CalMan 5 Release Notes and Discussion - Page 37

The gamma has been kept so all material from the last 50 year's is still usable.

It isn't just the age of the material. It's that there aren't enough bits available in consumer digital video encoding schemes to encode the range of light to dark in the material without a gamma curve. You have to take advantage of the way the eye perceives light to dark to get all of the latitude to fit. There would be banding at lower IRE's (hahaha) if the encoding was linear.
Edited by erkq - 2/26/13 at 5:17pm

To back this up here is a color comparator chart for a dE of 1.02


And here is one for a dE of 2.03


and lastly a dE of 3.02



The difference between a dE of 0.5 and 1.2 isn't going to be something you'd really be able to see.

Is it my eyes or is there Red in all the128s? A little Pink in the 191.
1.02 to 3.02 64 I can see differences.(3.02 is lighter)
191, between the Actual and Target, I can see differences on 2.03, and 3.02.
Hey, this could be a good test on how us Forum Members perceive what we're looking at! Good test Sotti!

The colors are all off in exactly the same direction for L*a*b*

Also to note, remeasuring the colors on my screen show 100% with a de of 1.3 for 1.03, and 1.95 for the 2.03, and 2.8 for 3.02.
The swatches where generated using calculated data, so they are all off the exact same amount, on a theoretically perfect rec.709 gamma 2.2 display.

So the actual dE varies a few points based on your monitor calibration, but you do get a good idea of how much error a 1, 2 or 3 is.
Edited by sotti - 2/26/13 at 5:20pm

make sure your computer's display is calibrated

edit: hmm, 4 posts from 1? now deleted.. forum is glitching a bit

Although since the colors are relative, not having a precisely calibrated monitor doesn't entirely invalidate the color comparator.

These look like shades of gray on my monitor. I understand that gray certainly has errors. But I was wondering if there's a source of the same type of swatches with color. That would be interesting to see. My color vision is a little compromised, but I saw some pink in the 191 target on the 3.02 series. My wife was able to see the 2.03 differences but said it is extremely subtle for her and she'd never know unless it was pointed out. For her the target and actual for 1.02 look identical.

Never thought of it.
My computer monitor has never been calibrated, since I've never had a reason to do so. So, we could be seeing a difference as far as we perceive B & W & Colours, plus our computer monitors may be throwing in their own flavouring!

The bottom swatches are the reference the top swatches show the error.

sometimes the bottom swatches get modified to show the relative error (like at 100% if we need to add more red we need to make the reference darker.

The color comparator is from CalMAN 5.1, it can show any thing we measure.

Guess I should download the new version.

imo. What you guys are talking about theoretically is correct, but we are dealing with consumer displays and consumer source material and its for that reason why I think its important to get your dE's as close to 0 as possible along with everything that does impact a real calibration.
The fact remains that with are consumer displays and consumer source material there are going to be variations that will affect a good calibration.

Case in point, with Calman's latest software release for a LUT Cube calibration there is a nice improvement over 5.0 in PQ. The dE's (Using dE as an example because of what you guys are talking about) are not much better than I was getting with 5.0, but there is no denying that the overall PQ is better. The one thing that has gotten better is that there is a lower max dE but average (0.8 with ver 5.0 verse 0.7 with 5.1) is about the same.

And then there is Joel's "color comparator charts" that as p5browne points out the differences for target and actual. And yes I understand the point about a moving picture and that any error up to the "yellow line" dE 5.0 is ok because you will probably not see it. Of -course most of the causal viewer's will not see much of a difference but some of us will, we are the hobbyist and the not so causal viewers that are pro calibrators and software programs bread and butter. So why not try for the best (lowest error) calibration or software even though "in theory" it may not be seen.

Kinda like the old saying, if a tree in the woods falls and there is nobody around, does it actually make a sound.

ss

Exactly why I use Color Comparator in combination with Color Checker to see which dE values are visible and which are not.

This tool was made for me in that it makes sense of my objective when I calibrate. My actual results have a visible target that my eyesight can assess.
As previously exploited by the algorithms, human eyesight is far from perfect and this tool shows exactly what we can 'get away with'.
I particularly like to make the patterns adjacent to one another to make the difference (if there is one) more obvious.

I suppose there is a certain academic satisfaction in getting all dE's in all areas of a 5x5x5 Cube below 0.5 but it won't necessarily improve the visible result.

In my opinion, the true test of the Cube is Color Checker because it tests points that have had to be interpolated.

Again you are correct but some 'Experts' are trying to convince us to use 'Electro Optical Conversion Function' (EOCF), as a more appropriate term rather than than the 'Gamma' we are all familiar with.

