Gamma 2.2 - Basic reference for most movies

Gamma 2.4 - very useful for movies with proper coloring and smooth transition to black (AKA Batman BD - with Adam West) Horrible for movies with already over-saturated colors (The Mummy - BD)

On any film that has more natural coloring and smooth transitions to black within the image, Gamma 2.4 is better for 3D effect and POP then 2.2, but... in the example above AKA The Mummy, Gamma 2.4 pushed the movie into an almost unwatchable state IMO...

I for one, love older movies for this reason, newer movies tend to be to harsh with the highlights vs black on screen and usually employ too much black crush. Older movies with little to no post processing seem much better to me. (better Coloring and highlights vs dark objects on screen)

So... Conclusion?

I use Gamma 2.2 for most of my viewing.

I use Gamma 2.4 when the software was mastered properly (AKA little to no post processing - Most new movies are not in this catagory range)

I've setup my projector using 2.22 (using a colorometer and HCFR software) and even with that my shadow detail is just what I would call acceptable. Any higher gamma, and I think I would loose way too much shadow detail. It did help greatly with making color rich and giving that punchy image though.

Just FYI, on my projector set at 2.2 gamma, it was actually averaging a gamma of about 1.9 and when I set it to gamma of 2.4 it was actually averaging a gamma of 2.0. I had to go into the custom gamma settings to get a true 2.22 gamma.

Have you measured your display, or are you just going by the settings that it shows in the user menu?

Dan

Dan,

What projector are you using? It makes a lot of difference when using higher gammas.

Exactly. Many projectors don't have a high enough on/off CR to use a gamma of 2.4 without black crush. Once you get to the 30,000 to one neighborhood, you can use it without crush. BUT and a BIG BUT. Gamma is a preference and only a preference. It depends on the viewing environment too. There is no correct answer.

Gamma 2.2 - Basic reference for most movies

I use Gamma 2.4 when the software was mastered properly.

This seems logically discontinuous.

Seeing as how reference gamma is more in the 2.4-2.5 neighborhood, it bothers me that it is still commonly stated that 2.2 is the standard gamma when it most certainly is not. That assertion comes from a mis-reading of the standards and wrongly assuming that the inverse of the encode gamma is what the display gamma should be. Not to mention the 1/2.2 encode gamma is long since out of date anyway, replaced with 709.

Gamma is a preference and only a preference. It depends on the viewing environment too. There is no correct answer.

I think that gives way too much ground to the willy-nilly subjectivists of the world. There is a reference we should shoot for when calibrating. When we run into limitations like low contrast ratios, yes subjective compromises gome into play, but certainly turning your gamma curve into spaghetti, or running a gamma of 1 is hardly what anybody should be doing in a properly calibrated system. I mean, sure, there are some big-name people who don't understand gamma at all and think everything is linear, but idiocy is not an excuse for throwing gamma to the subjective winds or just saying you should set it however you feel like it. There are standards for a reason.

Why don'tyou query Mr. Rogers of WSR on the subject or Joe Kane. gamma is a user preference. That how the CRT camera gamma was originally set. Assuming a perfect viewing environment and sufficient head room, i.e., on/off CR, increasing gamma leads to a more contrasty picture. there is a limit above which the average viewer will say too much contrast, probably 2.5 or 2.6. The optimum average would be somewhat higher than 2.4. IT IS A USER PREFERENCE. Pure and simple. Do some research.

Have you measured your display, or are you just going by the settings that it shows in the user menu?

Dan

I measure all my display devices. Same method as you.

This seems logically discontinuous.

Seeing as how reference gamma is more in the 2.4-2.5 neighborhood, it bothers me that it is still commonly stated that 2.2 is the standard gamma when it most certainly is not. That assertion comes from a mis-reading of the standards and wrongly assuming that the inverse of the encode gamma is what the display gamma should be. Not to mention the 1/2.2 encode gamma is long since out of date anyway, replaced with 709.

I understand what you are saying but... Not many digital displays allow even 2.2 gamma.

And I think they should get rid of CRT reference displays all together at this point.

The new reference should be the Pioneer Kuro Elites calibrated.

But thats just my opinion.

Then they would be mastering with a 2.2 gamma film camera, 2.2 gamma mastering display, and 2.2 (hopefully) Consumer display.

And then there would be a standard.

Dan,

What projector are you using? It makes a lot of difference when using higher gammas.

Epson 1080UB.

I agree with you.

Dan

Does it really have to be one or the other? I have a Sony VW-60 and have found 2.2 too dark to make out any details in the shadows clearly using The Mummy as just one example. I needed at least 2.1 or lower to distinguish many details that were in the movie but hidden with a higher gamma... in fact a 2.0 gamma seemed to reveal all the 'hidden' details and going any lower just made them pop out more... however it made everything pasty at the other end in the brighter scenes. Eventually (after consulting with some AVS members on the calibration forum) I ended up with a 'compromise' gamma and set 2.04 gamma at the lower end (10IRE) and have it cross 2.22 at 40IRE and flatten to 2.3 at 60IRE on up.... which according to HCFR ended up giving me an average gamma of 2.22. In the end I feel as if I've got the best of both... a gamma that gives me enough shadow detail and pop down low and adequate contrast and 3-D pop at the mid to brighter end. If there's something terribly wrong with doing it that way I'm clueless as everything I watch looks great to me. Besides, what else am I to do? A 2.2 gamma simply leaves me with very little shadow detail to the point of actually missing much detail altogether yet anything lower causes the bright scenes to look really washed out... is there another way to get the best at both ends of the grayscale? Please enlighten me... thank you.

Seeing as how reference gamma is more in the 2.4-2.5 neighborhood, it bothers me that it is still commonly stated that 2.2 is the standard gamma when it most certainly is not. That assertion comes from a mis-reading of the standards and wrongly assuming that the inverse of the encode gamma is what the display gamma should be. Not to mention the 1/2.2 encode gamma is long since out of date anyway, replaced with 709.THX now requires 2.2 gamma at the display for THX certification from what I gathered at CEDIA. But then this is the same group where the guy teaching at the booth on Sunday told me that the 30k:1 for the JVC RS20 was just marketing and didn't mean anything and that a G90 would do very poorly for black performance. I wonder if these THX guys have gotten confused about gamma, are requiring 2.2 because of some displays that have trouble doing more reasonably because they have relatively low on/off CRs, or something else. I sent an email to Charles Poynton mentioning this THX requirement and asking him what he thinks, but I haven't heard back.

