happycube

"baked in" rainbowing on DVD's is the result of bad comb filtering in the DVD mastering process - the analogue on LD's would be comb filter failure for whatever reason.  

There are encoding techniques that reduce rainbowing etc such as SuperNTSC, and Technidisc apparently never used any of them.  So there are areas where there's a lot more rainbowing than say the DC LD, like the Star Destroyer shot as it picks up the Tantive.

Here's a comparison of one of the test lines in every frame of most non-THX disks... - left is the Pioneer reference disk (frame 6000 or so), right is the Technidisc SWE:  http://imgur.com/FErYGpa - the Technidisc color area on the right is simply really noisy.  (A lot of LD's are)

Technidisc vs DC:  http://screenshotcomparison.com/comparison/146021 and http://screenshotcomparison.com/comparison/146024 - both of them were done with 3D comb filtering.

As for the JSC, it should be possible with a fancy player to reframe the picture to cut out the Japanese subtitles.  I bet VLC can do it, and mplayer/mpv with the right options definitely can.

That's a different thing - the PC1 PAC is reading the CD[-ROM] part of the disk (LD digital audio is literally CD audio's RF put through some filters), kind of like a CD rip (especially if it can pull audio)  ld-decode works on the analog video and audio portions.

The graphics card is used as a GPGPU (vector processor) that can run things like FFT's and other parallel operations much faster than a CPU can.

Audio noise implies that it's a bad disk.

LD quality isn't 100% consistent after all, and the plain fox copy appears to be a 1982 Pioneer Japan pressing - some of their early disks were less than great and I don't know if that first version is CAA or not.  A later pressing of the CBS/Fox release (especially post-1986 Pioneer USA) is probably better.

There's also the 1992 CLV/CAA pressing by Mitsubishi, which is probably the same video master.

I'll see what I can dig up this weekend on my end.

Huh.  If this has that many shifts a (semi-)automated tool might be needed.  But IMO I think the time compressed disks should simply be posted as raw 480i, with corrections left to interested tinkerers.

I got a GeForce 740 the other day ($80 refurb, 4GB DDR5) and started playing with CUDA - I converted the first bit of ld-decode and accelerated video FM decoding by ~4x - when the card is cool enough to run at full speed, that part gets to about 70% real time now. :)

With more work than I can probably afford to put in, and a Geforce 980 or somesuch that I can't afford yet, ld-decode could probably be made to work in real time.

(no updates on digital audio, but at least that's not too hard to capture from many players...)

EDIT:  Installed the Nova S Plus I got.  Looks like it'll work (with vmux=1 on the cxadc insmod line) but it has noise at 5mhz which will likely cause issues.  *sigh*

On another front, it might be possible - albeit with modified firmware - to use some of the cheap USB FX2-based oscilloscopes as ADC's for this.

(I hadn't checked this bit of the forum in a while - oops!)

I use the Hauppauge WinTV 34132 which uses the Connexant CX23880 chip.  None of that specific version are on eBay right now but the last two that sold were cheap!  There's a hacked driver for it I updated and improved, and it's usable on other cards with the same chip.

edit: There's an eBay seller with >10 Hauppauge Nova-S-Plus cards (CX23883, composite input w/o adapter) for $13 each (shipped) - ordered one to see how it works.

https://docs.google.com/document/d/1n5VArQenHNHE6e7Azq5iAt0L0Cr752LGJs7Vu7OYoRg/edit?usp=sharing is a (somewhat) live document with a lot of info scattered throughout. 

Good luck poita!  Recover quickly and well :)

Some Laserdiscs have more aliasing than others... the JSC could probably especially benefit from SR.

Probably the most effective way to set up a DSLR would be to lock the shutter open (either by hardware or software means) and then strobing LED lighting when you want to take a shot to prevent the jello-shutter effect CMOS sensors have taking video (which is why the shutter would still trigger for video)

Bonus points for removing the color filter and strobing R G and B LED's, one shot per color.  An ideal implementation of this could be quite effective for it's price, if a bit slow.

