Info: IOData GV-USB2 capturable range

[lollo]

Because my new capture PC hardware, as per the Hauppauge USB-Live 2 (https://forum.videohelp.com/threads/418254-Info-Hauppauge-USB-Live-2-capturable-range) I did an update of the drivers for my IOData GV-USB2 to latest available (thanks a lot to Alwin for providing me the 115 drivers; for some unknown reason I was unable to download the latest IOData drivers myself on their site with my seial number).

The card is now able to capture in the range 1-254 as well!





The new drivers at default procamp values are unbalanced, and the whites are clipped, so the contrast need to be reduced, something that I do not like because I want to reduce the procamp manipulations at minimum (I think Sharc already mentioned that in another thread). I do not remember if it was the same with the old drivers, having lost my data.

default procamp values:


customized procamp values:


For people curios about comparisons, here a frame match for USB-Live 2 and GV-USB2. GV-USB2 is unfortunately cutting 2 pixels on the left side compared to the Live-2 (I wonder why producers are not able to capture all available good data across different cards and different drivers, is not nice for us simple users!).
I did not equalize the pictures for overall look, but the capture quality is very similar (I still have a small preference for the USB-Live 2)
https://imgsli.com/NDEwMjg4

Last note, about the AGC behavior that Sharc found in the USB-Live thread with test signals, I see the same with the GV-USB2 with real recorded material:


Unfortunately, in my experience there are no cards free of (small or large) automatic level correction (see the behavior of an old AIW here https://www.digitalfaq.com/forum/video-capture/7775-ati-aiw-7500-a.html analyzed by master msgohan/brad). A balanced set-up in term of input levels should mitigate AGC effect, but it really depends on recorded material (and then input signals) a lot.

[Alwyn]

The new drivers at default procamp values are unbalanced, and the whites are clipped, so the contrast need to be reduced, something that I do not like because I want to reduce the procamp manipulations at minimum (I think Sharc already mentioned that in another thread)

Sharc says -5 (older drivers). What would be a reasonable reduction for these drivers?

[lollo]

Yes, I reduced the contrast by 4 or 5 steps to do not clip the whites.

I am using a real recorded signal for my set-up (for the whites assesment, the first spot you can see here https://www.youtube.com/watch?v=AjcEMQZeUcA, which has high level of whites and small objects in the background appearing and disappearing, that is my worst-case).

IIRC Sharc is using signal tests, that can be more conservative and safer (and never happening in real life).

In any case, with the GV-USB2 now capturing in the full range 1-254 I think that a small reduction of the contrast (or the brightness) is enough to stay in the capturable range of the card. If you prefer to capture in the 16-235 range is another story, but is difficult and you have to test case by case, tape by tape, segment by segment how to stay inside 16-235 (another reason for me to avoid stress on the procamp setting and capture inside the capturable range rather than in the "legal" range). Your choice

[Alwyn]

Fair enough-it's a tradeoff between just capturing with a slightly changed Preset Contrast (-5, easy to set up for newbys/casual users) or capturing full-range and then diving into complex level corrections with AVISynth in POST: a fair proportion of hobbyists aren't going to do that. And if I read things correctly, importing full-range into NLEs (and some of VDubs filters) is bad news due to crushing.

Re the AGC, it does look a bit concerning on your GIF but your Youtube tells a completely different story: it's almost impossible to see the AGC kick in, even when you are looking for it. So I think it is a non-issue for everybody except the perfectionists; like the levels pump from the AV4 S-Video outlets on the Pannys. Technically yes it's there, practically, quite acceptable and nothing to worry about.

Thanks for posting.

[lollo]

(-5, easy to set up for newbys/casual users)

Sorry, I was not clear. -5 is to stay inside 1-254 with no clippling on the whites. If you want to capture in 16-235 ramge, you need much more than that (at least on some tapes). IMO is more difficult to always stay inside 16-235 than shrink with a simple one shot commmand like levels(1,1.0,254,16,235,dither=true,coring=false)

Re the AGC, it does look a bit concerning on your GIF but your Youtube tells a completely different story: it's almost impossible to see the AGC kick in, even when you are looking for it.

That youtube video was captured with the USB-Live 2 with no AGC issues. The gif is extrapolated by a GV-USB2 capture of the same tape, with AGC in action.
As I said, there is no control over the AGC, and sometimes it happens, depending on many variables, for almost all cards. Best we can do is to have "proper" input levels in all aspects.

[lollo]

And while here, I also wanted to experiment the setting of the sharpness inside the procamp with the new drivers, changing from default 16 to 0.