--Darin

What I find its that lots of techie people read the specs ( gamma , white ref for example) and quote them as absolutes without realising they "reference" points rather than limits.

I've never seen a video enviroment that delivered an end to end gamma of 1 ( well I have but it looked awful and got changed after the tech guys did a bit more research).

1.1-1.2 is commonly the desirable end aim. 1 is used as a tech check for robustness ( although I usually whang the gamma up and down way beyond this anyway).

My HD1 does about 2.3 ( it looks like it needs a touch up actually as its dimmed) . It looks much nicer if I tweak the gamma on my HTPC to 2.5 and readjust the brightness ( about one tick).

Print film is display 1.7 , material 0.6 in the specs. Its usually always/ideally 1.1-1.2 on the screen to compensate for dark surround though.

If you can hit 1.1-1.2 on just about any display regardless of environment most people prefer it. I would say in the interests on avoiding this spiraling into some sort of argument try it and see what you prefer. I prefer it higher 1.1-1.2 . The panny plasmas for interests sake out of the box seem to actually hit closer to 2.2 when you select 2.5 on the commercial screens. 2.2 being more like 2.0.

THX video certification. What a joke. You can also quarrel with their audio certification standards. Its an artificial creation but accepted by many and many theaters are built to meeting the standard.

But I would rather look to ISF for video standards and even those standards are based on too much politics. THX and ISF are money makers first and foremost for their owners. ISF is no charitable foundation. It is a money maker and ego satisfier for Joe Silver.

Just like some churches are money making machines for Tammy Fayes and their husbands.

What I find its that lots of techie people read the specs ( gamma , white ref for example) and quote them as absolutes without realising they "reference" points rather than limits.

I've never seen a video enviroment that delivered an end to end gamma of 1 ( well I have but it looked awful and got changed after the tech guys did a bit more research).

1.1-1.2 is commonly the desirable end aim. 1 is used as a tech check for robustness ( although I usually whang the gamma up and down way beyond this anyway).
[/QUOTE]


Indeed, and this is a major thing that many seem not to be aware of, among them ISF and now apparently THX as well. Many people are very much wrongly assuming that the decode gamma should be the exact inverse of the encode gamma when this has never been the case. Poynton explains this well, but apparently not enough people read his book or pay attention, or even just his short gamma FAQ. It is good to hear that at least some in the pro field realize the end-to-end flow is not one, but a little bit greater than one.

I've tried not to pass judgement on THX's new foray into calibration since there are a handful of really excellent calibrators from I think LionAV who have been helping craft their classes, but given THX's past ideas about video with their optimode stuff, I do have a hard time accepting that they have any competence at all on the video front. Hopefully it's an entirely different group of people this time around. I'm trying my best to stay neutral until I know more, but this doesn't bode well so far.

1.1-1.2 is commonly the desirable end aim. 1 is used as a tech check for robustness ( although I usually whang the gamma up and down way beyond this anyway).

My HD1 does about 2.3 ( it looks like it needs a touch up actually as its dimmed) . It looks much nicer if I tweak the gamma on my HTPC to 2.5 and readjust the brightness ( about one tick).Have you ever measured gamma on one of the CRT mastering monitors? I'm curious about what they generally do for gamma, but it sounds like there might be some disagreement out there about what they are doing. I think it would be interesting to see what they put out for 80%stim and 20%stim compared to white since one of the differences between a 2.2 gamma at the display and higher gamma numbers is that 2.2 gamma results in less CR in that mid-range between 80% and 20%. I'm wondering if a 2.2 gamma is crushing the CR there a little bit compared to the CRT mastering monitors or whether the 2.2 gamma comes close to what those monitors do.

--Darin

I think if THX sets a gamma standard, it will be valid only for the purpose of meeting their criteria. Thus, since some people will insist on meeting am otherwise invalid standard, the indudty and HT viewers will be hurt rather than helped.

Have you ever measured gamma on one of the CRT mastering monitors? I'm curious about what they generally do for gamma, but it sounds like there might be some disagreement out there about what they are doing. I think it would be interesting to see what they put out for 80%stim and 20%stim compared to white since one of the differences between a 2.2 gamma at the display and higher gamma numbers is that 2.2 gamma results in less CR in that mid-range between 80% and 20%. I'm wondering if a 2.2 gamma is crushing the CR there a little bit compared to the CRT mastering monitors or whether the 2.2 gamma comes close to what those monitors do.

--Darin

Yes I've measured, but that is somewhat besides the point.
I'm not interested in what native gamma of the screen is: "gamma" is a covenient simplification of the actual response curve of the display. Most displays I've used come in closer to 2.5 out of the box , most domestic CRTs I've owned and calibrated were more like 2.5 ( albeit a bit lumpy) put of the box.

All the displays in a mastering suite are profiled and then lutted to whatever standard is required. I usually switch between print film and rec.709 on monitors that are variously sRGB or some sort of sony type phosphor masquerading as sRGB.

When I'm working with video I display at both 2.2 and 2.5 and pull it to extremes to check robustness. On display to clients I use 2.5 as invariably they describe 2.2 as flat although the room is fairly dark.

Like I say give it a whirl and see what you prefer, no-one should be demanding that you stick to one or the other.

People on this forum get too hung up on strict numbers: good video should be robust enough to cope with a reasonable variance in gamma. No two masters will exhibit the same "gamma" from shot to shot anyway. I see some imagery with nearly all of the negative content transfered to video without clipping and then some with just a sliver depending on the required result: what the relative gamma in these circumstances is something as relevant as trying to pantone match a formula 1 car as it drives past at 180.