There's always loss in the printing process, and a 70mm has much less loss than a 35mm print.  Also there's a good chance the process was done more carefully/slowly for the 70mm print, further increasing quality.

It wasn't troublesome for me until today.

(And a 925 should do just fine)

https://docs.google.com/document/d/1n5VArQenHNHE6e7Azq5iAt0L0Cr752LGJs7Vu7OYoRg/edit?usp=sharing 

I'm not sure the RF from the V4300D is as strong as the 2950/x03 type, but it'll do fine for at least PAL captures.  An amp circuit might be needed for best results (in general - I just need the time to design/build/test a simple one and this weekend isn't it!)

Once we get there, 640K should be enough for anybody.

But seriously, it's great to see all this coming together.

The BluRay has more detail, but this has soul.

How good's the lossless compression available to you?

The 2950/704/99 type have given better RF captures - they have an RF gain adjustment pot on the board unlike the late players.

It'd be nice to have a more capable ADC, but there's nothing vaguely in the price range that does the job... yet.  An amplifier would help too, I've been meaning to play around with some high bandwidth op amps one of these _'s.

Already posted this to the super8 thread but moving it over here where it's a bit more on-topic - here's how ld-decode compares with the X0 project capture:

http://screenshotcomparison.com/comparison/128827

ld-decode has more noise and comb filter aliasing, but considering the source player is a repackaged cld-s104 (v2800) and the total equipment cost, not too bad.

edit:  Made a version with some NR (0.25IRE chroma, 0.7IRE luma).  Still not a match for the X0, but a bit closer - http://screenshotcomparison.com/comparison/128842

And one more with Chroma LPF, which closes the gap on C3P0's hand etc http://screenshotcomparison.com/comparison/128845

Thanks, and LOL! :)

(and agreed - that's why I made sure to mention the film shot :)  I'll move the LD/ld-decode bit to althor1138's thread, since he's got it set up as well. edit:  got a comparison linked there with some NR turned on which closes the gap a bit...)

Thanks!  After doing adjustments for positions and brightness levels (the X0 was set to black at IRE 0), I can compare them directly.  ld-decode is a bit noisier and the comb filter misses a couple of spots, but it holds up alright I think...

http://screenshotcomparison.com/comparison/128827

Neither one is as detailed as the film sample, of course.  I'm too lazy to resize and color correct it, but viewing that .jpg at 1280x580 versus the LD screenshots reveals so much more detail especially on the droids.

team_negative1 said:

Thanks for the comments. To be fair, we did compare it to the bluray. And the Bluray is still better for detail.

 So much for my idea of nicknaming the SW blu's the "blurrys". ;)

There's a decent chance the 1997 Kuraray Faces pressing was also SuperNTSC:

http://forum.lddb.com/viewtopic.php?t=2773&p=38314

As an aside, I just looked at side 1 of the DC, pressing quality seems good but the mastering does have some oddities - no "white flags" for indicating start of frame (probably part of the MCA spec, but still odd) - and especially odd for a THX disk, not much of a VITS!  Just a bit with 100IRE and a drop to 0 in the middle.  No THX test signal line or standard NTSC test lines either.

Anyone got uncompressed X0 screenshots so I can some comparisons against my stuff?  Thanks!

It's a pity that the X0 project wasn't finished, but the film projects more than make up for it ;)

Very nice improvement!

No need for a branch (probably) - it'll be a matter of finding and implementing the best settings, maybe an extra comb filter switch or two.  It's probably time I look into EFM/CD decoding though - if my test mode trick for playing PAL disks actually works I'll need it.  (It sort of does, but there is color distortion that may or may not be from the NTSC->PAL conversion the disk I have uses)

I also might document the TBC output so that people can play with making new comb filters.

The biggest actual issue with the JSC is that it was simply too sharp for 1986 LD mastering/production technology.  It has a high color bandwidth but was done before SuperNTSC, so it is very hard to get rid of all the rainbowing, and the sharpness comes with noise.

But there's a lot in there, if it can all be dug out cleanly.