OK, one more time: IRE levels from AVI capture to DVD authoring

[sanlyn]

There was a time not long ago when I thought I had this IRE black-white level business fugured out. Or maybe I really did, but I'm now involved with a problem tape that's out of the ordinary.

I started with: capture a VHS tape to Huffyuv-compressed AVI (RGB24) via ATI capture card and VirtualDub.

I'm ending with: A DVD with wild contrast, thin midtones, stretched-out brights, gloomy darks. Really looks rather stark.

So at some point Im screwing up the dark and bright extremes and over-lightening the midtones. This happens on several VHS tapes, but not all. So here's my procedure (the same one I use with all my retail and home-made VHS tapes, but not always with the results I'm getting on the "problem" tapes):

Most of my retail VHS tapes are copy-protected. To work around Macrovision, I use the following gear and media to record VHS tape to my PC:
- Play a tape on a Panasonic, JVC, or old SONY VCR.
- Input to a Toshiba RD-XS34 DVD recorder, used only as a pass-thru device for its line-level TBC. Toshiba's TBC does a nice job of cleaning vertical lines and most VHS chroma bleed, and does a tad of grain filtering.
- Because none of my DVD machines will record copy-protected material, the signal goes thru the XS34 into my AVTools TBC. From the TBC, go thru a SignVideo PA-1 proc amp to adjust luma levels and maybe do a bit of color correcting.
- Record to a hard drive via ATI All-In-Wonder 9600XT thru VirtualDub to Huffyuv AVI.

Problem: No matter where I insert my PA-1 or AVTools proc amp -- whether at the start, the end, or anywhere in between in this circuit -- some VHS Macrovision problem tapes give me dark gray bars along the bottom and right-hand side of the image. The semi-transparent bars are a little bit thinner than the overscan border bars you often see on some graphics programs. The bars weaken and fade (and the color usually improves) after about 30 to 40 minutes of tape, apparently due to Macrovision fade-out near the end of the tapes.

First, don't blame ATI's capture card. I know ATI's MMC has settings for this problem, but it's not the same problem you get when capturing with MMC. Anyway, I'm not using MMC. With certain copy-protected tapes, the bars also appear when I record thru a full-frame TBC into any of my Toshiba or Panasonic DVD recorders. If I remove the proc amp, the bars disappear.

I have thrown 4 different proc amps into and out of this circuit. Proc amp in, gray bars appear. Proc amp out, gray bars disappear.

Unfortunately, some of these tapes really do need heavy luma and color correction. I have to perform all this post-capture; those Macrovision'ed tapes just won't leave my proc amps alone. When I work with various filters on the AVI's in Virtual Dub, I'm using level and gradation filter plug-ins to keep the luma and RGB levels within the 16-235 range. On my PC monitor, the darkest darks do look a little thin and the brightest whites aren't so very brilliant, but I realize this is due to the different view between my PC monitor and my tv. When I render these corrected AVI's to MPEG using TMPGenc encoders, TMPGenc's histograms show me that I'm well within the 16-235 range. When I edit the video in TMPGenc Editor and author in TMPGenc DVD Author, I see what looks pretty much like a "normal" color and contrast range (though TMPGenc's editor and authoring players aren't so precise that you can use them to judge color balance, etc. -- they weren't designed for it) -- and TMPGenc's editor players do make the image look rather dimmer and more gungy than any other player).

Throw the DVD onto my tv set thru any of 4 DVD players, and DUH! Crushed darks and grimy lower midtones, colors that should be auburn or blue look black, no shadow detail, midtones and flesh colors washed out, and brights look stark and unreal.

Right away, let's not get into monitor calibration on my PC's or tv's. Most of my tape transfers, all of my retail DVD's and cable recordings, etc., etc., look just fine. I have $600 worth of monitor calibation software and hardware that are constantly being used to check the monitors, and spent over $500 having my tv's color-corrected. Everything I watch looks great -- except for this handful of VHS tape transfers.

So, somewhere -- in VirtualDub or in TMPGenc encoder -- I'm screwing up the IRE levels. My direct-to-DVD recordings of these tapes look just fine on both my Toshiba and Panny DVD recorders, because I used proc amps to put the quietus on luma levels (but, unfortunately, proc amps + Macrovision + TBC = bottom and right-hand gray bars). Many old-time movie fans zoom-out a bit when viewing classic films on TV, because they want to see 100% of those old 35mm Technicolor frames. So, I either find a way to keep those gray bars away from my proc amp, or figure out what I'm doing in the correction process to screw up luma levels.

