Reply to thread

Message: <blockquote data-quote="tourettes" data-source="post: 661890" data-attributes="member: 10858">Unfortunately that data was gathered from the dev PC where the drifting is not happening (at least I haven't been able to reproduce it).EVR presenter's adjust shouldn't be responsible for the varying starting values, most likely the reason for that is the audio buffer value not increasing smoothly, but instead in blocks (10022, 40088). When we have only few ms of data the result will jump up and down because of that. At least this is what I think is the reason for the fluctuation in the beginning, before the multiplier settles down. I guess we could just filter out the first few seconds of audio buffer values (like you have proposed in the earlier posting). This woulnd't create more than few ns of drifting.Problem with the adjustment and possible (highly I would say) drifting it causes is quite nasty one. Currently we have no knowledge how much of audio data was slowed down or speeded up with the 1.003 / 0.997 multiplier. It depends on a random factor. We can see how long the EVR presenter was pushing the sync to up or down, but the audio resampling is done asynchronously compared to that, so with bad luck the audio might have been processed with the adjustment multiplier with 200 ms amoutn of data where as EVR presenter pushes only 50 ms correction time. The shorted the correction burst is the bigger the error will be.I'm currently checking on the HTPC if the adjusment is to blame (disabled it completely). If it is that then we need to figure out how to Can you see any open issues with the current code that could cause drifting? (I did commit few changes into SVN to allow those debug info to be displayed in ! stats). How should the current logic for reference clock calculation (assuming the current code is otherwise ok - at least the excel results look ok) be changed to take the adjustment into account? I can provide following kind of information quite easialy: xxx ms of audio stream was resampled into yyy ms. This would be done only when the adjustment is done, since if we do it for the bias as well I fear that we might be introducing some new drifting source again.Thanks already in the advance!</blockquote>

Verification