s-video artifacts?

[sanlyn]

It will make it more blurry, that's the side effect of using antialiasing

If you don't need it, remove it. Same with all the other filters. If you can get better results with fewer filters, not only will it be faster to process, better quality. "Filteritis" can be a bad thing

Agreed. The script I posted is probably OK without the daa() function, and could do with santiag alone. It's an effort making those drawn lines look solid, there's so much wiggling and blurring from bad line and frame output.

[zerowalker]

Well i didn´t use the Daa() as i didn´t even know what it did, instead i changed place and use Santiag(6,6) in the motion compensation script right before Dfttest, and it seems to work well;S

[davideck]

There were no observed improvements made to multiple types of errors. It behaved no different than a JVC recorder...

Very interesting. That JVC DVD Recorder is at the opposite end of the TBC performance spectrum from the Philips unit, as it is particularly inept at handling the types of timebase errors often present in VHS camcorder tapes. The timing disturbances in Mini-Me's clip of walking down the stairs could actually be exacerbated by the JVC and made noticeably worse. For this reason, the JVC is one DVD Recorder that I would not recommend as a passthru TBC. Even the TBC-1000 outperforms it.

If there was no difference in behavior between the Philips and the JVC, then your multiple types of errors must not have included appreciable timebase errors. Or perhaps you used your JVC TBC/DNR VCR as the source without turning off the TBC/DNR.

[sanlyn]

Most if not all of those are NTSC right?
As i need PAL, and it´s well, alot harder to get S-video and stuff here thanks to SCART;O

I don't know why I didn't notice this statement earlier. SCART is as good as s-video and, if I understand it correctly, isn't some of its functionality similar to YPbPr ?

[jagabo]

SCART can carry composite, s-video, or RGB. Some of the pins are shared so you have to specify what to output on the source device, and what to input on the receiving device. Not all devices support all three modes.

[zerowalker]

Yeah and we send composite on 99% of all vcr probably.

[zerowalker]

Has there been any "solution" for this?
Meaning, can it be that the output may be wrong (240p-480i thing), or something along the line?
And that Dscaler may be able to fix it with special resolution settings or something?

Or is it just permanent, meaning it´s just a bad composite to S-video?


Note:

I got the problem on Both Xbox and PS2, but both of them show the Checkboard pattern, which is pretty much the same every even frame.

And i also got it on my Toshiba (Stated in another Thread), but it´s more like a diagonal line pattern (if you use deinterlace it appears like that atleast).
It looks pretty much like Herringbone, but i can conclude that it isn´t Dot Crawl (Pal got similar dot crawl) if the device isn´t always outputting composite and upscaling to S-video for some weird reason.

I am pretty sure it´s permanent though, but i was wondering if there could be some grounding issue with the S-video part, but have hard time seeing that as they are at the same place as the Composite.

[pandy]

I got the problem on Both Xbox and PS2, but both of them show the Checkboard pattern, which is pretty much the same every even frame.

Sounds like CVBS on Y line - or use CVBS to capture or configure source for S-Video (Y/C shall be sent over independent wires).

[zerowalker]

CVBS = Composite?

The thing is here with consoles that i capture.

With S-video it looks Very Very Sharp, But i get the checkboard.
With Composite, it looks very blurry, watchable yes, but the details i get with S-video blows it away even with the artifacts.

(though for VCR capture i can´t afford checkboard of course, but just telling that composite looks much worse for me)

[pandy]

Issue is that capture card at S-Video Y input not expect to have Chroma (C) signal - pattern is a exactly C signal combined with Y signal.

So or You configure Source to produce REAL S-Video ie Y (Luma) and C (Chroma) signal on different wires or You should not use S-Video input - as a workaround You may try to remove C from Y with help of for example some FFT filter ( Avisynth Defreq(), FFTQuiver() ).

[zerowalker]

Issue is that capture card at S-Video Y input not expect to have Chroma (C) signal - pattern is a exactly C signal combined with Y signal.

