2013-12-10, 09:21
Quote:Perhaps at this late point in the initialization, changing the channel layout kills the upmix matrix?
The engine should adapt to the layout the sink returns. Please post log.
The engine should adapt to the layout the sink returns. Please post log.
2013-12-10, 22:00
That's strange, log looks ok. If you played a video with this channel layout: FL,FR,FC,LFE,SL,SR, would you hear sound on the backs?
ffmpeg mixes BC into BL/BL or SL/SR, so it should not matter if sides or backs are returned by the sink.
ffmpeg mixes BC into BL/BL or SL/SR, so it should not matter if sides or backs are returned by the sink.
2013-12-10, 22:56
I noticed the discrepancy between SL/SR and BL/BR as well. Without your last patch, it tried to init FL,FR,FC,BL,BR and failed. My diff added, it inits FL,FR,FC,LFE,BL,BR and the rears are working. So perhaps the upmixer has already decided to mix into BL/BR, and by initializing SL/SR now, these channels are lost?
"Regular" 5.1 works on all channels, regardless whether it's BL/BR or SL/SR. Just double-checked, since most stuff I have is "5.1 (side)", but the few things I have that are "5.1" work fine as well (and always did).
Edit: I could reproduce the missing rears now with regular 5.1 stuff. I set my output to "5.0" in XBMC settings, so WASAPI initialization would fail in all cases and it would therefore forcibly run into your loop. In this case, the rears are always lost, regardless whether I'm playing "5.1 (side)" or "5.1" streams.
"Regular" 5.1 works on all channels, regardless whether it's BL/BR or SL/SR. Just double-checked, since most stuff I have is "5.1 (side)", but the few things I have that are "5.1" work fine as well (and always did).
Edit: I could reproduce the missing rears now with regular 5.1 stuff. I set my output to "5.0" in XBMC settings, so WASAPI initialization would fail in all cases and it would therefore forcibly run into your loop. In this case, the rears are always lost, regardless whether I'm playing "5.1 (side)" or "5.1" streams.
2013-12-11, 10:07
You need to set channel layout to 7.1 for this test, then play FL,FR,FC,LFE,SL,SR. If you set channel layout in XBMC to 5.1, it requests BL/BR from the sink. It looks like SL/SR are not mapped by you device.
2013-12-11, 13:11
I have the channel layout set to 7.1.
Both SL/SR and BL/BR are mapped. The rear channels are only lost as soon as XBMC runs into your new loop.
Both SL/SR and BL/BR are mapped. The rear channels are only lost as soon as XBMC runs into your new loop.
2013-12-11, 14:58
What version do you run? I was not able to locate the git revision shown in you log in master branch. Does it include this commit? https://github.com/xbmc/xbmc/commit/6a6b...d036484a51
2013-12-11, 15:19
I'm using Git rev. e3a145, which I pulled a few hours before this month's merge window started, then applied your BC channel patch (PR 3744, which includes the commit you mention), and then this and this as per your request. So basically, I'm using the November monthly build with only the additional commits applied that you asked me to apply.
2013-12-11, 15:34
This confuses me because PR 3744 includes the other two changes you mentioned. When did you apply PR3744? Commits in this PR changed a couple of times before merge.
If the change I mentioned was missing, it would explain the behavior. If you have this change, I have no idea what's going wrong. From an engine perspective it does not matter it the engine requests FL,FR,FC,LFE,SL,SR and sink returns this layout or engine requests FL,FR,FC,SL,SR and sink returns FL,FR,FC,LFE,SL,SR. So assuming engine works correct the issue must be in the sink. But I don't see what could be wrong there.
If the change I mentioned was missing, it would explain the behavior. If you have this change, I have no idea what's going wrong. From an engine perspective it does not matter it the engine requests FL,FR,FC,LFE,SL,SR and sink returns this layout or engine requests FL,FR,FC,SL,SR and sink returns FL,FR,FC,LFE,SL,SR. So assuming engine works correct the issue must be in the sink. But I don't see what could be wrong there.
2013-12-11, 15:44
I applied PR 3744 on December 2 (around evening). I will reapply the current PR version tonight and report back.
2013-12-11, 21:16
Recompiled using the latest version of PR 3744 (and nothing else), now it's working as expected. So I guess I was missing one or more of the PR changes that happened after Dec 2.
Thank you for your effort!!
Thank you for your effort!!
2013-12-11, 21:59
huh 
glad to hear that. I was clueless on what still could have gone wrong.
glad to hear that. I was clueless on what still could have gone wrong.
2013-12-11, 23:39
So now that the BC channel and silence issue is fixed, this request remains:
Here's my proposal: http://xbmclogs.com/show.php?id=96156
It brings back the advancedsetting <audio><resample> as boolean. When it's set to "true", the "Limit sample rate" setting in XBMC is always visible/used, so we can use a fixed sample rate even when "Best Match" or "Optimized" speaker layout is selected.
Any chance to get this (or something similar) committed into XBMC?
(2013-11-30, 23:05)Shine Wrote:
- The advancedsetting <audio><resample> is apparently ignored now. Any chance to honor this advancedsetting again in the future? I'd like to be able to set a fixed sample rate while using best match speaker config.
Here's my proposal: http://xbmclogs.com/show.php?id=96156
It brings back the advancedsetting <audio><resample> as boolean. When it's set to "true", the "Limit sample rate" setting in XBMC is always visible/used, so we can use a fixed sample rate even when "Best Match" or "Optimized" speaker layout is selected.
Any chance to get this (or something similar) committed into XBMC?
2013-12-13, 10:22
We have just cleaned out all that mess from advanced settings. If a setting makes sense for more than a single user it should go to gui settings.
2013-12-15, 02:24
I just tested the 7.1 channel aac mapping and well the rear and side channels are swapped. And since the ticket was closed I thought the best thing is to post my results here ^-^;
Here is a test file.
Here is a test file.
