ATI 650 USB blown out highlights?

[waloshin]

Capturing mpeg 2 through the catalyst software I am getting major blown out highlights. There are parts of the he video where I cannot see faces as they are just white blurs. I did not get this problem through my Canopus Advc 55.

Any suggestions?

[msgohan]

No fix most likely. It's a bad design with AGC pumping. Possibly caused by false (and only partial) Macrovision detection.

http://www.digitalfaq.com/forum/vide...t-fix-agc.html
https://forum.videohelp.com/threads/...=1#post2149571
https://forum.videohelp.com/threads/...=1#post2032756
https://forum.videohelp.com/threads/...ol-(AGC)-Issue
http://forum.lddb.com/viewtopic.php?...f5dc5ee#p44457

[sanlyn]

The ATI 650 USB has specifically been not recommended for years in this and other forums. For example, quoting lordsmurf, it's "cards like the ATI 650 . . . have issues with luma flusing during recording, because if the AGC." (http://www.digitalfaq.com/forum/vide...html#post27432)

Or here:
http://www.digitalfaq.com/forum/vide...html#post48301

And other posts for years:
http://www.digitalfaq.com/forum/vide...i-600-usb.html
http://www.digitalfaq.com/forum/vide...html#post28120

The PCI version of the 650 is recommended, not the USB:

[lordsmurf]

At one point in time, before some of these things came to light, some of us were suggesting the 650 USB. I either didn't notice AGC issues in my tests, or the tests where not long enough. The 650 is why I altered some of my testing parameters regarding AGC.

I don't remember off-hand if the 650 PCI has it.

Pretty sure the 750 cards have the AGC problem, which is why I've never suggested those. In fact, I think it had several.

Not sure about 550 or 600 PCI, actually. I've forgotten. The 600 USB is fine.

[waloshin]

Thanks for your replies everyone. :-)

I see that the MX02 HD/SD is suggested what about the MX02 Mini?

[msgohan]

Have you done a search of the forum? Several users have posted their experiences with that device.

[waloshin]

I have but haven't really got a definite answer other than it is nice.

Are all the MX02s the same quality when capturing?

[metaleonid]

I don't recommend ATI 650 at all. It's the worst. You get AGC issues. It also suffers from chroma as far as I remember.

Lately I was using ATI 750 HD PCIe and USB versions to capture from VHS. These both cards do suffer from AGC as well. But not always. In my case it depended on particular VHS, particular VCR. When they didn't suffer from AGC, they did a really good job. If I were you, I'd get ATI 600 USB. I wouldn't recommend it for LaserDiscs, but for VHS is fine.

[bilbofett]

Quote:

Originally Posted by metaleonid

If I were you, I'd get ATI 600 USB. I wouldn't recommend it for LaserDiscs, but for VHS is fine.

Why not use 600 for laserdiscs? (I have both laserdiscs and VHS to capture via SVHS)
TIA!

[metaleonid]

Not full chroma which is fine for vhs but you want full chroma for LDs.

[bilbofett]

Quote:

Originally Posted by metaleonid

Not full chroma which is fine for vhs but you want full chroma for LDs.

I have the USB600 and also ATI-600 in PCI-E form...
Does the VC500 do "full chroma"?

What about clipping superblacks and AGC?

TIA!

[sanlyn]

The ATI 600's clip superblacks. The VC500 does not. None of the devices you mentioned above have AGC.

[lordsmurf]

Quote:

Originally Posted by sanlyn

The ATI 600's clip superblacks. The VC500 does not.

I re-tested this yesterday, comparing ATI 600 USB to VC500, and I'm not seeing it. If anything, the VC500 was more off level-wise, and needed more fine corrections for perfection. But seemed fine. To me, the 600 is still more preferable.

Perhaps the PCIe does, but t's not suggested anyway.

[sanlyn]

Quote:

Originally Posted by lordsmurf

Quote:

I re-tested this yesterday, comparing ATI 600 USB to VC500, and I'm not seeing it.

I see it all the time, and it's usually rather obvious. The example below is from videohelp. The white arrow on the histogram shows clipping at y=16.