Remember gamma itself is just a coarse simplification of the transfer curve: is really so people don't have to write masssive equations to describe the curves.

People routinely speak of film having a 0.6 gamma : its not of course , it has a chemically and subjectively designed curve a bit of which is somewhat close to 0.6.

One thing that is commonly accepted is you want an end to end gamma of 1.1-1.2 as most people seem to prefer that ( which is all standards really are anyway).

People routinely speak of film having a 0.6 gamma : its not of course..During (digital video) capture and or the post production color correction process, "video" tonal pre-emphasis often resembles a lazy "S" film like curve.

Display gamma is a mathematical formula which describes the relationship between the voltage input and the brightness of the image on your monitor screen and can also be described as the measurement of contrast that affects the midtones of an image.

Film gamma is defined as the average slope of the response curve in its linear (aka midtone) region.


http://www.colormatters.com/comput_gamma.html

Why do you continue to post defintions of terms to people who are using them in proper context, as if they asked what the term meant?

Is cutting and pasting your way to higher post counts some sort of personal goal?

The subject regarding an "ideal" gamma had been mentioned above. The first point made in my post is (which are my own words) factual in that the "encoded" gamma or tonal pre-emphasis if you will is NOT usually a simple (Rec. 709) power function. So to suggest a given display gamma (2.2) should approximate the reciprocal of the encode is not really reality for most DVD motion picture releases.

I understand what you are saying but... Not many digital displays allow even 2.2 gamma.

And I think they should get rid of CRT reference displays all together at this point.

The new reference should be the Pioneer Kuro Elites calibrated.

But thats just my opinion.

Then they would be mastering with a 2.2 gamma film camera, 2.2 gamma mastering display, and 2.2 (hopefully) Consumer display.

And then there would be a standard.

Well you soon will get your wish. The last of the Sony CRT stock is gone. We have a few BVM24 and BVM32 sitting in the warehouse doomed due to lack of replacement CRTs.

Yes the Kuro is seriously being tested by many facilities. We went with the Panasonics in editing areas because the first Kuro's were "sticky". The latest version has yet to be tested.

All in all our feeling is that Plasma is the most promissing CRT replacement for now. LCDs just can't cut it in the blacks plus the off axis viewing is intolerable.

The new Sony broadcast grade 32in LCD is a failure. They scrapped developement of the 42in version after poor interest at NAB 2008. Will Pioneer or Panasonic be the next reference grade monitor?

Remember a plasma is a lit phosphers machine. The fundamental difference is that the phosphers are deposted on glass just like a CRT but are ignited by a gas plasma instead of an electron beam. That's why a plasma can have blacks almost as good and sooner or later as good as a CRT machine.

Yes I've measured, but that is somewhat besides the point.
I'm not interested in what native gamma of the screen is: "gamma" is a covenient simplification of the actual response curve of the display. Most displays I've used come in closer to 2.5 out of the box , most domestic CRTs I've owned and calibrated were more like 2.5 ( albeit a bit lumpy) put of the box.

All the displays in a mastering suite are profiled and then lutted to whatever standard is required. I usually switch between print film and rec.709 on monitors that are variously sRGB or some sort of sony type phosphor masquerading as sRGB.My question was meant for what they do as setup for mastering things like Blu-ray, more than what they are native.
Like I say give it a whirl and see what you prefer, no-one should be demanding that you stick to one or the other.I agree, but am concerned about things like THX requiring manufacturers to make the gamma 2.2 at the display in order to get their certification. If the user can then change it then great, but as an example, I'm not sure if a person will be able to pick the THX mode in the JVC RS20 and then tweak the gamma afterward while keeping other things pretty constant or whether they will have to forego the THX mode altogether if they don't want a 2.2 gamma at the display.
People on this forum get too hung up on strict numbers: good video should be robust enough to cope with a reasonable variance in gamma. No two masters will exhibit the same "gamma" from shot to shot anyway.But if we set up our displays like the mastering monitors are setup while mastering that material, then we should at least get closer to what the person doing the mastering saw. Much like with color. And if people want to deviate from some gamma or some color standard then they can, but it would at least be nice to know what the standard and mastering monitors are doing (for both gamma and color).
Remember gamma itself is just a coarse simplification of the transfer curve: is really so people don't have to write masssive equations to describe the curves.Of course. And that is one reason I said I would like to find out what the contrast ratio between 80%stim and 20%stim normally is on the mastering monitors.
One thing that is commonly accepted is you want an end to end gamma of 1.1-1.2 as most people seem to prefer that ( which is all standards really are anyway).Yes and there is no good way for me to achieve that if I don't have any idea how the source was mastered or what the mastering monitor was doing. As I pretty much mentioned, in order to get higher than 1.0 gamma overall, the inverse gamma for the source needs to have a lower gamma number than the display gamma. Like 2.5 gamma at the display meaning 1/2.2 for the source to get a little over 1.1. But that leaves open the possibility for somebody to get confused and think that the display gamma was supposed to be 2.2 because that is the source gamma they hear (although inverted) and they figure 2.2 at the display will invert the source gamma and result in about 1.0, so it must be correct. I wonder if THX has made that mistake.

--Darin

Yes the Kuro is seriously being tested by many facilities. We went with the Panasonics in editing areas because the first Kuro's were "sticky". The latest version has yet to be tested.

All in all our feeling is that Plasma is the most promissing CRT replacement for now. LCDs just can't cut it in the blacks plus the off axis viewing is intolerable.

The new Sony broadcast grade 32in LCD is a failure. They scrapped developement of the 42in version after poor interest at NAB 2008. Will Pioneer or Panasonic be the next reference grade monitor?Do you know what kinds of display gammas these have? Basically, how they would be setup for mastering as far as gamma goes?

Thanks,
Darin

Do you know what kinds of display gammas these have? Basically, how they would be setup for mastering as far as gamma goes?
A Kuro will be a flat 2.2 gamma, though the ISF controls may allow you to adjust it higher, but they are rather coarse, so I don't know how successful you would be at getting a flat 2.4 response, for example.

