ffdshow tryouts project: Discussion & Development

[albain]

Thank you all for your encouragements, but we still lack of results

Here is an updated and working version (at least for AC3/DTS) of MPC renderer

Actually wasapi implementation is buggy : it may be due to the drivers or to an incompatibility butween wasapi with event driver mode (which is the easiest to implement) and directshow.

Anyway, here it is (http://damienbt.free.fr/MPCRenderer.zip)

There are 2 versions : one debug and one release (that should avoid glitches)

THank you for your feedbacks !

EDIT : I have an audio delay on this one but no more glitches. I will fix this later.

[Skinleech]

Hi albain.

Thanks for yet again more excellent work. Here are my results with the new renderer:

TrueHD: The sample played jerkily, and in PCM: http://pastebin.com/f34f6be21
DD Plus: As above: http://pastebin.com/f7d40474e
DTS HD MA (from m2ts): The file played, but I got no audio: http://pastebin.com/f702b8304
DTS HD MA (from mkv): The file played as PCM, but halli splitter loaded rather than your mpeg splitter: http://pastebin.com/f38172210

Thanks!

[albain]

Thanks but you have a xonar right ?

It seems that the xonar will refuse the arcsoft media types (!).

I believe that there is additional security rules in the audio driver that will refuse the connection if for example the process is not protected or something like that (because otherwise FFDShow will try different media types including the arcsoft's ones).

Sorry, but for now I need a radeon 5xxx owner.

Maybe this is fixable for the xonar : for that I am still trying to improve the dumper tool to retrieve more information from the *well known player*

[Skinleech]

Yep, I have a Xonar. Can't say I'm surprised that it's causing problems!

I will keep my eyes peeled on the thread.

[rica]

..for now I need a radeon 5xxx owner.

Now i'm here; gonna try. :)

[SamuriHL]

I'm awake now. Give me a minute to switch over to the ATI card. :)

[rica]

albain, your sample thd m2ts gives dolby digital but ffdshow output gives only 2 channels.

Btw i created some sample m2ts files via eac3to+tsmuxer but ffdshow never accepted to connect to MPCAudio Renderer (with those files) while your sample m2ts (thd) file works; dunno why?

Would you please remux your dtshd.mkv to m2ts?

[SamuriHL]

TrueHD: http://pastebin.com/pastebin.php?dl=d1931dda2
DTS-HD MA: http://pastebin.com/pastebin.php?dl=d5ef51a85

TrueHD shows me lights. DTS-HD MA shows me DTS. Neither give me audio. This is the debug version. I can try the release version now but I'm not hopeful.

[SamuriHL]

Release version made no difference.

[SamuriHL]

rica, converting the mkv's to m2ts is academic. Get tsMuxer.

[rica]

rica, converting the mkv's to m2ts is academic. Get tsMuxer.

I don't think you read my previous post, did you?

[SamuriHL]

I don't think you read my previous post, did you?

Not sure I follow. I remuxed the DTS-HD MA mkv that was posted around the 6th to m2ts and it's fine. What's the issue exactly?

[albain]

Something is wrong. I don't see many traces from MPC renderer. Is MPC renderer really used at the end ?
Please use graphstudio with the MPC splitter that I released in the last pack, thanks


Could you try with this one (http://damienbt.free.fr/MpcAudioRendererFilter_2.zip)

Thanks

[SamuriHL]

Those logs were with MPC. I've tried with GraphStudio with the same setup and it gave the same results, so, that's why I didn't post GS logs. I'll try with your new filter now.

[albain]

Those logs were with MPC. I've tried with GraphStudio with the same setup and it gave the same results, so, that's why I didn't post GS logs. I'll try with your new filter now.

Thank you !

[SamuriHL]

Same thing. TrueHD lights are on but no one's home.

http://pastebin.com/pastebin.php?dl=d79905c1e

[SamuriHL]

DTS-HD MA...it's an mkv converted to m2ts and played in GS.

http://pastebin.com/pastebin.php?dl=d518707c6

DTS light, no audio.

[rica]

Not sure I follow. I remuxed the DTS-HD MA mkv that was posted around the 6th to m2ts and it's fine. What's the issue exactly?

