What is best capture device

[DB83]

I have just noticed something. Look closely at user profile on, for example, reply #34

To the OP.


It would be of great assistance if, when you get the kit, you can upload some short samples expecially from the IO-Data

[dellsam34]

I have just noticed something. Look closely at user profile on, for example, reply #34.

It's hard to know who to trust in these days of virtual world, I can imagine how frustrating for the OP to decipher through the posts what is a good advice and what is not, I hope everything works out for him and get things going and I also hope as well that he would share his results here so other members can learn from them.

[boolian2]

I have just noticed something. Look closely at user profile on, for example, reply #34

To the OP.


It would be of great assistance if, when you get the kit, you can upload some short samples expecially from the IO-Data

Yes, I will when the stuff gets here next week

[lollo]

Your information simply is not correct. Quite a few Hauppauge cards have been documented online, sometimes with up to 5 wildly different variations.

Show me 1 single example for USB-Live 2

You're also missing firmware, chip revisions, and other hardware on the card. There is a complex relationship between chips, and a single non-main component change can cause issues.

I am not issing anything. Unfortunately for you I am an Analog Designer and a ASIC/ASSP architect, and have been working for Atmel, the provider for the 943 IC . No firmware change for USB-Live 2 has never been reported for malfunctioning.

A perfect example of this is the AVT-8710, green vs. black situation (and many other Cypress models). The FPGA was "the same", but also not at all.

DIfferent hardware, nothing to compare. FPGA here?

Aside from all this, Conexant isn't a high-end chipset. Some can be decent, many more can be crap. The issues is three-fold, not just chip model, but chip firmware/revision, as well as usage by the OEM.

Same as before, you are repeating this Conexant myth and legend since years, and I already showed you that is not true for the 23102. https://forum.videohelp.com/threads/401110-640x480-Vs-720x576-for-Hi8-capture-and-some...e2#post2615447 Again, have you sample to shows?

[DB83]

Ok Boys. This 'discussion about chipsets is not relevant now since the OP has made his purchase.

Better let the topic stay on track and await the OP's results.

[Alwyn]

I received my USBLive2 today from Hauppauge in Singapore and immediately applied the blowtorch in a shootout with the GV-USB2. The USBLive2 melted.

General comments on the USBLive2:

-I installed only the drivers from the Hauppauge website. There is a capture program available that captures into TS and then does DVDs but I didn’t install it.

-I had no trouble running Virtual Dub with the Live2.

-Capture computer is an i5 running Windows 10 21H2, 64bit.

-All captures were done using an LG GC990W VCR using composite out with no TBC (apart from one capture, as noted below). S-VHS VCRs are scarcer than hen’s teeth here in Australia and the one I recently bought locally started eating tapes during the tests today. Not Happy Jan!! In any case, it was no better than the 990, picture-wise. I couldn't use my Panny EZ-48 (s-video out) because it's stabilisers are so strong it would invalidate the comparison.

-All captures were done with Virtual Dub 1.9.11, PAL format, over USB2.

-I haven’t done any long captures to check audio sync, and for reasons that will become obvious, I’m not going to bother.

-Pre-capture, I tried to set the Brightness and Contrast to within the boundaries of the VDub histogram (IOW no overflow into the forbidden zones at the ends).

-The levels controls (brightness and contrast) were difficult to use. They were very sensitive sometimes, and on every occasion, jumped to another setting when I hit Capture.

-The USBLive2 Chemist video showed ugly magenta artifacts. I tried lowering the saturation to no avail.

-The USBLive2 RACWA video was so bad that I had to press my Sony HXD-890 into action as a passthrough, which, I must say, performed pretty well in stabilising the video for the Live2.

Verdict: The GV-USB2 beats the living daylights out of the USBLive2. The Live2 was so bad, I’m considering sending it back. I certainly have no use for it here. In a way, I'm glad I spent the money. I always had niggling doubts about the GV-USB2 compared to Live2. Doubts that I have no more.

Fire away, and if anybody wants any more tests done, please ask.

Note: Another 4 videos being uploaded to the next post.