While they may have very impressive black levels for a non-CRT display, there are a number of issues with Plasma technology that make them unsuitable for mastering content, so I certainly hope this does not become the case.

I don't see how you could have any ‘reference’ display that adds noise to the image, for one thing.

Just use a good VP such as the Sencore by Lumagen with the Kuro and create whatever gamma curve you want.

Joe Kane use gamma 2.5 because it is the reference.
MLill on HCFR use 2.45.

From my point of vue, 2.5 is ideal gamma if the post-production is correct. If not, it can help to use 2.2 or 2.3 gamma.

It also depends on the darkness in the room.

Film gamma is defined as the average slope of the response curve in its linear (aka midtone) region.




Which isn't actually linear at all: which is my point, gamma is an approximation of more complex curves.

Do you know what kinds of display gammas these have? Basically, how they would be setup for mastering as far as gamma goes?

Thanks,
Darin

What you are missing is the concept of "robustness" at the mastering stage. Its not about hitting some idealised target exactly with no room for error. Good mastering always allows a margin for error to still give pleasing results on the display.

Exactly the same as engineering tolerances.

Which isn't actually linear at all: which is my point, gamma is an approximation of more complex curves.Your point would be misleading Mr. D, the definition of gamma is an "approximation" for good reason..

http://www.asahi-net.or.jp/~rt6k-okn/ddp/digital.htm
consider the "gamma value", which is the gradient of the linear region on the gamma curve..

*When b < xij < a, the gamma is linear.

*When {xij} = xij, i.e. no low pass filter was used, this is just a hyperbolic conversion of gamma curve. The relation between xij and yij is shown in fig-3. You can see that the curve is very similar to the gamma curve of conventional film.

In fig-3, the curve of the log-conversion, which is available in most of IP software, is also shown. Note that there is no linear region in this curve. I believe that the linear region of gamma curve should be important to achieve the 'photo quality'.
• Straight-line portion. Sooner or later, the relationship between density and (log) exposure becomes more or less linear. This region is known as the straight-line portion of the curve, though it is seldom absolutely straight. The slope of this curve is known as ‘gamma’ and is an indication of the contrast of the material: a steep slope (higher gamma) means more contrast, a gentle slope, less.

This region is known as the straight-line portion of the curve, though it is seldom absolutely straight.


Exactly what I said thanks.

No problem..just didn't what anyone to change the definition of the term;)

What you are missing is the concept of "robustness" at the mastering stage. Its not about hitting some idealised target exactly with no room for error. Good mastering always allows a margin for error to still give pleasing results on the display.

Exactly the same as engineering tolerances.I doubt I'm missing anything with respect to that at all. I understand that there is margin for error, but there is still some typical point or small range. Just like with colors. My memory is that you were arguing that REC.709 primaries were the correct way to play things like Blu-ray back, and I think there is a good chance that there is more visible difference between 2.2 and 2.5 gamma than SMPTE-C and REC.709 primaries, yet here you seem to be arguing against us finding out what the mastering monitors typically are setup for as far as gamma. Why didn't you take the same tact along the lines of, "People on this forum get too hung up on strict numbers ..." with respect to color primaries? Or am I missing something here about what you did argue with respect to the color primaries?

Do you feel the same way about the white point? That is, people shouldn't want to know whether the mastering monitors are setup to D65 or something else because there is a margin of error and a robustness factor?

--Darin

I doubt I'm missing anything with respect to that at all. I understand that there is margin for error, but there is still some typical point or small range. Just like with colors. My memory is that you were arguing that REC.709 primaries were the correct way to play things like Blu-ray back, and I think there is a good chance that there is more visible difference between 2.2 and 2.5 gamma than SMPTE-C and REC.709 primaries, yet here you seem to be arguing against us finding out what the mastering monitors typically are setup for as far as gamma. Why didn't you take the same tact along the lines of, "People on this forum get too hung up on strict numbers ..." with respect to color primaries? Or am I missing something here about what you did argue with respect to the color primaries?

Do you feel the same way about the white point? That is, people shouldn't want to know whether the mastering monitors are setup to D65 or something else because there is a margin of error and a robustness factor?

--Darin


Nope you're not quite understanding me. Rec.709 is the standard but no-one who creates images would make them so flimsy as to fall apart if the end to end gamma was .1 off.

If you mean white point on the display some people will restrict it to a specific value whether they are dealing with video or print, however with the bit starved displays a lot of people are using these are often just set as high as they will go without clipping to maximise the contrast range: if you mean white point on the material , I certainly check my video with a 16-235 clip in place to make sure it looks okay as well as whanging the gamma away from the norm ( when I work I tend to use 2.2 as it discloses nearer black easier ( although I will whang the gamma anyway to check the blacks beyond this).

It depends what you are trying to do . I would assume a grading suite has a locked white point on their displays.

Like I said earlier in the hopes of this not turning into an argumentative thread: try 2.2 then see if you prefer 2.5.

Nope you're not quite understanding me. Rec.709 is the standard but no-one who creates images would make them so flimsy as to fall apart if the end to end gamma was .1 off.Would the images fall apart if the color was off by the same percentage?
If you mean white point on the display some people will restrict it to a specific value whether they are dealing with video or print, however with the bit starved displays a lot of people are using these are often just set as high as they will go without clipping to maximise the contrast range:I mean the color balance of the white point of the display at the point the person thinks the colors look correct when shown on that display. Might have been clearer to say the the grayscale.

Should people worry about setting their color balance to D65, or should they figure that people are too hung up on numbers? With color people can try different values and may decide they like something other than D65 more. The standards are basically there so that we know how the mastering monitors are setup. I'm basically just asking the same thing, but with respect to gamma. Standards for the source for gamma (like the inverse of 2.2) should have been made knowing what the mastering monitors would have typically displayed.

--Darin

Nope you're not quite understanding me. Rec.709 is the standard but no-one who creates images would make them so flimsy as to fall apart if the end to end gamma was .1 off.

