2013 my video capture device comparison screenshots

[vaporeon800]

I'll continue to edit these posts to add more stuff over the next while.

[Explanation of methods goes here]

Issues/features chart

The pictures are huge (1724x764) because of all the scopes I included. I recommend opening each in its own tab at full-size and flipping back and forth between them to easily see the difference. I use Firefox with the Tab Wheel Scroll extension set to the reverse direction from default so that I can just flip the wheel a bit instead of having to move the mouse and click.

The idea here is that you compare the images marked Reference to the images captured from each device. The most "accurate" device is whichever is most similar to the reference.

The Philips DVP642 player that I used for the DVD patterns unfortunately clips superblacks and the very top superwhites.

I added white borders around the images to align them rather than cropping, so you can see the area that each device captures. Some of them start at NTSC line 22, most at line 23, and the ATI 750 has the stupidest line range of them all because it starts at line 24. That means it captures the bottom half-line (525), and then extends outside the active picture area -- the bottom row of pixels is always superblack mixed with garbage chroma data!

Hmm. The VC500 & EZGrabber2 appear to have some issue with intraframe levels (that's my attempt at distilling the problem to two words). For example: when calibrated for DVD shot 3, the levels are wrong in DVD shot 1, and vice-versa.

The ATI 600 adds a slight green tint to everything, including greyscale patterns. The ATI 750 adds a slight orange tint to colors.

The HD AV Grabber is pure garbage.

[vaporeon800]

Reserved1... missed posting on the 3-year anniversary by seconds on the west coast. Shouldn't have doddled.

[Non-VHS screenshots go here]

For the simpler patterns, the PNG filesizes offer some measurement of how losslessly-compressible the videos are, relative to noise levels (DNR has been disabled, where possible -- some amount may still be in effect; need a good test). All were created with VirtualDubMod's built-in PNG save feature.

Source: DPS-470AV "10-bit Test Signal Generator" mode

SDI 10-bit → VirtualDub 8-bit


SDI → HDMI 8-bit (superblacks were clipped by the HyperDeck Shuttle 2's HDMI conversion; that's annoying)


Reference: SDI 10-bit → FFVideoSource 8-bit


Diamond VC500


Mygica EZGrabber2


ATI 600 USB


danno78 XCard (AVerMedia hardware)


Easycap STK1160


ATI 750 USB


HD AV Grabber


Source: VHPatterns2 test DVDR (DVE + Ice Age THX test patterns)

Reference: DVD → YUY2 (interlaced chroma)


Diamond VC500


Mygica EZGrabber2


ATI 600 USB


danno78 XCard (AVerMedia hardware)


Easycap STK1160


ATI 750 USB


HD AV Grabber

[vaporeon800]

[VHS screenshots go here]

This particular VCR's CVBS input has some issues, so we can't trust the measurements too precisely.

The VC500's luma is a lot more stable on these still patterns frame-to-frame than the ATI 600, while its hue is a lot less stable. But the ATI 600's saturation "wobbles" as well.


Diamond VC500


Mygica EZGrabber2


ATI 600 USB


danno78 XCard (AVerMedia hardware)


Easycap STK1160


ATI 750 USB


HD AV Grabber


----

Diamond VC500


Mygica EZGrabber2


ATI 600 USB


danno78 XCard (AVerMedia hardware)


Easycap STK1160


ATI 750 USB


HD AV Grabber

[vaporeon800]

ATI Radeon X1950 Pro


Here are VHPatterns2 captures made using the the "VIVO (Video In/Video Out)" feature of the ATI Radeon X1950 Pro (2006). I didn't include it previously because the GPU part of my card is pretty messed up, but I had reasons to install it in a machine again, so I figured I would grab some captures while it's in.

The "ATI T200 AVStream Analog Capture" name imparted by the driver implies that it uses the Theater 200 chip, but evidently ATI's device naming is not to be trusted. I'm not about to tear the cooler off to inspect the chip myself, but images in reviews all agree that it's marked "ATI Rage Theater 213RT1ZUA43G". The non-AIW version of the 9600XT (2003) uses "ATI Rage Theater 213RT1ZUA43", i.e. the same code minus the "G" revision at the end. Its driver names it correctly as "ATI WDM Rage Theater Video", according to images in reviews. Meanwhile, the "ATI Theater200 213T20ZQA12" as seen on the AIW Radeon 9x00 series used this same "ATI WDM Rage Theater Video" name and NOT "T200", at least in the initial driver versions. Yeah...

The 720x480 mode of the X1950 Pro is cropped and stretched; restoring the proper dimensions requires a downsize to 711x480 (i.e. the "capture window" of the card is non-standard). Since 4:2:2 images can only have even widths, I had to convert to 4:4:4 and then back after padding (VideoScope only supports YUY2). I carefully picked the chroma resampler, and there are no visible artifacts added. The resize down to 711 did add a tiny amount of ringing and a very tiny resolution loss on the most closely-spaced vertical lines. The luma rolloff that's immediately visible is part of the original capture, not an artifact of my resize.

The 9600XT VIVO mentioned above apparently suffers from the same stretch.

Note, there are other versions of Rage Theater: AIW 128 AGP 32MB (1999) is "213RT1ZUA32" and AIW 128 Pro AGP (2000) is "213RT1ZUA41". I suppose this different ADC explains why my 128 Pro captures are all the way down at a very-cropped ~689x480 stretched to 720x480 rather than matching the dimensions seen here. I had thought that perhaps a driver difference was to blame.