So or You configure Source to produce REAL S-Video ie Y (Luma) and C (Chroma) signal on different wires or You should not use S-Video input - as a workaround You may try to remove C from Y with help of for example some FFT filter ( Avisynth Defreq(), FFTQuiver() ).

Yeah i know that it sends same signal through both, but still the quality in resolution and sharpness is Alot better than composite.

How do you mean configue the source to produce Real S-video?
It´s a console (or DVD), i can´t really do anything, and i can´t really see anything within the machine as i don´t know what to look at.

[sanlyn]

Issue is that capture card at S-Video Y input not expect to have Chroma (C) signal - pattern is a exactly C signal combined with Y signal.

So or You configure Source to produce REAL S-Video ie Y (Luma) and C (Chroma) signal on different wires or You should not use S-Video input - as a workaround You may try to remove C from Y with help of for example some FFT filter ( Avisynth Defreq(), FFTQuiver() ).

I realize there might be a language barrier here, but what do you mean?

If yjr O.P. has this problem thru multiple input sources at capture, I'd say something's going on with the way the capture device handles s-video.

[pandy]

Capture device for S-Video configurations use 2 independent A/D converters, path for Y signal assume that there is no ANY Chrominance signal (or at least chroma is bellow noise floor/quantization level), similar case for C path - nothing around 4.43MHz (3.58MHz for U.S.) is expected.

So if You feed to Y A/D converter composite signal (CVBS) it will be gray-scale signal, with wide bandwidth but also with Chroma visible as interfering pattern - i assume this is case described by topic author.

So real S-Video source means only that source is able to output independent Y and C signal (or in other words - some sources can have S-Video output but on both outputs i.e. Y and C, CVBS signal can be avaiable thus S-Video capture shows color video however with combined artifacts due lack separation Luma from Chroma).
Usually for video sources without separate S-Video output (typical case for SCART products), video source need to be configured that output video signal is S-Video not RGB or CVBS. This was my pint when i've said: correctly configured video source.
For U.S. and countries without component video output standard, S-Video should be available as a higher quality (than CVBS) video output.

[zerowalker]

The problem is that it can´t be configured, and that the quality is still better even though composite is sent over S-Video.
Here is a picture.

comp left, S-video Right.
Big difference.

[pandy]

So i see only one solution - capture twice, one of the pass with CVBS, second pass with Y only, then separate Y plane from CrCb and combine. Y plane can be filtered with one of mentioned earlier filters - luckily chroma pattern have/or it is very close to one frequency.

[zerowalker]

I don´t really get what you mean.

Do you want me to capture one time with Composite, and the other time with S-video, them combine Luma from S-video and Chroma from Composite?

[sanlyn]

I think the way zerowalker is using his composite-output VCR, is that composite output is going into a DVD recorder used as a pass-thru device. The DVD machine converts composite to s-video by internally separating Y and C. The DVD machine has a dedicated s-video-only output (does not require an SCART adapter).

Also, I'm having a difficult time trying to accept cross-hatching as looking like dot dot crawl in point-resize enlargement. IMHO the crosshatching looks more like aliasing due to simple point-resize enlargement (it goes away when other enlargement methods are used), and this dot crawl doesn't seem to have any dots in it and occurs only on soft, non-straight edges. Much of it looks like CUE (chroma upsampling error) which appears most prominently in non-detail areas of deep red (*see below). The other point I'm having trouble with is any kind of recurring diagonal gray lines being referred to as herringbone noise. Ground loops or other power problems usually look like stationary or rolling gray bars ("hum" bars), but AM or FM interference has the herringbone or chevron pattern (>>>>>>>>>>>). Both types of noise can result from any number of sources, including failing or aging electronic components, poor device design, noisy power sources, connecting cables with poor insulation or faulty connectors, tape stored for long periods near devices with strong magnetic fields (such as loudspeakers or TV picture tubes), or any of an almost unlimited number of causes.

* The CUE I refer to doesn't mean YUV-RGB conversion as in Avisynth or VirtualDub, but differences in the way Y/C -YUV - YCbCr material is handled/interpreted by various devices.

