i just installed a hammerfall 9652 with oss driver and it works with -inchannels 8 -outchannels 8.
latency.pd measures 22.6ms latency. how could i get lower latency (<4ms)?
does it make sense to start pd with -32bit, if i have a 20bit AD/DA (Fostex VC-8)?
On Tue, 28 Jan 2003, smoerk wrote:
i just installed a hammerfall 9652 with oss driver and it works with -inchannels 8 -outchannels 8.
latency.pd measures 22.6ms latency. how could i get lower latency (<4ms)?
does it make sense to start pd with -32bit, if i have a 20bit AD/DA (Fostex VC-8)?
Yes, it always makes sense to use the -32bit flag. How did you measure latency ?
Normally it is adjustable with the -blocksize flag. With realtime mode you should be able to run it with -blocksize 64, which is the lowest latency you can get with the card.
Guenter
with -blocksize 64 i get even bigger latency. -blocksize 4 gives latency around 4ms, without dropouts (in realtime mode). i know that the smallest blocksize for the hammerfall is 64, i don't know why it does work with -blocksize 4.
there is a letency.pd path in doc/7.stuff/tools, which measures latency.
guenter geiger wrote:
On Tue, 28 Jan 2003, smoerk wrote:
i just installed a hammerfall 9652 with oss driver and it works with -inchannels 8 -outchannels 8.
latency.pd measures 22.6ms latency. how could i get lower latency (<4ms)?
How did you measure latency ?
Normally it is adjustable with the -blocksize flag. With realtime mode you should be able to run it with -blocksize 64, which is the lowest latency you can get with the card.
Guenter
Zitiere smoerk smoerk@gmx.de:
with -blocksize 64 i get even bigger latency. -blocksize 4 gives latency
around 4ms, without dropouts (in realtime mode). i know that the smallest blocksize for the hammerfall is 64, i don't know why it does work with -blocksize 4.
i think there is a weird bug (or feature) that multiplies the blocksize with the number of outchannels (or inchannels ?). So if you specify a blocksize of "4" in 8channel-mode, you will try to get a "real" blocksize of "32" which is clamped to 64. (as can be seen in the xrmectrl) If you would use 16 channels with "-bs 4" , you would still have a real blocksize of 64.
But maybe my hammerfall-pd is out of date....
mfg.as.rd IOhannes
On Tue, 28 Jan 2003, smoerk wrote:
with -blocksize 64 i get even bigger latency. -blocksize 4 gives latency around 4ms, without dropouts (in realtime mode). i know that the smallest blocksize for the hammerfall is 64, i don't know why it does work with -blocksize 4.
there is a letency.pd path in doc/7.stuff/tools, which measures latency.
Ok, taking into account the bug that Johannes reported this is the lowest latency you might be able to get.
Note that people normally only report theoretical latency. You might be able to reduce it by setting the samplerate to 48kHz.
Greetings,
Guenter
On Tue, 28 Jan 2003, smoerk wrote:
i just installed a hammerfall 9652 with oss driver and it works with -inchannels 8 -outchannels 8.
latency.pd measures 22.6ms latency. how could i get lower latency (<4ms)?
How did you measure latency ?
Normally it is adjustable with the -blocksize flag. With realtime mode you should be able to run it with -blocksize 64, which is the lowest latency you can get with the card.
Guenter
To start with, run ALSA instead of OSS, then what I would do is install the rpms from Stanford Universiy's Planet CCRMA site, you need to install the rpm that patches your kernel for low latency, then you should get more like 2ms latency.
On Tue, 28 Jan 2003, smoerk wrote:
i just installed a hammerfall 9652 with oss driver and it works with -inchannels 8 -outchannels 8.
latency.pd measures 22.6ms latency. how could i get lower latency (<4ms)?
does it make sense to start pd with -32bit, if i have a 20bit AD/DA (Fostex VC-8)?
PD-list mailing list PD-list@iem.kug.ac.at http://iem.kug.ac.at/cgi-bin/mailman/listinfo/pd-list
On Tue, 28 Jan 2003, Brian Redfern wrote:
To start with, run ALSA instead of OSS, then what I would do is install the rpms from Stanford Universiy's Planet CCRMA site, you need to install the rpm that patches your kernel for low latency, then you should get more like 2ms latency.
Scheduling is not the problem in this case, so the lowlatency kernel won't help. Then the ALSA support for pd does not take into account the peculiarities of the Hammerfall. (BTW, fixing this would be a useful task ... hint for Kjetil :)
The 2ms latency is a theoretical number which comes from 1/48000 * 64(samples) * 2(input+output) = 2.66.. ms) Taking this into account the 4 ms measured I/O latency are not at all bad. Do not forget that we have 64 samples jitter in this measurement too. (1.45ms at 44kHz, 1.33 at 48).
Greetings, Guenter
On Tue, 28 Jan 2003, smoerk wrote:
i just installed a hammerfall 9652 with oss driver and it works with -inchannels 8 -outchannels 8.
latency.pd measures 22.6ms latency. how could i get lower latency (<4ms)?
does it make sense to start pd with -32bit, if i have a 20bit AD/DA (Fostex VC-8)?
PD-list mailing list PD-list@iem.kug.ac.at http://iem.kug.ac.at/cgi-bin/mailman/listinfo/pd-list
PD-list mailing list PD-list@iem.kug.ac.at http://iem.kug.ac.at/cgi-bin/mailman/listinfo/pd-list
-
Brian Redfern -
guenter geiger -
smoerk -
zmoelnig＠iem.at