[alsa-devel] Timing Info
I'm using snd_pcm_readi() to collect samples and am wondering if/how I can get accurate timestamps for them. Seems like the snd_pcm_status() gives me this info but I'm unclear on the specifics. Can someone point me toward the correct field in the status structure that I should be using and what sample that timestamp corresponds to?
Does snd_pcm_status_get_tstamp() get me the timestamp for the first frame returned by the next snd_pcm_readi() call?
P
On Fri, Jun 11, 2010 at 6:00 PM, Paul Dugas paul@dugasenterprises.com wrote:
I'm using snd_pcm_readi() to collect samples and am wondering if/how I can get accurate timestamps for them. Seems like the snd_pcm_status() gives me this info but I'm unclear on the specifics. Can someone point me toward the correct field in the status structure that I should be using and what sample that timestamp corresponds to?
Does snd_pcm_status_get_tstamp() get me the timestamp for the first frame returned by the next snd_pcm_readi() call?
I know now get_tstamp() isn't what I'm looking for after reading some prior postings and other web sites. Can't seem to find a discussion of timing in the ALSA docs themselves. Maybe I'm being dense.
Looks like some processing of the trigger timestamp is what I need but I'm not following the logic of "triggers" in the simple use of snd_pcm_readi() that I've got. Before the first readi() call, the trigger time is 0. The second time, it's pretty close to the wall time for the first call. Perhaps that's the timestamp of the first sample of the first period I read and I need to add the period_time each time? Seems like that'd accrue error when recording over longer periods of time (which my application does).
What's the "timestamp mode". Does it generate a timestamp each period? How would I access that timestamp?
Thanks in advance for any assistance,
Paul
Paul Dugas wrote:
Looks like some processing of the trigger timestamp is what I need
Trigger events are starting or stopping of a stream.
What's the "timestamp mode". Does it generate a timestamp each period? How would I access that timestamp?
Set SND_PCM_TSTAMP_ENABLE to generate a timestamp whenever the hardware position is updated. Call snd_pcm_status to read the position and the corresponding timestamp.
HTH Clemens
On Sun, Jun 13, 2010 at 10:10 AM, Clemens Ladisch clemens@ladisch.dewrote:
Paul Dugas wrote:
What's the "timestamp mode". Does it generate a timestamp each period? How would I access that timestamp?
Set SND_PCM_TSTAMP_ENABLE to generate a timestamp whenever the hardware position is updated. Call snd_pcm_status to read the position and the corresponding timestamp.
So, does setting SND_PCM_TSTAMP_ENABLE change the meaning of the tstamp field in the status? I believe I read somewhere that with TSTAMP_NONE, tstamp was set to "now" when readi() was called. Does it change to meaning " the timestamp od the first frame returned"? If not, where do I look?
Much thanks,
Paul
On Sun, Jun 13, 2010 at 3:47 PM, Paul Dugas paul@dugasenterprises.com wrote:
On Sun, Jun 13, 2010 at 10:10 AM, Clemens Ladisch clemens@ladisch.de wrote:
Paul Dugas wrote:
What's the "timestamp mode". Does it generate a timestamp each period? How would I access that timestamp?
Set SND_PCM_TSTAMP_ENABLE to generate a timestamp whenever the hardware position is updated. Call snd_pcm_status to read the position and the corresponding timestamp.
So, does setting SND_PCM_TSTAMP_ENABLE change the meaning of the tstamp field in the status? I believe I read somewhere that with TSTAMP_NONE, tstamp was set to "now" when readi() was called. Does it change to meaning " the timestamp od the first frame returned"? If not, where do I look?
I spun up a test program but I'm not seeing what I expect. I'm still missing something.
I'm calling snd_pcm_status() and snd_pcm_status_dump() before each snd_pcm_readi() call. I see the "trigger_time" not changing (after the first readi call). The "tsamp" isn't increasing in consistent steps. I expect it to be consistently 10,666 usecs per the "period_time" value below.
The PCM device is configured like so:
Plug PCM: Linear Integer <-> Linear Float conversion PCM (S32_LE) Its setup is: stream : CAPTURE access : RW_INTERLEAVED format : FLOAT64_LE subformat : STD channels : 8 rate : 96000 exact rate : 96000 (96000/1) msbits : 64 buffer_size : 32768 period_size : 1024 period_time : 10666 tstamp_mode : ENABLE period_step : 1 avail_min : 1024 period_event : 0 start_threshold : 1 stop_threshold : 32768 silence_threshold: 0 silence_size : 0 boundary : 4611686018427387904 Slave: Route conversion PCM (sformat=S32_LE) Transformation table: 0 <- 0 1 <- 1 2 <- none 3 <- none 4 <- none 5 <- none 6 <- none 7 <- none Its setup is: stream : CAPTURE access : MMAP_INTERLEAVED format : S32_LE subformat : STD channels : 8 rate : 96000 exact rate : 96000 (96000/1) msbits : 32 buffer_size : 32768 period_size : 1024 period_time : 10666 tstamp_mode : ENABLE period_step : 1 avail_min : 1024 period_event : 0 start_threshold : 1 stop_threshold : 32768 silence_threshold: 0 silence_size : 0 boundary : 4611686018427387904 Slave: Hardware PCM card 0 'HDA Intel' device 0 subdevice 0 Its setup is: stream : CAPTURE access : MMAP_INTERLEAVED format : S32_LE subformat : STD channels : 2 rate : 96000 exact rate : 96000 (96000/1) msbits : 32 buffer_size : 32768 period_size : 1024 period_time : 10666 tstamp_mode : ENABLE period_step : 1 avail_min : 1024 period_event : 0 start_threshold : 1 stop_threshold : 32768 silence_threshold: 0 silence_size : 0 boundary : 4611686018427387904 appl_ptr : 0 hw_ptr : 0
Hmmm.... Not sure what to try next.
Paul
On Sun, Jun 13, 2010 at 10:10 AM, Clemens Ladisch clemens@ladisch.de wrote:
Set SND_PCM_TSTAMP_ENABLE to generate a timestamp whenever the hardware position is updated. Call snd_pcm_status to read the position and the corresponding timestamp.
I've re-read this a number of times and have been fiddling with code that dumps the status at various points but I'm still not grasping the concept here. Can comeone point me to an explanation of the delay, avail, and avail_max values? Maybe I can figure this out with that.
Paul
PS: Is there a better place to ask questions like these? People on #alsa at Freenode sent me here.
Paul Dugas wrote:
On Sun, Jun 13, 2010 at 10:10 AM, Clemens Ladisch clemens@ladisch.de wrote:
Set SND_PCM_TSTAMP_ENABLE to generate a timestamp whenever the hardware position is updated. Call snd_pcm_status to read the position and the corresponding timestamp.
I've re-read this a number of times and have been fiddling with code that dumps the status at various points but I'm still not grasping the concept here. Can comeone point me to an explanation of the delay, avail, and avail_max values?
The avail value tells you how many frames are available, i.e., can be written, in the output buffer. The largest value that this had since the last snd_pcm_status call is avail_max.
The delay value is the time that will elapse before the next frame that you write into the buffer will actually be output by the hardware. This is not only the time until the hardware reads this frame out of the buffer (which could be computed from avail) but also any processing after that.
HTH Clemens
participants (2)
-
Clemens Ladisch -
Paul Dugas