How to capture with no head-switching noise (certain VCRs)

[Brad]

The continuing saga that began with Which passthrough devices straighten out the bottom of VHS signals?. I've been meaning to start a thread on this since 2017.

A current discussion in a grave-dug thread has prompted me to post proof of the method. Just a mini write-up at the moment, not a real guide. But this is a simple procedure once you know where to look inside the machine.

With my recent usage of a SAA7133 capture card that can be forced to show the Vertical Blanking Interval, it's now easier to demonstrate exactly what's going on.



Internal adjustment of VCR potentiometer



Panasonic PV-S4670 service manual:


Our goal here is to intentionally deviate from the spec, so you don't actually need any of the tools/equipment mentioned in the service manual besides the appropriate-sized Phillips-head screwdriver. Unless you want to restore the original setting.

Warning: Metal screwdrivers are a bad idea for this, at least with the models like PV-S4670 that have a metal support bar right near the PG Shifter pot. I fried mine, and I found out that a guy who read my post fried a PV-S4280 and PV-S4566.

Models

I've only tried a handful of models so far:

1. Citizen JVHS3989B (garbage mono VCR)
2. Panasonic AG-1980
3. Panasonic PV-S4670

Unfortunately, the AG-1980 creates chroma artifacts toward the top of the frame if the HSW point is pushed outside the active image.

Typical JVC VCRs need not apply. They use some sort of auto-adjustment, set with the hard-to-find Jig RCU service remote.

Similarly, the Mitsubishi HS-HD2000U service manual says how to perform the automatic adjustment of "playback switching point". Nothing about how to manually force it.


(The attached sample is encoded in ULY2 = Ut Video codec.)

[dellsam34]

I wish capture cards can capture at a flexible resolution that can be set during capture so that the final output when crapped to say 720x480 only clean video will be left. For example if the junk lines are 8 on top and 12 on bottom and 7 pixels on left and 13 pixels on right, the capture resolution will be set to (720+7+13)x(480+8+12) or 740x500 then crap in VDub the values mentioned earlier and you will get clean legal 720x480 output frame.

[Brad]

"Crap"? Autocorrect?

I'm not sure I understand. Analog tapes will always have blanking on left & right edges at 720x480. The closest standard res is 704x480.

For the vertical direction, DScaler's Vertical Delay option with the special BT8x8 & SAA713x chip support lets you line up whichever scanlines you want into your 720x480 output frame.

I think certain Osprey cards using Fusion 878A chips let you perform the same task in their own software with their own drivers.

[oln]

The Bt8x8 tweaker thing in Virtualdub works on my Conexant Fusion878 based card, so I would presume DScaler would as well.

[dellsam34]

I'm not sure I understand. Analog tapes will always have blanking on left & right edges at 720x480. The closest standard res is 704x480.

That is because video capture hardware is locked to a certain resolution. For the horizontal resolution where no pixel structure is defined in the analog source, the hardware could capture at higher pixel count so that the number of pixels from where the real video in the scan line starts and where it ends is exactly 720 and this can be made to be adjusted while monitoring. Unfortunately most capture cards have pre defined sets of resolution and the only solution of getting around those borders without damaging the video is mask them in black to keep the resolution legal. I know there is nothing can be done about vertical scan lines, they can either be doubled or tripled, they cannot be fraction-ed.

The reason why I'm bringing this up is because capture devices are still being made most of them are crappy off course but this can be taken into account in a new design to solve the tape based video capture border syndrome, I'm aware of software solutions such as dscaler but I'm talking about hardware solutions at the source.

[Brad]

That is because video capture hardware is locked to a certain resolution. For the horizontal resolution where no pixel structure is defined in the analog source, the hardware could capture at higher pixel count so that the number of pixels from where the real video in the scan line starts and where it ends is exactly 720 and this can be made to be adjusted while monitoring. Unfortunately most capture cards have pre defined sets of resolution and the only solution of getting around those borders without damaging the video is mask them in black to keep the resolution legal.

I still don't understand. Filling the entire 720 width with truly "active" picture from the tape would change the sampling aspect ratio to some non-standard ratio that would vary with every source, sometimes from shot-to-shot within the same source.

Any player that assumes this material was captured according to spec would output an image with a horizontal stretch.

[dellsam34]

I still don't understand. Filling the entire 720 width with truly "active" picture from the tape would change the sampling aspect ratio to some non-standard ratio that would vary with every source, sometimes from shot-to-shot within the same source.

Any player that assumes this material was captured according to spec would output an image with a horizontal stretch.

The non standard resolution is intended for capturing purposes only, the final resolution will be legal such as 704x480, there will be a compromise to keep a legal aspect ratio by leaving some junk either vertically or horizontally but it won't be on all 4 sides.

[Brad]

I've added the following warning to the OP.

Metal screwdrivers are a bad idea for this, at least with the models like PV-S4670 that have a metal support bar right near the PG Shifter pot. I fried mine, and I found out that a guy who read my post fried a PV-S4280 and PV-S4566.

[dellsam34]

I only once came across a VHS tape where there was no head switch, it felt strange watching that in the capture preview window, I have no explanation to that but I will be posting a sample when I get home.

Edit: Here is the sample I was talking about, It was only cropped horizontally to 704, that's it.

I never cared for the head switch, it didn't bather me at all, I follow this procedure I posted at digitalfaq to go from the full frame on tape say 720x525 (625 PAL/SECAM) to the final 704x480(576) as the final master file. But when I started to upload to youtube I decided to crop to a bare minimum (based on the ratio 704:480) and resize to 1440x1080, My point was since I'm going to resize and screw up the image anyway may as well crop it clean and resize, I was shocked to how little was the damage, Here is two samples from the same source, This one from the master 704x480 de-interlaced and encoded to h.264, and this one is from the above file before encoding then cropped to a bare minimum and resized to 1440x1080 then encoded to h.264.
.
.
.
.