dtsutil is a little utility I wrote to extract the raw DTS bitstreams from PCM WAV recordings taken from the LaserDisc SPDIF output.

I know a couple of utilities already exist, but I had issues reliability extracting the entire movie's stream with them. I capture my LDs as one audio/video file with all disc sides in it. This means there are blank pauses/gaps between the SPDIF data in the files. This sometimes seemed to cause issues with the existing utilities I tried, so I wanted to try to make my own that will (hopefully) handle this reliably.

Here's the GitHub link if anybody is interesting in trying it: https://github.com/mikechambers84/dtsutil

It's an alpha version right now. I haven't tried a lot of discs, but I did a couple and had no issues. I use an ESI U24 XL with ASIO to capture the SPDIF output.

I will probably add AC3 support as well, but am starting with DTS and just threw this together today. I hope it's useful to someone else out there!

If anybody has problems getting it to work properly, please create an issue on the GitHub page so that I can look into it.

Usage example:

Will git it a try!

How long did it take to process Dark City entirely after TOSLINK capture?
Did you concatenate side 1 and 2 into a single WAV before extraction?

Julien

It took about 8 or 10 seconds to process the whole movie's stream.

And yes, they were in a single WAV. When I capture LD's, I just record everything into one file for all sides and then clip out all the frames in the video before the first disc frame, and between side changes in VirtualDub 2. During this step I also crop to 720x360 (for widescreen movies) and run the IVTC filter on it.

I don't have to clip out any of the blank audio between sides when extracting DTS though, I wrote this so that when it hits any samples without DTS packets, it just keeps seeking forward until it sees the next one and then it resumes extraction to the output file.

I am noticing an odd issue though. When I played the extracted DTS file directly in VLC, everything sounded perfect and the entire movie's length was there with no glitches in the sound. Now that I'm trying to mux together an MKV with video and audio in ffmpeg, it seems to miss packets and the audio is only around half the length of the movie in the MKV output and it seems to play through it too fast. Hmm.

I'm not sure if it's an issue with my extraction or with ffmpeg's muxing. Like I said, the raw .dts file sounds great alone in VLC. You can also see in the screenshot that's it's processed 1:45:07 worth of samples in DTS frames, which is the correct length of the video, and is also what VLC reports as the length of the .dts file.



This is the command line I used.

EDIT: I'm also thinking of writing another utility to stream the SPDIF input and save it directly as an AC3 or DTS bitstream. It would be nice to eliminate the intermediate step of making a WAV file.

Compiles instantly with GCC.



Will play a DTS LD + ZOOM U-44 capture to give it a try soon.

Julien

Regarding the issue I mentioned of it sounding like it's playing too fast (sounds like missing chunks of samples in anything but VLC) -- It doesn't seem to be an issue with my code at all. BeSplit's output has the same issue from the same source WAVs. The output is identical so my code seems to be fine.

There's possibly an issue with the original capture itself, however using the exact same audio device to capture AC3 bitstreams works flawlessly. I've transferred several AC3 movies now and muxed those streams with the video into MKV files. They're all perfect.

I'm not really sure what's up with this yet, but again, I think my code is good and I've got something else going on.

Julien, what software do you use to capture the DTS WAV? I'm using Reaper.

EDIT: Also, I've started adding AC3 extraction to my utility as well. I want it to be a 2-in-1 tool for LD caps.

AC-3RF decoding would be epic Mr. Chambers

Actually doing it direct from the RF output would be awesome, but unfortunately impossible with PC audio capture equipment.

The carrier frequency is 2.88 MHz, wayyyyy beyond what you can do with audio equipment... and that's not even getting into figuring out how they modulated the signal into it. Demodulator and SPDIF capture required!

Once you have that it's a pretty simple thing, identify the AC3 frame data in the PCM bitstream and just throw them into a raw .ac3 file. Similiar to DTS.

I'll be trying with Audacity.

Julien

Interesting. I tried that before Reaper, but it seemed to corrupt a lot of the data no matter what I did. What driver/settings are you using?

EDIT: I'm on Windows though. Audacity doesn't support ASIO out of the box, and the other drivers it does support are probably mangling the data.

Audacity couldn't sync with ASIO well, I went straight to Reaper.

First 60 seconds of Jackie Brown (1997) [15065 AS]:





Trying the DTS file with Foobar2000...

It plays as 44.11kHz/6ch/32bit bitrate 2470kbps DCA (DTS Coherent Acoustics), but it plays twice too fast?

Trying VLC...

It plays as DTS Audio (dts)/3F2M/LFE/44.1kHz/32bit, and it plays at the right speed.

Header issue?

Tried another 2min sample from 11:30 till 13:30:



Same result: VLC at the right speed, Foobar2000 at double-speed.

Tried PowerDVD 22 for good measure: it can't detect the media information but plays the DTS twice too fast as well.

I would assume Foobar2K and Power DVD trust the header data and get confused.
VLC skips the header and tries to guess the contents by itself?

This seems to be a known issue when extracting DTS from (basically) a CDDA stream: https://forum.doom9.org/archive/index.php/t-54738.html

Foobar2000 detects:


Isn't a bitrate of 2469kbps suspiciously around 2 x 1200kpbs from the LD?