Experts or academics? Whenever there is a disagreement between two things that could both be technically correct, I always go with what is most commonly used - taking into consideration the context. The English language is full of these, for example (color, colour). That is how we communicate with people in the real world, by being flexible. I'm not really interested in the pure science of it.

Exactly. When I setup CalMan and experiment with gamma, I am quite happy to read about it in the numerous articles exemplifying gamma and providing known references. If I was at a trade show and lucky enough to meet a veteran of the industry with 50 years experience I would be very happy to adapt and use potentially incorrect terms like IRE and gamma for the sake of communicating.
Edited by AVfile - 2/27/13 at 9:01am

Right, it's called minimizing the overall error budget.

Heck, for peace of mind if nothing else!

All the more reason to use % white or digital levels and STOP using IRE unless you are discussing analog video and you understand what IRE is.

% white and digital levels are SIMPLE... easy easy easy concepts. IRE is NOT simple, it is FAR more complex and has nothing to do with digital video.

When you discuss IRE you refer to a voltage... say 400 millivolts and how that voltage relates to how bright the TV screen is. Digital video doesn't have voltage (per se)... it has 1s and 0s. The 1s and 0s determine how bright the screen is. IRE makes the whole thing far more complex.

Back in the days during the transition between analog and digital video, DVD players, for example, had analog video outputs and they were often connected to analog video displays. In THAT context, the discussion of IRE was quite valid and you HAD to know whether IRE 7.5 = black or whether IRE 0 = black or you'd never get images to look right. Now that analog video is gone, trying to hang on to IRE is just dumb. Especially when % white and digital levels are so much easier and completely unambiguous.

This was in reference to the visibility of different levels of dE errors...

You are missing something VERY important/critical... you are looking at something intended to give you a side-by-side comparator using still images of test patterns. When you display MOVING VIDEO IMAGES and when you HAVE NO STATIC TEST PATTERNS AS REFERENCE perceiving differences on this scale (dEs between, say, 0 and 3.5) are essentially impossible to detect. In a movie, color will change more than dE=3 just from someone moving from one side of a room to the other... because the light changes in different locations in the room. So you have no static reference when you are viewing moving images. If you were to freeze the frame and measure a red dress on the left side of the room and use that as your reference, then the person moves to the other side of the room and you pause and measure that red dress again, it would be a MIRACLE if the difference wasn't dE=10 or higher when you calculate the difference between the 2 measurements. Not only THAT... but if the display is calibrated reasonably well, both red measurements would be accurate. And if the display's red primary was off a bit, the dress would still change in correct proportions when measured on 2 sides of the same room. So the red might not be perfect, but the change in red would still be pretty accurate. Try making an animated GIF out of the gray test patterns and have them move randomly over the screen while you are trying to evaluate their color... gets MUCH harder. Now make the test patterns into birds and animate their motion in front of a stationary forest image with many shades of green and brown with the forest 20+ feet away from the camera position so there's a complex pattern of trees, branches, and leaves and sky in the background. Kinda makes seeing differences in the stationary test patterns simple in comparison. So don't be patting yourself on the back if you think you can see differences in the static patterns if you focus on them long enough... they'd never be on the screen long enough in a movie to allow you to do that.

Hi Doug,

My second sentence in your quote of my post was in fact not an argument merely my submission / capitulation to the advance of the digital age and was not an attempt to continue my 'Luddite' tendencies.

I enjoy my amateur calibration and hopefully will continue to learn.

Incidentally, whilst digital technology is as you say a series of 0's and 1's, am I correct in saying its purpose is to emulate analogue technology (including the old fashioned voltage and presumably valves) and is not necessarily a completely new science?

What are your views on using the term gamma.
Edited by PE06MCG - 2/27/13 at 1:05pm

The point is that IRE can be converted to % digital if you know if the content has a pedestal, and you can convert percentage into IRE if you know if the analogue format uses a pedestal.

But if you only have % digital you have no reference for how to convert it to IRE without additional information.

Being that IRE is not used at all in signal chain for modern digital video, the label doesn't have any place in the nomenclature for discussing video content.

Wow, is this all the difficult? When I was just getting in to this and having trouble understanding something, RayJR reached out to help me and I talked to him on the phone. One of the first things he said was IRE is no longer used, it is an old term from analog days, we use % luminous to describe the gray scale patterns as that is what they are 10% of 100% white 20% of %100 white, 30% ...etc
I said, ok, that makes sense ..

Not to jump off the current subject, but...

I just received my Calman C3 package today and I am having a very difficult time. Is there advanced settings for the meter and the readings at all in the software that I don't know what to change?

