Capturing laserdisc using component inputs

[Avisynth_challenged]

I wanted to post my simple laserdisc capture plan to the Doom9 community for critique.

I have a laserdisc that I wish to backup to DVD. It is an extended cut of a movie that is exclusive to LD. This movie is available on DVD, but in a shorter theatrical version.

I have a plan to capture the LD to a computer using the following capture chain:


Stage 1: LD movie source from LD player, using composite video out,

Stage 2: send the LD composite video output to a Sign Video DR-1000 Image Enhancer device, to increase video detail,

Stage 3: send the DR-1000 composite video output to a Sign Video PA100 proc amp, to set proper black and white levels for 480i North American NTSC, and to make any needed color or hue adjustments,

Stage 4: send the PA100 proc amp composite video output to a Toshiba RD-XS35 DVD recorder with advanced Y/C separation and component YPbPr video outputs,

Stage 5: send the Toshiba RD-XS35 component YPbPr video outputs to a PC which has the PDI Deluxe capture card with component inputs,

Stage 6: capture YPbPr component video with VirtualVCR to Huffyuv (720x480, 29.97fps, YUY2 colorspace).


My philosophy for this project is to obtain a capture of the laserdisc which is as close to perfect as possible for black/white levels, saturation, and hue. (I plan to use Avisynth after capture to tweak it to perfection.) But there is one pre-capture issue I think I need guidence on.

The question is whether I should adjust the source video's gamma during the capture process (which I believe is possible using VirtualVCR's gamma adjustment, found in its built-in proc amp), or should I adjust gamma after capture, using Avisynth.

The movie I'm converting contains bright daytime scenes and dark nighttime scenes. I'm thinking I can tweak the gamma so that it matches the official (and shorter) DVD version, which I have as a reference (.d2v file) that I can examine with Avisynth's Histogram filter.

My initial tests show that the laserdisc version has crushed blacks in night scenes, when compared with the DVD, for proper settings of black and white levels (0-100 IRE). A careful gamma tweak, either before capture or after capture, should allow me to bring out hidden details in dark areas without disturbing the light areas much.

Anyway, I thought I'd post my LD capture plan and see if anyone out there has some comments, criticism, or suggestions on the gamma issue, or anything else. Thanks for reading.

[Blue_MiSfit]

Sounds pretty elaborate :)

I wonder if all that conversion hardware would actually make a difference versus using a good LD player, a good composite cable, and a good capture card, maybe with some avisynth tweaking to clean things up.

Post some samples (!)

~MiSfit

[Avisynth_challenged]

Sounds pretty elaborate :)

I wonder if all that conversion hardware would actually make a difference versus using a good LD player, a good composite cable, and a good capture card, maybe with some avisynth tweaking to clean things up.

Post some samples (!)

~MiSfit


No kidding... but then I figure that the Doom9 community can certainly appreciate the elaborate approach.

I'll definitely post samples when I begin this project in earnest. Right now I'm trying to sort out how to get the best image quality prior to A-to-D, so that only a minimum of postprocessing is required to fix remaining issues.

For example, I've used a VirtualDub filter for removing "dot crawl" areas on some Huffyuv encodes of analog captures... but unfortunately, while it did clean up the dot crawl quite nicely, it also resulted in a perceptible loss of image detail.

Then I read this great thread (http://forum.videohelp.com/topic239205.html) on Videohelp.com about how hardware preprocessing prior to A-to-D capture can eliminate common issues (such as Y/C separation to eliminate dot crawl, and luma/chroma correction) before the analog signal gets digitized, thus eliminating some postprocessing and filtering steps which would invariably diminish the overall quality.

So I'll try the hardware preprocessing approach and see what I can achieve. Thanks for responding :)

[laserfan]

I truly wish you luck with this. I have 1000s of LDs in my closet, of which maybe a few dozen (mostly concert-type discs) still have unique value.

I've studied the same posts over at videohelp but haven't had much success w/my own equipment (includes a Sony color processor).

You lost me though at the "send the DVD recorder output back to the PC" step. At some point the all that transferring (and cabling) has gotta outweigh the ability of amps/procs/avisynth/filters to enhance? But please let us know if you achieve "laserdisc transfer nirvana"!!! :)

[Avisynth_challenged]

I truly wish you luck with this. I have 1000s of LDs in my closet, of which maybe a few dozen (mostly concert-type discs) still have unique value.

Thanks... I truly appreciate the sentiment. I've lurked here for years before finally registering. There's tons of good info here, and it's cool to kind of participate in the discussion of various issues that are encountered in this frustrating hobby :)


I've studied the same posts over at videohelp but haven't had much success w/my own equipment (includes a Sony color processor).

Just out of curiosity, what were you trying to achieve in your captures? In what way(s) were your capture results deficient with respect to your capture goal(s)?


