WEB.DE DSL Flatrate f?r nur 16,99 Euro/mtl.!
> > http://produkte.web.de/go/02/
> >
> > _______________________________________________
> > Speex-dev mailing list
> > Speex-dev at xiph.org
> > http://lists.xiph.org/mailman/listinfo/speex-dev
>
> ____________________________________________________________________
> Hilfe f?r Haiti! Spende per SMS: Sende UI HAITI an die Nummer 81190.
> Von 5 Euro je SMS (zzgl. SMS-Geb?hr) gehen 4,83 Euro an UNICEF.
>
>
>