I'm not really understanding you either. The image won't fall apart if you use SMPTE C either. The difference between 2.2 and 2.5 is a fairly significant difference. I would say it's much more significant than the visual delta between SMPTE C and 709 primaries. No, the image won't look terrible either way, but there certainly is a difference, and there has been some confusion as to what the standard is, because the standards are not explicit on this. And because some bodies apparently seem to be telling people that 2.2 is the standard, it begs the question if content has also migrated away from 2.5, and whether this is done with intent or out of ignorance. Either way, it has an impact on my choice as an end-user in how to approach reference performance because I need to know how to match the original And you saying "well, gee, it's not really that big a deal pick whatever you want" isn't really an effective standard. That argument could go for white point "meh, don't measure, just do it by eye so it looks kind of whitish" or for gamut "things are confused, just pick what looks good to you" or even color decoding for that matter "meh, the difference between 601 and 709 isn't that big, you don't need to worry if your system gets it wrong." I mean, why not just leave the display in Vivid then?

Ideally you want and end to end gamma of 1.1. Most presentation displays I've seen in professional enviroments (film or video) aim for 1.1-1.2 ( its an approximation based on a curve that isn't really a gamma anyway remember).

2.5 x 0.44 gets you 1.1

Bear in mind the dark surround effect also plays a role: you may prefer to back off the gamma in a bright enviroment to make the lower tones more easily discernible you don't normally find bright enviroments in professional viewing situations. However in my experience most people's viewing conditions tend to fall into the dark category.

Not really sure what more you want guys , no point talking about white ref and chromaticity just beacuse you don't like the answer.

Ideally you want and end to end gamma of 1.1. Most presentation displays I've seen in professional enviroments (film or video) aim for 1.1-1.2 ( its an approximation based on a curve that isn't really a gamma anyway remember).

2.5 x 0.44 gets you 1.1If that 2.5 is what the mastering monitors usually are then great. I don't have a way to measure the end-to-end gamma since I just get the source as it is. So, if 2.5 gives 1.1 then 2.2 at the display with that same 0.44 source gives less than 1.0. Pretty much defeats the purpose of the whole wanting greater than 1.0 in the first place, as I believe I said multiple posts ago. So, if what we get is from mastering to 0.44 because the mastering monitor is typically 2.5, then THX's approach would result in an end-to-end gamma of less than the mastering monitor was doing typically and get rid of the 1.1. Ideally I would want to know the whole curve for a typical mastering monitor and not just a number, which is one reason I said I would like to find out what the ratio between 80%stim and 20%stim is typically.
Not really sure what more you want guys , no point talking about white ref and chromaticity just beacuse you don't like the answer.The point there is that it is rather strange for somebody who has said that people should set their primaries to REC.709 and not SMPTE-C for HD to then scold others by claiming they are too wrapped up in numbers (which is all the primary and white point standards are) and claim I don't get it because there is margin of error. The standards are a means to an end, not the end themselves. Just like we should want to know how the mastering monitors are setup for white point and primaries, we should want to know how they are setup for gamma. If you were consistent and just told people to try whatever color primaries or white points they feel like and ignore what the mastering monitors are doing typically, then it would have been a different story when you took basically that tact with gamma.

--Darin

My interest in this is more empirical than theoretical. I have tried a gamma of 2.4 on a couple of displays (of different types) and in both cases the image did NOT appear correct to my eyes. The image appeared darkened and unnaturally contrasty (if that's a word). I have experimented quite a bit and I find 2.25 looks about right, achieving the the best blend of shadow detail and image depth.

I have also measured several Sony CRTs and after optimizing black level, white level, and gray scale, they seem to offer a gamma very close to 2.2.

I just read this paper (http://www.w3.org/Graphics/Color/sRGB)(or at least the section on gamma)--which is nominally about CRTs used for computer monitors, but the points seem valid for CRTs in general--and it states quite explicitly:

"The ITU-R BT.709 transfer function in combination with its target monitor is attempting to achieve a viewing gamma of 1.125 by incorrectly assuming a CRT gamma of 2.5 and an LUT gamma of 1.0/2.222 [emphasis mine]"

"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."

"Exhaustive testing carried on at Hewlett-Packard on VGA computer monitors from many brands has shown the average CRT gamma to be indeed 2.2, with a standard deviation of about 0.2"

"There has been significant confusion derived from assuming the CRT gamma value is identical to the exponent in equation 0.4. This has led to many claims of CRT gamma values of 2.5 for video, 1.8 for the Apple Macintosh and 1.4 for SGI monitors. Unfortunately, it has been our experience that this misconception is not well founded in the actual physics of the displays and solid measurements." [emphasis mine]

I understand that Chris has been on something of a crusade recently in an attempt to disabuse everyone of the belief that 2.2 gamma is correct and that 2.5 is the correct reference gamma. This is fine. I have attempted a few crusades myself from time to time. Frankly, I don't have an opinion about the theory behind this, but this claim just doesn't seem consistent with my experience.

Another aspect of this debate that does not speak to real-world concerns is that many displays come out of the box with a gamma that is in the 1.2-1.8 range. Manufacturers do this I suppose to increase the subjective brightness of the display. Just going from 1.5 to 2.2 offers truly massive improvements to image quality compared to the relatively subtle difference between 2.25 and 2.4.

So, some questions: If correct CRT gamma is 2.5, why don't calibrated CRTs measure that way? Why does this HP/Microsoft paper repeatedly refer to the mistaken belief about 2.5 gamma? Why does a 2.5 gamma appear to darken the image? I am persuadable on this, but right now I don't see it (literally).