OK, restarting PC fixed the connecting issue :)

truhd > Dolby Digital light on > with sound (ffdshow output gives two channels only)

dtshd-ma > no light on > without sound

dts (core) > DTS light on > with sound (ffdshow output gives
two channels only)

EDIT: btw- those trials made with ex-filter; didn't see new update while writing :)

Now i'm trying the new one.

[mrcorbo]

I'm getting mixed results here on the formats I am able to test.

AC3 passthrough has the same behavior as before with me having to check "output strings to debugger" and run DebugView to get bitstreaming to work. One or the other of these does not work, it requires both. When it *does* bitstream, though, it now seems to be mostly glitch-free with maybe a couple of skips when first starting playback.

TrueHD, unsurprisingly, falls back to decoding and outputting PCM. With my sample, it is able to play back the decoded PCM audio as 5.1 channels (matching my speaker configuration) and sounds fine. No distortion, no glitches.

DD+ and DTS-MA also fall back to PCM, but play back with a lot of distortion.

Another odd behavior pops up when trying to test with a file with an MP3 audio track. FFDShow won't connect directly to the MPC audio renderer and instead inserts the WMAaudio Encoder/Decoder combo in between. Deleting these filters from the graph makes the FFDshow filter connect to the renderer directly. :confused: Either connected directly or with the WMAudio filters in between there is distortion of the audio during playback.

[rica]

Here are the results:

truhd > Dolby Digital light is on > with sound (ffdshow output gives 5.1 even it shows 2 channel output in ffdshow properties page) _this is maybe comes from using thd+ac3 file (created by eac3to) in tsmuxer?

dtshd-ma > no light is on > without sound

dts (core) > DTS light is on > with sound (ffdshow output gives 5.1 even it shows 2 channel output in ffdshow properties page)

Seems there is a conflict between the results of mine and samuri?

http://img7.imageshack.us/img7/4803/51079204.png

[SamuriHL]

Post logs. That's the only way to know for sure what's going on.

[albain]

Okay, there is too much delay between each buffer feeding on the renderer side (FFDShow takes too much time to parse the bistream) so the renderer should not sleep so much too

I have adjusted this time dynamically according to FFDShow parsing

http://damienbt.free.fr/MpcAudioRendererFilter_3.zip

Please try this one with then without debugview

@Rica : I suspect that the HD formats are refused by your audio driver. Please post the debugview trace to confirm this

[Brazil2]

Here is another DD+ sample:

(DD+ 7.1 Channel Test, untouched, From "The Sound of High Definition" Blu-Ray)
http://iknowu.net/files/public/Samples/00026%20-%20DD+%207.1%20Channe%20Test.m2ts
This sample seems to have problems. MPC-HC can play it fine but it stutters a lot with many other players.

MediaInfo reports the audio to have 5 channels:
Audio
ID : 4352 (0x1100)
Menu ID : 1 (0x1)
Format : E-AC-3
Format/Info : Audio Coding 3
Duration : 1mn 34s
Bit rate mode : Constant
Bit rate : 1 024 Kbps
Channel(s) : 5 channels
Sampling rate : 48.0 KHz
Stream size : 11.5 MiB (9%)

eac3to reports the audio to be AC3 5.1, the track not to be clean and can't extract/convert the audio:
M2TS, 1 video track, 1 audio track, 0:01:34, 60i /1.001
1: h264/AVC, 1080i60 /1.001 (16:9)
2: AC3, 5.1 channels, 640kbps, 48khz
Track 2 is used for destination file "EAC3.wavs".
[a02] The Nero decoder doesn't seem to work, will use libav instead.
[a02] Extracting audio track number 2...
[a02] Decoding with libav/ffmpeg...
[a02] Remapping channels...
[a02] Reducing depth from 64 to 24 bits...
[a02] Writing WAVs...
[a02] This track is not clean. <WARNING>
[a02] [libav] Dependent substream is not implemented. If you want to help, update your FFmpeg version to the newest one from SVN. If the problem still occurs, it means that your file has extension which has not been tested due to a lack of samples exhibiting this feature. Upload a sample of the audio from this file to ftp://upload.mplayerhq.hu/incoming and contact the ffmpeg-devel mailing list. <WARNING>
[a02] [libav] invalid header <WARNING>
[a02] The libav decoder reported an error while decoding. <ERROR>
Aborted at file position 1048576. <ERROR>

And TsMuxer simply doesn't recognize the audio track.