Of several examples posted at digitalfaq, the example below shows clipping at y=16, including black borders. The image shows discoloration and artifacts from the clipping, which is also explained in the text from the original post:



The above example was originally posted here: http://www.digitalfaq.com/forum/vide...html#post43771

The example below shows minimum luma levels readout stopping at y=16:



The above example was originally posted here: http://www.digitalfaq.com/forum/vide...html#post43950

Examples of no black clipping were posted using an 9600XT AIW and a Diamond vc500. A readout of the VC500's levels show min/max "y" values all the way down to Y=0 (black borders) is below:



The examples were posted in the same thread in post #43951: http://www.digitalfaq.com/forum/vide...html#post43951

The VC500 readout shows brightest y-level detail at y=240, which is not significantly useful detail but which can easily be recovered using any number of simple means in YUV and could have been tweaked during VirtualDuib capture. But clipped detail cannot be recovered or successfully brightened and often affects dynamic range and contrast levels in the final result.

[bilbofett]

Quote:

Originally Posted by sanlyn

The VC500 readout shows brightest y-level detail at y=240, which is not significantly useful detail but which can easily be recovered using any number of simple means in YUV and could have been tweaked during VirtualDuib capture. But clipped detail cannot be recovered or successfully brightened and often affects dynamic range and contrast levels in the final result.

So you adjust the ProcAmp with black borders present (or a fully black scene) until you get Y=0, and then check it again with a borderless (cropped) pic to get black=16? Because if it can never get below 16 that means there's crushed blacks? I still am adjusting to studying and understanding that histogram and if there's a smooth curve on the far level black level I guess that's good, and if its a sharp cutoff that's bad?

P.S. Thoughts on the ATI-750 USB? Crushed blacks? AGC?
Thank you Sanlyn!

[sanlyn]

Quote:

Originally Posted by bilbofett

So you adjust the ProcAmp with black borders present (or a fully black scene) until you get Y=0, and then check it again with a borderless (cropped) pic to get black=16? Because if it can never get below 16 that means there's crushed blacks? I still am adjusting to studying and understanding that histogram and if there's a smooth curve on the far level black level I guess that's good, and if its a sharp cutoff that's bad?

P.S. Thoughts on the ATI-750 USB? Crushed blacks? AGC?
Thank you Sanlyn!

Temporarily crop away borders and head-switching noise in the capture set up, and your levels should be natural after the borders are restored.

Clamping or clipping is usually a spike with no tail, not a sudden cutoff.

The 750 has horrible AGC.

[lordsmurf]

Quote:

Originally Posted by sanlyn

clip superblacks

That's a confusing phrase, because it's not video jargon. Googling shows it to likely be the invented phrasing* of our own msgohan, and this is his definition:

Quote:

the standard for video is that black is 16 and 235 is white. If your monitor and video card are set up for 0-255, your standard media player by default will assume that the content is using video levels: it will clip superblacks and superwhites, put 16 at 0, put 235 at 255, and expand everything in between to fit. The same goes for hardware players outputting RGB, which are less configurable if you want them to do something different.

NTSC digital video has always been a problem, because you're taking a smaller colorspace and sticking it into a larger one (0-255 RGB displays). That smaller YUV scale is generally referred to as 16-235, but I'm fairly certain that's actually an equivalency range (at least once upon a time), not a precise digital one. It's not uncommon to see anything in the 8-240 range as accepted as YUV values. This is separate from IRE, which also causes issues. In theory, the YUV is re-mapped to keep blacks black and whites white. The only real problem is when you don't remap, and everything in the 0-16 range is crushed to black, and 235-255 is crushed to white/highlight. The wrong IRE mapping can do some wild things, too.

* Note: Not necessarily a bad thing, and I may add it to our glossary,

Now, looking back at the referenced threads, a comparison between the ATI 600 USB and another card implies that the values weren't mapped 16-235, but instead a wider ranger (8-240ish?). At face value, yes, it's not ideal.

(Though I wonder if this is by design, not accident. As you often mention, as do I, too much crushing for "contrast" happens on Youtube/etc. Yuck. I want real, not fake color/contrast/sharpness.)

But we're also talking VHS/Video8-based. I have to tweak proc amps more than I care to as it is, because the tapes are rarely shot/broadcast with ideal values. After proc amp tweaks, either in hardware (TBC-3000/BVP4+) or software controls, or both, I'm just not seeing a discernible difference between the VC500 and ATI 600 USB. I even compared it to a wider range of cards, too: Pinnacle, Hauppauge, several ATI AIW drivers, EZcraps, Tevion, as well as several makes of DVD recorders.