VLC detects (while playing):


Julien

From the first frame in my DTS extract:



Based on https://wiki.multimedia.cx/index.php/DTS#14-bit_words and https://www.etsi.org/deliver/etsi_ts/102100_102199/102114/01.05.01_60/ts_102114v010501p.pdf, that would imply:

Sample frequency = 0b1000 => 44 100 Hz
Bitrate = 0b11101 => targeted bitrate "open"

Open probably leaves some room to the decoder to improvise?

I played a little game and replaced 5,175 times all 0x7FFE8001FC7CE00263A0 with 0x7FFE8001FC7CE0026260 which is supposed to force 1,280kbps (or 0b10011). But that didn't work, I might be a few bits off!

Checking with a DTS CD wav file, the bitrate decoded by Foobar2000 is declared as 1,255 kbps (similar to what VLC detects when decoding).

Then I tried WAV2DTS from https://forum.fanres.com/thread-2255.html:



Again, VLC plays it normally and find 2:00 length.

It plays well as well in Foobar2000 but 2:00 is detected as 1:45 at 1,411 kbps, not good either. However when played side to side they are identical... and Foobar2000 display goes up to... 2:00/1:45!

Clearly some players trust fully the headers while other try to figure out the contents regardless of headers.

dtsutil header is 0x7FFE8001FC7CE00263A0.
wav2dts header is 0x7FFE8001FC7CDFF262C0.

Julien

Thanks for experimenting! That's interesting, I'll need to dig deeper into the header and see .

While BeSplit outputs a dts with the exact same file size as mine, I also see a few differences in the header. I wonder what's going on there. So yeah, these other extractors are making some mods to the header.

But again, BeSplit's output also acts funny for me in anything other than VLC for me as well. I'm going to give wav2dts a try too and see how that does.

I'm going to figure out what changes I need to make and then hopefully upload a fixed version that also includes AC3 support some time in the next few days.

I went ahead and replaced all dtsutil headers 7FFE8001FC7CE00263A00D3A8009EF7B to wav2dts headers 7FFE8001FC7CDFF262C00D3A8009EF7B.

File plays at 2:00/1:45 just like the wav2dts version.

More than this, a binary diff between the 2 shows they're identical bit-for-bit.

Only the header seems to be the part to modify to get to the magic 1,235kbps that seems to correspond to the 1,200kbps 14-bit expanded to 16-bit.

Next step: sampling the full Jacky Brown DTS soundtrack on 3 sides in 3 WAVs



All done:

Side 1
Extracted DTS frame: 141070 (Time 00:54:35)
Side 2
Extracted DTS frame: 145407 (Time 00:56:16)
Side 3
Extracted DTS frame: 115980 (Time 00:44:53)

Total size for the full Jacky Brown DTS 5.1 soundtrack (at fixed bitrate) is 1.34GB.
If my calculation is right, that's about 1,202kbps, which is probably a rounding error for 1,200kbps.

Julien

This might be useful: https://github.com/hypercube-software/DTS-parser

Will decode the header for you.

I ran Side 3.dts through DTSParser and the "fixed" output is binary identical.



It does detect the 20bit/44.1k/5.1ch as well as the "open" bitrate.

But I think we're just 20 years late to the game





The "fixed" header seems to be: 7FFE8001FC7CE00262C00D3A80

However Foobar2000 still see it as 1,411kpbs and believes the track will be 47:46 when it's actually 54:35...
And it plays double-speed

Julien

Good tips and info, thanks Julien!

I was hoping to get back to this over the weekend, but things got busy unfortunately. I will definitely work on it soon though.

I will re-extract a WAVDTS from a DTS CD to play with it, check headers and how VLC/Foobar2K detect it.

I had to go back to the 2002 version of https://www.etsi.org/deliver/etsi_ts/102100_102199/102114/01.02.01_60/ts_102114v010201p.pdf to find Table 5.8 that was removed later:





So I guess the best would be to convert 14-bit DTS to 16-bit DTS (padding 2 bits to 0) and set the standard 1.411Mbps rate?

Julien

Here you go (as recognized by Foobar2000):

• WAV DTS- 35.5MB [3:31] - 44100Hz/2ch/16bit/1,411kbps (using Foobar2K dts plugin)
  
• WAV2DTS output = 31MB [3:05] - 44100Hz/6ch/32bit/1,411kbps - plays normally
• DTSUTIL output = 31MB [1:45] - 44100Hz/6ch/32bit/2,470kbps - plays too fast
• BeSplit dtswav = 31MB [3:05] - 44100Hz/6ch/32bit/1,411kbps - plays too fast
• BeSplit dtswav -fix = 31MB [3:05] - 44100Hz/6ch/32bit/1,411kbps - plays too fast

DTS headers are respectively:


BeSplit -fix doesn't seem to change anything.

So far the best candidates are the dts plugin from Foobar2000, getting the running time and replay correct by parsing the whole data inside the WAV.



... and VLC -- not trusting the header but computing running average values -- will play any of them properly.

Julien

Not all DTS soundtracks seem equivalent.

Used the same sampling chain for Eagles: Hell Freezes Over (1994) [ID4065GF] and the WAV-DTS can be played in Foobar2K or VLC.

But if I use wav2dts the output is not readable (Foobar2K) or just plays silent (VLC).
I get better results with dtsutil.

Julien