eac3to - audio conversion tool

[rickardk]

Originally Posted by madshi

All DTS 5.1 movies: C L R Ls Rs LFE
All DTS 6.1 movies: C L R Ls Rs LFE Cs
Most DTS 7.1 movies: C L R LFE Lsr Rsr Lss Rss
Shoot 'em Up movie: C L R Ls Rs LFE Lw Rw
Hi-Def 7.1 demo: C L R Ls Rs LFE Lsr Rsr

Does this mean that Shoot 'em Up is handled incorrect with v2.49?

[xkodi]

Originally Posted by rickardk

Does this mean that Shoot 'em Up is handled incorrect with v2.49?

no, it is handled correctly i tested it before a minute. only 7.1 channel tracks with order "C L R Ls Rs LFE Lsr Rsr" are handled incorrectly and there is only one such track from "Qtec Hi-Definition Reference Disc". so, all known movie track are handled perfectly correctly.

p.s.
you can always do the following test to check:

decode the track with Sonic (only 5.1 channels) to wavfile1
decode the track with Arcsoft (all 7.1 channels) to wavfile2

split the channels of the wavfile1 and wavfile2 with Wavewizard. compare wavfile1_ch0 with wavfile2_ch0 and if they are byte by byte identical then eac3to handles the track correctly.

[nautilus7]

Originally Posted by xkodi

- Dolby Digital Plus 7.1 in M2TS is not detected by eac3to

i am preparing Dolby Digital Plus 7.1 samples and will upload shortly, so you can do your magic

I have sent these samples (we have the same disc) to madshi a long time ago, but he has forgot to add support for them. If he doesn't have them anymore, i have samples cut and ready from the first time, so you can count on me for that.

[madshi]

Originally Posted by xkodi

is there any chance to patch that info in the header and for example make "C L R Ls Rs LFE Lsr Rsr" of the Hi-Def 7.1 demo to "C L R LFE Lsr Rsr Lss Rss", so it decodes correctly?

Possible, but difficult. I think I'll rather post a warning if such a track is decoded. If the problem really occurs with a real movie (and not just a demo track) I might look into that.

Originally Posted by xkodi

- on my samples Nero outputs 24bit, but i think this is wrong because if set Arcsoft to 24bit there is noise, so i believe these samples are actually 16bit and Arcsoft is correct

No. AC3 and E-AC3 tracks do not have a fixed bitdepth. If there is noise when decoding the tracks to 24bit with ArcSoft then this is a clear bug in ArcSoft. The native output of any AC3/E-AC3 decoder is floating point. Which is much more than 24bit.

Originally Posted by xkodi

- Dolby Digital Plus 7.1 in M2TS is not detected by eac3to

I know. I don't think it's important because there's not a single movie in the world (at least as far as I know) which uses 7.1 E-AC3. There's only a Dolby demo. Or am I wrong?

Originally Posted by rickardk

Does this mean that Shoot 'em Up is handled incorrect with v2.49?

I think it's decoded correctly. At least the volume is identical between Sonic and ArcSoft. It seems that ArcSoft gets along with the speaker setup used by Shoot 'em Up, although it's not the one used by most other movies. The only track where I've seen a volume difference between Sonic and ArcSoft is that one demo track.

Originally Posted by nautilus7

I have sent these samples (we have the same disc) to madshi a long time ago, but he has forgot to add support for them.

No, I didn't... I just decided that there are more important things to do.

[nautilus7]

Originally Posted by xkodi

split the channels of the wavfile1 and wavfile2 with Wavewizard

You do know that eac3to can output mono wavs, right?

Originally Posted by madshi

I don't think it's important because there's not a single movie in the world (at least as far as I know) which uses 7.1 E-AC3. There's only a Dolby demo

Originally Posted by madshi

No, I didn't... I just decided that there are more important things to do.

That's why I didn't bugged you with this earlier.

[tebasuna51]

Originally Posted by madshi

I'm basing my channel doubling on what DTS decoders give me for 6.1 tracks. I'm remapping the channels so that I get this:

FL, FR, C, LFE, BC, BC', SL, SR

That should be correct, right?

Yep, but this is for next version? Or only for dts-hd/eac3?

Now with:

eac3to seven_channel.wav output.wav -extensible

the 7th channel is doubled to 8th channel without remapping then is always wrong.

[xkodi]

Originally Posted by madshi

Possible, but difficult. I think I'll rather post a warning if such a track is decoded. If the problem really occurs with a real movie (and not just a demo track) I might look into that.

i agree, it doesn't worth the effort until there is only one demo track with that channel order and warning is perfectly good, after all the same track is available in TrueHD too.