That's just my take on some of these disturbances. Frankly I don't think it's possible to get an absolutely clean VHS transfer to digital video, even with lab-quality equipment (and digital video isn't all that perfect, either). Considering the way tape playback systems operate mechanically and electronically, it's a wonder they work at all.

[pandy]

I don´t really get what you mean.

Do you want me to capture one time with Composite, and the other time with S-video, them combine Luma from S-video and Chroma from Composite?

yep - if You insist to have "sharp" Luma and color and Your current configuration behave so oddly then yes - this is dirty hack to deal with software/hardware limitations.

[pandy]

That's just my take on some of these disturbances. Frankly I don't think it's possible to get an absolutely clean VHS transfer to digital video, even with lab-quality equipment (and digital video isn't all that perfect, either). Considering the way tape playback systems operate mechanically and electronically, it's a wonder they work at all.

I don't want to start discussion about perfect capture - luckily VCR (AFAIR all systems) use same rule to record video on a tape - FM modulated carriers for SEPARATE Chroma and Luma signal thus on tape ALL video more or less recreate S-Video idea. Only question is how source was configured and how VCR deal with video internally.

[zerowalker]

Okay it seems maybe both of you has missunderstood me here.


The picture above, and what i am Currently talking about is ONLY in Consoles where they have the Cheap hack.

What i mean is, why does S-video look so much better in resolution (it´s like 320x240 to 640x480), when it isn´t even S-video, but composite transfered over both?

When i am going on the VCR as sanlyn is talking about.

VCR -> DVD, Composite to S-video, i don´t get any improvement, it´s just worse.
Though for dvd material, S-video looks much better than Composite, IF you ignore the crosstalk (Which i can´t ignore for important stuff).

But i can add that i am also curious to why i get the Herringbone crosstalk in my DVD, as a random VCR-DVD combo doesn´t get it. Concluding something is wrong, probably cheap splitter.

[zerowalker]

And as an answer to the Perfect capture.

I know that with analogue sources it can´t be Perfect, as it´s different each time, as noise is added, you can´t capture Only what´s Really on the VHS, it´s just how it´s built (or atleast so i understand it?).

But what i, and probably you guys want is the "Best". Meaning, a good VCR, with Good transport, and S-video(True S-video).
Then the rest is up to manually make it "perfect", or saving it lossless for archival.

PS:

Does anyone know if NV-SV 121 has the S-video artifacts?

[pandy]

S-Video due hardware separation (independent wires) don't need to separate chroma and luma - usually this is made in a way that luma have something named "notch filter" with center on chroma burst frequency (4.43 or 3.58MHz most common however precise value depend on current color standard). So softness is introduced by filtering signal, there are some com filters trying to subtract chroma from luma however it is very difficult to separate those two combined signals.

But if You are interested then:

http://www.jim-easterbrook.me.uk/pal/

http://www.techmind.org/vd/paldec.html

and check chroma trap

http://books.google.nl/books?id=6dgWB3-rChYC&lpg=PA845&ots=1JAhHCPxQ8&dq=chroma%20trap...20trap&f=false

[sanlyn]

Very well, that's sophisticated material. But how to implement the fix? Modifying circuitry is a large undertaking, and the software filtering discussed assumes some technical knowledge not available to hobbyists. How would a one use this in a home environment?

[jagabo]

What i mean is, why does S-video look so much better in resolution (it´s like 320x240 to 640x480)

The difference is nowhere near that large. You're being fooled by the chroma noise (rainbow artifacts). If you eliminate the chroma noise (ColorYUV(cont_u=-256, cont_v=-256)) you'll see there's only a tiny difference in sharpness in the luma channels. The composite cap does have an echo though. The sharpness difference won't be noticeable on VHS caps where the horizontal resolution is only about 360 lines (across the entire frame). And the echo probably won't be noticeable because of the low resolution.

[pandy]

(ColorYUV(cont_u=-256, cont_v=-256))

or Limiter (0, 255, 128, 128)

[pandy]

Very well, that's sophisticated material. But how to implement the fix? Modifying circuitry is a large undertaking, and the software filtering discussed assumes some technical knowledge not available to hobbyists. How would a one use this in a home environment?

