tolerance on audio playback duration

[yanj-github]

Driven from issue #118
tolerance on video playback duration will be changed from 20ms to 50ms for video.
At the moment the duration tolerance is shared both by audio and video observations and this change is going to impact to audio as well.
It might be better to separate out the tolerance for audio, which might require different value.
For video we are capturing every 8.33ms (on 120fps) so there will be a difference between captured time vs where the frame actually appeared.
Audio is different, it is captured accurately with a little difference (couple of milliseconds).
We will need to run more tests on the device to work out what tolerance we need on audio duration check, but what I can see so far is that 20ms for audio duration passes most of the test.

[jpiesing]

Feb 20th meeting
Expect to be within 20ms based on the way we measure audio. 50ms is also OK.
Parameter is purely an observation & doesn't impact test observation.

@FritzHeiden Please consider including perhaps 2 audio tests in what is run at the HbbTV plugfest so we have some data to evaluate this.

[jpiesing]

March 5th meeting
Discussion

• 50ms brings a risk of passing something that should fail
• 20ms would reduce the risk of passing something that should fail
• 10ms would be possible to measure

If we change it to 20ms, do we risk breaking anything? Requires work on both test runner and OF as tolerance is common for both video and audio.

There's probably more useful things for people to be doing.