Re: Synchronizing audio with other events ...

Hi Joe,

Thanks for pulling up the reference. I'm happy to see precision sync on the radar.

Just for completeness, here is a related recent thread on the issue of 
"playbackTime" in ScriptProcessorNodes.

http://lists.w3.org/Archives/Public/public-audio/2013AprJun/0186.html

Best,
-Kumar

On 31 May, 2013, at 7:58 PM, Joseph Berkovitz <joe@noteflight.com> wrote:

> Hi Kumar,
> 
> This is an outstanding gap in the API that hasn't yet been remedied, although there is a blueprint (capturing a conference call discussion) for doing so here, as part of Bug 20698:
> 
> https://www.w3.org/Bugs/Public/show_bug.cgi?id=20698#c23
> 
> …Joe
> 
> 
> 
> On May 31, 2013, at 7:24 AM, Srikumar Karaikudi Subramanian <srikumarks@gmail.com> wrote:
> 
>> Hi all,
>> 
>> Which API features do I use to get (for ex.) MIDI and audio events synchronously triggered to within 1ms  of precision, given a precise enough MIDI scheduler implementation? 
>> 
>> With the current state of the API spec, it appears that the precision to which I can synchronize scheduled MIDI and audio is limited by the buffer size used by the audio API implementation. Unless I'm missing something (in which case I'm happy to have a solution), this appears to be a gap in the API - one that can be filled pretty easily.
>> 
>> Regarding synchronization of audio events with visuals at 60fps, current implementations of the Web Audio API update the AudioContext.currentTime frequently enough that checking "currentTime" within a requestAnimationFrame callback is adequate for most purposes (though not ideal). At a sample rate of 44100Hz, 15ms corresponds to 735 samples and current buffer sizes in implementations are 256 samples or fewer, afaik. By the same token, I would need currentTime to update multiple times every 1ms to get that precision with MIDI, which is too much to ask.
>> 
>> However, if along with AudioContext.currentTime I also have the DOMTimeStamp of the time at which the sample corresponding to currentTime will exit the audio subsystem, then the precision to which I can schedule MIDI and audio will not be limited by the audio buffer size any more. Indeed, the buffer size can be 2048 samples for all I care (though event-to-sound latency will suffer at that size). This will also permit better synchronization with visuals.
>> 
>> Is there currently any way to calculate this DOMTimeStamp value or is this indeed a gap?
>> 
>> A minimal test case for this might be - map a key on the keyboard to trigger, with a delay of 1 sec, MIDI 'C' on channel1 simultaneously with an oscillator-based note and check whether the time difference between the onset of the two sounds is a constant independent of when the key is pressed.
>> 
>> Thanks.
>> -Kumar
> 
>