[Alwyn]

Next set of comparison videos, GV-USB2 and the USBLive2.

[lollo]

LG GC990W VCR using composite out with no TBC

Hauppauge USB-Live 2 cannot perform well in this conditions. It needs a Y/C signal because its comb filter is poor, and a lineTBC in the VCR or in passthrough mode, as we said several times.

However, thanks for your experiment!

[dellsam34]

It's still a legit comparison since the test conditions are the same, It clearly shows that the USB-Live 2 is to be avoided even if a good VCR with line TBC is used, Also composite connection cannot be avoided sometimes if capturing other video format such as uMatic, LaserDisc and Betamax.

[lollo]

It clearly shows that the USB-Live 2 is to be avoided

Sure, look to my captures

edit: I stop here, there are many users of that card here and in digitalfaq and doom9 forums, their results, and mine, can be easily checked. I am not a defender of USB-Live 2, and do not wish to convince anyone

[Sharc]

@Alwyn: Any specific reason why you capture to RGB rather than YUV color space?

[dellsam34]

there are many users of that card here and in digitalfaq and doom9 forums, their results, and mine, can be easily checked. I am not a defender of USB-Live 2, and do not wish to convince anyone

Perhaps there is indeed few versions of them or at least quality consistency issues, Or perhaps you haven't tried it in the same conditions as the above poster. Either way a good capture device should give good results in all conditions not just with a good VCR, S-Video connection and line TBC. That's at least how I see it.

[oln]

I think the conexant chip usb dongles are just bad in general when used with raw VCR output rather than there being much difference between revisions/versions. Both due to horizontal wiggling and bad Y/C filtering on unstable input . So for that use case the I-O data beats the hauppauge handily. Which is admittedly something I find a bit strange given that Conexant/Sypantics isn't some random bottom of the barrel company otherwise, so idk why they weren't able to develop something that handled it at least a little better than they do compared to video decoder chips from other vendors. There is also some odd differences in behaviour when using them (or at least the Diamond VC500) on linux compared to windows, the Y/C filtering seems less bad, and how the horizontal wiggling presents is a tad different (though not good still.) Afaik the venerable ATI AIW cards don't really handle unstable input all that well either, though maybe not quite as bad.

When it comes to using a card to capture the stable S-Video output from a tbc or dvd-recorder letting that do the stabilizing it's more about not having other problems and giving a transparent representation of the input. In my experience at least the actual image quality in these cases does not differ all that much between somewhat decent devices. Of course in these cases it may also be viable to use component or HDMI/SDI and skip the chroma encoding/decoding part between tbc and similar and capture entirely if the device has those outputs.

Ideally we would have more capture dongles with TBC/jitter correction and buffering built in but those are few and far between so we have to make do with what's available.

Main issue I've noted with the Diamond VC500 besides not handling unstable input is the lack of audio input level control in the windows drivers (works fine on linux so definetly a driver thing), not sure to what degree other conexant usb dongles have that problem. The I-O data does not have this issue, driver seems higher quality.

[lordsmurf]

you are repeating this Conexant myth

It's not a myth. You should know better than to accuse me of parroting myth. I detest myths.

edit: I stop here, there are many users of that card here and in digitalfaq and doom9 forums, their results, and mine, can be easily checked. I am not a defender of USB-Live 2, and do not wish to convince anyone

As a quick aside, you and I are disagreeing again, but it is friendly. I hope it stays that way.
In other areas, like Avisynth or upscale, no disagreement. Just the cards.

Hauppauge USB-Live 2 cannot perform well in this conditions. It needs a Y/C signal because its comb filter is poor, and a lineTBC in the VCR or in passthrough mode, as we said several times.
However, thanks for your experiment!

I agree with this...

It's still a legit comparison since the test conditions are the same, It clearly shows that the USB-Live 2 is to be avoided even if a good VCR with line TBC is used, Also composite connection cannot be avoided sometimes if capturing other video format such as uMatic, LaserDisc and Betamax.

... and this.

@Alwyn: Any specific reason why you capture to RGB rather than YUV color space?

This always makes me raise an eyebrow.