Originally Posted by madshi

No. AC3 and E-AC3 tracks do not have a fixed bitdepth. If there is noise when decoding the tracks to 24bit with ArcSoft then this is a clear bug in ArcSoft. The native output of any AC3/E-AC3 decoder is floating point. Which is much more than 24bit.

ok, but if the E-AC3 track is made from 16bit WAV file then it is a waste of space to output to 24bit or am i wrong?

i am still testing and not sure, but i think that on one of the sample ArcSoft do not output noise when is set to 24bit. is there any chance that ArcSoft can detect from what source WAV file the E-AC3 track is made - 16bit or 24bit?

Originally Posted by madshi

I know. I don't think it's important because there's not a single movie in the world (at least as far as I know) which uses 7.1 E-AC3. There's only a Dolby demo. Or am I wrong?

yes, i also have only demo files.

Originally Posted by nautilus7

You do know that eac3to can output mono wavs, right?

yes, i know, but to be honest totally forgot about that, thanks to remind me.

[deathlord]

madshi,

I've chosen this solution because of the way 6.1 PCM is stored in Blu-Ray: It's stored as 7.1 PCM where the back surround channels have identical data.

I'm not sure if I understand correctly. Are you sure these are tracks were originally recorded in 6.1 and the back channel was doubled without applying any attenuation?

I wish I knew what receiver manufacturers are doing, or what DTS/Dolby recommend.

I doubt there is an agreement on the 6.1 to 7.1 conversion.
However, I find the idea proposed by tebasuna51 (here), based on the distribution of acoustic power, very convenient.
For 6.1-->7.1 it would mean:
Back_new(l,r) * cos(30) = 0.5 * Back_old
==> Back_new(l,r) = 0.5 * Back_old / cos(30)
taking sqrt:
Back_new(l,r) should be 0.76 * Back_old

In theory we should also subtract the newly added x-component from the side channel, but I think this can be neglected.

[xkodi]

Originally Posted by xkodi

i am still testing and not sure, but i think that on one of the sample ArcSoft do not output noise when is set to 24bit. is there any chance that ArcSoft can detect from what source WAV file the E-AC3 track is made - 16bit or 24bit?

i confirm that there are E-AC3 tracks that ArcSoft output without noise when set to 24bit and that there are E-AC3 tracks that ArcSoft output without noise only when set to 16bit. ArcSoft decodes the DTS-HD MA correctly only when the right parameters are set, so maybe it is the same with the E-AC3 track and somehow ArcSoft finds the resolution of the source WAV file from which the E-AC3 track is made. probably, the best is madshi to look at that, when there are no more important things to do, i will be happy to provide the samples when the time comes.

[EPiPH0NE]

@madshi

Are you going to be adding support for the rest of Arcsoft decoders ie., TrueHD/EAC3?

[kartman_canada]

Hi, all...

I've installed ArcSoft's TotalMedia Theater 15-day trial... and have eac3to v2.50. I'm trying to rip my copy of Ice Age 2. I thought this one would be easy as it's not seamless branched and it's only MPEG2 instead of x264! It's a WinXP Pro SP2 box.

The curveball here is that the audio is DTS MA... I've had success with AC3 but DTS-MA is new for me...

eac3to log:

eac3to v2.50
command line: "E:\Video_Encoding\eac3to\eac3to.exe" "E:\Ice_Age_2\BDROM\BDMV\STREAM\00001.m2ts" 3: "E:\Ice_Age_2\BDROM\BDMV\STREAM\test3.wav" -down2
------------------------------------------------------------------------------
M2TS, 1 video track, 5 audio tracks, 1:30:36
1: Chapters, 25 chapters
2: MPEG2, 1080p24 /1.001 (16:9)
3: DTS Master Audio, English, 5.1 channels, 24 bits, 48khz
4: AC3, French, 5.1 channels, 448kbit/s, 48khz, dialnorm: -27dB
5: AC3, Spanish, 5.1 channels, 448kbit/s, 48khz, dialnorm: -27dB
6: AC3, English, 2.0 channels, 224kbit/s, 48khz, dialnorm: -27dB
7: AC3, English, 2.0 channels, 224kbit/s, 48khz, dialnorm: -27dB
[a03] Extracting audio track number 3...
[a03] Downmixing multi channel audio to stereo...
Loading white noise (needed for dithering)...
[a03] Decoding with ArcSoft DTS Decoder...
[a03] Writing WAV...
[a03] Creating file "E:\Ice_Age_2\BDROM\BDMV\STREAM\test3.wav"...
Video track 2 contains 130338 frames.
eac3to processing took 15 minutes, 50 seconds.
Done.

Everything is good except that the demuxed video finishes as 1:30:36 but I get 1:31:08 for the audio no mater how I try to demux/convert it. Until I figure this out, I've stopped but the plan was to continue with MeGUI to encode in x264 and 2-channel AAC and mux into MP4 for PS3 playback.

