[PD] knacks at low latency

[n++k]

[Johannes M Zmoelnig <zmoelnig at iem.kug.ac.at>]
| maybe it would be of general interest, that we are using a
| rme-96 PST (or was it PSD, stup abbr) interface (that's the one with
| 8-tdat-I/O + 2 analog I/O + spdif I/O)
| 

Oh i have the same card. You have to bear in mind this interface is not
able to handle all buffersize and buffer parameters.. I think it's
something like
the total buffersize is minimum 2048 bytes, maximum 8192. The fragments
sizes thus
are something like the buffersize divided by the number of fragments..
Don't know
if that explains your problems (I haven't been trying yet to push the card
to its limits) but you may want to try 'round' numbers.


| mfg.fdsa.brt
| IOhannes
| 
| On Wed, 30 May 2001, IOhannes m zmoelnig wrote:
| 
| > hi all!
| > 
| > what is the lowest possible latency that anyone got under linux ?
| > with alsa-0.9.0b4 + kernel-2.4.3(+ low latency patch by a.morton) we
| are
| > only coming down to
| > 982 samples (which is almost 22.2 ms under 44.1).
| > this works for "-frags 6"
| > decreasing the fragsize even more, will produce knacks, unless you
| > constantly move the mouse inside the pd window (seems like a tool for
| > live-gigs...)
| > 
| > what exactly does the "-fragsize <n>" option mean ? anytime we changed
| > this parameter, it behaved the same (and very badly indeed), meaning
| > 4210 samples of latency and knacksing all the time, no matter which
| > fragsize we choose.
| > 
| > mfg.sd.asdf
| > IOhannes
| > 
| > 
| 
| 
| 


--
  n
++k