Comparing I-O Data GV-USB2 vs. ATI 600: Flagging, Jitter, and Frame Drops

[Darryl In Canada]

I'm currently testing my I-O Data GV-USB2 (hardware ID USB\VID_04BB&PID_0532&REV_0005) to see how it handles flagging and horizontal jitter (waviness) compared to my ATI TV Wonder 600 USB. As part of this, I'm also running a comparison on dropped, duplicated, and inserted frames over a continuous 6-hour tape capture. I'm planning a comparison video between the two devices at @VideoCaptureGuide on YouTube.

Waviness test
For the waviness test, I’m using my most problematic tape: a 40-year-old VHS recorded in SLP mode. I digitized the tape using the following workflows:

Workflow A: Panasonic PV-VS4821-K (S-VHS VCR) -> S-Video -> GV-USB2 -> USB -> Windows 7 with AmarecTV -> HuffYUV

Workflow B: Panasonic PV-VS4821-K (S-VHS VCR) -> S-Video -> ATI 600 USB -> USB -> Windows 7 with AmarecTV -> HuffYUV

The Panasonic VCR does not have a built-in line TBC, so both capture devices are facing a difficult challenge.

Based on the attached sample clips, the GV-USB2 seems to be doing a noticeably better job maintaining geometry with this difficult tape than the ATI 600. While it is an improvement, the raw output is still unacceptable. Taming the timing errors in this capture will ultimately require adding a passthrough device like the Panasonic DMR-ES15.

I have two questions for the VideoHelp community:

1. Is there any software tool, Avisynth/VapourSynth plugin, or established metric that can quantitatively measure the level of waviness/jitter over the course of the video? I regularly use Python for log and video analysis, but I'm looking for a reliable way to measure horizontal line variance to be as scientific as possible in this comparison.
   
2. Apart from handling timing errors and frame drops/inserts, what other specific metrics or visual comparisons would you recommend to properly evaluate these two devices against each other?

[Attachment 91927 - Click to enlarge]

[lordsmurf]

Tearing (flagging) will vary widely based on the tape itself, and to the playback VCR. Each tape, each VCR, each capture card, will create a unique combination that gives different results.

This test really doesn't say anything about the capture cards, and everything about the failures of pre-tape workflows. The condition of the tape, strength of even lack of TBCs, the quality of the VCR.

Such a test is mostly misleading.

All it concludes is that some form of TBC is required, to not have bad quality. Some will be less bad at times, but with another tape/VCR be worse. But it's like comparing cat poo to dog poop. Both are still poop.

[Alwyn]

@Lordsmurf, if the ATI was replaced with an Elgato you'd be the first one to sledge the Elcrapo.

In this test, the GV is better than the ATI. Just admit it.

Some will be less bad at times, but with another tape/VCR be worse.

This is complete nonsense. To completely poo poo this test because other tests could be worse or better is silly. It starts building a database. Every test adds information that informs the reader.

[The 14th Doctor]

Well in my own test using a Pinnacle 510 and IO-Data to compare if one handles time base wiggles more than the other, I can conclude that they are indentical.

This was with the worst tape I have signal wise (perfectly fine condition wise, no mold, not sticky). The tape is a VHS copy of a camcorder shot wedding from 1987.

This tape has terrible issues when played in my JVC units (VS30U or MV45U) with it's internal TBC enabled. So I'd either have to A. Capture it with an ES15 in it's place or B. Capture it with a Panasonic AG-1980, which does not have the same errors present here.

I've done other camcorder shot tapes from the same time period, and probably filmed on the same camera. Those tapes, being first gen "straight from the camera" tapes, do not have such issues.


[Attachment 91935 - Click to enlarge]
[Attachment 91931 - Click to enlarge]
[Attachment 91932 - Click to enlarge]
[Attachment 91933 - Click to enlarge]

Please ignore the difference in colors, that isn't my point here. The Pinnacle's colors are also not like that by default, I've manually adjusted the proc amp by +10 brightness and -18 contrast to prevent crushed blacks and clipped whites. Also ignore the Pinnacle was captured using composite, wanted to do a dual capture.

[Sharc]

Apart from handling timing errors and frame drops/inserts, what other specific metrics or visual comparisons would you recommend to properly evaluate these two devices against each other?

