[maxleung]

I'm reading conflicting reports and opinions on what is the correct gamma for viewing movies.


CRTs are typically 2.5 gamma, and is what we've been used to for many decades. However, with the advent of digital displays (LCD flat panels, digital projectors, etc.) there seems to be a lot of variance. Most LCD monitors for computers are 2.2 gamma - just look at the reviews on behardware.com, xbit-labs, etc. and they produce gamma charts of 2.2. My own DLP projector, the Benq 8700+, has a default gamma of 2.5. My new LCD monitor, the excellent NEC 20WMGX2, has a gamma of 2.2 and is close to D6500K right out of the box (but still needed some calibration, oh well)!


Watching movies on my projector (with a brand new bulb) is a different experience from my LCD - shadow detail is slightly muffled on the pj, but it has a more 3D look to it and video artifacts are not very noticeable. The LCD has loads of shadow detail, but there is quite a bit of video noise that is more apparent. I don't believe LCD overdrive is an issue here.


My guess is that the lower gamma on the LCD is the main culprit here.


I've read in the discussions here on AVS a while back that most DVDs are mastered with Sony Trinitron studio CRT monitors which have a 2.5 gamma. Yet why do most of the new digital projectors come with 2.2 gamma? And why do many of the esteemed ISF calibrators who post here use a 2.2 gamma if DVDs are mastered for 2.5?


This has me concerned - when people report they see "great shadow detail" while watching a trendy projector (last year it was the Pearl, this year it is the JVC RS-1), I wonder if it is merely the 2.2 gamma they are seeing. This will fool them into thinking the projector is superior, when another display they have been watching has a 2.5 gamma!


Of course, I have read here that at the lower end of the gamma curve, good projectors will have a linear output response in the lower IRE end of the scale, and then follow the standard gamma curve for the rest - this could also explain the enhanced shadow detail in some displays.


Anyways, I hope you guys can clear this up for me, and maybe when people gush over a new projector, they can tell us what the gamma is set up for it.

EDIT: I was wrong about the dE of my NEC being 1.0 before calibration. That would be AFTER calibration. My bad!

 

[jvincent]

I have done a bunch of reading on this in that last week or so. I'd recommend the following article, plus the gamma articles on Poyton's site.

http://www.w3.org/Graphics/Color/sRGB


Distilling all of opinions, the answer is that there is no single, correct gamma. The gamma you choose is heavily dependant on your viewing conditions and to a certain degree personal preference.

 

[jimwhite]

I don't buy it.... if the video is delivered to us having been "encoded" with a certain gamma, then by-golly our playback device should at least have the default set to the correct gamma to properly "decode" the delivered video. If you then slect a user gamma because you have ambient problems, so be it... but to say there is no standard is ludicrous... what's the point of calibrating? Just eyeball it...

 

[ChrisWiggles]

Reference gamma should be 2.5 if you want to see what the engineer saw on his CRT.


Note that this is not the exact inverse of the encode gamma. This is why people often get so confused and think it should be 2.2.


There are many good reasons why one may want to deviate a bit from 2.5, shadow detail is a good one. Many CRTers (myself included) tweak gamma for a bit better shadow detail because CRTs have weak ANSI.


But bottom line, theoretically the reference display gamma is an assumed natural 2.5.

 

[Victor]



Quote:
Originally Posted by ChrisWiggles /forum/post/0


Reference gamma should be 2.5 if you want to see what the engineer saw on his CRT.

Isn't ideal gamma 2.2 according link you provided and Poyton?

"Using the actual power function fit value for the 709 transfer function of 1.0/1.956 and maintaining the display gamma of 1.125, we can solve for the ideal target monitor gamma of 2.2. This is consistent with the CRT gamma value proposed in this paper."

2.5 is "phisical" CRT gamma

 

[maxleung]

Ok. It sure would be nice if displays had different gamma options that would also tell you what the gamma curves are like.


I'm still very surprised that some of the higher-end projectors use a gamma of 2.2. When people say "this projector is so amazing you will see the flaws in the DVD master!" I shake my head - they are probably seeing the effect of a low gamma on a source mastered for a higher one!


Also, the reports of macroblocking in certain player models can be attributed to this very gamma issue - something I verified last year when I calibrated my projector with a 2.2 gamma and saw macroblocking everywhere. It's not necessarily your player that is at fault!

 

[ChrisWiggles]



Quote:
Originally Posted by Victor /forum/post/0


Isn't ideal gamma 2.2 according link you provided and Poyton?

"Using the actual power function fit value for the 709 transfer function of 1.0/1.956 and maintaining the display gamma of 1.125, we can solve for the ideal target monitor gamma of 2.2. This is consistent with the CRT gamma value proposed in this paper."

