Hi all I am wondering how one can reduce the latency of PD. I have made a simple patch that connects some inputs on my soundcard to the outputs, and played piano through it. The latency seems to be at least 2-3 ms. Yet in Ardour when using Jack at a period size of 64, the latency is much less.
I am using the same window manager for both pieces of software and they are both running with low latency and pre-empt kernel patches. I have tried using PD's -rt flag to no avail. This is using PD's alsa support on PD 0.36.
Suggestions welcome.
Thanks Luke
On Tue, 19 Aug 2003, Luke Yelavich wrote:
Hi all I am wondering how one can reduce the latency of PD. I have made a simple patch that connects some inputs on my soundcard to the outputs, and played piano through it. The latency seems to be at least 2-3 ms. Yet in Ardour when using Jack at a period size of 64, the latency is much less.
I am using the same window manager for both pieces of software and they are both running with low latency and pre-empt kernel patches. I have tried using PD's -rt flag to no avail. This is using PD's alsa support on PD 0.36.
Try using OSS emulation, this should give you better latency figures. With ALSA the latency is at least 10 ms theoretically, due to the resolution of gettimeofday and how PD's ALSA part is implemented.
Guenter
Hallo Guenther, guenter geiger hat gesagt: // guenter geiger wrote:
Try using OSS emulation, this should give you better latency figures. With ALSA the latency is at least 10 ms theoretically, due to the resolution of gettimeofday and how PD's ALSA part is implemented.
Do you know some theoretical figures for Pd's Jack latency?
ciao
Frank Barknecht _ ______footils.org__
On Tue, 19 Aug 2003, Frank Barknecht wrote:
Hallo Guenther, guenter geiger hat gesagt: // guenter geiger wrote:
Try using OSS emulation, this should give you better latency figures. With ALSA the latency is at least 10 ms theoretically, due to the resolution of gettimeofday and how PD's ALSA part is implemented.
Do you know some theoretical figures for Pd's Jack latency?
Yes, it theoretically adds one jack buffersize worth of latency. But first I have to check the problems that Luka reported. I am on a SBLive here, which won't let me do buffersizes smaller than 512 ....
Guenter
At 03:31 AM 20/08/2003, guenter geiger wrote:
Yes, it theoretically adds one jack buffersize worth of latency. But first I have to check the problems that Luka reported. I am on a SBLive here, which won't let me do buffersizes smaller than 512 ....
I would be happy to test any code with Jack if you like. Using a Terratec DMX 6Fire 24/96 card and am able to reliably use a buffer size of 64.
Luke
At 06:41 PM 19/08/2003, guenter geiger wrote:
Try using OSS emulation, this should give you better latency figures. With ALSA the latency is at least 10 ms theoretically, due to the resolution of gettimeofday and how PD's ALSA part is implemented.
I managed to get it nice and snappy with the -audiobuf flag set to 12. This is using -alsa. The -rt argument makes no difference in this situation.
Thanks for the suggestion anyway.
Regards Luke
-
Frank Barknecht -
guenter geiger -
Luke Yelavich