Inserted frames not being reported? (VirtualDub)
 

I was yesterday years old when I learned that two identical tape captures of mine did not sync up perfectly...

So apparently, back at square one, the Pinnacle-710 USB is inserting frames- and moreso, not telling the user of when they occur (unlike the Dazzle, which immediately informs when there's frames being inserted/dropped). After my findings on my two tape captures, I went ahead and captured some tapes that had DVD counterparts, so I could have a "sync-point".

Two tapes were successful, they matched perfectly down to the frame. However my third experiment was with a tape that was jittery, and could only be fixed with an ES10. Since all my previous captures were in VirtualDub2, I figured it's possibly the software. I thought a jittery tape would be a good experiment to see if the TBC is doing it's job.

After going frame by frame, I found regardless of capture software, there's still a matter of inserted frames.

So........ what to do from here? My goal is a 100% authentic grab from the VHS, without any inserting of frames. Some tapes I have don't have DVDs, so I can't line them up to see if the sync is perfect. Let's hear this discussion.

My specs:
JVC SR-V10U + HR-S9600EU
Pinnacle-710USB
Datavideo TBC-1000
VirtualDub2 (formerly), VirtualDub 1.9.11

Attached: VirtualDub timing settings, capture details, and a video project with both captures matched to an earlier frame. Shows the end of the clip with the delay.

Home videos or commercial tapes?

Frames are inserted for a reason.

Are the inserted frames always at the same point in the tape? If not you could try multiple captures then stitch together pieces as necessary to eliminate/replace the inserted frames. You should be able to use the audio track wave forms to check sync.

I'm not expert but I've always had The Resync Mode on No 3.

Where's Sanlyn's VDub guide when you need it? :cool:

A frame is inserted by the capture software in order to keep synch between audio and video when the card did not send a frame inside the expected time slot.

It may be reported or not depending on the capture card and/or the capture software. Just as example, I found the pair Hauppauge USB Live-2 / AmarecTV beeing accurate in reporting, while sometime ATI USB 600 /VirtualDub pair does not report correctly.

If a frame is inserted and properly reported, it is marked in the captured avi file by any capture software. You can jump to it opening the file in VirtualDub and using the bindkeys "?" (previous) and "^" (next).

If you see the inserted frames exactely at the same position across multiple captures, the problem is in the tape.

If a frame is inserted and not properly reported, you can use a large set of AviSynth script to find them (basically you search for a difference = 0 between current frame and previous frame). Some example here http://avisynth.nl/index.php/CategoryDuplicate_Frame_Detectors, many other specific scripts in dedicated forums.

If a frame is dropped and not reported, there is no way to find it, except a boring step-by-step check (difficult to find it in any case) or a comparison with a different source. On my side, for a sanity check I often use a DVB dump of the same program when available or a DVD release. Be careful that a DVD release may use a remastered video and the frame alignement is sometime lost (i.e. I have seen recent DVD releases missing frames compared to previous releases; probably the frame was so bad in the original source that was simply removed).

Yeah same, I've always gotten audio drift if disabling the resync stuff.

Also note that although it's pretty rare, both the datavideo and ES10 CAN drop/insert frames in some extreme cases (an stupid example is if feeding NTSC video into either in PAL mode.) You can also get some weird extra frames when the JVC TBC acts up depending on what's next in the chain, don't remember quite how the datavideos react to it but with the ES10 you can get these weird blended/mixed frames (guessing you did not use the TBC on that tape if you needed to use the ES10 tho?). May be best to re-check if you get the same result on the specific part of the tape if you try again.

If the 710USB has some buffering internally it's possible it could too, but idk.

My NTSC DMR-ES15 and DMR-ES25 always drop/insert frames internally with VHS input; a few per hour. At one point I owned two DMR-ES15s and I captured the same VCR signal from both simultaneously. They inserted frames at different spots. I also confirmed that the DMR-ES25's inserted frames appear on its DVDR disc recording, in the same spots as the passthrough output (my ES-15s have broken drives).

It doesn't happen when they're fed a signal from a DVD player.

Do you also mean that capturing the same tape without the DMRs (if possible) there are no inserted frames?. I know that is not easy to distinguish between the 2 cases, just want to understand if the DMRs add frames by their intrisic nature or if they do in correspondance of some "event" that will produce the inserted frame even capturing without them.

