My 2nd captures: Luminence levels, distortion?
 

Hello! In 2020 I bought a nice digitizing workflow from this board, but pandemic fallout took my attention away. I'm now re-learning the equipment and software. I've made some test captures, and would love some feedback and critiques!

Workflow is: JVC SR-MV40 VHS -> TVOneTask 1T-TBC -> Pinnacle 710-USB -> VirtualDub 1.9.11.

(I used Lagarith codec rather than Huffyuv, based on LordSmurf's advice about my specific laptop.)

I've attached three 10-second clips of 1997 commercials. All from the same cassette, taped off TV at EP speed.

The last time I tried this (with a different tape), the captures had low black levels; Sanlyn pointed me to his VDub Settings guide. This time I followed his guide, checked the luminence histogram before capturing -- and found that I didn't need to adjust the contrast/brightness at all, it was already perfect.

Question: Is that a red flag? Is it normal to find tapes that are already perfect, luminence-wise? Or does this mean that I have a setting wrong, or misunderstood the histogram?

More relevant question: How does the luminence level look in my attached clips?

Also, there's a thin row of distortion at the top of the image. This originally was much larger, but I turned on the VCR's "Calibration" setting and it reduced to what's seen here.

This distortion did not appear on my previous captures (from a different tape), but those were done more than a year ago and the VCR has lain dormant since then.

Question: What's causing this? The tape? The VCR? And is there a quick fix, either in the equipment settings or in post-capture restoration/repair?

And, final question: Is there anything else wrong with these captures that I'm not experienced enough to see? Colors, image quality, audio, etc?

Thanks very much! I look forward to your thoughts!

No, it's a good thing.

Looks good.

Combination of tape and VCR. It's called "top curl". You could also call it flagging / tearing, but this is a mild case of it.

You can't fix it in post, except by cropping it out. If you added a Panasonic DMR-ES10 / ES15 / etc. in passthrough and disabled the VCR's internal TBC, that would probably fix it, but that also disables the VCR's DNR at the same time and then you've got to contend with the downsides of the Panasonic.

The VCR's TBC is also failing to fully correct at other points; see the wavy borders in the attached screenshots for example.

I didn't listen to the audio, but the picture quality looks like mush, which is typical of JVC playing EP. Can you do a capture of one of these same clips with Picture Mode set to EDIT for comparison? Should have more texture (but will also reveal more video noise).

The histogram has gaps (calling lollo). Don't know the best post to link for this, but there was some discussion here: https://forum.videohelp.com/threads/...ht#post2624623

The capture device seems to only do 16-235?

The histograms shows gaps and spikes. If you look to second picture is evident that there age gaps (vertical black lines), which is not good at all (banding and posterization).

Attachment 14581

Attachment 14582

Be sure that you do not touch the procamp levels of the capture card for debug purpose. Eventually try to capture with another sofware, i.e. AmarecTV.

In addition, the captured video is strange: I am not familiar with NTSC videos, but the structure of the fields and the the frames is wierd. It can be a source problem from the broadcast, or a capture problem, or something I am missing.

Attachment 14583

You can try to remove the TBC from the signal path and check what happens.

Thanks for contributing. :)

That shot (and the other one like it) were just sped up in editing.

Currently, each cut appears as a field from scene A and a field from scene B. Add:
The cuts are now clean when viewed as frames, the 3:2 pulldown clips are still normal, and the sped-up clips are progressive with some blending.

Thanks to both of you for the replies!

Well -- I'm thoroughly confused. Long story short, I've played that tape a few more times, and the top curl appears and disappears, gets larger or smaller, and it doesn't seem to matter what VCR settings are off/on or whether the TBC is active.

(Also, I may have accidentally turned off the VCR's TBC when I made the captures above. Ugh.)

So! Below are two more captures of the same Wal Mart commercial. (Both with TBC on!)

- The first one shows no sign of the top curl. I have no idea how that happened. Picture mode is NORM.

- The second one has the picture mode as EDIT, as msgohan requested. I made this capture five minutes after the other one, and changed no other settings -- but the top curl is back, intermittent, and huge. What the heck.

I got frustrated and tried another tape. And the top curl vanished! Is it possible that the first tape is just bad?


Either way: the 3rd-4th clips below are a commercial from this new tape, captured in NORM and EDIT picture modes.

The preview histogram for this new tape looked fine too, but since both msgohan and lollo2 pointed out gaps in the last clip's histograms ... I'm really starting to suspect that I don't know what a "good" preview histogram looks like.

