GV-USB2 capture range

[theseeker2]

My ATI USB 600 finally gave up the ghost and I've moved on to a GV-USB2. I wanted to continue using a Windows XP-based computer for capture, so I installed the version 1.11 GV-USB2 drivers and am still using VirtualDub for capture. I did have to install GraphStudio in order to get the procamp to work while capturing with VirtualDub, but that wasn't too tough and it's working just fine.

I was surprised when capturing with the GV-USB2 to see values in the red on the histogram on both the left and right sides. With my ATI USB 600, the 0-16 range was always completely dark, and the 235-255 was visible but red.

Based on this, does the GV-USB2 allow the full 0-255 range for capturing? This is different than what I read before I purchased it. I figured it would be 16-255 like the ATI USB 600.

I've done a few initial captures and adjusting the brightness and contrast when faced with the entire 0-255 range is a little more complex than I'm used to! Any tips or other experience you can share with this card would be much appreciated.

[lollo]

I was surprised when capturing with the GV-USB2 to see values in the red on the histogram on both the left and right sides. With my ATI USB 600, the 0-16 range was always completely dark, and the 235-255 was visible but red.

The reds in the histogram are not a problem, as long as you stay in the capturable range of your card. You do not have to capture in the 16-235 range, but in the capturable range of your card!

ATI-USB 600 is limited to 16-235, so you have no choice: https://www.digitalfaq.com/forum/video-capture/9865-canopus-advc-110-a.html#post62640, and you have to shrink the input levels to 16-235 to avoid to crush the blacks and clip the whites.

With the IOData GV-USB the range is a little bit extended, so you have less limitations.

In post processing, if and only if needed, you can shrink the captured levels to 16-235, i.e. for RGB conversion required by some filters and for some display option.

[Sharc]

In post processing, if and only if needed, you can shrink the captured levels to 16-235, i.e. for RGB conversion required by some filters and for some display option.

Take note of the many posts and warnings by poisondeathray about the "mistreatment" of huffyuv for example by many video apps and NLEs (causing RGB clipping for Y<16 and Y>235)

[lollo]

Sure. And you can also add the (in)famous YUV-RGB cube.

Actually, OP question was about capturing, so all the rest comes later. He was wrongly worried abot the "reds" in the Vdub preview histogram, which is not as issue, as long as the card can capture them. That's all I wanted to say. Post-processing is another story.

[Sharc]

Sure. And you can also add the (in)famous YUV-RGB cube.

Actually, OP question was about capturing, so all the rest comes later. He was wrongly worried abot the "reds" in the Vdub preview histogram, which is not as issue, as long as the card can capture them. That's all I wanted to say. Post-processing is another story.

Agree. I referred to your "In post processing .......you can shrink......"

[lollo]

Anticipating a bit, after capturing @theseeker2 will have more questions for sure

[theseeker2]

Anticipating a bit, after capturing @theseeker2 will have more questions for sure

Haha, you know it!

First question is what do people generally prefer for the sharpness setting on the GV-USB2? The default sharpness is 16, but I've seen a few videos where people are capturing VHS with sharpness set to 0 for this card. I wish Brad had a GV-USB2 on hand back when he did his comparisons in 2013 at https://forum.videohelp.com/threads/360704-2013-my-video-capture-device-comparison-screenshots.

I'm having some trouble with my JVC VCR right now, so it might be a little bit before I can get some sample captures posted, but I will definitely do that as soon as possible.

[theseeker2]

Ok, here are some test captures.

This is using the procamp defaults:
set_adrift_default.avi

This is with brightness and contrast adjusted to take up the entire 0-255 range and sharpness set to 0:
set_adrift_level_adjust_sharp_0.avi

I tried making a capture with the range 16-235, but it was super washed out so I decided not to include it.

Here's a screengrab from the procamp default video:

[Attachment 85811 - Click to enlarge]

Here's a screengrab from the procamp adjusted/sharpness 0 video:

[Attachment 85812 - Click to enlarge]

And here's an Image Slider:
https://imgsli.com/MzUzNDEy

Let me know what you think! Thanks.

[Alwyn]

This is with brightness and contrast adjusted to take up the entire 0-255 range and sharpness set to 0:

I'm interested... how are you actually achieving capture range?

[Sharc]

Looking at the screenshots only (may look at the .avi later du to temporary network issues):
The first looks ok. Few harmless out-of-gamut RGB pixels (cyan on the right) when decoding with the standard YUY->RGB conversion matrix. These will however "disappear" for TV playback using the limited->limited conversion equations. So nothing to worry.

The second has less "bad (means clipped) RGB" even for the standard conversion but it looks a bit washed out - no surprise, luma swing shrunk too much (darks elevated).

[Sharc]

Network is back. So here your 2 versions (using the standard Avisynth Rec.601 conversion "YUV limited->RGB full"). The Cyan pixels on the right indicate clipped RGB (percentage as number).

[lollo]

As Sharc said, the second capture is washed out, because you streched the levels to unnecessarly fill the available range. This is not recommended, and you are going against the nature of the source (and potentially introducing gaps in the histograms, not in this sample).

Once more, you act on the procamp just to stay inside the capturable range of your card (or to stay in 16-235 range if you prefere), and do all the rest in postprocessing (with caution, expanding too much will result in the same problem).

Concerning the sharpness parameter I prefer the default, because in this sample (and also in general with the GV-USB2 and the USB-Live2) is not causing halos.

P.S.: now that you have the GV-USB2. you can also compare its captures with the ATI 600 USB

[Alwyn]

you act on the procamp just to stay inside the capturable range of your card

How are you working out what the "capturable range" is? By trial and error with AVISynth's histogram after the event?

[lollo]

How are you working out what the "capturable range" is? By trial and error with AVISynth's histogram after the event?

Yes, I do not see any other method (even because I found the same card having different captures ranges, i.e. VC500 across all its revisions). You can also use Vdub histograms in real time, checking the "accumulation" on the borders.

But is something you do just once, finding a source segment with deep blacks and high whites.

[Sharc]

One could also prepare a DVD with well defined YUV testpatterns, play it in the "passthrough" DVD recorder for example and capture it, or play it and visualize the levels/histogram in realtime via a filter graph (Graphedit) + script + SW player (PC).

Something like this:


[Attachment 85820 - Click to enlarge]


Capture devices use at least 10bit ADCs, so its a matter of the manufacturers internal scaling how they map and position their "defaults" into the 0....255 8bit range, how they do or do not clamp or clip. Hard clipping at Y=16 and 235 is a "poor man's solution" for pseudo legalizing, IMO, and such hard clipping is difficult for compressing.

[theseeker2]

This is with brightness and contrast adjusted to take up the entire 0-255 range and sharpness set to 0:

I'm interested... how are you actually achieving capture range?

By expanding the histogram via contrast adjustment to take up the entire 0-255 range... just as lollo warned me not to do a few posts later. Lesson learned!

[theseeker2]

Network is back. So here your 2 versions (using the standard Avisynth Rec.601 conversion "YUV limited->RGB full"). The Cyan pixels on the right indicate clipped RGB (percentage as number).

So based on those videos you posted (thank you by the way), there's some black clipping at the beginning of both of my samples (below y=16) and some white clipping at the end (above y=235). Is that correct?

It does appear that there is more clipping in the levels-adjusted sample than the default, and I agree that the default looks better and not as washed out.