How do you address YC delay problems? Y/C delay cable length calculation?

[aramkolt]

Hello all, I've noticed that some of my captures and live display have a fair amount of YC delay meaning that the chroma signal is displayed far enough to the right of the luma signal that it is noticeable. My understanding is that each pixel of delay is about 74ns (nanoseconds) and when I adjust my TBC/Framesync which has a digital readout of YC delay, it comes out to -296ns to get them to roughly match, meaning that luma is arriving 4 pixels before the chroma signal. Since this is VHS, it is unclear to me if the content may have been recorded that way as most of what I have been using for testing are not first generation tapes. It's also possible that the VCR it is coming out of isn't ideally adjusted, however, most VCRs don't seem to have much in the way of a user adjustable Y/C delay in regular menus or even discussion of it service manuals. The VCRs that I've seen that do have YC delay adjustments tend to be broadcast machines that have combo line/frame TBCs such as the SVO-5800. I looked for YC delay in the AG1980 manual and while there are quite a few adjustments you can make, YC delay doesn't seem to be one of them, or at least not one described in the service manual.

I've read that one could lengthen cables of S-Video's connection of either the luma or chroma to make the delay match that way, but is there an actual calculation on how much cable should result in how much delay? If I entered the correct number of zeros on a "speed of light" calculator (since electricity travels at a similar speed) it would take something like 20 meters of cable to delay the signal by one pixel horizontally which doesn't seem ideal when you're dealing with 4 pixels of delay - signal losses and hassle would make this unreasonable to do in practice.

I also understand this can be addressed in post processing with Avisynth, just seems like a rather annoying step if it's possible to correct in hardware at the time of capture. Some of my hardware allows for this, but in testing a bunch of devices for performance, most of them will not let you adjust YC delay in hardware, but if you were using one of the recommended TBCs like the TBC-1000 or AVT-8710, they do not have YC delay settings.

I also would imagine that most people have never really thought of YC delay as being something they could address in their captures and probably think it represents chroma smearing or bleed which it very well may not be.

Curious how others handle that problem when it arrises, and does anyone know if YC delay is more of a problem in formats that have separated luma at the output like component and S-Video? Perhaps composite has some immunitybecause it is all sent over a single wire as a single waveform?

[Sharc]

In coaxial cables signals propagate with roughly 2/3 of speed of light. Depends on the cable design (dielectric constant, geometry etc.).
For a luma sampling rate of 13.5MHz your cable length will be about 14.8 meters per pixel horizontal delay.

In the attached testfile the center row is perfectly Y/C aligned. Rows above and below are offset by +/-1, +/-2, +/-3 pixels
I would address such shifts in post processing, together with chroma sharpening for VHS sources. Your 4 pixels shift may for example be something like 2 pixels shift plus 2 pixels smear caused by chroma subsampling upon digitization. And yes, VHS dubbing aggravates the shifts, including vertical shifts.

[pandy]

Unequal cable length may lead to propagation differences but normally this should be not the case - in video circuitry most propagation differences between Y and C is made in filters introduced in signal path (so called group delay https://en.wikipedia.org/wiki/Group_delay_and_phase_delay?useskin=vector - it may be not linear and frequency dependent) - normally this should be addressed by encoder/decoder (digital encoders offers possibility to delay/advance C vs Y).

[timtape]

I believe VHS color can only resolve 40 lines horizontally whereas for the luma it's 240 lines horizontally. A color line is a pretty wide line. Are we talking a displacement in luma lines or color lines?

[Cornucopia]

In analog, EVERY horizontal line is discreet (vertical resolution), so luma or chroma, both should resolve to ~240(NTSC)/288(PAL) visible lines (per field).
Where chroma is weak compared to luma is in the horizontal resolution.
Similar for any medium (broadcast, tape).


Scott

[timtape]

Where chroma is weak compared to luma is in the horizontal resolution.

Scott

Yes only 40 lines, as I said. Blurry compared to the luma's 240.

[Sharc]

The cable delays the continuous analog VHS signal coming out of the VHS player. This has nothing to do with resolution or color subsampling. Sampling and quantization is done by the subsequent digitization only. There are no pixels in the analog domain. So one can actually add any delay. Its not tied to a raster. A cable in the C wire delays the sync+luma, a cable in the C wire delays the modulated chroma.
Pixel is just used a reference measure here, instead of talking nanoseconds or inches on the monitor.
Whatever, I would do the corrections in post, solving chroma sharpening and temporal/spatial misalignment issues overall.

[Cornucopia]

Yeah, but "lines" horizontally makes no sense, from an analog video perspective.


Scott

[BW37]

Yeah, but "lines" horizontally makes no sense, from an analog video perspective.


Scott

I couldn’t agree more and it always confused the heck out of me yet…


[Attachment 84318 - Click to enlarge]

It was always the standard of the industry meaning number of vertical lines that can be resolved horizontally across the screen. Talk about obtuse!

BW

[lollo]

I do not understand the confusion.

The horizontal number is the quantity of information inside a line:



384 for PAL TV signal at 5MHz
240 for VHS

The vertical resolution is the number of lines: 480 or 576 (for PAL 625 pieces of waveforms each during 64usec, where only 576 contain video information)

No pixels here so far, we are in the analog domain.


The next step is how to digitize such a signal. For PAL TV applying Shannon the magic number is 384x2 = 768(x576) pixels (sound familiar?); in practice there is never that amount of information so 704/720 pixels are more than enough -> 704x576/720x576 pixels (sound familiar?)

For VHS 352(x576) are more than enough but we prefer to adhere to the 601 spec and use 720x576

Just as simple as that!

[timtape]

Yeah, but "lines" horizontally makes no sense, from an analog video perspective.


Scott

I couldn’t agree more and it always confused the heck out of me yet…


[Attachment 84318 - Click to enlarge]

It was always the standard of the industry meaning number of vertical lines that can be resolved horizontally across the screen. Talk about obtuse!

BW

In that specs sheet perhaps the most important technical spec is omitted: the horizontal color resolution of only 40 lines. I guess they didnt want to make it too plain that (compared to the broadcast TV standard of that day) in SVHS the color was still just as weakly defined as it was in VHS, and other consumer VCR formats. So they only mentioned the improvement in luma resolution.

[pandy]

I think term "lines" refereeing to resolution "wedges" visible on many older patterns like famous 'Indian head'

[aramkolt]

Thanks for the responses, after more testing, it does seem to vary so much from tape to tape that having different cable lengths and the possible losses associated with that is probably just better to have a TBC or decoder with Y/C delay features - it's just interesting that it doesn't get talked about too much as I image it is relatively common on non-first generation tapes, or low production run manually dubbed commercial tapes.

What's interesting to me is that when I do a YC delay with a CVR250, the captured file via SDI output shifts the luma and not the chroma in the frame, so the main image that you see basically changes where the apparent "black bars" are on either side that usually consist of a total of 16 pixels which is kind of interesting. Suppose it makes sense, but I somehow thought the picture would remain more or less centered - I guess chroma is still in the center, but luma is not haha.

Furthermore, I had one capture where there were almost no black bars on either side which kind of suggest that somehow that tape had a "longer" window of horizontal picture data, or less time between sync pulses to display more information horizontally. Didn't even know that was possible.