New calibration disc - Page 2

GetGray,

I can test it, but won't be able to check for accurate video levels.

Bill, you are not alone. I assume that you are getting some pretty strange color values when you capture a frame and then test it in MS Paint or Photoshop, right? As a reference, if you have one of the commercial test discs, capture a similar or identical frame from it and perform the same test. Your results should be very close to the results you got from Scott's disc, equally bizarre.

Scott was able to get more accurate results using DGIndex and measuring the color levels directly from the encoded video (I think) rather than capturing a frame and testing it in a paint program - results that make a LOT more sense. I have no idea what is going on with TT, but I noticed that all of the gray levels seemed to track just fine, but when I tested color windows, the results were way out in left field.

I found some of my notes on the values using TT during some of the early color window tests:

First, here's some reference 75% window JPEGs, loaded directly into MS Paint (note that the 75% values are approximately correct):

75% red window measures 180,0,0
75% green window measures 1,180,1
75% blue window measures 0,1,180
75% cyan window measures 1,181,182
75% magenta window measures 179,0,180
75% yellow window measures 181,180,1

Next are Scott's 75% windows, played back through TT using VMR9, captured, and then loaded into MS Paint:

75% red window measures 234,0,0
75% green window measures 0,210,0
75% blue window measures 9,7,248
75% cyan window measures 0,203,204
75% magenta window measures 224,0,224
75% yellow window measures 188,188,0

and last are the DVE Pro 75% full fields (title 12, chapters 15 - 20), also palyed back through TT using VMR9, captured, and loaded directly into MS Paint:

75% red window measures 238,0,0
75% green window measures 0,213,0
75% blue window measures 6,7,253
75% cyan window measures 0,205,204
75% magenta window measures 228,0,229
75% yellow window measures 191,190,0

The reference color JPEGs, though not perfect, are very close to perfect. The slight discrepancies could be simple rounding errors in the testing procedure AFAIK. Scott's captured values and DVE Pro's captured values are BOTH way off from what they should read, though they both read similarly, and since DVE Pro should be reference, Scott's should be equally accurate.

Once again, I believe that Scott got much better results using DGIndex, so I have to assume that the error is in TheaterTek....This concerns me as to using TT for any input calibration (and I even wonder what it is doing to my normal DVD playback), and as things stand now, I will only use the patterns played from a hardware DVD player or from Jeff's built in patterns (when I use a PC to generate them).

Thoughts?

I have found that TT does not capture what I'm seeing when I use VMR9 when doing a screengrab. ZP seems to capture things accurately. For some reason TT grabs a level-expanded version so what you capture is way off, but what you're seeing is not.

Well that sucks! I had understood that the screen grabs in VMR9 should be right, and it was the grabs in Overlay that would be completely wrong. Also, I noticed that all of the gray percent stimulus patterns seem to grab correctly, while just the color windows are out to lunch. Chris, do you have any idea why this happens? And what might be the difference between the way that ZP is grabbing as compared to TT?

Since I don't use ZP, could you also check the color windows for accuracy using ZP?

no I don't know why this happens. I think Vern Dias suggested setting up a print to file and doing a printscreen grab and that should work, but I haven't bothered setting that up.

I've been meaning to look at the levels on this disc, but I've been really swamped with work right now, unfortunately.

Peter Finzel does a disk that is a lot easier to use and navigate compared to DVE, and that's even considering that it's in German!

http://www.peterfinzel.de/pftd.htm

Worth every Euro IMHO.

There's a link here in German so you may have to go via a translator:

http://www.online-translator.com/srvurl.asp?lang=en

And there's a PDF in English to explain it all, though most is self evident.

Looking forward to your disk Getgray.

Gary.

Bob and Chris -
I got similar results from Scott's disc as what Bob posted. My grayscale results (both mixed and window patterns) were dead-on. The primary and secondary windows were off. To test, I used a program called ColorCop from www.datastic.com (a great little freeware color sniffer). Since I use VMR9, I used ColorCop to check the levels as they were rendered. I also compared screen grabs from Windows' "Print Screen" functionality.