And here I was for years, thinking I had this down pat until I run into this handful of crummy tapes ...

[edDV]

...
Right away, let's not get into monitor calibration on my PC's or tv's. Most of my tape transfers, all of my retail DVD's and cable recordings, etc., etc., look just fine. I have $600 worth of monitor calibation software and hardware that are constantly being used to check the monitors, and spent over $500 having my tv's color-corrected. Everything I watch looks great -- except for this handful of VHS tape transfers....

A very complex chain you have. My immediate question is how you are monitoring the proc amp changes? You need to tap in a "known good" monitor into each processing step to see where things are going wrong. A "known good" monitor is your TV* not the computer monitor.

If a monitoring tap isn't possible at each step, establish a known good path (all processing off) using a known good source (e.g. commercial DVD, cable box or test signal generator), then add a processor (or software process) one at a time to find the offender.


* TV/HDTV sets differ from pro monitors as they contain numerous correction processors (e.g. AGC, black level, chroma saturation processors) that hide signal errors. Pro monitors show the signal as it is 'warts and all' so you can correct it with the proc amp.

[sanlyn]

Thanks, edDV, but I've tried all your suggestions. Example, one test I made was to use this same input/output chain but lead straight to another DVD recorder instead of my PC -- along with several variations in between. I watch what the proc amp does, watch my PA-1 proc amp's level meters, etc., and make a 9000 bitrate recording. That looks pretty good to me on both my calibrated tv's. I copy that recording to my PC and check it with TMPGenc, convert part of it to AVI and check it in VirtualDub, and on and on. If it looks good on my tv, it never fails to look 'normal' on my PC or any of my 3 monitors. The big problem? With this handful of copy-protected VHS tapes, using the proc amp never fails to produce the gray borders described.

Consider this: there are two main variables between the two ways of recording: the IRE and color scale problems occur when I record directly to AVI without a proc amp's corrections. That said, I'm beginning to suspect either a colorspace problem or one with IRE at some post-capture processing step.

But now you mention it, edDV -- one or two steps in this capture process seem related to the bad output. The first step concerns using the Toshiba RD-XS34 as a pass-thru line-level TBC filter. As I said, it does a great job cleaning up crooked lines, various temporal noises, and chroma problems. But its output is, indeed, IRE 0 - 110, not IRE 7.5-100. I normally fix that luma/RGB problem with a proc amp in-circuit. But Macrovision as found on this handful of tapes has a problem with proc amps.

So I'm recording a source that apparenrly has 7.5 IRE on tape, comes out of the VCR as 0 IRE, apparently comes out of the Tohiba at 0 IRE, and without a proc amp it gets recorded at that IRE to AVI by VirtualDub.
I really don't wanna get into multiple IRE conversions while correcting color and eliminating noise, but does anyone think that one of the following would work after the major cleanup filtering is accomplished?

(a) Before rendering the AVI to DVD, use AviSynth's "Levels" command to convert the RGB of the AVI from 0-255 to 16-235 for TV? After AviSynth fixes the RGB levels, I can feed the AVI to TMPGenc for rendering to MPEG.

or,

(b) Take the 0-255 RGB AVI and render it to MPEG using TMPGenc's famous CCIR switch? When turned off, this switch directs TMPGenc to make no IRE or RGB level changes (output in this case would be 0-255 RGB at 0 IRE). When turned on, the switch directs TMPGenc to render with CCIR601 RGB levels (i.e, output would be 16-235 RGB and 7.5 IRE -- but does TMPGenc seriously clip the level extremes?).

How do other folks handle this IRE problem with AVI-to-MPEG? I'm sure others must have encountered it. Anyone prefer the (a) AviSynth level fix over the (b) TMPGenc level fix? I remember seeing old forum posts on this problem, but I can't seem to find any them after hours of browsing.

[sanlyn]

Problem: My LCD monitors (lordy, I despite LCDS's!!) are Samsungs. They're calibrated with Colorvision Spyder2. Never try to judge an LCD by eye(!). The ATI graphics drivers are Cat 4.5 and they're ignoring all .icm color profiles. Got in touch with Colorvision; they say, use Cat 4.7 or later. Also, DO NOT install ATI Control Panels if you use calibration software on your monitors.

Got out my old CD backup with Cat 4.7, uninstalled current drivers, installed Cat 4.7 on all three PCs (no Control Panel), ran Spyder2 calibrations on everything.

Voila!