You lost me though at the "send the DVD recorder output back to the PC" step. At some point the all that transferring (and cabling) has gotta outweigh the ability of amps/procs/avisynth/filters to enhance? But please let us know if you achieve "laserdisc transfer nirvana"!!! :)

The last step is when the detail/luma/chroma optimized composite video signal gets separated into 3-signal component (YPbPr) video. With the PDI Deluxe card's ability to capture YPbPr, I can (hopefully) achieve a clean Y/C separated video, and also avoid the bandwidth limitation of composite video output, versus component.

I'll keep this thread updated with any progress. Thanks to anyone who reads and/or responds. :)

[*.mp4 guy]

You should use the PDI Deluxe cards internal gamma adjustment, since it is done before the video is converted to 8bit yuv, and will therefore be higher quality then if it was performed later on.

[FlimsyFeet]

Well I'm by no means an expert but I do have some comments.

Stage 1 is the most important part of the chain, but you don't say what LD player you are using. I know there is a big difference in quality between the basic level models and the high end models that were only available in Japan.

Stage 2 sounds like a good idea.

I'm not sure if stages 3 and 4 are necessary - doens't the PDI deluxe already have "advanced Y/C separation" and levels and colour adjustments built in?

Finally, I've founf VirtualVCR is not very good at keeping audio sync so I reccomend VirtualDub instead.(mind you, this was with VHS which kept dreopping frames, so maybe you won;t have a problem).

[2Bdecided]

I find the typical doom9/AVIsynth philosophy (fix it all in software) and the typical videohelp.com philosophy (fix it in all hardware) quite at odds. You should use the best of both.

Be aware of cascading multiple stages of hardware 8-bit video processing. You can introduce horrible banding onto gentle smooth clean colour gradients. Shots of the sky, and sunsets/sunrises, often show this up.

Cheers,
David.

[scharfis_brain]

Id go this way:

- get a decent capture card.
(I found my SAA 7134 based one pretty good)
- play back the LD
- enable a live-view histogram of your capturing software (VirtualVCR)
- adjust the brightness/contrast/gamma/saturation sliders of the capture card's driver, until you maximized the range of the histogram WITHOUT clipping things.
(you have to do some trail and error to fit this range correctly in 16...235 luma range)
- capture in 704x480 (NTSC) or 704x576 (PAL) resolution and YUY2 and HuffyYUV compression.
- filter out remaining dot crawl or rainbowing with one of the various existing AVIsynth filters. (THe output f the SAA hip is fairly crips, so you won't loose harpness here)
- apply IVTC.
my comments on your planned process:

Stage 1: LD movie source from LD player, using composite video out,
OK
Stage 2: send the LD composite video output to a Sign Video DR-1000 Image Enhancer device, to increase video detail,
This can be done using AVISynth in a much better controlled way. most standalone-boxes usually only apply unlimited unsharp masking which enhances noise and haloing around edges.
Stage 3: send the DR-1000 composite video output to a Sign Video PA100 proc amp, to set proper black and white levels for 480i North American NTSC, and to make any needed color or hue adjustments,
This can be done using your video capture cards brightess and contrast controls. (also much more controlled due to a live view histogram)

Stage 4: send the PA100 proc amp composite video output to a Toshiba RD-XS35 DVD recorder with advanced Y/C separation and component YPbPr video outputs,
You really should compare the Y/C separation of your capture card and the one of you DVD-Recorder.
If the recorders seems better to you, use it.

Stage 5: send the Toshiba RD-XS35 component YPbPr video outputs to a PC which has the PDI Deluxe capture card with component inputs,
Only if you found the recorder to be better, else just use plain composite input.

Stage 6: capture YPbPr component video with VirtualVCR to Huffyuv (720x480, 29.97fps, YUY2 colorspace)
OK, but look for the left & right overscan.
If you capture 720 pixels width left & right there must be approx an 8 pixels wide border. If there is no border left & right, choose 704x480, if possible.

[laserfan]

Just out of curiosity, what were you trying to achieve in your captures? In what way(s) were your capture results deficient with respect to your capture goal(s)?I'd really hoped to improve the colors, but especially Black & White (!) i.e. Contrast. I mean, I had no expectation for improving sharpness, but especially given that I've wanted to transfer Concert videos I was looking for more "pop".

Gotta print-out and sleep with scharfis_brain's post under my pillow. Maybe I will assimilate something out of it. He's brilliant! :)

BTW for a while I thought about buying a better LD player (mine's only a CLD-3080 that I modded for RF AC3 out) but was stopped by the notion I might be throwing good money after bad (10s of thousands of $ on discs already). :o

[Avisynth_challenged]

I did some tests with the PDI Deluxe capture card and thought I'd post a followup to this thread.