Correct. (Using a test tape I recorded with burned-in frame counter.)

Thanks! An additional side effect of the usage of DVD-R recorders in passthrough mode, then. However, this can be considered very minor.

My biggest gripe in all of this is the fact the software isn't telling me- and both Inserted/Dropped values stay at 0. With my Dazzle, it had the courtesy to tell me if a tape was inserting anything, that way I can go back and resume capture from a certain point.

Perhaps, it's worth sticking to the Dazzle if it's gonna tell me when frames are being inserted?

Commercial tapes are my tests- I'm lining them up with their DVD counterparts.

I've always been against the resampling mode, primarily due to the fact that if the video is perfect, the audio shouldn't be taking a hit by slowing down or speeding up. In my eyes, a perfect preservation capture means an authentic representation of the source, both in audio and video.

I'm not sure about dropping frames, that's usually dependent on the setup of the user. Inserting frames, I wouldn't have an issue if the software told me it was doing so (usually, you want to only use a DMR passthrough to correct a problem that isn't fixable on your main gear (VCR+TBC) ). Less processing on the video being captured.

There is no way for any capture device to know when a device upstream (like DMR-ES15) is inserting or dropping frames. The only way to know is by manually analyzing, like you're doing by comparing multiple captures or to a digital source. As far as your capture device is concerned, it's receiving a stable signal from the DVD recorder's output.

Those cheap Dazzles easily insert/drop.

ES10/15 silently inserts as well. Again, not a TBC. This is a negative side effect. These are ideally only used for tapes where the net effect is improvement, not as a TBC replacement. These units have downsides, fail rates, artifacts, side effects. But it was lower cost, right? You get what you pay for.

The Pinnacle does not drop/insert without reporting. I would not use it, if that were the case.

sanlyn's VirtualDub guide has errors. Many times, different cards need different settings. For example, if you try to resync on this Pinnacle, it will actually cause audio problems. Resync should be disabled.

I've never seen an ATI 600 USB not correctly report.

Retail tapes may look cleaner, in general, but are actually some of the worst to deal with. Those tapes were usually not recorded, but made by contact.

Capture devices have no idea if upstream caused errors. It only know what it's doing. If the signal was made decent by insert/drop beforehand, then it considers the signal good for capture. Over and over and over, I discuss why certain TBCs are not good, why "also has TBC" isn't actually a TBC (internal needs, not yours), and why non-TBCs are often worse yet. This is prime example. I do often forget about the drop/insert issue, but it's top 5 on the list.

How can I see the framedrop details capturing with AmarecTV?

You select in the menu the option to write a report file:

Attachment 14834

then you read the gerated file. An nserted frames are reported as
while a drop frame is reported as
A summary is written for every frame and at the end
where the second number represents the dropped frames and the third the inserted frames (first I do not know)

I do when I compare my captures (AmarecTV + Haupauge USB Live-2) with friend's capture (VirtualDub + ATI 600 USB). I always talk about the pair capture sofware + capture card, because I am not able to differentiate between them.

Some discussion here: http://www.digitalfaq.com/forum/vide...g-options.html

To the OP: the script Capture Glitch Analyzer (detect dropped frames) is difficult to run, but can be a test for your case if you have doubts about the reporting https://forum.doom9.org/showthread.p...31#post1462931

I would like to challenge you. Presented here are samples of 2 captures via Direct Stream Copy; one PAL, one NTSC.

In the PAL capture, you can see an inserted frame from #96-#97. The head switching noise at the bottom remains the same, so it's not a fault of the master dubbed to this particular cassette.

In the NTSC capture, from frames 53-95, there is a very strange field distortion (???) that is present. I rewound the tape, and the issue does not appear again, so it's not the tape.

If this is a fault with either the Pinnacle 710 or the TBC-1000, what are the steps to take from here? Just to reiterate, VirtualDub is not reporting dropped frames or inserted frames. It has (once) on a super static tape, so I'm sure it knows when to insert when it sees a problem. But this- this is just something I have to comment on. It could be an issue that is being overlooked by even the pros here.

I see some weird combing on the left side of the video.

So you repacked the tape and the issue is gone.