Basically, all I know is that it shouldn't have red on the left or the right. (Meaning, levels outside the 16-235 "safe zone" mentioned in sanlyn's VDub settings guide.) If all the lines are in the blue zone -- I assume it's A-OK to capture. Right? Or is there more to it than that?

How are the luminance levels in these two new clips? Are there gaps/spikes in these histograms as well? If so, how should I eliminate them?

Also: which looks better to your experienced eyes, NORM mode or EDIT mode?

Thanks again!

There are 2 different aspects:
- Luminance levels: almost no card can capture full 0-255 range, many do 16-25X. In order to do not crush blacks and whites, if Y<16 or Y> 25X levels exist on the video to be captured, the procamp of the capture card can be used to stay inside the card range. If you prefere to stay in 16-235 range instead of the card range is ok, but if your card and your video have "matching" levels and range you are introducing unnecessary "digital" processing through the procamp
- Luminance pattern: while capturing the video you are using the Analog to Digital converters of the card, and the obtained digital video shows a Histogram. When using the procamp inside the card you are acting on a 8-bit enviroment not able to shrink or expand the range with accuracy. Hence the gaps/spikes (it may also happen at default settings). Have a look here: http://www.digitalfaq.com/forum/vide...html#post81202
Gaps in the histogram cause banding and posterization. Some post-processing can fix them.

You last captures "car-sample" show a clip at Y=16, so all details in that range are lost. Gaps/spikes in the histogram also present:
Attachment 14593

No difference between norm and edit:
Attachment 14594

Not acting on procamp if not necessary (use it only to avoid the levels that your card cannot capture, if present in the video, lesser evil). Never expand a range using procamp.
If it happens at default setting you cannot avoid it. Here a sample from another user (HushPower IIRC) with a IO Data GV-USB2:
Attachment 14595

Long story and many different opinions. It also depends on your tapes. Have a look here: https://forum.videohelp.com/threads/...gs#post2644788
Some comparison here:
https://www.youtube.com/watch?v=X1jpyNRGNGY
https://www.youtube.com/watch?v=X0FDfvNTEss

If I compare your "norm" and "edit" capture they are very similar. I prefer edit, because I see some more detail on the newspapers in the background (but it can be just different noise). Judge yourself: https://imgsli.com/OTIyNTA

Thanks for the response, lollo!

Since my second capture (car-sample) is clipped at Y=16, and the preview histogram looked fine to me, it seems that my initial suspicion is correct: I have no idea what a "correct" preview histogram looks like!

In this case, what should I have done to avoid this? Turned the Brightness slider up? (Sanlyn's VDub settings guide says "Brightness adjusts black levels; Contrast adjusts brights".)

Is there a guide for learning how to correctly read a preview histogram, and how to use it to set my luminance levels properly?


If I understand you correctly, you're saying that the usual cause of histogram gaps/spikes is having changed the proc amp's settings. Yes?

When you say "the settings," do you just mean brightness/contrast? Or other settings?

As far as I know, all of my capture settings for all the videos I've posted have been at "neutral" or zero. Which means that in this case, these gaps/spikes are being caused by something else, and there's no way to prevent them. Correct?

You mention that "some post-processing can fix them." What, specifically, should I do?


Examining my workflow, it looks like I have three opportunities to change the brightness/contrast:

1) physical buttons on the outside of my external TBC box.
2) sliders that control the Pinnacle device's settings: accessible under "capture filter" in the VirtualDub "Video" menu.
3) sliders that control VirtualDub's capture settings: accessible under "levels" in the Video menu.

Which of these would you recommend I use? Does it matter? I've only been using the third one. (Though again, it's been at Zero for all the examples I posted.)


Thank you for all your answers! They've been very helpful as I slowly learn how to work my equipment. :)

Maybe, but it may happen at default setting also. For example my Hauppauge USB Live-2 always produces spikes in the histograms and I have seen many captures of IO Data GV-USB2 producing gaps.

Whatever it causes the Y range to stay inside 16-25X range (brightness is the best candidate ;))

Any post-processing will re-format the levels in the Histogram. See some example here: https://forum.videohelp.com/threads/...It#post2615773
In your case a simple denoise with TemporalDegrain2() fill the gaps in the histogram:

Attachment 14596

Attachment 14597

2) and 3) should be the same. Concerning 1) be sure that your TBC is not altering the input levels. Do all your experiment without it, and once all is clear for you, put it back and redo.

Be aware that some card crush the blacks before the signal enter the procamp, so you can't avoid it.
If you will be able to avoid the crush you could be able to recover some detail. Here an example when avoiding to clip the whites (Y>25X) with the same card as yours: http://www.digitalfaq.com/forum/vide...html#post79910

Thanks for that link. Interestingly, it looks like when that user tuned his proc amp settings correctly, it also eliminated many of the gaps in his histograms! Perhaps the same will happen for me.

