DGDemux development
Re: DGDemux development
Umm. I don't know how to test it properly yet. But I don't see much difference vs 1.0.0.8. Can you do 500mb RAM buffer for each stream for test?
PS: Minimize bug is fixed now, thank you! Thanks for other improvements too.
PS: Minimize bug is fixed now, thank you! Thanks for other improvements too.
- Bullwinkle
- Posts: 338
- Joined: Thu Sep 05, 2019 6:37 pm
Re: DGDemux development
Nothing like that can help. The OS manages physical writes to the disk from the buffers. Many have benchmarked this and found increasing beyond 1MB file buffer brings you nothing; me too. Are you chasing unicorns? What problem do you have?
"Umm. I don't know how to test it properly yet."
Let us know when you figure it out.
"Umm. I don't know how to test it properly yet."
Let us know when you figure it out.
Re: DGDemux development
Actually I don't have any issues for now. All looks great. I thought about improving things which works good, you are right, maybe I'm chasing unicorns.Bullwinkle wrote: ↑Mon Dec 02, 2019 1:40 pmNothing like that can help. The OS manages physical writes to the disk from the buffers. Many have benchmarked this and found increasing beyond 1MB file buffer brings you nothing; me too. Are you chasing unicorns? What problem do you have?
- Bullwinkle
- Posts: 338
- Joined: Thu Sep 05, 2019 6:37 pm
Re: DGDemux development
It's called hoomin error, hoomin paranoia. Try to learn from it.
Now we have to focus on the demuxed streams, checking basic functionality, AV sync, etc.
Waiting on an important upload right now...
Now we have to focus on the demuxed streams, checking basic functionality, AV sync, etc.
Waiting on an important upload right now...
Re: DGDemux development
Nit pick comment but could you change the try icon? I would like to have both DGDecNV and DGDemux in my taskbar without having to guess which is which. Thanks and great work.
Re: DGDemux development
The reported BD disk has been uploaded to this FTP. And an explanation of duplication procedure is placed aside.
Thank you for effort!
Re: DGDemux development
Another "Browse" bug
- Attachments
-
- Video_2019-12-03_175929-2019-12-03.rar
- (1015.36 KiB) Downloaded 352 times
Re: DGDemux development
Thanks, redbtn, for pointing that out.
I really need to know which MPLS you used for that 10ms thing.
I really need to know which MPLS you used for that 10ms thing.
Re: DGDemux development
OK, zqslzwzw, I have this fixed. I want to research a few corner cases versus the DTS spec before releasing a fix.
@redbtn
Fixed your new browse bug, too. MPLS for your 10ms thing? Third time asking!
@redbtn
Fixed your new browse bug, too. MPLS for your 10ms thing? Third time asking!
Re: DGDemux development
Here is version 1.0.0.10:
* Fixed demuxing of DTS-HDHR streams.
* Fixed GUI corruption after canceling 'browse output' operation.
http://rationalqm.us/dgdemux/DGDemux_1.0.0.10.rar
* Fixed demuxing of DTS-HDHR streams.
* Fixed GUI corruption after canceling 'browse output' operation.
http://rationalqm.us/dgdemux/DGDemux_1.0.0.10.rar
- Bullwinkle
- Posts: 338
- Joined: Thu Sep 05, 2019 6:37 pm
Re: DGDemux development
Even Rocky has limits!
Re: DGDemux development
Iron man 2, mpls 00847. DTS-MA track.
Before 1.54.34.500 it perfect, after has shift +10ms. I guess there is place where last m2ts starts. Maybe there are 2 identical frames and 1 should be deleted, but for some reason it wasn't.I fail to find duplicate frame, but there starts difference +10ms (vs eac3to), and stream length is +10ms vs BDInfo information. I'm not 100% sure that it's issue, but if you look at it, it will be great.
Re: DGDemux development
Great report, redbtn. Thank you. Now I can get on it tomorrow.
