Info: Hauppauge USB-Live 2 capturable range

[lollo]

My old Asus Win10 laptop stopped working, so I switched to a new Win11 laptop.

While there, for my Hauppauge USB-Live2 model 1341 / model610 I installed the latest drivers 6_0_119_39299 (driver10_6_0_119_39299-win7-8-36075.exe).

Big suprize! Now the card is able to capture in the range 1-254 (it was limited to 16-254 with old drivers).





Note that the card is the old model, Hauppauge is now selling USB-Live 2 with a new hardware, see this thread: https://forum.videohelp.com/threads/417100-Improve-Hauppauge-USB-Live2-capture-quality.
I ignore how the new card behaves behaves, because I do not own it.

edit: comparison between drivers drivers6_0_119_36075 and drivers6_0_119_39299, same capture card (Hauppauge USB-Live2 model 1341 / model 610), same laptop (Asus Zenbook S 16 UM5606WA_UM506WA), default procamp values:





at default procamp values the latest drivers perform better!

[Alwyn]

Updated! Thanks.

[Alwyn]

@Lollo, just tried; VDub is still showing a "wall" well inside both the left and right edges of the histogram, even with minimum brightness and maximum contrast.

Can I only see the real edges in an AVISynth histogram, and based on comments from PDR, Jagabo et al re swapping to RGB and back, is it a good idea to capture 0 to 254 anyway? Should we just stick to the safe range?

[Sharc]

... and based on comments from PDR, Jagabo et al re swapping to RGB and back, is it a good idea to capture 0 to 254 anyway? Should we just stick to the safe range?

The simple answer is yes, stay within the safe range.

More detailed:
From a pure capturing point of view exploiting the full 0....255 luma range (8bit realm) is optimum in the sense that one obtains minimum banding/highest SNR (signal-to-noise ratio) / maximum dynamic range.

BUT1:
Practically, there are pitfalls.
A subsequnet processing (e.g. some filters in avisynth) may make a RGB conversion by using the limited->full range standard expanding matrix. Hence all data in the "forbidden" zone get clipped, means an unrecoveralble loss. This can silently happen "under the hood", without warning. So when you capture to the full range you should know what the consequences are. For this reason I personally always recommend to stay within the safe (limited) range, unless one knows exactly what one is doing and what is happening.
Also, some well known NLEs expand the range (depending on codec) upon importing the capture. If so, the damage is there. I refer to the posts and explanations of pdr and jagabo. I assume you have seen these.
Also, take note that even when you stay in YUV with your processing there is a least a final YUV->RGB conversion when you play the video for viewing. Your monitor cannot display YUV data, only RGB. So the result depends on the player. TVs normally accommodate the full range. Viewing the clip on your PC depends on the conversion matrix (renderer, player settings...)

BUT2:
Even staying within the safe YUV range is no warranty for unclipped i.e. undistorted RGB. I (and others) have commented this several times with examples. I have posted an example recently in the "1-TBC...." discussion.
Keywords: Out-of-gamut RGB / RGB block within the YUV cube / Conversion matrices / levels / legalizing / color distortions.
Staying nicely within the limited ranges helps however to avoid worst cases and is - as a rule of thumb - good practice. Anything else just calls for troubles - at least IMO.

P.S. By the way you can convert your YUV to RGB and display the RGB parade (waveform) using avisynth. You will see which components get clipped at 0 or 255 respectively. Also, NLEs usually can show the RGB parade.

Unnecessary YUV<->RGB back-and-forth conversions should be avoided anyway due to rounding errors (8bit integer realm), even when everything is "legal". These rounding errors are however normally harmless compared to the damage due to clipping.

[lollo]

Can I only see the real edges in an AVISynth histogram, and based on comments from PDR, Jagabo et al re swapping to RGB and back, is it a good idea to capture 0 to 254 anyway? Should we just stick to the safe range?

My recommendation is always the same: capture inside the "capturable" range of the card not in the 16-235 area, to reduce to minimum the shrinking performed by the (poor) proc amp of the cards.

Now that the USB-Live 2 with the new drivers is able to capture 0-254, setting the proc amp at default value (middle of the scale) guarantees that the levels of 99% of the captures will not be crushed nor clipped.
If you want to be really obsessed, put back the Y level at 255 (if present) at 254 lowering by a minimal step the
brightness to get it captured

Later in post-processing, if you need a filtering requiring a RGB conversion or for final rendering or for whatever reason you need to shrink the 0-254 levels to 16-235 you can do it with the power of AviSynth plugins (dithering, coring and so on), much much better that what you can do with the proc amp of the cards.

edit: on the other side, all the considerations explained by Sharc are valid and pertinent. If you know and understand what you are doing follow me for higher quality outcomes, if not follow Sharc

[Alwyn]

Thanks both.

Probably a bit off-topic, but what controls the Histogram left and right limiter in VDub? Apparently not the digitiser if the Live2 can go to 0 and 254.


[Attachment 88275 - Click to enlarge]

[Sharc]

No idea. Did you apply a limiting tweaking filter? (coring true vs false) or similar?

[lollo]

Crop the bad borders before checking for histograms

[Alwyn]

Thanks both, I'll continue in a new topic to save a derailment here.

[Sharc]

Here my result with a full range test picture (including "superwhites" and "superblacks" 0.....255) with the Hauppauge USB-live2 model 122000 LF Rev C1, driver version 6.0.119.39299 of 26.10.2021

[Alwyn]

Thanks Sharc, you're on VDub 2.3. That is probably it. I'm on the original 2.0 (and the 1.x series were the same); it doesn't show anything outside the 16-235 range.

[Sharc]

It was just a quick test. The histogram looks a bit strange (?) to me. Normally I am using Avisynth or a DS filtergraph. Anyway it shows values outside the limited range, below 16 with the driver 6.0.119.39299 of 26.10.2021.

[Sharc]

FWIW here a full range B&W ramp, captured with the USB-live2 model 122000 LF Rev C1 + driver 6.0.119.39299 of 26.10.2021.
Y<16 and Y>235 are included in the capture.


[Attachment 88283 - Click to enlarge]

Note that the non-expanding "PC Matrix" was used to display the Y<16 and Y>235 in the B&W picture correctly.
Otherwise one would get "full black" and "full white" at the dotted lines already and the YUV data Y<16 and Y>235 getting ignored (lost), i.e. shifted out of range and forced to RGB(0,0,0) and (255,255,255) respectively.
See post #4.

[Sharc]

Here another observation for a full range YUV source:
The USB-Live2 (using the driver as quoted before) reduces the high levels automatically from 255 to 235, slowly "darkening" the capture in steps until the max luma settles at 235. See the histogram.
The GV-USB2 keeps the levels rock stable.

I leave it to everyone to judge whether the "AGC action" of the Live2 is desirable or rather not

Edit:
After discovering more "levels self adjustment" issues (in addition to my earlier reported impedance mismatch of the chroma wire) my Hauppauge USB-live2 model 122000 LF Rev C1, Product of Malaysia, with driver 6.0.119.39299 of 26.10.2021 on W11 goes back to the drawer. Rest in piece .
(Does not mean that it is a useless device as some of these issues may not become immediately or annoyingly visible for "natural" home video material ).