Then you're right, it wasn't the tape.

That's (one of) the issue(s) though. If these are flaws being introduced by the gear, and we're not preserving an authentic representation of what's on the tape, then could we really call it a faithful preservation?

Even if the inserted frames don't throw a video off sync, the accuracy part irks me so much, it's been a battle for nearly 4 years- because if these tapes just... cease to exist, you'd want the file exactly as it was. That's why I feel it's important for even the professionals here to be aware of this.

Lollo, could you check this please? I did a capture with VDub and Amarec and VDub reported 75 inserted frames:

Attachment 14837

Amarec showed this in it's report:

It seems that the second number "(+)73" is inserted frames and the last is dropped frames.

Don't worry, through the ES-15 I got zero for both. :)

Edit:
The above test was done with my GV-USB2

I ran a test with my other capture box, a Startech USB3HDCAP and VDub reported 2 dropped and 32 inserted. Amarec reported Zero of both, even though there were obvious freezes and jumps in the video.

At least for my 2 capture sticks, it seems that VDub is more consistent for dropped and inserted frames.

Repacking can affect drops/inserts. But don't go running and repacking your tapes -- it goes both ways. And you take added risk or needlessly FF/REW tapes, especially these days.

Can TBC-1000 insert? Yes. It's a strong TBC, but errors can be stronger. Rarely happens, but happens.

Yes, there is a mistake in my previous post. Obviously the number close to "+" sign refers to inserted frames, while the number close to the "-" sign refers to dropped frames. Sorry for that!

Yes, I found that is more the pair capture software / capture card in cause than the software itself. In your specific case, it is also possible that the signal delivered to the capture card is freezing and jumping itself, because apparently is a not so stable tape needing a lineTBC correction (your VCR does not have lineTBC IIRC); it may have happened in a test and not in the other, i do not know.

Because you are familiar with GraphEditNext (I remember some discussion on VH forums), you could also experiment to capture the same segment with it and compare. In this case you build "your own" capture software, rather than using VirtualDub or AmarecTV, and only rely on the correctness of the report of the Microsoft filter AVI Mux, which is quite accurate.

With my Hauppauge USB Live-2 I build a graph like this when I do not display the incoming video on the PC and use a TV connect to the VCR to see the video:
Attachment 14841

or like this if I only use the PC (no audio preview while capturing):
Attachment 14842

- replace the Hauppauge DS Filters with the Video filters of your capture card, and eventually the rendering display art according to your PC

-configure AVI Mux for INTERLEAVE_CAPTURE https://msdn.microsoft.com/en-us/lib...(v=vs.85).aspx

- when you stop capture, do not close the GraphEditNext window, but click on properties of the AVI Mux filter to read dropped frames (the capture "software" so built is not able to insert frames, so it just drops frames (or better, AVI MUx drops frames) when the incoming flow of audio and video packets is not perfectly stable.

Just for doc purposes, these are the graphs used by VirtualDub and AmarecTV (yes, they also use a graph internally ).

VIrtualDub:
Attachment 14843

AmarecTV:
Attachment 14844

Yes, it's just a cheapo VCR, but the tape is appalling. Nice and "steady" though through the ES15.

Thanks very much for the info on the graphs. I will look into that later (I only use Graphstudio these days to control the Proc Amp on my capture cards).

Is that JVC SVHS with TBC on? That looks a lot like the effect you get when the tbc in the JVC VCRs is acting up, which at least in my experience on PAL they are pretty prone to if there are issues in the non-visible area like dropouts or on e.g unstable camcorder recordings but YMMV. When the VCR TBC does this I've noted one tends to get inserted or weird blended frames depending in setup, and even if not there will be vertical jitter, even with my datavideo TBC. I find using a panasonic dvdr and TBC off is better in such cases but as noted even that can give drops. Haven't noticed the TBC in the Panasonic NV-HS1000 acting in this sort of way (though it can do this left-right wavy thing instead in some cases), don't know when it comes to the later panasonics - in either case the in-vcr tbcs don't work well on very bad tapes, nthg gen tapes etc.

Hodgey, that behaviour is not unusual with the lineTBC of JVC VCRs. Rarely the even or the odd fields are shifted by a certain number of scanlines (I experimented 1 to 4, 5 in just few cases).