The PDI Deluxe card has composite video in, s-video in, and component video in. So I basically tested all those inputs to see which one yielded the best capture.

As my source, I used the SMPTE colorbar pattern. The results are presented in the screenshots below, labelled TEST 1, TEST 2, and TEST 3.

No video preprocessing was done to the LD source for TEST 1. For TEST 2 the LD signal was converted to Y/C separated s-video by being passed through a Toshiba RD-XS35S DVD recorder. For TEST 3 the LD signal was converted to YPbPr component video by being passed through the same Toshiba RD-XS35S DVD recorder.

And now for the results.

TEST 1
Source: Composite video from laserdisc player:
http://i64.photobucket.com/albums/h183/partingshot/cbars_composite.png

TEST 2
Source: LD composite video passed through Toshiba RD-XS35S DVD recorder, Y/C separated s-video sent to PDI Deluxe:
http://i64.photobucket.com/albums/h183/partingshot/cbars_svideo.png

TEST 3
Source: LD composite video passed through Toshiba RD-XS35S DVD recorder, YPbPr component video sent to PDI Deluxe:
http://i64.photobucket.com/albums/h183/partingshot/cbars_ypbpr1.png

It looks to me that the best result is from Test 3. Comments are welcome.

[2Bdecided]

Wow - it looks like it's really sharpening the colour transitions. Does it work that well on real content? Some chroma decoding tricks only work on still areas/images.

btw, the levels seem different - especially on the last three bars.

Cheers,
David.

[laserfan]

I agree w/2B, Test 3 is a WOW!!??!! But the implication is that the Toshiba is the performer of the magic here!?

You ought to be able to see a radical improvement in the quality of real video program content--do you?

[FlimsyFeet]

Out of interest what does it look like if you record onto DVD (at the max quality level) then rip onto your PC?

(Thinking for those whose capture cards don't have componet inputs - like mine, whether the better comb filter makes up for the detrimental effects of MPEG-2 compression)

[scharfis_brain]

seems like the dvd-recorder applies CTI (chroma transition improvement), which is similar to warpsharpening.
Also it applies some sort of spatio-temporal noise reduction, which IMO should be switched off, if possible.

[neuron2]

You're going to have to find an image host if you want to post several large pictures like that.

[Avisynth_challenged]

Hi,

I've attached 3 jpegs to this post showing the raw YPbPr capture results of a NTSC LD title.

The big gripe for me is the right-edge haloing, which I cannot seem to avoid during capture.

I'll see if I can record an LD as an MPEG2 file with the Toshiba DVD recorder, and post that result later.

Thanks to all for your comments and feedback.

http://i64.photobucket.com/albums/h183/partingshot/jp01.jpg

http://i64.photobucket.com/albums/h183/partingshot/jp02.jpg

http://i64.photobucket.com/albums/h183/partingshot/jp03.jpg

[PatchWorKs]

...and what about capture through a DV-in camera ? ;)

[*.mp4 guy]

Thats very good Y/C separation, I don't think your going to get much better then that, I would definately guess that thats better quality then any pc capture card short of high end production equipment will be able to deliver.

[laserfan]

The big gripe for me is the right-edge haloing, which I cannot seem to avoid during capture.Looks frankly sensational to me. Can I send you all my LDs? Only a couple thousand of them.. ;)

[Avisynth_challenged]

Out of interest what does it look like if you record onto DVD (at the max quality level) then rip onto your PC?

(Thinking for those whose capture cards don't have componet inputs - like mine, whether the better comb filter makes up for the detrimental effects of MPEG-2 compression)


I finally did a test for this. Here's what I did, followed by the results...

1. Recorded a laserdisc to DVD-R using Toshiba DVD recorder. I was able to manually set the recording bitrate, and I set it to the maximum allowable value of 9200Kbps. (According to the manual, this setting would result in a recording time of about 60 minutes per DVD.)

2. Finalized the DVD-R

3. Ripped the DVD contents to a PC

4. Opened the VOB files with VirtualDubMPEG

5. Took two screengrabs and exported them as JPEGs (I tried to do lossless PNG, but the image hosting site I used seemed to have a problem with displaying them when hotlinking.)

Here are the screenshots. I selected them on the basis of fast motion content which would show macroblocking for low quality MPEG encodes.


IMAGE 1 (B-frame)
http://i64.photobucket.com/albums/h183/partingshot/t2mpg1.jpg

IMAGE 2 (B-frame)
http://i64.photobucket.com/albums/h183/partingshot/t2mpg2.jpg

[FlimsyFeet]

Thanks for doing that. Actually I was more interestred to see if with that method you still got the good y/c performance, as shown in the captures of the colorbar pattern, rather than the motion macroblocking. Sorry I didn't make that clear.