How to correct vertical shifts in VHS capture (incomplete field starts)

[MLS221]

I've started an attempt to capture a series of VHS tapes, originally recorded on an old Philips VHS Camcorder (Philips VKR-6810).

After adding/replacing hardware component after hardware component in my setup, I still cannot get rid of vertical field bouncing because of field-start issues. So I hope for your advice on getting rid of these issues (and maybe some suggestions for further quality improvements).
Sample video: https://we.tl/t-X97f9Oi3KL

The affected fields start with one or more (max ~5) black lines, whereafter the correct signal picks up again. This causes a (vertical) offset for that field.
- this results in the video bouncing in a vertical direction for a few frames (some parts of tapes once every 1 - 5 seconds); very distracting.
- one bounce mostly is limited to a few successive frames.
- the vertical offset can be different each time (jittery) and can be different for the two fields that make up a frame (see picture; notice the baseball hat is clearly split vertically).

These glitches happen with all my tested VCR's (Philips VR1100/Blaupunkt RTV-950/...) and the signal problems can also be seen on a CRT TV connected straight to the VCR's output.

My current work-around is to de-interlace QTGMC after capture and use Adobe Premiere Warp stabilization (only panning). It works quite ok if you cut the timeline for each original video cut, but of course this is not the way to go.
One thing I noticed is that if I pause the video and go frame by frame, the problem is gone and I have a rock steady image. I was almost thinking of creating an automated 'send *next-frame* IR-command anddd capture' apparatus

Is there anyone who has a more realistic idea of how to fix this (either in the capture workflow, or in post)?
Besides a hardware-based solution, I can imagine an Avisynth script that:
1. splits the fields odd/even
2. detects if a field has unexpectedly got a few black lines at the top (prevent false positives somehow)
3. cuts out those first lines that are completely black and shift the result up; pad to match original frame size
4. merges odd/even again
5. Crop/border the rough edges

My hardware setup (All on S-Video):
VCRs:
- Philips VR1100 Digital TBC/NR (Recommended JVC-Clone) - both with and without TBC enabled; gives the cleanest signal
- Blaupunkt RTV-950 (Panasonic NV-FS200 Clone) - more noise and slightly more dropouts (same issues).
TBCs:
- Kramer FC400 TBC
- even tried a full field Video Mixer WJ-AVE5 (same result)
Capture:
- Terratec AV350MX USB (ATI600USB) or
- also tried a Panasonic DVD Recorder (same v-shifting issues, with/without TBC)


All help is welcome

[MLS221]

In an attempt to implement the script I proposed above, I'm looking into Autocrop and then padding odd/even fields to equal sizes.
However, it seems Autocrop can only sample one specific frame (or set of frames) to find the borders (black lines) for a clip.
If it could do it for every individual frame/field of the clip (preferably only top side), I would have a start...

[Brad]

On the post-processing side, if they are random, have you tried something like a five-capture Median?

On the capture side, I think certain hard-to-find Pioneer DVD recorders are generally considered to handle this vertical jitter the best.
http://www.digitalfaq.com/forum/video-restore/7577-reducing-vertical-jitter.html

[manono]

One video sample is worth much more than all those words and the picture

[MLS221]

On the post-processing side, if they are random, have you tried something like a five-capture Median?

On the capture side, I think certain hard-to-find Pioneer DVD recorders are generally considered to handle this vertical jitter the best.
http://www.digitalfaq.com/forum/video-restore/7577-reducing-vertical-jitter.html

The Pioneer recorders could be an option, thanks for the suggestion. Does specifically the "Pioneer DVR-530H" excel in this V-sync correction, or are the renowned Panasonic ES10 units also capable of this?

I'm unsure about the Median filters; they distort the image severely in my tests.

[MLS221]

One video sample is worth much more than all those words and the picture

You can (could) find a link to the video sample in the opening post.

[manono]

You're right and I apologize

I'm unsure about the Median filters; they distort the image severely in my tests.

Are you referring to the same median filter as vaporon800, where you make multiple captures of the same video and it selects the best frames from among them? Here's the link to the discussion:

https://forum.doom9.org/showthread.php?t=170216

[MLS221]

Are you referring to the same median filter as vaporon800, where you make multiple captures of the same video (...)

Ahh, now I understand the type of median! Thanks for clarifying.
I'm afraid the glitches are quite consistent in terms of their frame nrs (and position in the frame), so there is little clean signal available, but I will certainly try it with a short clip to verify!

[oln]

The Pioneer recorders could be an option, thanks for the suggestion. Does specifically the "Pioneer DVR-530H" excel in this V-sync correction, or are the renowned Panasonic ES10 units also capable of this?

I'm unsure about the Median filters; they distort the image severely in my tests.

I know at least the the successor models I use (440, 540), which do have nearly the same chipset, are not as capable as the ES10 and related, though they don't have the side effects of the panasonics either. (Use with VCR TBC off)

I've experienced tapes where my JVC SVHS camcorder (JVC GR-sx22e) was the best at avoiding vertical jitter, but it's TBC is a bit hit and miss.

Both the virtualdub deshaker plugin, and some of the avisynth stabilization scripts may be helpful, I've mostly used the former. You can set it to only pan vertically, ignore borders etc.

[MLS221]

Hi everyone, I just wanted to give an update before closing this thread as solved:

The tip of vaporeon800 that pointed me to a post of hysteriah was of great help.
I PM'ed hysteriah and after some helpful sharing of experience, I decided to go for the Pioneer DVR-530H-S as pass through machine.
Remarks: The DVR-630H does a similar thing to the signal. IMHO the S just stands for a Silver model and is only printed at the back of the machine.

The vertical jitter is resolved instantly and my captures are now usable without any software fixes or even extra TBC-ing.

My final H/W setup:

Philips VR-1100 (NR-on: cleanest with least chroma issues), although Blaupunkt RTV-950 is not that much worse.
|
SVIDEO (SCART)
|
Pioneer DVR-530H-S
|
SVIDEO (DIN)
|
Kramer FC-400 (TBC does not make a lot of difference anymore after adding the Pioneer)
|
SVIDEO DIN
|
Canopus AV350MX (ATI600USB)

Thanks everyone for your input.