In the posted sample, for example, the frame 100 is corrupted, because the even field is shifted up 1 line by the lineTBC. It can be easily seen moving frame-by-frame in VirtualDub . It can be corrected with a simple AviSynth script:
Attachment 14847


edit: frame 2, 23, 26-30, 34 and many others are also affected

Isn't that how dvd recorders work? They have to insert or drop frames to keep audio and video in sync no matter how bad the input signal is.

Is it possible to upload a short example of your test video with the burn in frame counter so I can create one for myself.


I also noticed the field shifts. Had the case before as well.
I would first turn off the autotracking, if that doesn't help try a Sony/Pioneer DVD recorder (Sony 680,870,205...Pioneer 630,640,560...) or best another player.

It seems that the VCR is not the right player for this cassette. That's why many have several players to choose the best one for each cassette.

There is no such thing as the perfect VCR, perfect TBC, perfect capture card, there were too many different VCRs, VHS tapes that recorded to VHS. That is why it was often recommended to use the VCR that recorded the tape. Only these devices rarely have an S-Video output, line TBC.
Whether you should follow this recommendation depends on your capture hardware and the condition of the cassette and the content. Line-TBC is not necessary if you want to use a DVD recorder as passthrough.

The field shift is not a problem at all, and can be easily fixed; few lines of an AviSynth script will rebuild the correct frame without artifacts or side effects.

The problem is that I experienced the shift happening 1 time over >10.000 frames, while in the sample of thestarswitcher it happens too often. His capture can be fixed to its original quality as well, but it requires hours or days to find and fix individual errors. BTW, I use an automatic script to find the vertical shift that was developed to recognize generic bad fields https://forum.doom9.org/showthread.php?t=183582

As you properly stated, in the case of the OP, the best will be to change player (switch to a Panasonic VCR) or to disable JVC lineTBC and use an ES-10 or ES-15 instead. I am always reticent to use my ES-15 because the loss of details in bright areas, but in the case of the OP is the only choice if he does not want to spend a lot of time in fixing his capture.

Yes and no. Any ingest device has to drop/insert in order to maintain the ingest rate. If something is untimed, it has to make the decision to drop/insert to maintain. The obvious fix is to time the source, though some uneducated users blame the ingest process/device. Noting that some bad devices exist, but it's more about bad drop/insert decision-making, sometimes bad external forces like drivers.

But it really has nothing to do with audio. Video is here, audio is there. They should meet, but don't have to. And that's the problem of analog ingest, of course. It is what it is.

DVD recorders can lose sync, too. It depends on how the recorder is programmed, chips used, etc. So yes, it can drop/insert, and yet maintain sync. But it's not really to maintain sync. That's a byproduct.

Remember, DVD recorders were not made to transfer tapes. These were PVR devices, record from antenna or cable. Videotape capturing was an edge case use, in terms of the overall market using DVD recorders.

Yep. :congrats:

Yes and no. There is no "one VCR to rule them all", but there are VCRs that are usually best. General rules exist for a reason. JVC S-VHS decks usually best, followed by Panasonic S-VHS decks (more quirks), and then the list quickly dips off to non-recommended VCRs.

No. This advice was based on fallacy. The "original recorder" is almost never the correct player, unless it happens to be a recommended player. Even when a tape is misaligned, you can misalign a better player to match (VHS/S-VHS only, not Video8/Hi8).

Tape condition only goes so far. Consumer format analog videotapes (VHS, etc) have intrinsic flaws of the format. So even a "good" VHS tape will have the flaws. Those flaws cause later issues at capture, and with quality in general. Those flaws need to be adjusted with TBCs, both line/field and frame.

False. Line TBC is redundant with some DVD recorders. Not any random DVD recorder. You know this, I know this, but readers may not. We must choose words carefully sometimes. :wink2:

Slightly off topic, but is that Tugs? Not sure about international versions but it was a UK production (TVS, when their stock was high in the late 80s) and had an uncredited Patrick 'Voice of the Apocalypse' Allen on VO duty? It used fair chunks of the know-how and team from the original Thomas The Tank Engine series. Not thought about Tugs for years.

Mr Allen had the famous for the job of being probably the last voice Brits' would have heard should the Cold War have turned 'hot'. Like this lovely little thing about how to bury a fatally irradiated loved one in the garden.