I've not bothered just chopping the tail off the audio yet... more puzzled at the moment. What happens if I try to mux the audio as it it stands?

[madshi]

Originally Posted by tebasuna51

Yep, but this is for next version? Or only for dts-hd/eac3?

Now with:

eac3to seven_channel.wav output.wav -extensible

the 7th channel is doubled to 8th channel without remapping then is always wrong.

Hmmmm... I'll recheck that.

Originally Posted by deathlord

I'm not sure if I understand correctly. Are you sure these are tracks were originally recorded in 6.1 and the back channel was doubled without applying any attenuation?

I don't know what the studio did. But what I can say is that there are some movies where the studio advertises 6.1 on the backside of the Blu-Ray (e.g. The Descent, just to name one) and which is 7.1 on the disk with the back channels being identical. I don't really think that the studio did attenuation because I guess that some Blu-Ray players may only output one back channel and ignore the other one. But I don't really know...

Originally Posted by deathlord

I doubt there is an agreement on the 6.1 to 7.1 conversion.
However, I find the idea proposed by tebasuna51 (here), based on the distribution of acoustic power, very convenient.
For 6.1-->7.1 it would mean:
Back_new(l,r) * cos(30) = 0.5 * Back_old
==> Back_new(l,r) = 0.5 * Back_old / cos(30)
taking sqrt:
Back_new(l,r) should be 0.76 * Back_old

It sounds reasonable. But I'm not sure if that's what Dolby/DTS and the receiver manufacturers are doing. I'm not really an expert in audio theory... I guess I'll check what the ArcSoft DTS decoder does when I ask it to output 7.1 for a 6.1 track. The decoder upconverts then. I guess that they use a "DTS authorized" method for that. So whatever they are doing might be what I'll implement in eac3to, too. Not sure yet, though...

Originally Posted by xkodi

i am still testing and not sure, but i think that on one of the sample ArcSoft do not output noise when is set to 24bit. is there any chance that ArcSoft can detect from what source WAV file the E-AC3 track is made - 16bit or 24bit?

Originally Posted by xkodi

i confirm that there are E-AC3 tracks that ArcSoft output without noise when set to 24bit and that there are E-AC3 tracks that ArcSoft output without noise only when set to 16bit. ArcSoft decodes the DTS-HD MA correctly only when the right parameters are set, so maybe it is the same with the E-AC3 track and somehow ArcSoft finds the resolution of the source WAV file from which the E-AC3 track is made.

I've already answered all this. Let me copy&paste:

"AC3 and E-AC3 tracks do not have a fixed bitdepth. If there is noise when decoding the tracks to 24bit with ArcSoft then this is a clear bug in ArcSoft. The native output of any AC3/E-AC3 decoder is floating point. Which is much more than 24bit."

Originally Posted by xkodi

ok, but if the E-AC3 track is made from 16bit WAV file then it is a waste of space to output to 24bit or am i wrong?

You need to stop thinking about (E-)AC3 as having anything to do with a specific integer PCM bitdepth. That's just not the case. The (E-)AC3 encoders convert the original PCM data to something completely different before encoding the data. Any resemblence to a specific integer PCM bitdepth is lost during this process.

Originally Posted by EPiPH0NE

Are you going to be adding support for the rest of Arcsoft decoders ie., TrueHD/EAC3?

I see no reason to add support for the TrueHD decoder, as the libav decoder is virtually perfect. I might consider adding support for the (E-)AC3 decoder later. But xkodi's reports don't really encourage me to do that. I don't like wasting hours on integrating a buggy decoder. FWIW, integrating the ArcSoft decoders into eac3to is a pain in the ***. So I will do that only if it's really worthwhile...

[madshi]

Originally Posted by kartman_canada

Everything is good except that the demuxed video finishes as 1:30:36 but I get 1:31:08 for the audio no mater how I try to demux/convert it.

Well, is the audio in sync with the video at the beginning and end of the movie? That's the only thing that counts.

[xkodi]

@madshi

two interesting DTS-HD MA samples, both decode perfectly with eac3to, but the first one is the only 16bit, 6.1 channels DTS-HD MA track i found so far and the second has some weird channel order i think it is the same as Shoot 'Em Up

Internal Affairs II, DTS-HD MA, 6.1, 16bit, 48kHz:
http://rapidshare.de/files/39868411/....ma.dtshd.html

Pan's Labyrinth, DTS-HD MA, 7.1, my tests made me think it has interesting channel order:
http://rapidshare.de/files/39868422/....ma.dtshd.html

Originally Posted by madshi

I might consider adding support for the (E-)AC3 decoder later. But xkodi's reports don't really encourage me to do that.

i found a workaround for the noise problem - if i set the bit depth in ArcSoft decoder control panel before connect the graph there is no noise and the decoded sound is normal. other than that i don't know how to test the ArcSoft (E-)AC3 decoder, because of that not integer thing that you have explained.