Here's what I measured a direct view CRT (not BVM) at
http://members.aceweb.com/dlarsen/PICS/log-log.jpg

If I had to assign 1 # to this, it wouldn't be 2.5
Dave

So, some questions: If correct CRT gamma is 2.5, why don't calibrated CRTs measure that way? Why does this HP/Microsoft paper repeatedly refer to the mistaken belief about 2.5 gamma?From what I have heard behind the scenes there is some disagreement about what the CRTs actually measure. I can't remember all the details, but I believe Charles Poynton has said 2.5 while somebody else has said 2.2 and I was told that one of them was asked why there was a disagreement. I don't remember which one of them was asked, but heard that their initial answer was that the other person didn't know what they were talking about (but said it with a laugh). I think there was then something about depending on how the monitor was setup and there is the factor of fitting a curve to a number. I would personally like to see what each gets for points along the curve to look at things like the ratio between 80%stim and 20%stim, since it could be that they are typically one gamma number in part of the curve and a different gamma number at a different part of the curve. I haven't personally had the opportunity to measure a CRT mastering monitor (either calibrated for mastering or not) and is one reason I'm curious about what others have measured.

I sent an email to Charles Poynton about gamma and asked what he thought of THX going with 2.2, but haven't gotten a response.

--Darin

If I had to assign 1 # to this, it wouldn't be 2.5
Thanks Dave. I don't think anybody would assign 2.5 to that even if they had to pick one number.

--Darin

To throw even more mud into this debate, a widely-quoted BBC study claims that reference CRT gamma is 2.35. Although a little high IMHO, this seems more reasonable to me than 2.5.

http://www.ebu.ch/CMSimages/en/tec_doc_t3320-2008_tcm6-50985.pdf

"Using the actual power function fit value for the 709 transfer function of 1.0/1.956 ..."
Thanks for posting this quote.

1.0/1.956 = 0.511

Also, this graph shows that Rec709 encoding is closer to 0.50 than 0.45.
gamma.gif (http://archive2.avsforum.com/avs-vb/attachment.php?attachmentid=25025)

For Rec.709 and the encode side, The bottom ~1.8% of the EO curve (where the black details are) is not a non-unity power function. It's linear. Good luck in making anything close measurement wise down there. The power function is 0.45 but offset on a linear segment.

Dave

And what that means, is good results on a wide variety of displays under less than ideal viewing conditions. That is why tweaking the display gamma can lead to a more pleasing picture in your own environment on your display. The good masterers provided the flexibility for you to do this.

For Rec.709 and the encode side, The bottom ~1.8% of the EO curve (where the black details are) is not a power function. It's linear.
You could say that it is a power function with an exponent of 1, which is linear.

You could say that it is a power function with an exponent of 1, which is linear.

Yes, you could say that. I'd say it's linear. I stand corrected and edited my post;)

Here's what I measured a direct view CRT (not BVM) at
http://members.aceweb.com/dlarsen/PICS/log-log.jpg

If I had to assign 1 # to this, it wouldn't be 2.5
Dave

Based on the measured Direct CRT range above, I would say an average of 2.23 is about right.

Based on the measured Direct CRT range above, I would say an average of 2.23 is about right.As I said above gamma is defined as an "average slope" and IS most accurate in the midtone (~ 15-75 IRE) region. The end-end objective is to linearize your dispaly response there. i.e. resulting in a transparent correlation that happens to corresponds to the bulk of the image content.

Mr. D makes a good point regarding real world tolerances..black (non CGI) content is not always adjusted for (16, 128, 128) YCbCr.. and white is not usually (235, 128, 128) either. Real world images nominally have a very low APL. To this day content is still tweaked by artist/engineers while viewing industry standard (Sony & Ikegami BVM) that are designed with a perfect 2.20 response. A well designed consumer CRT should have the same transfer response.

As Dave mentioned in the past to me, gamma IS "subjective" and interacts with viewing ambient as well.

thomas

"Exhaustive testing carried on at Hewlett-Packard on VGA computer monitors from many brands has shown the average CRT gamma to be indeed 2.2, with a standard deviation of about 0.2"Sony was also involved with these same test. Sony' results validated the HP findings of ~2.2 as well.

..and pro (BVM) $20-30K direct view reference CRTs are "designed" for a ruler flat 2.20 response all the way down to 1%. See link below:
http://www.displaymate.com/ShootOut_Part_2.htm
The Gamma for the Sony CRT agrees perfectly with the 2.20 standard value. (CRT monitors from Ikegami, another major brand of professional studio monitors, also have a Gamma of 2.20 according to their Director of Engineering.)

Sorry, if this is off-topic, but in this context, why is DCI specified to Gamma 2.6 ? Is there a different mastering process/some reason ?

Sorry, if this is off-topic, but in this context, why is DCI specified to Gamma 2.6 ? Is there a different mastering process/some reason ?
DCI does not use Rec709 encoding. DCI encodes with 1/2.6 gamma, and decodes with 2.6 gamma.

Yes, yes, but what is it mastered/controlled on ? CRT with Gamma either 2.2 or 2.5, what this disussion has circled around or with another 2.6 device ? And why did they chose 2.6, not 2.2 ?

Also one more clarification question: if we say for a REC709 playback Gamma around 2.4 to 2.6 one needs a display with high on/off, does that mean becauses of sources with 2.2 displayed at higher Gamma in order to avoid black crush ?

Because DCI is speced with rather low on/off, if read the paper correctly, how does that fit in on Gamma of 2.6 and the on/off demand requested, shouldn't DCI be more like 2.0 with such low on/off ? Or has that to do with more than just on/off, like the 14Bit source ?

Just trying to better understand the complete picture.

Yes, yes, but what is it mastered/controlled on ? CRT with Gamma either 2.2 or 2.5, what this disussion has circled around or with another 2.6 device ?
DCI projectors use a 2.6 gamma.
And why did they chose 2.6, not 2.2 ?
I think they chose 1/2.6 & 2.6 because it has less noticeable contouring artifacts than 1/2.2 & 2.2.
Also one more clarification question: if we say for a REC709 playback Gamma around 2.4 to 2.6 one needs a display with high on/off, does that mean becauses of sources with 2.2 displayed at higher Gamma in order to avoid black crush ?
Rec709 content is encoded with 1/2.0 (approximately), so if it is decoded with 2.6 (for example), the end-to-end gamma is 1.3, which causes levels near black to be much darker than their original luminances, and thus they are closer to the display's black level (which is presumably not absolute black). In other words, this process reduces contrast in the shadows. A higher-on/off display may allow more shadow contrast to remain than a lower-on/off display allows.
Because DCI is speced with rather low on/off, if read the paper correctly, how does that fit in on Gamma of 2.6 and the on/off demand requested, shouldn't DCI be more like 2.0 with such low on/off ? Or has that to do with more than just on/off, like the 14Bit source ?
DCI content is encoded with 1/2.6, and when it is decoded with 2.6, the end-to-end gamma is 1.0, so this process allows levels near black to stay the same as their original luminances. In other words, this process by itself does not reduce the contrast in the shadows. However, before the content was encoded with 1/2.6, there may have been other processes that reduced contrast in the shadows. In addition, the display's on/off may limit the amount of shadow contrast.