Better ask to thestarswitcher about his captures ;)

Ah, yeah, with a forum handle like that, it's going to be Tugs. It was the Star fleet (recourse to Wikipedia) and a switcher is a small tug aparently.

Back on topic! :P

Not entirely accurate. The JVC line TBC behaves differently than the Panasonic field TBC (multi-line). The advantage here is this...

The single-frame/field line TBC doesn't have the advantage of the multi-line field TBC, which can look back and ahead. So missing, truncated, and damaged lines are simply null inserted. The JVC doesn't understand missing/incomplete/damaged lines, so you get jitter. This is a flaws of JVC, more than line itself. But it affects other line TBCs. Panasonic line TBC in ES10/15 doesn't usually suffer here.

This is usually only nth gen tapes, but not always. It's actually most common with cheaply made contact retail tapes. I've also seen it on SP VHS tapes made in cheap Panasonic VCRs.

This is a reason I got my first Panasonic AG-1980, way back when.

Correct, but this is not the problem here. The whole field is shifted up, because a stability problem in the incoming signal that the lineTBC of the JVC VCR (but it also happens on lineTBC of the Camcorders) is not able to fix aligning to previous field. But, as I said, the whole integrity of the shifted field is preserved, and can be restored back.

The only case where I found a potential problem is when the shift counts 5 scanlines; you may miss then the first scanline of the rebuilt frame if the black border of the captured frame is not big enough to hide it; but that's something that is not visible if you do not slowly move frame-by-frame inside the video.

When I look at the sample, the insertet frames are the lesser evil than the shifts. You can correct a lot of things in post-processing, including a non-constant audio delay caused by dropped frames if you invest enough time. Is that worth the effort, if by adapting the hardware to his tapes the problem can probably also be solved.

My experience is different. I don't like the picture from the JVC players. If I were to follow your recommendations (JVC+Datavideo) I would invest a few thousand dollars/euros in hardware that doesn't match my tapes recorded on Sanyo & Siemens VCRs and that's 100+ tapes.

How many user would adjust the alignment of their recorder to fit a video cassette? In order to set it back correctly, you need more experience and hardware in dealing with video recorders.

Well, since here in the forum only the Panasonics ES10 & 15 are recommended anyway, a video recorder with line TBC is not necessary. Because then you can not use the special capabilities of these devices.
Also one of the reasons why I don't understand the recommendation for JVC's SVHS recorders with tbc, if the users want to use an ES-10 for jitter correction. You pay for functions of a video recorder that you then do not use at all for recording.
It's not so simple. This may be for users who have no experience with other devices but if you are a little or even a little more concerned with the matter, there are alternatives to the recommendations.

I can only repeat myself, which hardware to use depends on the cassette and its contents.

I do not understand your reply. Why asking that? I wrote "(changing/adding hardware) ... in the case of the OP is the only choice if he does not want to spend a lot of time in fixing his capture", so we agree there ;)

I just wanted to confirm your statement that it is possible to correct this in post-processing but it is not worth the effort, which is what you said.
Not that anyone else comes to the thought that this is the recommended way and makes false hopes.


Bear with me lollo2, I read in English and write in German with the help of an online translator due to time constraints.:wink2:

No problem at all, Bogilein, there is no need to apologize. We all appreciate and understand your post even if you use online translator :)

Gonna be replying out of order;

You better believe it :D . I'm convinced it's one of the most beautiful pieces ever put on television, and it needs to be preserved (the film elements got accidentally junked)!! I run a page all about preserving the show, mainly highlighting behind-the-scenes materials from crew members- if you wanna check it out!

Really appreciate the script here! For field shifting, I use Telecide() but I'm under the impression this one is more precise, in the means of actually repairing the individual frames opposed to the whole.

As far as Tugs is concerned, I have so many different tapes loaned here from friends and collectors to combine the absolute best elements from each, to an eventual full-scale series restoration. Of course if one tape is jittery, we can rectify it by using other copies. That being said, for other videos, if this field-shift is a quick turnaround, I'm sure it would be extremely useful for my projects!

I do want a (PAL/NTSC) Panny's, but I really need to run into a tape that will truly justify the cost.