+- Kodi Community Forum (https://forum.kodi.tv)
+-- Forum: Development (https://forum.kodi.tv/forumdisplay.php?fid=32)
+--- Forum: Kodi Application (https://forum.kodi.tv/forumdisplay.php?fid=93)
+--- Thread: Testing audio engine ActiveAE (/showthread.php?tid=170338)
RE: Testing audio engine ActiveAE - franzem - 2013-10-13
some more Cue file test files: http://we.tl/BOQ5cGWGHE
RE: Testing audio engine ActiveAE - FernetMenta - 2013-10-13
are those not working cue files?
RE: Testing audio engine ActiveAE - franzem - 2013-10-13
They should work perfectly, just download them for testing.
RE: Testing audio engine ActiveAE - FernetMenta - 2013-10-13
Got it. Thanks!
RE: Testing audio engine ActiveAE - hochopeper - 2013-10-16
A general question, I'm an audio geek that loves xbmc and I was thinking maybe there is a way to increase test coverage for audio output. I wanted to put a proposal to you gents before I start writing some code and see if there is something I've missed.
I'm no good at C and a bit of a hack with Python but I can make things that seem to work ok
I thought if I generated a set of test files of some arbitrary waveform at a variety of sample rates (and maybe bit depths?) I could set up a Python script on the host OS with PyAudio to read the input and trigger xbmc to playback the test file and read the loopback recording and test it for bit perfect and ensure that things are silent when they should be.
Some notes:
- Should be able to test this for multichannel analogue looped back to ensure that difference is only noisefloor/distortion.
- For spdif we can only decode PCM so no multichannel tests there.
- For hdmi there is no hdmi input possible so no test there.
For me, initially the PCM spdif test would be my focus.
Would this be of help for you guys when testing for bugs in the audio playback chain?
RE: Testing audio engine ActiveAE - meridius - 2013-10-16
Hi all
Has the lip sync been fixed in xbmc 13 with 24hz and dts and dts-hd audios. As with xbmc 10,11 and 12 you needed to delay the audio by about 175ms to 200ms.
Never seemed to have any problems with 24hz and other audio formats. It's been a bad bug for a few years now and tried other software and never had any problems with audio sync.
Thanks for the hard work.
RE: Testing audio engine ActiveAE - baijuxavior - 2013-10-16
It is not fixed. I am using advancedsettings to set a delay of 200ms and it solved the problem to some extent.
RE: Testing audio engine ActiveAE - FernetMenta - 2013-10-16
This is not directly related to the audio engine.
RE: Testing audio engine ActiveAE - stupid-boy - 2013-10-16
(2013-10-16, 06:14)hochopeper Wrote: ...and test it for bit perfect and ensure that things are silent when they should be...
quote is not exactly intentional and containing key sentence.
in general, you can't. there is no way to pass some signal multiple times thru some adc and to get same result. same for dac. also passing something thru dac-adc will lead to a lot of errors. no bit-perfect here.
human ears can accept and correct a lot of errors some digital system will catch. in reverse, human will catch for example difference in playback speed, while such digital system will perfectly accept this ( if it is not intentionally made to catch it ).
for this subsystem it is important to be well accepted by human ears, not to pass some extreme digital tests. sound is finally analogue, newer mind that we modern people always try to digitalize it.
another example, current stereo upmix feat is not bit-perfect in any way and will fail any tests. even that, it sound great and most of us love it in its current imperfect form.
sorry
RE: Testing audio engine ActiveAE - meridius - 2013-10-16
(2013-10-16, 13:04)baijuxavior Wrote: It is not fixed. I am using advancedsettings to set a delay of 200ms and it solved the problem to some extent.
thanks for the post but its not very handy when only dts tracks are out of sync and other formats are ok and some are not out of sync as much as 175ms.
I was hoping the new ActiveAE audio was going to fix this as this bug has been going on for over a couple of years.
maybe there could be some sort of options added to the audio menu to be able to delay different types of audio files
like say
eg
DTS -175ms
DTS-HD -175ms
Dolby Digital AC3 default
Dolby TrueHD. -50ms
something like that and other formats could be added, then you can delay each type of format for your system to make it work better. does anyone think this is a good idea to add this to xbmc13 and activeAE ?
RE: Testing audio engine ActiveAE - Martijn - 2013-10-16
No, options are never a good idea. Let's first see where this difference comes from and then see if it still needs options.
RE: Testing audio engine ActiveAE - baijuxavior - 2013-10-16
I had never experienced the lipsync issue in older versions, and in Gotham upto the beginning of August. I don't know if it is related to ActiveAE. Unfortunately I deleted all my old Gotham builds.
RE: Testing audio engine ActiveAE - FernetMenta - 2013-10-16
As long as nobody gives a precise description of the issue it most likely won't get better. Some say the problem has been around since Eden, others it was introduced later, some only DTS, other every format. And others don't have any problem at all. You have not even mentioned whether you use PCM or passthrough and if it would make any difference. Would it make any difference if you start video when refresh rate is already set to 24hz?
This kind of info helps, just saying that there is an issue is useless.
This is a development thread. Please contribute with valuable information here.
RE: Testing audio engine ActiveAE - hochopeper - 2013-10-17
(2013-10-16, 16:03)stupid-boy Wrote: quote is not exactly intentional and containing key sentence.
in general, you can't. there is no way to pass some signal multiple times thru some adc and to get same result. same for dac. also passing something thru dac-adc will lead to a lot of errors. no bit-perfect here.
human ears can accept and correct a lot of errors some digital system will catch. in reverse, human will catch for example difference in playback speed, while such digital system will perfectly accept this ( if it is not intentionally made to catch it ).
for this subsystem it is important to be well accepted by human ears, not to pass some extreme digital tests. sound is finally analogue, newer mind that we modern people always try to digitalize it.
another example, current stereo upmix feat is not bit-perfect in any way and will fail any tests. even that, it sound great and most of us love it in its current imperfect form.
sorry
I totally agree. The analogue loopback would need a threshold based on the performance of the equipment used, it's why I wasn't terribly interested in that case, I'm mostly interested in spdif passthrough testing. As a way to test the base case. Dynamic Range Compression or upmixing won't ever be bitperfect but the data into it should be if we have passed a bit perfect test on a bitstream test, which is all I was planning to target with the test proposal that I was going to code up. The other issue is that if something fails a test it's always going to be necessary to read the debug logs to check if it is a xbmc intended behaviour or possibly a driver/OS issue.
I'm an EE and working on some audio DSP hardware in my spare time as a hobby for active surround sound speaker system and I've done a pretty thorough background study on speaker and crossover design, distortion perception, audible preferences and algorithms for DSP to suit those preferences. I'm certainly unconcerned about the output of the DAC looking rather different to what was on the recording, I have no analogue crossover so I'd better hope it doesn't!
I'm going to write the code I suggested anyway, the purpose of my proposal was that I was wondering if it might be of use for people testing Active AE rather than just saying 'well it sounds ok to me' and also testing that the underlying pass through of PCM data is 100% might mean that it's quicker to identify bugs.
Chris
RE: Testing audio engine ActiveAE - FernetMenta - 2013-10-17
@hochopeper any help on this is highly appreciated. There are feature requests for mixer functions which definitely require this kind of testing.