Thx Erik, but to answer my first question more directly, DCI is not mastered/controlled on any CRT or so in between with a different Gamma, as this disucussion here on studio CRTs measured between Gamma 2.2 and 2.5 has circled around. It's just encoding and decoding, always at 2.6. As there is no Gamma change in between, with such a low on/off spec it is absolutely ok to use Gamma of 2.6 as the outcome is 1.

With contouring do you mean banding effects of a 14Bit source or any effects of the special DCI DLP chips ?

Thx Erik, but to answer my first question more directly, DCI is not mastered/controlled on any CRT or so in between with a different Gamma, as this disucussion here on studio CRTs measured between Gamma 2.2 and 2.5 has circled around. It's just encoding and decoding, always at 2.6. As there is no Gamma change in between, with such a low on/off spec it is absolutely ok to use Gamma of 2.6 as the outcome is 1.
That's correct.
With contouring do you mean banding effects of a 14Bit source or any effects of the special DCI DLP chips ?
Yes, contouring, or banding, is affected by various factors, such as the bit-depth of the source, the bit-depth of the display, and the gamma.

Ok, but then it probably more due to the display, if contouring is the reason for Gamma of 2.6, as I really cannot believe in banding of a professional 14Bit master, derived from the original scan running at such high bitrates....

Ok, but then it probably more due to the display, if contouring is the reason for Gamma of 2.6, as I really cannot believe in banding of a professional 14Bit master, derived from the original scan running at such high bitrates....
DCI content is 12-bit. This web page explains why 12-bit was chosen.

http://www.cinematography.net/edited-pages/ColorBitDepth.htm

* Gamma 2.6 best follows the threshold of human perceptual discrimination for luminance at cinema levels
* All viewers could see 8 bits all the time
* Almost all observers could see steps at 10 bits
* Some observers (40%) could see steps at 11 bits
* Essentially no observers (1%) could see steps at 12 bits.
* Observers learned to see steps and became more sensitive with repeated viewing

Quote:
* Gamma 2.6 best follows the threshold of human perceptual discrimination for luminance at cinema levels
* All viewers could see 8 bits all the time
* Almost all observers could see steps at 10 bits
* Some observers (40%) could see steps at 11 bits
* Essentially no observers (1%) could see steps at 12 bits.
* Observers learned to see steps and became more sensitive with repeated viewing

And sadly we are stuck with 8 bit HD. J6P can see the banding/steps without special training:(
Who was the knucklehead who decided 8bit was good enough for the consumer, thanks for making decisions on our behalf.

Yes, 12 Bit, just reconsulted the DCI paper V1.2, Sorry, must have picked up the 14Bit while reading about DCI projectors over in the DCI section of the high end forum.

Edit: and indeed interesting is the remark => "Gamma 2.6 best follows the threshold of human perceptual discrimination for luminance at cinema levels" which obviously answers my question.

Yes, 12 Bit, just reconsulted the DCI paper V1.2, Sorry, must have picked up the 14Bit while reading about DCI projectors over in the DCI section of the high end forum.


The DMDs are driven at 15 bits because of potential banding from PWM.

And sadly we are stuck with 8 bit HD. J6P can see the banding/steps without special training:(
Who was the knucklehead who decided 8bit was good enough for the consumer, thanks for making decisions on our behalf.

What saves the consumer 8 bit format is noise. Noise masks contouring very effectively. A very low noise or noise free 10 bit format would be much better overall. The same argument holds for 4:2:2 or 4:4:4 compared to 4:2:0.

And sadly we are stuck with 8 bit HD. J6P can see the banding/steps without special training:(
Who was the knucklehead who decided 8bit was good enough for the consumer, thanks for making decisions on our behalf.
Yep. I also wonder who decided to stay with the restricted color space and 4:2:0 encoding. And while we're at it, why isn't 1080p50 and 1080p60 supported, either? Ok, ok, maybe I'm going over the top, but a new format isn't born every day (or year). So now that everyone is trying to replace DVD with Blu-Ray, why didn't they spec Blu-Ray to be future proof from the get go? :(

I guess one of the problems was that Blu-Ray was originally developed as a recording technology for MPEG2 HDTV broadcasts. Maybe that's where all the limitations are coming from. But HD DVD could have raised the bar - and Blu-Ray would have been forced to follow suit (as they did with adding all the new video and audio codecs to the spec). Well, too late now...

I think they chose 1/2.6 & 2.6 because it has less noticeable contouring artifacts than 1/2.2 & 2.2.I was glad to see they used a higher number like 2.6. Seems that with 2.6 and their 12 bit log or whatever it is that they have enough levels in the bright images and the 2.6 should help with banding type issues down low as we get higher on/off CRs. I don't think we'll be stuck at the relatively low on/off CRs of DCI forever, especially if DCI content moves into the home market. They do have levels in their equations that come out to high CRs, like (50, 50, 50) which looks like it is a level that comes out to about 95k:1 CR from white. As it is most of those just clip at the higher black level of DCI projectors now, but they could go further down in the future. Might be a while though.