I did a little experimenting with my preview histogram's sliders, and it appears that it doesn't display "red" levels on the extreme lefts and rights (like the screengrab in Sanlyn's guide does). It shows nothingness, blankness, in those areas.

I ran the sliders up and down to their extremes, and couldn't make any red appear. Instead, the edges of the blue "bell curve" just disappeared off into these blank areas.

(This partly explains my problem; I thought "no red = good capture". Sigh.)

So, how should I be adjusting the sliders? Should I just "pull in" the edges of the blue until none of it is obscured by those areas?

I can't supply screengrabs right now, but let me try this:

At default settings, playing that commercial, my preview histogram looked like this -- the left side cuts off sharply at the edge, and lots of room on the right:

|-----\ |

So, I should adjust the sliders to pull the left edge in, and move its right edge over, giving me something like this?

|/--------\|

I am confused here. If you have no red it should be good, i.e. no crush at Y=16. Are you sure you are capturing and evaluating the same video that shows crushed blacks (car-sample1-norm.avi)?

I do not use VirtualDub for capturing, but AmarecTV, with a different procedure for evaluating histograms in real time. Maybe some VDub user can support better than me...

Correct. It's confusing me too! Right now my best guess is that some piece of hardware (presumably the Pinnacle) is masking out the extreme blacks/whites before they reach the computer.

I'll try some tests, adjusting the levels until the entire "blue blob" of the preview appears to be inside the left/right edges, and re-capture that same commercial.

What software are you using to examine my post-capture histograms?

AviSynth:

or

Read jagabo's post here; about halfway through he shows "a capture device that crushes at Y=16 and Y=235 will yield a histogram that looks like..." That seems to be what yours is doing. So the goal is to avoid a "spike" at either end, but remember to always crop the borders when evaluating and uncrop before actually capturing. (The borders are supposed to be "blacker than black".)

I want to respond to the earlier posts when I can look at the samples on my PC (just on mobile now).

I've never liked the term "crushed blacks", as it implies the capture card is doing something wrong. The problem is the source has illegal values outside the legal 16-235 range. Anything in the illegal range is forced legal. So your 0-15 values are all forced 16, same for 236-255 as 235.

Is that good? No.
But is that bad? Also no.

If you want to restore the values, then it needs to be done with a proc amp before conversions. Ideal is an actual proc amp. While many suggest the card proc amps, I'm fairly certain most of those manipulate the post-captured footage, and don't make analog domain adjustments. Few exceptions, and Pinnacle may be one of those (as it is a fairly complex card, better than most).

Some cards allow 0-255, but you still have to deal with it before conversion to a delivery format, especially disc formats intended for TVs.

That isn't necessarily true. The source can have perfectly legal voltages that the capture device's AGC decides to turn into illegal digital values. Your statement assumes that the capture device is accurately translating all levels at all times.

Ah, yeah, also true. :congrats:

I don't deal with a lot of wonky AGC cards these days. Oversight. :o

The Pinnacles USB 500,510,700,710 with the Marvin chip can only record the range of 16-235.
You have to adjust the contrast slider in the range of -20 and maybe also with the brightness slider.

The range of 0-16 can be neglected, but the super-white range of 235-255 should be recorded in any case.

Most USB capture cards can record the range of 16-255.
But there are also capture cards that can record the full range 0-255.

My recommendation in connection with a Pinnacle Capture card would be a Pioneer/Sony DVD Recorder from the year 2007 and later (e.g. Sony RDR-680, Pioneer DVR555.....) to use as Passthrough device. Here you can adjust (AGC on) the levels and the output via S-Video delivers the range 16-235.

But you should not forget that not only the capture card can be responsible for the clipping it could be the vcr or tbc,too.

In the end it always depends on your cassette and the content how you have to tune the hardware for capturing.

Great post, also all true. :congrats:

My apologies for abandoning the thread for a week. Work is crazy right now!

Thanks to everyone who commented with advice, links, histogram screengrabs, etc. It's all been very helpful!

I spent some time yesterday playing tapes, watching the preview histogram as it shifted around in real time, and experimenting with the various sliders. I think I understand how to pull everything into the safe "blue" range now.

I've uploaded three test videos; the same commercial as my previous post ("car sample"), now with hopefully correct luma levels. I think they look right. But as we've established upthread, what do I know?! ;)

Would someone (lollo2, msgohan, etc) please check me, and let me know if I've done anything wrong here?