I think the conexant chip usb dongles are just bad in general when used with raw VCR output
something I find a bit strange given that Conexant/Sypantics isn't some random bottom of the barrel company otherwise,

- Top line products, bottom line products.
- Quality grade, reject grade.

Not unusual at all. This happens quite a bit in the semiconductors.

Ideally we would have more capture dongles with TBC/jitter correction and buffering built in but those are few and far between so we have to make do with what's available.

I would suggest those don't exist at all, and what we have for "capture card with TBC" are somewhat laughable imitations.

[DB83]

Well to throw some Gin in to the Tonic, like I mentioned earlier in this thread I am a USB-Live2 user of some 9 years.

Chances are that the product I linked to is not identical to what I acquired back then.


Are my captures perfect ? Hell no. I posted some mpg samples with a raw feed and the same filtered with that great 'TBC' aka ADVC300 some months ago.


Some other lossless samples have also been uploaded in other topics.


But are they 'good enough' for me since it is only moi that will eventually see them ? Si, Yawhol, Oui (and any other language you care to think of)


And many of us would think twice about acquiring another card 'on a whim'

[lollo]

As a quick aside, you and I are disagreeing again, but it is friendly. I hope it stays that way.
In other areas, like Avisynth or upscale, no disagreement. Just the cards.

Of course we are "friend" in this virtual world, why should not be the case

[Sharc]

The Y/C separation of the USB-live2 seems indeed unacceptably poor in Alwyn's tests - unless the dongle has simply been overdriven by a too high input signal level which caused nonlinear distortion (clipping), means additional and uncontrolled Y/C crosstalk.

For low-res VHS with little luma-chroma spectral overlap it isn't very difficult to separate luma from chroma adequately by means of basic lowpass and bandpass filtering (sometimes called notch filtering). Higher quality recordings like S-VHS, Hi8 etc. footage will however suffer from a loss of details and sharpness due to tight notch filtering (G-Data?), or show dotcrawl and rainbows if these filters are too loose (USB-live2?). One won't find anything in the specs.
More sophisticated 2D and 3D comb filters allow an almost perfect separation of luma and chroma, preserving sharpness and resolution.
Comb filters which adapt between 2D and 3D operation to reduce certain artefacts are found in good DVD recorders. Improved Y/C filtering plus the line TBC functionality help to obtain acceptable capture results even with inferior USB dongles. Hence the recommendation to use USB dongles always in conjuction with adequate DVD recorders in passthrough and to connect the dongle via S-video to avoid its (inferior) Y/C filter.
Unfortunately the Y/C filtering principle is often not specified at all (dongles) or only obscurely described in manuals and technical specifications.

Consumer VCRs:
Optimum comb filtering requests that the color subcarrier frequency and the scanline frequency are synchronously locked.
Consumer VCR gear record the color in the so-called color-under format where the color subcarrier is shifted from its standard NTSC or PAL frequency downwards to the lower end of the spectrum. Locking of the color subcarrier with the scanline frequency is then lost, jeopardizing optimum comb filtering. For playback the VCR player shifts the chroma back again using a free-runing oscillator, but synchronism with the original scanlines is lost.

Well, I guess you know all this ....

[Alwyn]

Sharc: @Alwyn: Any specific reason why you capture to RGB rather than YUV color space?

Lordsmurf: This always makes me raise an eyebrow.

I didn't realise I was doing so, Sharc. I too raised an eybrow. Checking the VDub settings, the only option I can find to select the colour space, using Huff, is here:


[Attachment 65509 - Click to enlarge]

It appears to me that RGB is not selected.

I normally use LAGS or MagicYUV, which produce YUV colour using the same settings.

Why then is Virtual Dub's HUFFYUV producing RGB colour when it appears I don't have it selected?

What settings should I use in VDub to produce YUV with HUFF?

[Alwyn]

In the interests of fairness, here is the RACWA GV-USB2 with HXD-890 passthrough. The Live2 is much closer to this (using the S-Video input to both sticks, as alluded-to by Lollo).

I've also attached captures of the RACWA video with my Startech USB3HDCAP.

Note: all videos with "HXD-890" in the name are Composite from VCR to 890, then S-Video from 890 to capture stick.