2.5 is "phisical" CRT gamma

No, Poynton assumes CRT's natural 2.5 gamma to be the display gamma. Note that the encode gamma is not the inverse of this, i.e. we don't have unity gamma from end to end. That is to say, the original NTSC gamma of 1/2.2 gives the mistaken impression to many that display gamma should be the inverse of that, or 2.2. The reality is that though the display gamma is not specified, it is assumed to be 2.5 which is the natural response of a CRT.


The paper linked by someone else above seems to be in error, because it seems to assume CRTs have a natural gamma of 2.2. While CRTs vary and it's certainly possible to have a CRT with a gamma of 2.2 (or higher than 2.5) it's generally assumed to be that the gamma of CRTs is about 2.5, which is natural. CRTs behave physically according to a 'five-halves power law' or 2.5.


Poynton is fairly clear in assuming a natural 2.5 gamma for CRT displays, including in his fairly quick Gamma Faq: http://www.poynton.com/PDFs/GammaFAQ.pdf


I'm not a physicist, but I am inclined to believe the attribution of the 5/2 power law for CRTs. Therefore, I clearly side with Poynton despite numerous internet sites that claim that a CRT's natural gamma is 2.2.

 

[laric]



Quote:
Originally Posted by ChrisWiggles /forum/post/0


No, Poynton assumes CRT's natural 2.5 gamma to be the display gamma. Note that the encode gamma is not the inverse of this, i.e. we don't have unity gamma from end to end. That is to say, the original NTSC gamma of 1/2.2 gives the mistaken impression to many that display gamma should be the inverse of that, or 2.2. The reality is that though the display gamma is not specified, it is assumed to be 2.5 which is the natural response of a CRT.

Thanks Chris, totally agree with that... but who am I.


Also, encoded gamma not being the inverse with addition of linear segment "mistery", plus some "strange" implementations on some projectors can be really confusing.


--Patrice

 

[ChrisWiggles]

Also, just FYI the wikipedia entry on gamma correction is really quite terrible and rampant with ommissions and wrong facts.


And just to reiterate before, while gamma is confusing, it is subjective too. I don't particularly have problems with conservative deviations from an "ideal" 2.5 display gamma for various very good reasons, as long as you are aware of what you're doing. However, there are a lot of displays (especially digitals) with very low gammas that really make things look bad and lead to the kind of observations that Maxleung laments.

 

[maxleung]

Thanks for the clarification and links Chris. I had forgotten about them when they were posted at AVS a year or two ago.


Maybe a calibrator can speak up and explain to us why they prefer 2.5 or 2.2 gamma? I know that UMR assumes 2.2 gamma in the Accucal software, and I'm curious why he and others would prefer that when it appears most DVDs (and I assume hidef content) are mastered with 2.5 gamma playback in mind.

 

[jvincent]

Just when I though I had it......


Let me try and summarize.


1. Viewing gamma is assumed to be 1.125

2. Encode gamma is 2.2

3. Display gamma should be set to 2.5


Now, if I look at the ColorHCFR application, my measurement options are (ignoring the last two for now):

a. Display Gamma

b. Display Gamma with black comp.

c. Camera Gamma


So, should I be targetting Display Gamma with a value of 2.5?

 

[Bear5k]

For folks interested in the last major discussion on this subject, I would point them to this thread.
http://www.avsforum.com/avs-vb/showthread.php?t=771150


Per Poynton's book, you reverse the camera encoding in the display for an end-to-end gamma of 1.25 (0.5 camera * 2.5 display). The problem is that the gamma curve at the camera does not approximate the 0.5 power curve that Poynton says it does (at least not with any of the estimation methods I know that are defensible). CalMAN uses the defined curve because of the guiding idea of reversing the camera encode. To us, this is the most defensible method to offer to our users. It also happens to emphasize shadow details, vs. the crushed blacks that plague many displays.


What you do with encoding intent, CRT floating black levels, and ambient light all make this fairly confusing.


Bill

 

[umr]



Quote:
Originally Posted by maxleung /forum/post/0


Thanks for the clarification and links Chris. I had forgotten about them when they were posted at AVS a year or two ago.


Maybe a calibrator can speak up and explain to us why they prefer 2.5 or 2.2 gamma? I know that UMR assumes 2.2 gamma in the Accucal software, and I'm curious why he and others would prefer that when it appears most DVDs (and I assume hidef content) are mastered with 2.5 gamma playback in mind.

The idea of a single number for gamma is not really a good one in practice. The goal is a certain look to an image. This is defined by the relative contrast and is dominated by the lower 10 percent of the image in most cases. Remembering that the gamma function is part of how 8 bit video works well is key. Using a gamma significantly greater than the 2.5 target curve tends to result in artifacts like banding caused by insufficent bit depth. Much less than the 2.2 target curve and you will have a washed out image.


My software can display any gamma curve you choose as default. I use 2.2 as more of a lower limit and a guide, but in the end it is the look of the image that matters the most. Shooting for a specific numerical result when it is the lower portion of the curve that dominates the solution makes little sense to me.