I have a 55 LGLW5600 and the first thing I found is on the warm setting I am getting 9000 on the temp reading. I was fairly sure this was in the high 6's from what others had said. This setting leaves me no where to go on the 2 or 10 point grayscales. I have very little Red in the 30% and zero Red in the 80%. The only way to balance those 2 out are add a ton of red contrast and brightness and then I am wrong the rest of the calibration. I have tried everything I can think of at this point. I have raised and lowered from top to bottom on the contrast, then tried clicking up and down on the Brightness, and messed with the Backlight too. Nothing made a difference to give me any red on my charts to start with.

Does anyone know if there is a setting the meter needs to be on besides the normal LCD and manual pattern stuff? Anyone have an idea what I could try to get started with some red so I don't destroy my picture right off the bat?

Kind of at a loss after just the first day...

Personally, I prefer EOTF since that's the terminology used by ITU folks.

Gamma is the name of the exponent in the power-law EOTF, it is not the name of the function itself.

I find that the C3 and C6 on the LGs always seems to show elevated Reds. This is why I profile from an i1Pro. Also check which Menu Item you're choosing for the Meter, and try different choices. Don't have my Menu choice available here at work, but can post tonight when I get home. (Something to do with White LED - usually 2 choices - one that will give weird readings, and the other correct readings.)
The other problem we find with the LGs, their White Balance settings don't meet their own specs, and have to be re-calibrated to spec. Warm - x=0.313 y=0.329 to give the 6500K. Usually the settings they have give the typical 9500K found on many of todays TVs that try to bring in customers Brightness choices.

I believe I have it on LCD Direct LED for the meter setting. The other one said CFFL or something.. I am still trying today and getting the same stuff. Hopefully I figure this out.

Digital video doesn't obviate gamma... at this stage in digital video, gamma is still a necessity. 8-bits (minus 0-16, but including 236-254, 255 is "reserved" and not used, ever (supposedly) but the only way 8-bits is enough is if you "bend the line" into a gamma curve so there are more bits allocated to highlights to prevent contouring. Someone else (don't recall if it was this thread or a different thread) seemed to indicate gamma was necessary to provide more bits to avoid contouring in SHADOWS... that's not true. In fact, a gamma curve like we use, allocates fewer bits to shadows... just replace % white on the x-axis with appropriate digital bits... the flatter the "curve" the more distance along the curve is covered by each bit. At the right side of the gamma curve, the slope of the line gets steep and there are many more bits allocated to highlights than other parts of the luminance scale.

So 8-bit digital video is figitty and fussy... if it is not PERFECT through the entire playback chain, you end up with visible problems in the image (most often, contouring). 10-bit digital video is NEARLY foolproof, but it could be screwed up but it would have to be a much more egregious error than would upset 8-bit video. Once 12-bit video becomes standard/normal, it would become almost impossible to introduce video problems. Obviously, you could still induce problems if you were trying, but it would be pretty difficult for visible problems to exist if you were reasonably competent at creating the content. At the point 12-bit video becomes the norm (perhaps with Samsung-like internal processing done using 18-bits so controls are useful over their entire range of settings), we could, in theory, do away with gamma and use a linear relationship between digital levels (or % white). That means 50% white would be 50% stimulus (gamma in the range of 2.2 to 2.3 puts 50% stimulus at about 75% white). And legacy content (everything made prior to the "switchover date" could carry a "gamma flag" telling displays to apply an appropriate gamma curve while new content would have either no flag or a "no gamma for this content" flag. It would then be simple to manage.

So gamma COULD disappear in the future, but now is much too soon to get rid of gamma, we just don't have enough bits in current consumer video to get good results without gamma.

Yep... that was me... got it backwards.

In any case, this is the main point and a good one.

Thanks for reply Doug,

So things are changing which is as it should be with any science.

However, I wonder why the Black Level control on my display continues to be called Brightness.
Surely this should be an obvious target for correctness, or is there a good reason for this apparent error in terminology?

I gave it another shot and something just isn't working. To rehash.. I have the C3 meter and Calman 5 with my LG lw5600 LCD.

When I set the 2 point grayscale with 30-80%, I have no red in the 80% and only a small bit in the 30%. Blues are in the 112-15 range. My initial color temp readings are in the 8900-9000 range. Everything out there says the colors are really good on this tv and I believe they are. Just not sure why I have such a bad reading for temp and reds. They are basically missing through the whole 2 and 10 point grayscales. Turning contrast down has not helped at all. I have tried contrast from 94-72. The color clipping pattern with a contrast of 90 shows red stopping at 245 and green and blue at 247-249.. As everyon would figure, adding red in all the % points does nothing but kill the picture and I am left with nothing but pinks instead of whites..

Any ideas for me. I can get into the SM without a problem, but figured the controls are there without it, so I wouldn't yet.

The meter is brand new so I wouldn't think it is the problem..
Edited by eghill1125 - 2/28/13 at 2:57pm