Is any other 7.1 DD+ sample with speakers recognition available ?

[rica]

@Rica : I suspect that the HD formats are refused by your audio driver. Please post the debugview trace to confirm this

Ok but please remind me how to get debugview traces?

Or samuri?

[madshi]

This sample seems to have problems. MPC-HC can play it fine but it stutters a lot with many other players.

MediaInfo reports the audio to have 5 channels.

eac3to reports the audio to be AC3 5.1, the track not to be clean and can't extract/convert the audio.

And TsMuxer simply doesn't recognize the audio track.

Is any other 7.1 DD+ sample with speakers recognition available ?
You should simply forget that 7.1 DD+ ever existed. It's totally useless and AFAIK has not been used on *ANY* Blu-Ray disc out there (except for a Dolby demo disc). And that's why no software (which includes eac3to) can handle 7.1 DD+, because it's simply not worth it to add support for it.

[tal.aloni]

Brazil2,
1. most importantly: this track is untouched from the Blu-Ray.
2. madshi already said it.
3. Try TSRemux, working for me.

[albain]

Ok but please remind me how to get debugview traces?

Or samuri?

You have to create an "allowDPRINTF" DWORD value set to 1 in
HKEY_CURRENT_USER\Software\GNU\ffdshow

Then run debugview (http://technet.microsoft.com/en-us/sysinternals/bb896647.aspx) while you play the file

[rica]

You have to create an "allowDPRINTF" DWORD value set to 1 in
HKEY_CURRENT_USER\Software\GNU\ffdshow

Then run debugview (http://technet.microsoft.com/en-us/sysinternals/bb896647.aspx) while you play the file

Thanks.

[SamuriHL]

Okay, there is too much delay between each buffer feeding on the renderer side (FFDShow takes too much time to parse the bistream) so the renderer should not sleep so much too

I have adjusted this time dynamically according to FFDShow parsing

http://damienbt.free.fr/MpcAudioRendererFilter_3.zip

Please try this one with then without debugview

@Rica : I suspect that the HD formats are refused by your audio driver. Please post the debugview trace to confirm this

I'll try this at half time. Football is on so can't interrupt that. :D

[albain]

No problem, I hope that your team will win so that you get enough luck to make this new build working :-)

[SamuriHL]

No problem, I hope that your team will win so that you get enough luck to make this new build working :-)

Thanks. They're losing now. And they're losing to my wife's favorite team. SIGH. :D I'll get to testing soon and get you some new logs.

[albain]

I saw that in 5 days the passive radeon 5750 is released, I will get one of those as soon as available

[SamuriHL]

Oh, I see how it is. String us along til YOUR card come out. :p ROFLMAO! :D

[albain]

Yes I like to share the work, 31 betas this is a record lol

[SamuriHL]

Definitely not a problem for us testing it. You make it easy. Compared to the stuff I do at work, this is child's play. :)

[rica]

Here are the logs (created using previous MPC audio renderer):

dtshd sample log (i had to terminate the graphstudio)

http://www.mediafire.com/?4t5yzvgkmlm



thd sample log:

http://www.mediafire.com/?5ntv4zwezzm

[SamuriHL]

Ran a quick GraphStudio of a DTS-HD MA m2ts track. No change. DTS light, no audio. Did not get to try it without debugview running.

http://pastebin.com/pastebin.php?dl=d3f99f7d

[mrcorbo]

I managed to get AC3 to bitstream with this renderer without DebugView running a couple of times (I checked/unchecked output debug strings and got it to work once with and once without) but the results were inconsistent and now I can't reproduce at all. I will load Realtek-sourced HDMI drivers to see if the results are different. When bitreaming worked it seemed glitch-free.

AC3 with output strings to debugger checked:

Result: Stream is recognized, audio signal rapidly cuts in and out constantly.

http://pastebin.com/d60e8a57f

MP3 decoded to PCM:

It now connects directly to renderer, but will not play.