With all that, the results were consistently off (no measurement deviation here). Maybe someone who isn't on Andrew's s*t list might want to send him a friendly note once Scott does another pass through himself? I doubt I would get very far with him on the color neutrality of TT. I'll give ZP a try later, but there are a few too many things on my plate right now, so it's going to take me a while to do any signficant testing.

A quick question: does anyone know where in the rendering pipeline DGIndex takes its shots? Is it at the end like the print screen does, or is it earlier?

Later,
Bill

It is my understanding that DGIndex decodes the mpeg vob file itself....

Jim

Hmm, interesting bill, I'll have to look into this, hopefully early this week. I'll see what I come up with.

Interesting comments on TT's inaccuracy when taking screengrabs of color windows - I had this issue near the beginning of the year with DVE and Avia, yet later on the issue disappeared (I had perfect results very similar to the reference Bob cited). However, the issue may have come back and I was unaware of it - I will have to test this soon.

Are you guys using FFDShow? Is this with the NVIDIA 6x00 series cards? DXVA on? YUV Mixing? I recall doing my tests months ago with NVIDIA Purevideo only (DXVA, no YUV Mixing) and had good results, I believe with (now old) NVIDIA 71.89 display drivers.

I wonder if we would get better results using one of those DirectX frame grabbing programs (HyperSnap DX or whatever it is called)?

Perhaps the best thing to do if we can't trust the screen grabs is to use a colorimeter/spectroradiometer to measure directly off the display device, comparing a screengrab from TT and from a reference picture using MS Paint.

Max - I was testing using my deskotp system, which uses an R9600. DXVA was enabled, and no processing was used (NvVPP, FFDShow). I did not use YUV mixing, though, nor renderless (Plain vanilla VMR9).

Later,
Bill

Are you using a current version of TT?

Partly why I'm curious, because I do remember max's testing of TT levels some time back, and they weren't off if I recall.

Chris - I have the latest patch release (2.1.1) that is on the TT website support/downloads page. If there is another version buried in the forums, I have not searched it out and loaded it.

One other inconvenience is that TT reverts to overlay when displaying static images.

Later,
Bill

Ok, yeah just checking, because the older version was way off on the color at the default settings, you had to adjust to get the color right, but andrew changed it now so that it should be spot-on at default.

According to Andrew, TT 2.1.1 reverts to the system default renderer when displaying static images. In the case of XP, this would be VMR7.

Version 2.2 will use the renderer of choice, which in my case will be VMR9.

I've been out of the loop with TT for a while, what's the deal with 2.2? What are the new goodies?

Bob - I should be able to get a print screen off of VMR7, but I can't. ColorCop also reports "black" for all areas of static images - the only difference is border area vs. rendered area on the screen (0 vs. 16, respectively).

Any info on when 2.2 will be out?

Wow, and with that, we are WAY off topic here...

Later,
Bill

Bill, I guess it all depends on what you are calling a "static image". When I asked Andrew, I had specified playback of picture files (JPEGs, GIFs, etc.), and here is his exact quote:

Scott, would you consider making a version for CRTs that had the 75% plus 100% color windows?

Also, if you were willing to set up a website to offer the ISO file with a link to PayPal for a donation I think it would make it very attractive for on line buyers. Although PayPal does charge 3% to payees.

It's not just TT, but all Windows players. The problem seems to be caused in part by below black.

I was able to see BB on many players. If you see it- then that has to be absolute zero. When I do a screen capture it removes below black and tries to compensate for it. In any case, I don't think we can rely on the accuracy of screen captures for verifying IRE level accuracy

What do you mean by this?


In TT I get this as well, it's an expansion to the PC 0-255 range, though it's not what's actually seen as rendered video. I'm going to explore this hopefully soon to try to get actual screengrabs that reflect exactly what I see, and measure those. ZP does this without a problem it's just I have to spend 30 minutes every time figuring out how the heck to use ZP, and the ARs are all screwed up .


