[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

DEC-talk server incredibly flaky




your results are better than mine -- on My amd 64 box running FC3
and Alsa --- and with an appropriate DMIX plugin defined as the
default device, the default test program that comes with dtk5.0
produces horrible sounding speech.



>>>>>   <peter.rayner@cea.fr> writes:
    > T. V. Raman writes:
    >> 
    >> But first check if the dectalk is using OSS emulation
    >> which is what I suspect it is doing.
    > Yes indeed.  And it may b a bit worse than this.  As I
    > noted in my first post about this new machine, sound was
    > appearing at double speed and double pitch under OSS.  ALSA
    > fixed this.  Now I've found out how from the nice
    > documentation on OSS emulation under ALSA.  he sound device
    > I'm using (an INTEL I82801 seems, from reading the comments
    > in the driver, to only support 16-bit *stereo* samples.
    > DEC-talk produces mono.  Thus the sound hardware expects
    > twice as many samples as it's getting, all consistent with
    > the speed and pitch.  ALSA fixes this by inserting a
    > "plugin" layer in its OSS emulation to map what's being
    > sent to the device file (/dev/dsp0) to what the hardware
    > likes to see.  One can test this by turning off the plugin
    > per application.  If you do this you get back to speech
    > sounding like a duck on speed.
    >> 
    >> Also, check if you have any .asoundrc file in your home
    >> directory that will also influence how your personal alsa
    >> environment functions.
    > No file there.
    >> Finally, run some regression tests without emacspeak of
    >> the speech server --- send it some commands in a pipe and
    >> see what happens.
    > Standalone tests so far going fine.  Not clear evidence but
    > most of the hangs seem to happen when starting chunks of
    > speech, e.g. changing buffers and the like.  It doesn't
    > seem to crash in the middle of reading a window etc.
    > 
    > I'll keep looking and reporting but others might want to be
    > aware of potential excitement with the combination of
    > DEC--talk and the I88201.  cheers Peter
    > -- 
    > Peter Rayner LSCE - CEA de Saclay Orme des Merisiers, 91191
    > Gif/Yvette work: +33 (1) 69 08 88 11; mobile: +33 (6) 75 46
    > 56 52; fax: +33 (1) 69 08 77 16 mail-to:
    > peter.rayner@cea.fr
    > 
    > -----------------------------------------------------------------------------
    > To unsubscribe from the emacspeak list or change your
    > address on the emacspeak list send mail to
    > "emacspeak-request@cs.vassar.edu" with a subject of
    > "unsubscribe" or "help"

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    TVRaman
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc

-----------------------------------------------------------------------------
To unsubscribe from the emacspeak list or change your address on the
emacspeak list send mail to "emacspeak-request@cs.vassar.edu" with a
subject of "unsubscribe" or "help"