There a many, many, many parameters to test. To name just a few:
- Capturable range: Does the device capture 1....254? Does it apply a clipping limiter?
- Levels stability: Does the device show some AGC issues? Does it do some "self adjustment"? If so, how does it manifest?
- Ease and stability of reproducible "proc-amp" adjustments
- Frequency response: What is the true resolution? Check moiré and oversharpening and aliasing effects etc. Snakeoil vs facts.
- Noise vs details
- Color reproduction accuracy
- Effectiveness of Y/C separation in case of composite video. Impact on useful resolution/frequency response
- Pure Hardware tests, like impedance of the analog ports over the video frequency range. Should be 75 Ohm resistive.

Some - non exhaustive - examples and ideas are given here:
https://forum.videohelp.com/threads/418973-Info-IOData-GV-USB2-capturable-range
https://forum.videohelp.com/threads/418254-Info-Hauppauge-USB-Live-2-capturable-range
https://forum.videohelp.com/threads/420275-Capture-device#post2794542

To really test a capture device one has to take the VCR, tape etc out of the equation. One should rather inject well defined analog test signals into the card's analog inputs. This often exceeds the possibilities of hobbyists though. Too many examples of odd "comparions" exist, like comparing a piano with a violin and concluding that the piano burns longer ....
Lastly, be critical about spreding general conclusions. Accept that the test and conclusion applies for a particular scenario (HW, test setup, driver version, OS .....)

(And finally, please, not another dispute about Vdub vs AmaRecTV and dropped/inserted frames )

[The 14th Doctor]

Yeah this is one of those tests were the answer is "it depends" this isn't a clear cut and dry answer. Everybody's setup is different.

Heck we are dealing with analog media, it's never consistent, it's not digital. This isn't the same deal as ripping a DVD where the results are always the same.

From the difference in VCR and other factors, there is always going to be a slight difference. Even between two captures of the same tape on the same machine. Heck even two copies of the same retail release are going to be different in terms of head switching noise, dropouts, and frame positioning.

I feel like what we are trying to do is making this headache of an old media format as less of a headache to deal with as possible, by using high quality gear, trying to get as consistent results as possible.

[lollo]

I'm currently testing my I-O Data GV-USB2 (hardware ID USB\VID_04BB&PID_0532&REV_0005) to see how it handles flagging and horizontal jitter (waviness) compared to my ATI TV Wonder 600 USB. As part of this, I'm also running a comparison on dropped, duplicated, and inserted frames over a continuous 6-hour tape capture. I'm planning a comparison video between the two devices at @VideoCaptureGuide on YouTube.

Nice initiative, any solid test and comparison is welcome and contributing to the overall picture of the analog capturing world.

Is there any software tool, Avisynth/VapourSynth plugin, or established metric that can quantitatively measure the level of waviness/jitter over the course of the video? I regularly use Python for log and video analysis, but I'm looking for a reliable way to measure horizontal line variance to be as scientific as possible in this comparison.

A reliable and scientific way to measure waviness/jitter doen not come to my mind. I use sometimes some metric tools to compare the videos (in addition to the key aspects properly mentioned by Sharc in his post), such as VqmCalc, SSIM and similar; they give a set of parameters about the measured deviations.

The problem is that you need a reference and neutral video/image to which compare the 2 results, and is not the case here. Maybe you can capture with the ATI in time base corrected condition and compare against the workflow without TBC and do the same with the GV-USB2 and have some conclusion. In this way you have a pseudo-reference to extrapolate some information.

Tearing (flagging) will vary widely based on the tape itself, and to the playback VCR. Each tape, each VCR, each capture card, will create a unique combination that gives different results.

Sure, but given a common starting point (the signal generated by a VCR, or better a set of specific VCRs), the rest of the chain under analysis can tell something.

Such a test is mostly misleading.

I disagree. The proposed test is a "limit" condition for sure, and nobody will capture in that condition, but is an indication of how a card can be more resilient (GV-USB2 has no time base correction) in real life situation.

To really test a capture device one has to take the VCR, tape etc out of the equation.

I do not entirely agree. Sure, the first step is the test of the card with an ideal signal, to judge its intrinsic characteristics. But it is not representative of the real life, so the ultimate test is while providing a stable, good, and time base corrected signal as is present in our (best) workflow, generated by a true VCR with its weaknesses. Some of the aspects that are covered in the ideal case are irrilevant or not applicable in real captures (i.e. the impedence mismatch of the USB-Live 2 you measured).

[lordsmurf]

Too many examples of odd "comparions" exist, like comparing a piano with a violin and concluding that the piano burns longer

HAHAHA! So true!