The DVD recorders all give worse performance, save for the RCA's Zoran chipset (which was on par with the ATI AIW). Some of the PCI Hauppauges tended to be worse than EZcraps at the YUV stuff!

My point is this: I'd like to see what you do ... but I'm not.

Even if I could, I'm wondering if it actually has a drastic impact on quality, as you're suggesting. And I don't think it does, for this sort of source. Only Laserdisc has a decent argument here.

BTW, what shows the histograms? I forget. Is that VirtualDub or Avisynth? I use real-world testing, and have never relied on charts/graphs except when calibrating monitors.

This much I can say with certainty: the ATI AIW is easily the best of them all.

Another consideration is this: I've long suspected that the ATI 600 USB series card had some minor variations over the life of the card, which has resulted in the mixed findings to date. There was, after all, several package rebadges (Diamond, VisionTek, ATI direct, others?), and that often means a hidden sub-model existed with minor hardware variances. To date, I've been unwilling to smash apart the USB cards in the name of research. Some reports to date have been worse than your finds, while others mirror my own.

The VC500 has also been accused of this: http://www.digitalfaq.com/forum/vide...perblacks.html
Possibly a driver/OS issue, not the hardware.

Interesting that we have differing findings.

More testing is needed.

Have you actually tested this card yourself, or just relied on others posting test caps? Because as I mentioned at http://www.digitalfaq.com/forum/vide...html#post43956, there are often other variables at work.

Quote:

Originally Posted by sanlyn

The 750 has horrible AGC.

650, too.

[sanlyn]

I've seen the phrases "superblack clipping" and "superwhite clipping" since I started capturing video with Windows 3.1. I recall t hat I first saw that the notion mentioned on an ancient website that discussed the use of proc amps for capturing. I doubt that msgohan invented it, since readers at Videohelp and doom9 have been using the terms for years.

Quote:

Originally Posted by lordsmurf

It's not uncommon to see anything in the 8-240 range as accepted as YUV values.

That won't happen with the ATi 600 USB or with the Hauppauge USB Live 2. They both clip sub-16 blacks. You can get 16-240, but not 8-240. If you did happen to get 8-240 (and that doesn't include zero-black borders), you can rescue sub-black detail using any number of filters or contrast technqiues to raise black levels across specified ranges. Or you can let sub-16 values go dark, as often they don't contain useful information anyway. In my own case, when working with hi=def and digital recordings and backing-up retail DVD's, I don't see YUV values outside of 16-235. I do see frequent and obvious clipping on a lot of that material, though.

Quote:

Originally Posted by lordsmurf

In theory, the YUV is re-mapped to keep blacks black and whites white. The only real problem is when you don't remap, and everything in the 0-16 range is crushed to black, and 235-255 is crushed to white/highlight.

Agreed, but clipped values can't be remapped successfully. Clipping at y=16 means that anything below 16 is converted to 16. Anything above 235 is converted to 235. Once clipped, detail is gone. You can remap y=16 to y=32, but that won't retrieve any formerly sub-16 detail because that data has been destroyed. Ditto for brights: you can remap y=235 to y=220, but that won't retrieve any over-235 data because anything over 235 has been destroyed. Any remapping would have to be done prior to the capture device. An external proc amp would solve that problem. Fiddling with the capture device's own internal proc amp controls apparently won't help -- the material appears to have already been clipped by that point in the capture process, as the posted images indicate.

The clipped images I posted aren't the only example of difficulty in restoring (or simulating) snappy contrast during post-processing of clipped blacks. I recall another example posted at http://www.digitalfaq.com/forum/vide...html#post42190 and following posts.

[lordsmurf]

I just don't remember seeing the term "superblack clipping" having been used until recent years. It's somewhat ambiguous as to what it really means. For one thing, it suggests a "blacker than black" state, and that's just not the truth of the matter. And "blacker than black" refers to something else entirely (sync). When something clips, everything outside the range is simply dropped. So "clips at 16" is more accurate, or even "clips to black" (referring to the non-black darks).

The proc amp was very likely referring to "blacker than black", and resulting sync issues. That's easy to do. It likely had nothing to do with 0-255/16-235 colorspaces.