GCD - Gamut Calibration Disk - Page 18

100/75 is better than 100/100 for plasmas, 75/75 is even better but HCFR does not calculate targets for that automatically. You have to use the spreadsheet targets and enter them manually in a dummy reference sheet.

Hello Zoyd,
Is there a mixup in the definition? In the previous post, the numbers were defined as amplitude/saturation. Your answer would mean that 100%amplitude/75%saturation is best for plasma. I always thougth that the recommendation was 75% amplitude and 100% saturation. Those are the patterns on the AVSHD disc.
Can you please confirm.

Yes, sorry - I reversed amp/sat

Just an idea that "Amplitude" should be changed to "Luminance". Also IRE doesn't exist in the digital realm it's an analog term why not simple use "% of stimulus" or the actual proper term used to lable the gamma and grey scale patterns. What you guys think?

When I use 75%A/75%S to calibrate my Samsung D series plasma can I just use 75% of Rec 709 in chroma pure and match my CMS's RGB to as close to 0% without matching the target values in the spreadsheet provided by Zoyd? Would both methods produce accurate results?

Recently I made this calibration comparing a greyscale taken with patterns from GCD and AVSHD. Below 60% there is a small deviation.
These measurements were taken with fields. When I use windows, it is the same.
As AVSHD is longer around, and GCD is new on the block, I have the impression GCD has an error.
Can anybody check to confirm? Thanks.
Fields AVSHD - GCD.pdf 2346k .pdf file

Visca, thanks for the reply.
Did a quickie with Fields Gamma. Same comparison as before.
Looks like a case of rounding up or down to bits.
Fields Gamma AVSHD-GCD.pdf 2376k .pdf file

Hello Visca, I am using a new Display3PRO. I am quite sure it is not a meter problem. The difference is too big. Also the finding is the same , on different days and with different patterns. Below 60%, GCD is alsways a bit lower than AVSHD.
Would be interesting if you and Alluringreality could discuss this and bit settings.

What difference do you consider "too big"? I don't see anything in those comparisons to worry about.

I mean too big to be a random error. The difference between GCD and AVSHD is repeatable.
Whether it is something to worry about is another matter. The difference is small. I just wondered why there is a sudden jump in the greyscale at 60%. I thought bits are bits and expected AVSHD and GCD to be the same.

Right you are, I was looking at the gray scale and should have scrolled down farther. From those numbers it appears 10-50% GCD is shifted ~1 bit high relative to AVSHD.

Calculated offsets

I recently did some calibrations with version a5.0.0. Is everything where it should be for that version? Anything I should be concerned about?

Our initial discs were rounded to the nearest whole number this way, but somewhere along the way I decided to switch to using equal steps. I don't remember which version it was changed on, but it's in the release notes that we changed the rounding to match Digital Video Essentials. Except for the first value above black, I believe all the grayscale luma values are equally spaced. The way gamma is usually handled I thought it made more sense to eliminate the unequal step in the middle of the grayscale. Anyway, the values we use for the encode could be pulled from the images at:
http://www.avsforum.com/t/948496/avs-hd-709-blu-ray-mp4-calibration/3390#post_22305141

One bit grayscale differences are generally measurable, but it's sort of an academic discussion since most user controls cannot adjust that finely. There are far more obviously perceptible differences in how various technology displays will create slightly varying on-screen image from similar measurements. Basically home TV calibration is not nearly accurate enough for one bit to realistically have much bearing on actual results.

There is no color science rationale for choosing equal levels+bit shift or normal rounding. As long as the calibration software uses the same levels both are exactly accurate. The spreadsheet uses GCD levels, I'm not sure what HCFR or the other packages use.

Without knowing how various software calculates 5% steps, I simply doubted if software commonly breaks gamma calculation down to 8-bits. If the software simply uses typical rounding for their gamma calculation, the steps in the measurements would be expected to be equal. My thought was that 5%*(235-16)=10.95, which is very close to 11. The portion near black is not really expected to follow a simple gamma function anyway, and it's also common to question the accuracy of measurements near black. Basically I couldn't come up with a reason to stay with typical rounding, and since DVE was the primary alternative at the time it made sense to just match their choice. In my opinion there are far more glaring problems with display calibration, and even if everyone was using the exact same software and model of display, I think realistically it's still questionable if display to display variation would commonly be calibrated to the sort of accuracy in discussion. I know you were just commenting on the math, and like usual I'm harping on my opinion that calibration discussion generally seems to assume a far greater level of display to display accuracy than realistically possible with current calibration methods and the differences in how various displays operate.

I totally agree. My viewpoint is it's useful to identify systematic errors of this nature (however academic) just so folks have a complete understanding of their implications and can then file it under "unimportant" if they choose.

I don't know if there have been any recent changes, but the versions of ColorHCFR on my computer appear to calculate display gamma according to the listed percentage. For example a target gamma is indicated for 50%, rather than 49.77% (luma 125) or 50.23% (luma 126). Of course when you use the ColorHCFR pattern generator the output is actually an 8-bit value, and I think they use typical rounding for the generator, but regardless if their patterns round or round-down, the ColorHCFR generated patterns don't exactly match the calculation.

So, this is purely academic then. Good to know.

However, at the moment, is AVSHD709 (final release, version 2D) or GCD technically right (at least for use with CalMAN 5; Generic Calibration Disc)?

^^^^^i guess the only way to really find out is to compare them with a accupel, quantum or possibly a video forge pattern generator. If i had either of them, i would love to do it.

The question assumes that one is technically correct. If the calibration software bases its calculations around values that cannot actually be represented by 8-bit content, I'm not sure how one or the other is absolutely preferable. When black is defined as 16 and white as 235, it's not possible to exactly represent 5% or 10% grayscale with 8-bit luma values, since each 8-bit luma value between black and white represents 1/219 of the range. I suspect that most sources are going to simply round for their values from 0% to 50%, rather than using a round-down like DVE. Personally I have a hard time explaining how 10% equals 22 steps for all except 50% to 60% is only 21 steps, but that's what you have by rounding 8-bit values to the nearest whole number.

Oxcart with one square wheel and one round wheel.
If in the stone age, someone would have said that it is a small difference, then we would still be driving around like that now.

Joke apart, I think that progress is always valuable if it is reasonable relative to the cost involved. In this case, it seems best that the calibration software use bit level for their gamma calculation, and that calibration software and patterns are co-ordinated. I don't think it is that difficult if people such as Visca, Alluringreality, Tom and Sotti and others put their heads together and agree on a standard. From what I understand from another post from Sotti, there is an EBU (?) standard but let those people mentioned work it out. It would reassure users here if they would come out with a joint statement that they worked it out and implemented it (if required), and also inform what that standard is.

For software that assumes integer stimulus levels (which I believe HCFR does), I've calculated the %error mismatch in luminance that both the AVSHD and GCD disks will generate assuming pretty typical values of 2.2 gamma/120 cd/m^2 peak white. Again, software which assumes typical rounding will be perfect for GCD and software which assumes "round down" will be perfect for AVSHD.


Edited by zoyd - 12/17/12 at 11:06am