Does anyone have any preservations of these? I see many of them by a Jonathan Froes on youtube but not sure if he's here or if he'd share the original scans (as opposed to the youtube compressions).

(2019-04-22, 02:55 AM)digitalfreaknyc Wrote: Does anyone have any preservations of these?  I see many of them by a Jonathan Froes on youtube but not sure if he's here or if he'd share the original scans (as opposed to the youtube compressions).

i'm not sure if Jonathan Froes has a Dolby Penthouse reader to read and transfer the DD tracks.

I could be wrong, but the DD tracks look a bit like QR codes. I imagine it might be possible to parse them manually with a tool if you manage to photograph them all.

(2019-04-22, 05:48 AM)TomArrow Wrote: I could be wrong, but the DD tracks look a bit like QR codes. I imagine it might be possible to parse them manually with a tool if you manage to photograph them all.

I can imagine that they are just kind of a Bitstream, like every frame has a certain number of bits. Because I think these DD are a bit older than QR codes...
You would need a decoder, after translating the forms into the correct Bitstream.

All the scans I've seen (other than the DTS logo in front of the 35mm Jurassic Park) just use the optical stereo backup tracks for the sound.

You're best off syncing the DD or DTS track from DVD versions, though granted these will be the home theater mixes.

So I've taken a picture of a DD track from Wikipedia and done some initial observations. Note that I'm assuming that this is a real picture of a DD track which it may well not be, I don't know.



So, the overall dimensions in terms of bits are 76 * 76. That's 5776 bits.
The DD logo in the middle always stays the same and has dimensions of 12 * 12 bits, which is of course 144 bits.
So the effective amount of individual bits per sprocket hole is 5632.
There are 4 sprocket holes per frame and 24 frames per second, so that makes 540672 bits per second or exactly 528 kibibits per second.
Wikipedia says that DD on film has a bitrate of 320 kbps, so it's in the right ballpark and likely has some 1/3 of the bits reserved for error correction. Which of course makes sense, because the tiniest dust particles could completely ruin the tiny bits of information.

I also found this: https://www.youtube.com/watch?v=3np6ZAP75e0

It's the optical audio + the dolby digital code between the sprocket holes. Unfortunately in very bad quality due to Youtube compression. I wrote a comment asking if I can get the uncompressed variant. That would certainly be a great test dataset for trying to figure this out. The guy also has uploaded some 35mm trailer scans btw, might be of interest to some. I'll make sure to invite him here if he replies.

So yeah, here's the main things we don't know:
- How exactly the information has to be decoded and how the error correction is applied.

Writing a program converting the image data to bits should be relatively easy if the scanned data is in good condition and not distorted. I could possibly manage to do that. But without knowing how to decode the information, it's a bit useless so far.

With that said, the AC-3 patent is expired, so afaik it's not even illegal to try and reverse engineer now. Also, if some of you can find any papers on this particular subject (like patent papers), it would likely be very helpful. There are AC-3 specification papers that are very detailed, but I don't think they deal with the AC-3 coding on film prints specifically, at least I couldn't find references to it.

Overall conclusion - I think it should be doable, but depending on whether we can find further documentation, it may be a rather long-term effort and need some greater minds than me for the reverse-engineering.

Done a bit of additional math: If you take 320 kbps and you add 0.65*320kbps, you get 528 kbps. This implies that roughly 2/3rds, or precisely 65% of the original signal are added additionally for error correction. Of course this is just a wild guess and might not be entirely accurate since there may be details about the implementation I don't know and can't take into account.

Tom, good analysis! Agree that is doable - think we have now a way to "scan" an analog laserdisc, so...

Said that, I guess that it would be much easier to capture the AC-3 track from laserdiscs; sure, they were only released up to 1999/2000 (and few up to 2001), but *it seems* they retain the theatrical mix, and some think they are direct ports of the same mix used for theatrical prints - just encoded at 384kbps instead. Same could be said for DTS - apart that in this case the laserdisc used a different codec, yet there are good chances the original mix were the same. Sure, those fact could not be confirmed, and using a theatrical print to get the original AC-3 audio (along with Cinema DTS discs) is the only way to be 100% sure to get the original theatrical audio... also, get a print is not always possible, and is not that cheap, so a laserdisc capture is the best alternative (cost/quality) IMHO.

Sure, all good points, but hey, if the option exists, why not explore it.

Also, further information: Back then Dolby Digital was actually called SR-D on prints that had SR-noisereduced Dolby Stereo along with Dolby Digital. A different original term seems to have been Dolby Stereo Digital. That may help finding documents about it.

If we can actually find documents explaining how the format works, decoding it should be very trivial and just a matter of someone taking the time to write it.

Unless they use some form of encryption with keys hidden in hardware, but not even DTS does that and this kind of stuff only came much later, so my guess is that that isn't the case.

"trivial" is a biiig word in this case - unless you are a genious, I guess!

Sure, if the option will exist, it will be the best. Until then, and even after then until a print of a given title will be available, the laserdisc option is still good.

Yep, t was Spectral Recording Dolby Stereo Digital:



as AC-3 on laserdisc was Dolby Surround AC-3 Digital:



(Dolby loved longest names, huh?)

Now, it will be much more interesting to get SDDS tracks!

I found this great PDF which goes into a bit more detail and basically confirms what I already found out: https://www.hps4000.com/pages/digital/do...igital.pdf

Sadly it doesn't go deep enough to tell how exactly the error correction is applied, so the search continues.






As for me calling it trivial: If the algorithm already exists, all that is needed is to code it. Reading the individual bits from a well-scanned data block should be programming work for a few days max, then if the decoding algorithm is known, implementing that wouldn't be too much of a big deal either.

It seems that it all gets a little complicated though by the apparent fact that SR-D prints can actually include software updates for the Cinema Processor. Crazy!