Recording YPbPr on VHS

[jmac698]

Yep, it's true, you can record component on a consumer VHS. Of course you need to specially prepare the video to do it, and a computer to play it back as well.

So here is a crazy experiment you might find interesting. What I did was separate the colors into different tracks much like Betacam SP does. I also split the screen in half to double the resolution. By my calculations, I'm getting about 480x480 resolution here in 4:2:2 color resolution.

I've also discovered a way to record component video live onto a VCR, or with any capture card, with very simple equipment. I can also record/capture 480p component video with any capture card (with only composite in). This doesn't help eurpoeans with their SCART.

[jmac698]

Btw, I just came up with a way to do any avisynth processing live. This is great! Not only can I watch my component video encodings live, but I can calibrate to colorbars or do tweaks to video realtime, and then capture it. That really enchances the usability aspect.
Use iuVCS, go to filters, select ffdshow decoder, in ffdshow check avisynth, in that tab write your script.
For example just write one line, tweak(bright=-10, cont=1.1) you will see the tv input change instantly when you click Apply.
Very cool!

[admin]

What are the specs of the system you're running this on?

[jmac698]

AMD XP 2GHz, quite an old slow system. Basic tweaks are easily realtime.
I have a script which can overlay the measurements of colorbars realtime on the incoming video, and tell you the error. It can also automatically calibrate as best as it can with you not changiing any procamp settings.
I discovered somrething else too. The perfect procamp settings aren't really perfect. You can take advantage of the headroom you get from capture cards and make white 254 and black 0, and processing scales it back to the proper range.
What's the point of that, you ask? Well, as I discovered in my VCR comparison post, there are slight errors in the middle values of brightness being too bright and skewed, so I use a larger range which spreads out the errors, then when I shrink the range back again, the errors are shrunk too. The result: better linearity of the input.

I also discovered a way to generate true RF, composite, svideo, or component signal directly with your computer with nothing but a cable. I can now directly generate timebase errors and scientifically test timebase correctors, rf demodulators, etc.

Also I found out a feature of bt878 chipset called Ultralock, which seems to be a line TBC - I didn't know this was already built in to my capture card. I'll have to see how well it works. I never noticed that it cleaned up jitter that much, but maybe you just need a registry hack to turn it on - in which case this search for TBC just became trivial - wouldn't that be great news?

I also thought of a few ways to read the raw video signal, including the horizontal sync. This means a "true" software TBC may be possible! I just can't believe no one has done this yet. One way involves fiddling with GraphEdit and using a component called "Sample Grabber". Another way is mis-calibrating a chip register to shift the picture far over to the right, so that you can literally see the sync. You can actually read a blacker than black code of 0 here and use that as a basis for a simple straigtening of video.

Lots of new developments! This is what basic research gets you.

[admin]

Interesting stuff you're working on.

Quite a few things cooking here too -- details of which will become obvious throughout the year.
Stick around a while, and you'll notice the changes.