[Yraen]

madshi,
"Program too big to fit in memory"
I get this error no matter what I run, be it on files, a folder, -test or just eac3to by itself for help. This is about the third time I've gotten this error. The first time(s) I've attributed it to something I did, but this time I was just trying to list the contents of a folder after successfully muxing/converting another folder. No matter what I do after this, this is all eac3to will return. I solved it in the past by deleting eac3to and replacing it. Is there something else I can do on my end to fix it when this happens or is deleting it the best course of action?

I've currently got the bad eac3to in a separate folder and I'll keep it until I know if there is anything you want me to check it for. The current one I just unzipped back into my 2.50 folder works fine.

edit: Took a closer look at the files, eac3to.exe weighs in at 141,284 bytes instead of 1,484,288 bytes as it should. Not sure why this happened.

[ramki067]

Hi madshi,
I tried converting from .dtshd file which i have to .ac3and its giving me an error as below:

C:\Documents and Settings\ramki\Desktop\tools\eac3to_v2.48>"C:\Documents and Se
ttings\ramki\Desktop\tools\eac3to_v2.48\eac3to.exe" "C:\Documents and Settings\
ramki\Desktop\tools\eac3to_v2.48\12-24kbps-32tone-10.dtshd" "C:\Documents and S
ettings\ramki\Desktop\tools\eac3to_v2.48\12-24kbps-32tone-10.ac3"-64
The format of the source file could not be detected.

I'm attaching the .dtshd file, please help.

12-24kbps-32tone-10.dtshd

Regards,
Ramki.

[madshi]

Originally Posted by xkodi

the second has some weird channel order i think it is the same as Shoot 'Em Up

Yep. The speaker config is "C L R Ls Rs LFE Lw Rw".

Originally Posted by xkodi

i found a workaround for the noise problem - if i set the bit depth in ArcSoft decoder control panel before connect the graph there is no noise and the decoded sound is normal.

Ok, in that case there should be no problem when integrating the decoder into eac3to.

Originally Posted by Yraen

edit: Took a closer look at the files, eac3to.exe weighs in at 141,284 bytes instead of 1,484,288 bytes as it should. Not sure why this happened.

So the files on your harddisk suddenly go corrupt? I'd say you should urgently do a scandisk... Other than that I don't know what to say. I never had such a problem yet...

Originally Posted by ramki067

I tried converting from .dtshd file which i have to .ac3and its giving me an error as below:

The format of the source file could not be detected.

This file is "DTS-HD Low Bitrate" which is not supported by eac3to yet. These files do not have a conventional DTS core. They only consist of DTS-HD packages. That's not a stream type which eac3to supports right now. But this type of audio track is usually only used for PiP audio. So is this DTS track PiP audio? Or where did you get it from?

[Yraen]

Originally Posted by ramki067

Hi madshi,
I tried converting from .dtshd file which i have to .ac3and its giving me an error as below:

Along with what madshi said, that gui has a known bug. The -640 (I'm assuming you dropped the 0 when copying) needs a space in front of it, just for future reference.

Originally Posted by madshi

So the files on your harddisk suddenly go corrupt? I'd say you should urgently do a scandisk... Other than that I don't know what to say. I never had such a problem yet...

I would have thought that also except for two bits of information I accidently left off, probably due to lack of sleep. I've ran scandisk and it finds no errors. And one time this happened was on a separate computer. It probably is something I'm doing since no one else has reported anything like this.

[ramki067]

Originally Posted by madshi

This file is "DTS-HD Low Bitrate" which is not supported by eac3to yet. These files do not have a conventional DTS core. They only consist of DTS-HD packages. That's not a stream type which eac3to supports right now. But this type of audio track is usually only used for PiP audio. So is this DTS track PiP audio? Or where did you get it from?

Yes madshi, it is used for PiP audio which is used in our product. Its actually a file from our Customer. We have many such files which needs to be converted to .wav file ultimately. Is their any such tool for converting this low bitrate .dtshd file or can you please include support in the next release of your application?

Thanks,
Ramki.

[madshi]

Originally Posted by ramki067

Yes madshi, it is used for PiP audio which is used in our product. Its actually a file from our Customer. We have many such files which needs to be converted to .wav file ultimately. Is their any such tool for converting this low bitrate .dtshd file or can you please include support in the next release of your application?

I'll have to check if the ArcSoft audio decoder can decode such tracks. I'm not sure about that.

If I may ask - just for my interest (but not really important): Who is "we" and what is "our product"? You don't need to answer if you don't want to.