I've noticed that the silence at the beginning of discs, ie at the FBI piracy warning, that have a CD audio track have some form of silence that resembles a keep alive of sorts. It's not pure 0's but it looks like it has been crafted to be silent without being 0. It looks like random 8bit values > 0 < 6

It is digital and repeatable, thus technically part of the digital track but, my player never starts outputting over optical SP/DIF at the same sample. I can press stop and play 10 times, recording 10 different files and they will all be identical after a certain amount of samples but, each file might have a few more or less in front.

I used to use something called "Scenarist" for DVD authoring, many, many years ago and in a learning document it mentioned that DVD will read the headers of all files and start outputting from frame/sample 0. That doesn't appear to be the case with LaserDisc so I'm wondering how it knows when and where to start outputting sound at any point. For instance, if you skip 15 minutes of side A of the LaserDisc how does it know where to start reading again? Does a player ever drop 2 bytes and accidentally swap channels?

This would probably a good question for people who work on the Domesday project/ld-decode, paging @happycube

Yes they do have a TOC, telling about the number of sides, if CX decoding should be enforced, etc.

Spec is https://webstore.iec.ch/publication/3720 but I have never found a copy online.

Julien

Well, I read every other page of that as well as the 1991 and 1997 amendments and I'm sure some of it could be helpful somehow. I guess hacking on the hardware and datasheets themselves is what is required to figure out if every sample has been copied/read.

Even if one could find the Domesday hardware, it doesn't look like it would help, at least not out of the box. If I ever get a hold of a dead/trash player I'll hack on it. Until then, I'm forgetting about trying to rip every sample.

P.S. the specs are "out there", in several places in fact. Also, they're not very helpful... have you read them? I couldn't imagine if that's all you had to go on to build a player, the trial and error would be horrendous.

Each video field on a PAL or NTSC LaserDisc has several lines of digital information in the vertical blanking interval. There are codes for CLV/CAV positions, CAV stop frames, chapter numbers, safe fields to pause on, information about what audio channels are in use, and so on - see vbidecoder.h from ld-decode. So after moving the pickup, a player can tell where it is on the disc within a couple of fields. To find a given location, you make a guess about how far to seek, then see what location you ended up at - if it's not where you wanted to be, refine the guess and try again.

The digital audio track is separate from this - it's encoded pretty much exactly like a CD, with audio broken into blocks, then interleaved into a different block structure for error correction with multiple channels of extra metadata being woven in (see Simon's code in ld-process-efm for exactly how this works). The metadata includes a table of contents in the lead-in, a timestamp on each block, and other details like whether preemphasis is in use. LD-G subtitles are also encoded as a metadata channel in the digital audio track, as with CD+G.

So you've actually got two sources of timestamps while playing - the VBI lines in the video and the digital audio. in theory, at least - there are quite a lot of discs that are missing the digital audio timestamps for some reason, contrary to the spec, and real players don't seem to care. And many NTSC discs have two copies of the VBI data since there are both "MCA" and "Philips" formats on different lines. And some of the details of how exactly digital audio timestamps are meant to work aren't defined very well by the spec...

Because the digital audio track is decoded in blocks, it's not possible for it to just drop 2 bytes - you either get a complete block of stereo audio samples at the end of the decoding process (a "frame"), or you know it's invalid.

After looking through some of the ld-decode files, particularly https://github.com/happycube/ld-decode/blob/master/tools/ld-process-efm/Datatypes/sector.cpp, it appears that the digital audio is as you say, pretty much exactly like a CD. Since my player starts outputting a variable amount of samples much less than the block/frame size to the SP/DIF, it seems impossible to collect/read all samples without a hardware modification. Thank you for the links, this information clears up my interest in the matter

Out of curiosity, how is the RS correction data still present in the source file? What is used to read the file?

The RS correction data is baked into the CD/LDD format. The source 'file' is the EFM waveform that's in the Laserdisc RF, which is then demodulated and descrambled, and then there are two layers of error correction (and a third in the data CD format, which is used on Laseractive disks at least - I forget if it's on the Domesday disks or not)

While there can be a TOC in the CD/digital sound data (which is a side channel and will not show up in the audio), there's no TOC per se on the Laserdisc, chapter search on the player works by scanning the disk for the chapter flags in the vertical interval.

Thank you, I didn't understand that the "Doomsday Duplicator" was the source via USB 3.0. For me, that's a _LOT_ of money for those last few samples of silence . Although I guess you have to use that if you want some form of data other than music.

What's the LD scene like in Afghanistan?

Chapter, time and frame number is encoded in the vertical interleave. True TOC was added to discs in the late 80s, embedded into the PCM track. This include total run time, total chapter count and location of the chapters, and LD+G subtitles. If you do a chapter search on a disc without TOC, the player has to keep reading the RF signal to track the location. With TOC, the player knows exactly where the chapter marker is and search speeds are improved.