The key point to remember in calibrating any display is that it is the look that matters in the end. These tools are just a good way to get you in the ballpark of a correct image, but in the end we are measuring a small subset of what is really going to be displayed in a real image. Without good tools you may be too far off to get it right, but judgement should always come into play in a quality calibration.

 

[jimwhite]

This may be a terribly naive question, but.... Ignoring subjective desires for a moment, it would seem that a properly encoded linear grey scale ramp, when played back on a device with adjustable gamma, would show various linearity curves as gamma is adjusted, say from 2.0 to 3.0, and the gamma setting which yields the most linear grey scale ramp as measured at the display surface would be the technicaly correct gamma, would it not??

 

[Bear5k]



Quote:
Originally Posted by jimwhite /forum/post/0


This may be a terribly naive question, but.... Ignoring subjective desires for a moment, it would seem that a properly encoded linear grey scale ramp, when played back on a device with adjustable gamma, would show various linearity curves as gamma is adjusted, say from 2.0 to 3.0, and the gamma setting which yields the most linear grey scale ramp as measured at the display surface would be the technicaly correct gamma, would it not??

Jim - the problem is that the luminance information is implied, rather than explicit, in the encoding. How it gets turned into real-worl luminance levels is via the gamma look up tables (LUTs). If a grayscale ramp were created that gave not only the implicit signal level in some form (digital code values, % Stim, IRE, whatever), but also what the normalized luminance level ought to be, then you could validate your gamma in this fashion.


Bill

 

[Bear5k]



Quote:
Originally Posted by jvincent /forum/post/0


So, should I be targetting Display Gamma with a value of 2.5?

If you have a display that let's you change gamma and you have an environment where a 2.5 gamma won't destroy your shadow detail, then go for it. For the vast majority of displays out there, though, you have almost no control over gamma, and it is merely a reference value to help explain what you are seeing, much like the locations of the primary colors in your gamut.


Realistically, though, when you embark on tweaking gamma, you want detail on what your gamma is at each point in the grayscale, rather than one overall number. Few displays that we have seen have consistent gamma values, using whichever method, throughout their grayscale.


Bill

 

[maxleung]

Thanks guys. I can see why people send in their JVC DILA projectors to William Phelps for gamma tweaking - he can tweak the LUTs in the projectors directly.


It would be nice if all display devices could do that. I wouldn't mind tweaking my LCD monitor's LUTs and have a linear scale in the lower portion of the grayscale, and then following the 2.5 gamma for the rest. Then I can have my shadow detail but keep macroblocking in lighter areas down!

If I invested >$1000 on the high-end NEC LCD monitors I'd be able to do this - they have 10 bit LUTs that are supposedly tweakable by NEC's software utility.


My Benq 8700+ has 3 usable gamma presets - the Film preset crushes shadow detail, but the Graphics preset appears to bring them back, yet also keep a 2.5 gamma. I'm not sure if it merely increases the brightness or if it actually has the lower grayscale end on a different "curve".

 

[tbrunet]



Quote:
Originally Posted by ChrisWiggles /forum/post/0


Reference gamma should be 2.5 if you want to see what the engineer saw on his CRT.

Color correction artist make use of reference grade direct view CRT', and at ~$40K each, I seriously doubt the design engineers from (Sony/Ikegami) intentionally created a response or OETF of 2.2 by mistake?

Quote:
Originally Posted by Raymond Soneira /forum/post/0

http://www.displaymate.com/ShootOut_Part_2.htm

They're all just paper standards, but it's the de facto standards that really matter: what is actually being used to create professionally produced content.


I measured a professional Sony CRT monitor to have a ruler straight perfect power-law Gamma of 2.20 and the Director of Engineering for Ikegami assured me that Ikegami CRT monitors behaved in exactly the same way.


Professional CRT studio monitors provide the standard Gamma of 2.20 by using analog signal processing in the front-end electronics.

 

[umr]



Quote:
Originally Posted by Bear5k /forum/post/0


...For the vast majority of displays out there, though, you have almost no control over gamma, and it is merely a reference value to help explain what you are seeing, much like the locations of the primary colors in your gamut.

...

That is not what I find.


Most displays I see allow some control over gamma it just requires understanding how to do it. Typical parameters that can change this are picture modes, contrast level, gamma, gamma tables and black level options. Sometimes they are located in the service mode and other times they will be in the user mode. A few displays allow custom gamma curves in the service mode. I have used these at times to fix gamma problems.

 

[krasmuzik]

gamma is also adjustable in DVD players


I have an old JVC HTiB DVD player that has a "theater distance" button right under the more oft used "DVD menu". I have not measured its response curves (1|2|3

|off) but it is futzing with the gamma.


If you use HTPC those DVD players often have gamma adjustments as well.