The three videos are the same clip, captured using three different VCR settings: "NORM/AUTO" image setting, "EDIT" video setting, and NORM/AUTO setting with Calibration turned on. If anyone has an opinion about which setting provides the best picture, I'd love to hear it!

Thanks again!

Share your slider settings.

The captures now show histograms indicating that there is no "crush" of black levels:

Attachment 14667

Attachment 14668

Attachment 14669

The best capture in my opinion is the capture with edit=on; for example the details on the face of the girl are better in this mode:
Attachment 14670

Here some image comparison:
edit vs norm
https://imgsli.com/OTM0Nzc
edit is better

edit vs norm_calib
https://imgsli.com/OTM0Nzg
edit is better

norm vs norm_calib
https://imgsli.com/OTM0Nzk
very similar

In my captures the edit=off mode is better, in yours edit=on is better. Another evidence that we should not blindly follow any general guideline but always experiment with our own tapes and hardware!

Comparing the old capture with crushed blacks and the new capture I see an improvement. However is difficult to understand if details in the range Y<16 are now present in the clip with adjusted levels because the crush is avoided, or if they are visible just because thr higher brightness. You have to experiment yourself and to judge yourself (for example, increase brightness in the "crushed" clip and see if they appear or not):
Attachment 14671

clipped_edit vs edit
https://imgsli.com/OTM0ODE

-- merged --

Addendum: your "brightness increase" was probably able to bring some more details in the dark areas (not sure, as I said before, it must be checked). But as you can see, the video now is too bright imo. It was ok to (probably) avoid the crushed blacks, but now in postprocessing, you should bring the levels back to the "original" look. It can easily be done in AviSynth with Levels or ColorYUV without clipping anything.

Brightness +3, Contrast -16. All others at 0.

Though I'm under the impression that I'll need to change these settings for every tape.

-- merged --

Lollo, thank you for the histograms and the image sliders. These are helpful!

I have no real way of brightening the older clip to compare it with the new clip/check for missing details. (I imagine there's a way in AviSynth, but it's still very much a foreign language to me.)

Regarding my new, "too bright" clips: should I have lowered the Brightness slider by one before capturing? Or were my settings correct for capture, and post-processing is the accepted method for fixing this?

So -- thanks to you, I now have a better understanding of setting the levels before capture, and of the best settings for my VCR. I still have two problems left:

1) according to your screengrabs above, I'm still showing gaps in my histograms, correct? But this can be fixed in post-processing. (You suggested denoising.)

2) the videos are too bright. But this can also be fixed in post-processing. (And maybe with slider levels beforehand?)

Could this mean that I'm ready to stop doing tests and start capturing real videos?!

Yes to both questions.
Now you have all the elements for "good" captures.
Enjoy your captures!

Hooray! Thank you, this is exciting. ;)

Do you have any thoughts on this question of mine?

"Regarding my new, "too bright" clips: should I have lowered the Brightness slider by one before capturing? Or were my settings correct for capture, and post-processing is the accepted method for fixing this?"

I do not know. Use the brightness slider at its minimum change able to stay inside 16 Y range for a dark sequence.

With post processing you can correct the parameters. Of course, is better to have the clip at its "best" before postprocessing.

Start using AviSynth or VapourSynth, you'll be surprised by its power. If needed, ask here ;)

You're saying that, in general, I should only make the *minimum* changes necessary to the sliders in order to get the preview histogram fully inside the 16-235 safe zone. And nothing else. Correct? That makes sense.

In your screengrabs of the histograms from my most recent clip, I see that most of the graph is clumped over on the left. That's okay? I don't need to play with the levels to try to "center" that curve?

I'll get around to AviSynth at some point. One mystery at a time... ;)

Correct. If your card can capture 16-254 (for example) I would stay inside that range, rather than 16-235.

That's the "nature" of your video. You could offset the Y channel (-> increas/decrease brightness) at your preference to shift left or right the histogram, but in general the original pattern is the best. There is nothing to "center"!

Yes and no.

You need to establish a baseline that works with most tapes. The deviate from there. Don't start from 0 at all times if you know 0 is bad.

So is +3 / -16 the baseline?

You're asking me? Heck if I know.

I've only ever tried this on two tapes. The first (in my OP) worked fine at 0/0. The second one worked fine at +3/-16 (though given Lollo's comments it sounds like +2/-16 would have been better). So I don't have enough data yet to establish a baseline.

Sounds like I should start keeping track of which settings work best for each tape going forward, and trying to eventually glean a baseline that way.

I'll look at mine sometime soon, share my settings. I've always pre-adjusted with the AVT-8710. If the tape needed better proc amp, I'd use another workflow with another capture card.