Yeah this is one of those tests were the answer is "it depends" this isn't a clear cut and dry answer. Everybody's setup is different.

It's not even about "everybody". You can take your same setup, and it will react a % different each time to the same tape. Or very different with other tapes in your own collection.

Even condition and versioning of VCRs, TBCs, capture cards, all matter.

It's compounding variables. It's hard to make useful tests.

Sure, but given a common starting point (the signal generated by a VCR, or better a set of specific VCRs), the rest of the chain under analysis can tell something.

Very often, the something is "your VCR sucks". Not the model, just yours in particular. Maybe not even in general, just for this one tape. That's why tests have to be exhaustive, multiple VCRs for starters.

I disagree.

You always disagree with everything. I don't know why.

"The sky is blue." ---- I know you'll disagree!

[lollo]

I disagree.

You always disagree with everything. I don't know why.

"The sky is blue." ---- I know you'll disagree!

I disagree when I read something that in my opinion is wrong (and you did not write "The sky is blue", but rather "The sky is green"). That's life!

That's why tests have to be exhaustive, multiple VCRs for starters.

That's what I wrote, "a set of VCRs"

[Sharc]

Some of the aspects that are covered in the ideal case are irrilevant or not applicable in real captures (i.e. the impedence mismatch of the USB-Live 2 you measured).

Disagree. It has undesired side effects. A high impedance termination of the chroma port raises the level of the chroma compared to the luma (worst case by 6dB). This increases the chroma->luma crosstalk in the cable. Even though this level unbalance may be compensated by the device the crosstalk on the luma channel is not eliminated. Even with a well screend S-video cable the effect is measurable and visible as fine crosshatch in saturated colors. Now one can say use an ideal cable. Sure, but now take the recommendation to use the SCART S-Video out (in PAL) to avoid the luma pumping (Bogileins test). The SCART-adaptor has open wires inside which produce measurable crosstalk (I think I have demonstrated this earlier). So its more a question of generously ignoring it than relegating it to irrelevance, IMO
And it is an example of testing devices vs overall systems/setups. Testing tires or driving a car ("real life") and concluding on the quality of the tires is not the same. Not playing down the value of overall system tests, but it may mask issues.

[lollo]

This increases the chroma->luma crosstalk in the cable.

Which in my experience has no effect on the captured image.

Even with a well screend S-video cable the effect is measurable and visible as fine crosshatch in saturated colors.

Never seen that in a significant way for VHS/S-VHS captured material. Maybe if you use an ideal pattern generator or similar and low quality cable you may notice it.

take the recommendation to use the SCART S-Video out

I do not use and will never use any SCART adaptor, but only high-end S-Video Monster cables connected to S-Video outputs and inputs, so for me is irrelevant.

Testing tires or driving a car ("real life") and concluding on the quality of the tires is not the same.

I do not fully agree (yes lordsmurf ) on the example. It is more testing snow tires in icy conditions that will be used in the desert. An ideal signal will never be used in real life.

[lordsmurf]

in the cable.

Gah, I didn't even think about cables. Yep, yet another variable. So many variables.

[Alwyn]

Catastrophists in action again, I see.

One of the few sensible things said so far:

Nice initiative, any solid test and comparison is welcome and contributing to the overall picture of the analog capturing world.

[Sharc]

I already regret to have answered to the OP's 2nd question.

[Darryl In Canada]

Thank you all for the advice. And to Sharc, your list is exhaustive and I am exhausted before even begining to think about how to trackle all these tests. They are awesome ideas but some might be beyond me.

I agree with all that there is no definative test. But at the very least, I want to be able to back up my advice to newbies that the GVUSB2 is an acceptable choice. Right now I am basing it on the experts in the forum. Just want to see for myself that its true.

[Sharc]

Thank you all for the advice. And to Sharc, your list is exhaustive and I am exhausted before even begining to think about how to trackle all these tests. They are awesome ideas but some might be beyond me.

I agree with all that there is no definative test. But at the very least, I want to be able to back up my advice to newbies that the GVUSB2 is an acceptable choice. Right now I am basing it on the experts in the forum. Just want to see for myself that its true.

Fair enough

[lollo]

But at the very least, I want to be able to back up my advice to newbies that the GVUSB2 is an acceptable choice. Right now I am basing it on the experts in the forum. Just want to see for myself that its true.

While there, you can also compare the GV-USB2 versus the ATI 600 USB both in ideal conditions. You will be pleasantly surprised while following point by point excellent Sharc's checklist