[DB83]

For the purpose of unit comparison, here is an unfiltered huffy capture which I have uploaded here before. (Also done with vdub not vdub2)

Feel free to rubbish this as well


The VHS was selected since it is one that I own that is mv protected.

[Sharc]

Sharc: @Alwyn: Any specific reason why you capture to RGB rather than YUV color space?

Lordsmurf: This always makes me raise an eyebrow.

I didn't realise I was doing so, Sharc. I too raised an eybrow. Checking the VDub settings, the only option I can find to select the colour space, using Huff, is here:


[Attachment 65509 - Click to enlarge]

It appears to me that RGB is not selected.

I normally use LAGS or MagicYUV, which produce YUV colour using the same settings.

Why then is Virtual Dub's HUFFYUV producing RGB colour when it appears I don't have it selected?

What settings should I use in VDub to produce YUV with HUFF?

InVdub2 it would be: Video -> Compression ... -> Select the Huffyuv v.2.1.1 -> Press the Pixel Format button -> Select the format there -> (Save as defaults)
Don't know if it is similar in earlier Vdub versions as I am normally using AmarecTV.

[Alwyn]

InVdub2 it would be: Video -> Compression ... -> Select the Huffyuv v.2.1.1 -> Press the Pixel Format button -> Select the format there -> (Save as defaults)

The preference for capture according to The Lord is VDub 1.9.11, which I am using. There is no such option:

[Attachment 65521 - Click to enlarge]

This is the screen for VDub 2:

[Attachment 65522 - Click to enlarge]
Which one should I use for YUV?

[Sharc]

InVdub2 it would be: Video -> Compression ... -> Select the Huffyuv v.2.1.1 -> Press the Pixel Format button -> Select the format there -> (Save as defaults)

The preference for capture according to The Lord is VDub 1.9.11, which I am using. There is no such option:

[Attachment 65521 - Click to enlarge]

This is the screen for VDub 2:

[Attachment 65522 - Click to enlarge]
Which one should I use for YUV?

Try
4:2:2 YCbCr (YUYV, YUY2), or
4:2:2 planar YCbCr (YV16)
But as I said I am not really using Vdub for capturing, so others may give you better advice.

Also, keep an eye on the levels. It appears to me that most of your captures have an elevated black level. Means that even the black left and right pillars look greyish.
The best in this respect is Video 3 RACWA-Live2 HXD-890 trimmed.avi, IMO.
While to a certain degree a matter of personal preference, high black levels tend to make the colors look washed out.

[Alwyn]

Thanks Sharc, re the black levels, do you mean bring the Virtual Dub histogram in towards the centre or spread it out further?

[Sharc]

The Y waveform should go close to 16 for black (dark scenes) and close to 235 for white (bright scenes), means in between the brown bars of the waveform histogram in YUV color space, or between 0 and 255 in RGB.
The "correct" settings depend on the source. You can play with brightness and contrast settings, for example.

[lollo]

In the interests of fairness, here is the RACWA GV-USB2 with HXD-890 passthrough. The Live2 is much closer to this (using the S-Video input to both sticks, as alluded-to by Lollo).

Thanks Alwin!

Sorry, I did not make a comparison earlier because the captures were not good and comparing a bad capture (gv-usb) versus a bad bad (worse) capture (usb-live2) made no sense. In addition I did not want to play the role of the USB-Live 2 defender, because I recommended both cards in post number 6.

Considering "Video 3 RACWA-Live2 HXD-890 trimmed.avi" and "Video 3 RACWA-GV with HXD890 trimmed.avi" we can now say someting:

frame 74 image slider comparison: https://imgsli.com/MTEzNTIz

frame 390 image slider comparison: https://imgsli.com/MTEzNTI0

• The 2 captures are very similar and is difficult to prefer one over the other relying on the general look

• USB-Live 2 captures 1 more line on the top, but is useless because it does not contain complete video data

• The capture from USB-Live 2 has sligthly more contrast and looks sharper, GV-USB looks sligthly softer