This was my point - it not trivial task to separate chrominance from luminance and sometimes We need to accept some limitations or imperfections when color is decoded.
Without chroma burst we can't also use S-Video Y to decode color in software.
Unless issue can't be fixed by hardware perhaps CVBS need to be sufficient.

btw
http://forum.doom9.org/showthread.php?t=162832

[jagabo]

Zerowalker, the evidence of all your chroma/luma crosstalk problems is spread over several threads and many posts. I suggest you summarize all your evidence in one post, maybe even start a new thread with it, so everyone can easily analyze all your situation.

[zerowalker]

Zerowalker, the evidence of all your chroma/luma crosstalk problems is spread over several threads and many posts. I suggest you summarize all your evidence in one post, maybe even start a new thread with it, so everyone can easily analyze all your situation.

First the answer to the First about the Resolution.
I see, then it´s probably as you say, but i really like that the S-video doesn´t Blur the text anyway, but a Console gives full resolution which a VHS doesn´t.


Now to the other Quote.

Yes i have been kinda spreading, but there is some difference between them, and that´s why i have been asking here also.

When i first captured Xbox, one posted in my other thread and linked this one, here it was all the same (checkboard noise).
Then the discussing stopped, and i got another "problem" with the DVD, but it´s not the same look as the checkboard noise.

I then started the question about the Xbox checkboard noise again to ask why S-video looked to much better for me Even with noise.
Then you answered that it was more an Illusion thanks to the colors, but that there is some improved Luma sharpness.

I can´t really post the different stuff and let people analyze, as with the consoles atleast, it´s as you say, just composite over S-Video, not much to do except dot crawl filter, though it works better than using composite for me atleast.

The DVD is another issue but similar, and don´t know what to make of it, it´s probably some other cheap way of Y/C splitting, but not as "Clean" as the consoles which seems to completely just send composite.

[ybarra]

Hate to dig up an old thread, but I ran across this a while back...

The checkerboard pattern that the OP and several others have shown is a defect in virtually all of the JVC SVHS decks. I have an old 7600U as well as a "newer" VS101 and both have the exact same problem with S-Video OUTPUT. It doesn't matter what the source is whether it's an actual tape, or a video passed through the JVC, but anything coming out of the S-Video output of the deck itself will always have the checkerboard noise pattern. I first noticed it years ago when trying to record to DVD as it drove the mpeg encoder crazy on my DVD-R resulting in odd compression blocking. When I switched to composite, the problem magically went away. Later I moved to capture via DV (firewire) pass through because I didn't have an actual capture card. The DV compression didn't freak out over the checkerboard like the mpeg did, but it did archive most of it as I can see the noise in some of my older captures. Switching to composite gets rid of the checkerboard again. I finally got an actual capture card and have done some direct captures to uncompressed avi and sure enough the checkerboard is there. Without stripping the noise, it adds about 50% in file size to my h.264 encodes because the encoder is trying to preserve the checkerboard. If I capture direct using composite, boom...clear pic with no checkerboard.

Summary, almost all prosumer JVC SVHS decks have this checkerboard noise coming from the s-video out ports regardless of what setting you have in the menu. The only way to avoid it is to use the composite output with a slight sacrifice in the signal quality (especially from SVHS sources) as has already been pointed out. Personally, I prefer the slight signal sacrifice to the artificial grain the checkerboard creates.

[zerowalker]

Yeah sadly the only thing you gain with S-video in these cases or a very little bit sharper picture, but i think it depends on how the capture card handle S-video vs composite. But the signal is the same in S-video and Composite, the only thing they have done is send Composite in S-video Luma/Chroma pins, which is why we get these checkboard patterns, so there isn´t really any gain, and just loss as we get artfacts which Can be removed, at cost of detail. With Composite it may be a bit softer, but you don´t lose any details compared to the fake S-video.

I find it weird that JVC has done this, as if i have understand it correctly, JVC is the Top Notch in the VCR industry.