Generates the following log (pasted here since it's relatively short).

[9880] TffdshowDecAudio::StartStreaming
[9880] TffdshowDecAudioInputPin::NewSegment
[9880] TffdshowDecAudio::NewSegment
[9880] CMpcAudioRenderer::CheckAudioClient
[9880] CMpcAudioRenderer::CheckFormatChanged
[9880] CMpcAudioRenderer::CheckAudioClient First initialization of the audio renderer
[9880] CMpcAudioRenderer::InitAudioClient
[9880]
[9880] WAVEFORMATEXTENSIBLE :
[9880] subFormat : OTHER
[9880]
[9880]
[9880] wSamplesPerBlock : 0
[9880] Valid bits per sample : 0
[9880]
[9880]
[9880] WAVEFORMATEX :
[9880] wFormatTag :
[9880] PCM
[9880]
[9880] cbSize : 22
[9880] nAvgBytesPerSec : 176400
[9880] Channels : 2
[9880] Bits per sample : 16
[9880] Samples per second : 44100
[9880] nBlockAlign : 4
[9880]
[9880] CMpcAudioRenderer::InitAudioClient format supported
[9880] CMpcAudioRenderer::GetBufferSize
[9880] CMpcAudioRenderer::InitAudioClient failed (0x80070057)
[9880] CMpcAudioRenderer::Run Error on check audio client
[9880] TffdshowDecAudio::BeginFlush
[9880] TffdshowDecAudioInputPin::EndFlush
[9880] TffdshowDecAudio::EndFlush
[9880] TffdshowDecAudioInputPin::NewSegment
[9880] TffdshowDecAudio::NewSegment

True HD, EAC3 and DTS-MA all had the same result as before.

[albain]

Thanks, something is still wrong in the implementation

Another shot ?

http://damienbt.free.fr/MpcAudioRendererFilter_4.zip

@Mcorbo This one should also fixed your progressive cuts. For now I haven't work on PCM (it should be not very difficult)
Thanks

[SamuriHL]

Will give it a try soon. My game should be done within the hour.

[whurlston]

I saw that in 5 days the passive radeon 5750 is released, I will get one of those as soon as available

Wow. I was about to order a 5750 tomorrow. Looks like I'll wait a couple of days.

[SamuriHL]

Wow. I was about to order a 5750 tomorrow. Looks like I'll wait a couple of days.

Yea, that would definitely make a good card for an HTPC.

[SamuriHL]

Went for TrueHD in GS this time. THD lights came on, no audio as before.

http://pastebin.com/pastebin.php?dl=d3d4bf7e7

[GaBBa-Gandalf]

Sorry that I have to ask again, but for me it isn't really clear...

Does the hd-sound (dts-hd & truehd) works perfectly with the ati 5xxx series?

GaBBa

[SamuriHL]

Not yet. The audio renderer portion is still an issue. However, the bitstreaming support in the latest ffdshow tryout should be good. Just gotta get it to the receiver.

[albain]

. The delays are still incorrect, I have to work on this.

I will work on this tomorrow

thanks again for your tests

[SamuriHL]

Thanks for the work on this.

[mrcorbo]

Thanks, something is still wrong in the implementation

Another shot ?

http://damienbt.free.fr/MpcAudioRendererFilter_4.zip

@Mcorbo This one should also fixed your progressive cuts. For now I haven't work on PCM (it should be not very difficult)
Thanks

I tried the Realtek drivers with the previous renderer and the results were similar (slightly worse, actually).

Now I'm back to the ATI-sourced drivers and with this latest renderer the stuttering is gone when using DebugView. Also found that DebugView only needs to be running when playback is started to get bitstreaming to work. Once the stream is playing I can close DebugView and it will continue to work.

Without Debugview, playing back the graph in GraphStudio will still not bitstream on the initial attempt. However, if I hit stop and then play again bitstreaming works. This is repeatable.

[albain]

Hi,

here is a new one (http://www.mediafire.com/file/4yniyzztiwj/MPCAudioRenderer.zip) that should solve these buffer delays problem.

There are 2 versions : 1 debug and 1 release version

Thanks

[SamuriHL]

Checking now.