• Considering the details and the noise on the objects the captures are similar, and the fact that the frames are interlaced hide a little bit the edges; however, the small black square below the window in frame 74 is less defined in GV-USB capture; the same for the whole square window of the plane in frame 390. In addition the vertical lines on some of the objects and in the background for frame 74 are less defined and straight for GV-USB.
  IIRC you own a Panasonic ES15, why has not been used? For sure, you need a better VCR.
  Concerning the sky and the airport runaway it is not possible to well discern between noise and details (if any); I have the impression that GV-USB has less chroma noise but reduces some of the blacks spots on the floor

• A temporal denoise after deinterlacing in post processing will be very useful for this kind of material

• The histograms are similar, but USB-Live 2 extends a little bit more its luminance in the low side. I ignore what procamp setting you used, just be sure to do not over do, adjustement is probably not necessary for this shot

frame 74 comparison:

[Attachment 65531 - Click to enlarge]

frame 390 comparison:

[Attachment 65532 - Click to enlarge]

USB-Live 2 histogram

[Attachment 65533 - Click to enlarge]

GV-USB histogram

[Attachment 65534 - Click to enlarge]

And finally, the real comparison is on video, more than img comparison, which says a lot but can hide problems on motion. Here a side-by-side video slightly h264 compressed to save space: usb-live2_vs_gv-usb.mp4

edit: fixed link for mp4 file

edit 2: I share the same impression as Sharc about black levels, I briefly mentioned talking about the histograms. Looking at frame 390, the black text and the inside of the hangar on the left are better in USB-Live 2 capture. I am also seeing now that the characters in the text are sligthly better defined in USB-Live 2 capture as well.

[Sharc]

Sidenote: For levels (luma) checking I prefer the waveform display of the histogram. In Avisynth this means Histogram("classic") rather than Histogram("levels").
And as long as the waveform ist not clipped at 0 or 255 (8 bit) one can still apply the corrections in post processing.

[lollo]

I agree. I focused on "levels" because I have seen in the past spiky histograms for USB-Live 2 and histograms with ugly holes for GV-USB, so I wanted to check that, and is easier with "levels"

edit: adding comparison showing better levels in USB-Live 2 capture, but easily fixable in postprocessing for GV-USB


[Attachment 65537 - Click to enlarge]

[Alwyn]

Thanks Lollo, interesting comparisons. The Live2 is definitely "in the picture" when used with some sort of stabilisation.

Re the levels, the Live2 jumped around a lot when I was trying to set them, as well as change as soon as I started capturing.

IIRC you own a Panasonic ES15, why has not been used? For sure, you need a better VCR.

I do, but I didn't use it so that the brightness issue didn't influence the captures. As for the VCR, the 990 produces video as good as the HR-S5700 I recently bought, but which now eats tapes. I can't afford another.

[lollo]

Thanks Lollo, interesting comparisons. The Live2 is definitely "in the picture" when used with some sort of stabilisation.

They are both good cards when used with Y/C signal and a lineTBC in the path. I slightly prefer the USB-Live 2, but many of my friends have GV-USB and their results show the same quality as mine; that's why I recommended it as well.

In addition, both work flawless with Windows 10, no need to work with old OS.

The shortcoming of the USB-Live 2 is its requirement for Y/C and lineTBC correction, while the GV-USB performs better in that area. But in my world there is no place for a capture without a Y/C signal and a lineTBC correction, so the question does not arise .

Thanks again for all your effort, I really hope that our considerations help other users.

Re the levels, the Live2 jumped around a lot when I was trying to set them, as well as change as soon as I started capturing.

I know, especially with VirtualDub. Try AmarecTV, or better open a graph in parallel to the capture sofware to change on the fly the procamp settings on module "Hauppauge Cx23100 Video Capture".

I do, but I didn't use it so that the brightness issue didn't influence the captures.

I also own an ES15 and I use it rarely because the problem you mentioned. Only when the lineTBC in the VCR is not enough because playing tapes in really bad conditions, choosing the lesser of evils.

edit: what's your feeling about the Sony HXD-890? I never tried it, but it seems it has weaker line correction but no problem on the levels. Maybe oln can help here...