Chroma bleed on capture?
 

hi all,

i'm noticing light chroma bleed on all my captures. i believe at least some of this is happening either at the card or the frame tbc level.

my workflow is

ag-1980p (recapped by deter) -> s-vid -> avt-8710 -> s-vid -> aiw breakout cable -> aiw 9000 -> vdub @ 704x480 YUY2

the attached photo is an avt-8710 test pattern frame from the capture which clearly shows chroma bleed (i think).

is it possible to eliminate this at the capture level? is this something anyone has seen before? or do i need to just do the regular troubleshooting thing of replacing components one at a time =\

thank you!

No photo attached to your post.

Is the same unusually bad chroma bleed present without the AVT in circuit?

i don't know if it's exactly "unusually bad" -- i can confirm that there is chroma bleed present without the avt in the circuit but that might be "normal" vhs bleed.

trying again with the attachment.

That's a test pattern with 100% white, 0% black, and 75& color panels. I don't see anything unusual. You should test with a capture of real video and compare the video to AVT in circuit -vs- AVT out of circuit. The test pattern itself is not conclusive. The image itself has been converted is lossy compressed jpg in RGB. Hopefully you're not capturing in RGB.

i'm capturing in yuv422. unfortunately i don't know how to capture a frame in a lossless format with yuv422 pixels. png and bmp support rgb/bgr only afaik.

if you zoom on the test pattern you can see the bleed -- at least i think that is what it is?

see attached for a zoom. this is a png screenshot of a video which will have rgb24 pixels -- but it's already been converted to rgb for display by the player anyway.

er, sorry, no it hasn't.

but it seems unlikely to me that the yuv422->rgb24 conversion would cause this kind of artifacting.

Most people capture analog as 4:2:2 YUY2, 720x480 using lossless compression such as huffyuv or Lagarith, which works in Virtualdub.

Test patterns are all very nice, but your zoomed image shows herringbone as well as inter-panel bleed and a lot of lossy compression noise and smearing. If you're using a newer AVT unit, they are known to have quality issues. A totally clean unit is difficult to come by. That AVT test pattern isn't a reliable tool anyway, it's of poor use in judging real captures.

i am capturing 4:2:2 YUY2 704x480 with huffyuv. my understanding is that 704x480 is the native capture resolution of my video card (aiw 9000).

there is no compression anywhere in the chain for that image afaik.

so perhaps an avt issue?

understood that whatever that issue is might not have a huge impact on the final capture quality.

i am using a newer avt unit -- it seems like there's no bulletproof option for a frame tbc? at least one that is easily obtainable.

I see nothing unusual here.

All TBCs have minor quirks. Remember, the TBC is processing the video, it's not transparent. It's not a magic box where you input bad video, and it spits out clean video.

The Cypress/AVToolbox TBCs have a very modest checkerboard pattern. Many are almost imperceptible, some are more obvious. The other equipment in your workflow, including wires, and even the power coming into your devices, have some affect on it. It's also not a model-line thing (all AG-1980P VCRs, for example), but that specific unit.

That's what you're seeing. You can make it transparent -- I've done it -- but it's not an easy task. And luck has a hand here.

Newer AVT units have a number problems. I've not tested anything since 2011, as I didn't really want to waste my time. I know the problems persisted at least into 2014/2015. I've considered re-testing a 2016 model, as it is 5 years later now, and I don't read as many complaints now, but not if it'll cost me money. I'd need either 100% refund, or somebody to send me one at no cost (and I can returned it when done).

FYI: The first 4 numbers from the serial are the year of production.

thanks for the info -- the checkerboard pattern looks like some kind of crosstalk to me.

i've done some tests with the avt out and the image is very obviously worse with no avt (though the chroma bleed looks roughly the same).

Analog tape has chroma bleed, some a little, some more depending on the tape. The AG-1980 does some cleanup and at least keeps it from looking worse. Post processing can clean it even more. High saturation levels on input and sharpening can increase it. Whether or not it's excessive in your case would depend on a sample capture of a few seconds. The AVT test panel doesn't really tell you much unless it's really a serious problem. Also, keep in mind that chroma bleed and chroma shift are two different things, although you can have both at once.

I bought my problem unit in March of last year.

I may re-contact them. What's always been lacking is before/after chipset info. Somebody here started, but it never went that far. I was going to assist, way back in 2012 or so, and life got in the way. I wanted to give a solid bug report to them. At least 2 members here submitted bad units, and I did take photos of the bad unit chips.

i just took a look at this on my unit and unless i am missing something i don't think that is the case for me.

Hmm.. interesting. :hmm:

i picked up another unit from b&h just to see if there's a difference -- same serial format and behaves identically.