Re: DGDemux development
How to get the duration of a BD disk so precisely by BDInfo or other tools? I also tried BDInfo 0.7.5.5 (GUI), however it reports the duration as hh:mm:ss without the precision of milliseconds.redbtn wrote: ↑Tue Dec 03, 2019 5:32 pmIron man 2, mpls 00847. DTS-MA track.
Before 1.54.34.500 it perfect, after has shift +10ms. I guess there is place where last m2ts starts. Maybe there are 2 identical frames and 1 should be deleted, but for some reason it wasn't.I fail to find duplicate frame, but there starts difference +10ms (vs eac3to), and stream length is +10ms vs BDInfo information. I'm not 100% sure that it's issue, but if you look at it, it will be great.
Re: DGDemux development
1.54.34.500 is the time at which the gap between the last two M2TS files occurs. At each gap, processing is done to see what adjustments are necessary. In this case at that gap, DGDemux sees an accumulated desync that is below the threshold (approx. 5ms for DTS HDMA) for requiring correction. Therefore, no audio frames are deleted or added at this gap. AV sync during the last M2TS appears spot on to me when playing an MKV made from the demuxed files in MPC-HC.redbtn wrote: ↑Tue Dec 03, 2019 5:32 pmIron man 2, mpls 00847. DTS-MA track.
Before 1.54.34.500 it perfect, after has shift +10ms. I guess there is place where last m2ts starts. Maybe there are 2 identical frames and 1 should be deleted, but for some reason it wasn't.I fail to find duplicate frame, but there starts difference +10ms (vs eac3to), and stream length is +10ms vs BDInfo information. I'm not 100% sure that it's issue, but if you look at it, it will be great.
So, it appears that eac3to deletes a frame at this point and DGDemux does not. As I have mentioned several times, DGDemux and eac3to use completely different algorithms to decide when to delete audio frames. The decision at each gap will depend on what has been done at all the previous gaps.
Bottom line is I don't see a problem here.
Re: DGDemux development
Thanks for investigating! DGDemux is smarter than eac3to, even on non UHD BDs, it's good! So, it seems I have no issues for now anymore. I'll check UHD seamless disks if I find them. But it will be much harder to find bugs.Rocky wrote: ↑Wed Dec 04, 2019 9:55 am1.54.34.500 is the time at which the gap between the last two M2TS files occurs. At each gap, processing is done to see what adjustments are necessary. In this case at that gap, DGDemux sees an accumulated desync that is below the threshold (approx. 5ms for DTS HDMA) for requiring correction. Therefore, no audio frames are deleted or added at this gap. AV sync during the last M2TS appears spot on to me when playing an MKV made from the demuxed files in MPC-HC.
So, it appears that eac3to deletes a frame at this point and DGDemux does not. As I have mentioned several times, DGDemux and eac3to use completely different algorithms to decide when to delete audio frames. The decision at each gap will depend on what has been done at all the previous gaps.
Bottom line is I don't see a problem here.
Thank you for your hard work!
Re: DGDemux development
Great to hear, redbtn. You've been really helpful in DGDemux development.
Ah, that BDInfo report feature. I never noticed that before. Cool!
Ah, that BDInfo report feature. I never noticed that before. Cool!
Re: DGDemux development
Thank you very much. I never notice this before.
If I have not scan the whole disk before entering this feature, will the "Length" be detected accurately?
And I have not figure out how to make it on the command line version.
Re: DGDemux development
Today I find another DIY disk with demux problems by DGDemux. The duration of the obtained audio files deviate from that of the video stream.
But none of other softwares such as eac3to, tsmuxer, mkvtoolnix and ffmpeg can demux it correctly. I am not sure whether it is necessary to upload it for your analysis. This disk has only one main m2ts and can be played normally by potplayer.
Re: DGDemux development
Sure, go ahead and upload it. Happy to look into it for you. With only one M2TS the gaps processing is not involved, but there may be some other issue that can be fixed.
Re: DGDemux development
Why DGDemux GUI window becomes inactive after clicking "Demux"?