I think you meant level accuracy, IRE are not present on the disc.

Not sure why you guys are having these issues. I use MWSnap and capture from TT VMR9 renderless with none of these issues.

MWSnap also has a color picker tool, and I just reviewed one of my captures from the DVE pro crossed greyscale step pattern. I just displayed the PNG with Windows Picture and Fax viewer and used the picker to confirm each bars color values.
Every single one of the bars are spot on, perfect and equal RGB values.

The system I just checked this on is not my HT system, but the system I use for daily work at my desk. Card is an NVidia 5950 Ultra running the 71.72 drivers.

I'll do some additional checking later today.

addenda:

One other thing, apparently the white and black level contols have no effect whatsoever on the values that the color picker gets, which tells me that the color controls are applied later in the chain, probably via a LUT arrangement and do not affect the actual values in the video buffer.

On my system using DVE Pro crossed grey steps, video black is exactly 16,16,16 and video white is 235,235,235, regardless of the settings of the video control panels white and black levels or the levels in TT.

VMR9, renderless, ffdshow using Lanczos2 resize and DScaler sharpen, YUV mixing enabled.

I'll check both of HTPC's later today.

Vern

Vern, everything we measure in the gray scale, from BTB to WTW, measures perfectly. It is when we measure color windows (red, blue,green, cyan, magenta, or yellow) that the values are way off. I know it really shouldn't make any difference, as the greyscale patterns are made up of the same colors as the color windows, but for some unknown reason they don't measure correctly. If you have the time, please try measuring all six of the 75% or 100% color windows and let us know what you get.

If you are using DVE Pro, the patterns are not actual moving video, but instead are still frames, so you have to make the color changes in TT, then change chapters, and then go back to the chapter you are measuring in order for the changes to take effect. This is one of the problems with using a HTPC with DVE, and one which Scott corrected in his disc by using infinitely looping MPEG2 video rather than stills - none of the silly chapter changes in order to have the color controls kick in...

It would have been an easy task to simply copy DVE's "still frame" method, as we wouldn't have had to worry about levels being effected by MPEG2 encoding, but we really wanted to have the changes take place in real time, and thus Scott valiantly struggled with the encoding process until he was able to encode without screwing up the levels. Believe me, this was no easy task! Scott spent countless hours within a couple of weeks time span before finally succeeding, but his efforts were very well spent!

I am well aware of the issue with still frames and VMR9.

Even though I can clearly see the crushing on the screen, if I crank up the white level, the value shown by the MWSnap color picker is still 235,235,235.

So be aware: The values obtained from the color picker are NOT always the same as what is being sent to the display.

Here are my values for the DVE Title 12 Ch 15-20

Your DVE values ....... my DVE values

75% red window measures 238,0,0 ..... 183,15,16
75% green window measures 0,213,0 ..... 14,181,13
75% blue window measures 6,7,253 ...... 15,16,184
75% cyan window measures 0,205,204 ..... 13,181,181
75% magenta window measures 228,0,229 ..... 182,15,184
75% yellow window measures 191,190,0 ..... 182,181,13

These were taken direct from the screen, using MWSnap 3.0.0.74. Card is an NVidia 5950 Ultra running the 71.72 drivers.

VMR9, renderless full screen or window are the same, ffdshow using Lanczos2 resize and DScaler sharpen, YUV mixing enabled.

Vern

And you are using TT 2.1.1, not 2.2 or ZP, right?

I'll pick up a copy of MWSnap and try the same thing myself. It seems as if MWSnap, though it doesn't pick up color changes in the player, might be a better tool for measuring color values encoded in the video, though the values you reported from DVE Pro really should have been more accurate than that, don't you think?

Thanks, Vern.

If DVE Pro was mastered with analog signals, then it might not be. The mid-IRE values really take a beating for rounding errors when converted into (ahem, proper) digital values.