Hi James,
Im asking you becuae looking thru the archives you asked a similar question a long time ago.i'm using an open source VoIP application with thealsa driver. My card is the onboard intel8x0.
My problem is figuring out the patterns I amgetting with the alsa driver when transmittingpackets.For instance when I use the GSM codec with a packetssize of 80 msecs and 132 bytes, packets aretransmitted at alternate intervals of 64/96 msecs.When i use G.711 packets of 240 bytes are transmitted
every 32 msecs with an extra packet every 15 pkts.
With G.711 it is obvious that 256 bytes are added to driver sending buffer every 32 msecs.However this does not match for GSM because it needs only 132 bytes every 64/96 msecs.
Can the driver change the amount of data it gets from the soundcard to suit the codec?
Help appreciated,Brian.James Courtier-Dutton [EMAIL PROTECTED] wrote:
Brian Furey wrote: Hi all, I have an intel810 onboard soundcard.I am using the alsa driver with a VoIP session. The intel8x0.c file has a minimum period byte size of 32 bytes with the minimum no. of periods being 1.The min and max rate is set to 48k. How can I find out what actual(runtime) size period the alsa driver is dealing with? Does it use the minimum size as the period size? Brian. Brian, I am working on a VoIP setup. I am updating the asterisk alsa console driver so that it actually works! The current driver is stuck round about alsa api 0.5.xThe period size that the sound card is actually using does not really matter, if just effects latency. The bigger the period size, the higher the latency.Just set the period size to t
he
smallest the sound card can do, and then just read and write to it.I found that PLAYBACK and CAPTURE directions can have different period sizes, so it is better to open separate handles for playback and capture.In the config setup, you set the buffer and period sizes, and before you set them, you can retrieve the current min/max period and buffer sizes.For playback, it is best to have a certain minimum buffer being full most of the time, due to network jitter, and this buffer can act as the jitter buffer. The actual size of this is probably best found out from trial and error (I have not finished testing this bit yet).For capture, just poll for input, and then read whatever is in the capture buffer and transmit it. You can experiment with different methods of early buffer reads, but again, I have not finished testing with that, so I can't give you any definite answers.Another thing to test with could be sampling and
playback at 48k.Most sound cards work at 48k, and this would reduce the period_time. (period_size stays the same, but as one is using a higher rate, the period_time decreases, and thus latency.) The problems with that is that most VoIP is at 8k, so some resampling is required.CheersJames---This SF.Net email is sponsored by: IBM Linux TutorialsFree Linux tutorial presented by Daniel Robbins, President and CEO ofGenToo technologies. Learn everything from fundamentals to systemadministration.http://ads.osdn.com/?ad_id=1470alloc_id=3638op=click___Alsa-devel mailing list[EMAIL PROTECTED]https://lists.sourceforge.net/lists/listinfo/alsa-devel
Yahoo! Messenger - Communicate instantly..."Ping" your friends
today! Download Messenger Now