Very similar results, no effect on the halos (they are in the source), a little bit more definition when sharpness=0 (the red star in the bottom left). May be just speculating



Comparison with slider: https://imgsli.com/NDEwNDkx

edit: the previous was a wrong conclusion; equalizing the videos for same brigthness there in no more definition in the small details (the red star in the bottom left), anda the sharpness 16 image seems a little little bit sharper.
So I would stay with defaul 16 sharpness value and not 0, but the differences are so smal that is a futile exercise

https://imgsli.com/NDEwNzky

[Sharc]

Sorry, I was not clear. -5 is to stay inside 1-254 with no clippling on the whites.

Yes. In my calibration tests the -5 contrast proc-amp correction sets reference white (Y=235) of a synthesized test source to be captured as Y=235 ("spot on"). If your source includes superwhites >235 these will become captured whites >235 in a 1:1 relation. But as an 8bit source cannot be >255 this also means that the captured superwhites Y (236 ... 255) will not get clipped (but are still superwhites of course).

Despite of this I still recommend "noobs" to capture within the "safe" Y 16 ....235 to avoid unwanted (unexpected) clipping and to leave the headroom/footroom alone, as one may not always scan a full 2 hours video to find its maximum Y excursions. Also, halos have the tendency to spread into the >235 range, and when clipped present a challenge to compression efficiency. Pros and cons. Just know what one is doing, not to argue about what is "best".

[lollo]

If your source includes superwhites >235

Almost all my tapes have whites in the 235-255 range, sooner or later.

But as an 8bit source cannot be >255 this also means that the captured superwhites Y (236 ... 255) will not get clipped.

Only with a shrink of the captured range through the procamp, otherwise we lost data (see accumulation in the histogram of the capture with the default setting of the GV-USB2).

But OK, we are repeating the same concepts over and over againg, by now the readers should be familiar with the topic and maybe we should not bore them anymore

[Sharc]

In my tests I could not see that the I/O Data GV-USB2 has any level instabilities ("AGC issues" or "auto levels adjustments" - however we call it). I tried default settings, contrast reduced by five steps to avoid any clipping, and even rducing the analog input signal to about 50% (inserting a ~6dB attenuator). Rock stable video levels. Driver is 1.1.0.93 of 28.06.2010. Quite opposed to the USB-Live2 behaviour with its "gradual auto adjustments".
Of course the analog input signal which is delivered by the source device needs to remain stable for the tests.
I can only speak for myself and my devices of course.

[Sharc]

Based on a few more tests, here my speculation:
Hauppauge (USB-Live2) wanted to avoid complaints of clipped superbrights (invisible Y>235 depending on the decoding matrix). So when superbrights are found, it shrinks these in steps to be within the luma <235 range. With sources all in the 16....235 range nothing happens. But when Y>235 pop up in a video the downturn starts gradually. The problem is that the rest of the video remains at that lower level and hence the brights look too dark, as it seems not to recover (turn up again, in rare cases it does it though in an ugly step), so reference white stays at Y=219 worst case instead of 235.
My conclusion for my current Live2 and driver versions.

See also the discussion before:
https://forum.videohelp.com/threads/418254-Info-Hauppauge-USB-Live-2-capturable-range

[lollo]

Hauppauge (USB-Live2) wanted to avoid complaints of clipped superbrights

Most of my USB-Live2 captures have whites in the 235-254 range, even with the latest drivers. You have examples in the thread you linked. Honestly I do not know what may trigger the behaviour you describe in your experimets.

Not really important I suspect, because you're capturing in the 16-235 range anyhow

[Sharc]

Honestly I do not know what may trigger the behaviour you describe in your experimets.

Maybe the same what you found in post#1 for the GV-USB2 (which I can't confirm)?
Single and occasional short excursions will not trigger it, but when it persists a few seconds the downturn starts with a delay for the Live2 here.
Anyway my version is labelled 122000 LF Rev C1 2615. Seems to conflict with your earlier statements that there were no revisions and changes ever of this device

[lollo]

Anyway my version is labelled 122000 LF Rev C1 2615. Seems to conflict with your earlier statements that there were no revisions and changes ever of this device

I have the same as yours (Hauppauge WinTV-Live 2 USB-Live 2, model 1341 SL-1341-V1.3-5LANG-U model 122000), and I have always seen this "model" across years and many users.

Only recently a new version has been introduced on the market, and we described that it in another thread. It uses completely different drivers, because the internal USB bridge chip is different, so no confusion is possible. I ignore how it behaves, because I do not own it (but the captures shown in that thread were ok).

edit: link to the thread https://forum.videohelp.com/threads/417100-Improve-Hauppauge-USB-Live2-capture-quality