--Darin

"Exhaustive testing carried on at Hewlett-Packard on VGA computer monitors from many brands has shown the average CRT gamma to be indeed 2.2, with a standard deviation of about 0.2".Here is the (Sony/HP) study you are referring to Tom:

Microsoft and Hewlett-Packard initiated a collaborative effort several years ago to develop and standardize a common RGB color space
http://download.microsoft.com/download/1/6/1/161ba512-40e2-4cc9-843a-923143f3456c/MSColorCaseStudy.doc
• Sony provided studies that proved optimally calibrated CRTs in their native state are 2.2 gamma and have HDTV primaries. [Katoh and Deguchi, "Reconsideration of CRT Monitor Characteristics"]

I understand that Chris has been on something of a crusade recently in an attempt to disabuse everyone of the belief that 2.2 gamma is correct and that 2.5 is the correct reference gamma. This is fine. I have attempted a few crusades myself from time to time. Frankly, I don't have an opinion about the theory behind this, but this claim just doesn't seem consistent with my experience.


It's really not a crusade, but there are many people who have taken up the assumption that a display gamma of 2.2 is the standard, thus is defined as such somewhere. I've had a number of conversations with ISF calibrators and others who are certain that 2.2 is written down somewhere as THE standard to which displays should be calibrated. This is simply not true, and where possible I've attempted to shake them of this certainty.

This doesn't mean that 2.2 isn't a perfectly good gamma to use, subjectively, in many systems.

But people shouldn't be going around stating that 2.2 is defined in standards when it isn't. My understanding has been that a higher gamma(than 2.2) is normal for CRTs, and that's my own experience. This isn't always true, and it depends also on how the CRT is set up.

I mean, to be fair I could say that there isn't really a standard gamma defined in the standards, because really it's implicitly assumed to be a CRT's gamma. I take Poynton's side on the gamma issue, and that demands an end-to-end gamma that is slightly greater than one. This means that we need a display gamma that is higher than 2.2, assuming an encode gamma of 1/2.2. Now, this also ignores the fact that the 1/2.2 encode number is obsolete, replaced with Rec 709 gamma, but the curve remains effectively similar.

My problem is that people see 1/2.2 defined as the ENCODE gamma, assume a 1.0 gamma end-to-end (which is wrong) and then say that the DISPLAY gamma should therefore be exactly 2.2. But again, this doesn't mean that a display gamma of 2.2 is at all bad, but the understanding needs to be established first before we have people teaching willy-nilly that 2.2 is defined as standard display gamma when that simply isn't true at all. Maybe it will become true, and it may even become de-facto established as such, but as of right now it's not true and it shouldn't be taught that way.

Chris-

What does your CRT measure? Can you post up your measurement data? (Both native and how you may have changed it with your setup)

Note that the data that Tom referenced (from HP/Microsoft) and the graph that Erik provided, shows that if fit to a single power #, the Rec.709 encode curve fits closer to 0.51 rather than the 0.45 that it would be if it didn’t have the linear tail. This would yield 1.96 unity degamma for Rec.709. If 2.2 was the display degamma, this would yield 1.12 for end-to-end for Rec.709 with a linear tail on the encode side- Not 1.0 as you assert. The curve doesn't "remain effectively similar". If one has to assign single #s.

Note also the section where Poynton discusses the 1.1-1.2 end-to-end range and 1/2.2-1/2.5 in his Gamma FAQ. The heading is titled NTSC, not ATSC. 2.5 degamma with a Rec.709 encode curve fit of 0.51 would yield end-to-end of 1.28. Outside of the recommended 1.1-1.2 range cited in the Gamma FAQ for dim environment. In Ch 6 of “A TECHNICAL INTRODUCTION TO DIGITAL VIDEO” he states: “If you view your monitor in a dim surround, use a lower value of about 2.2.”

Seems to me that rigorously advocating and starting threads that claim 2.5 as the 'correct standard' for Rec.709 degamma may be worse than those who advise 2.2 if you give the 1.1-1.2 overall recommendation of Poynton for dim environment any heed.
http://www.avsforum.com/avs-vb/showthread.php?p=13393930#post13393930

This would yield 1.96 unity degamma for Rec.709. If 2.2 was the display degamma, this would yield 1.12 for end-to-end for Rec.709 with a linear tail on the encode side- Not 1.0 as you assert. The curve doesn't "remain effectively similar". If one has to assign single #s.


How did you get 1.96, .5*2.5= 1.25 .5*2.2=1.1

Again, I don't think 2.2 is at all an unreasonable gamma. I tweak my gamma down away from 2.5 to about 2.3 because of my own subjective preferences, due to ANSI CR limitations on an AC projection CRT display. On an LC unit, I'll leave gamma alone, or maybe just bring it down to about 2.4, almost no change at all, because of improved ANSI CR.

My only point is that it is misleading for people to assert that 2.2 is defined by standard when it isn't.

Note also the section where Poynton discusses the 1.1-1.2 end-to-end range and 1/2.2-1/2.5 in his Gamma FAQ. The heading is titled NTSC, not ATSC. 2.5 degamma with a Rec.709 encode curve fit of 0.51 would yield end-to-end of 1.28. Outside of the recommended 1.1-1.2 range cited in the Gamma FAQ for dim environment. In Ch 6 of “A TECHNICAL INTRODUCTION TO DIGITAL VIDEO” he states: “If you view your monitor in a dim surround, use a lower value of about 2.2.”

I don't have this earlier text of his, but in Algorithms and Interfaces he is quite clear in describing display gamma (Yd) as 2.5, and rec 709 encode gamma (Ye) as effectively ~.5, with a desired end-to-end gamma of 1.25. He includes also in the margin a small note describing the confusion about gamma and that many video engineers are not fully aware of the intent to have end-to-end gamma greater than 1. Now, we can agree or disagree about what that end-to-end gamma should be, and in what viewing environment, but Poynton, at least as far as this text goes, seems quite clear that end-to-end gamma is 1.25, display gamma is 2.5.

Regardless, however, I am more interested in what the gamma of the mastering bay display is, especially as we move away from CRT BVMs. If more and more content is being mastered at a lower gamma than 2.5, it makes sense to adjust to that de-facto change. And again, as I've said many times before, 2.2 gamma can be a perfectly fine choice. But just as I would object if people said that 7500K is greyscale standard simply because it looks a little more blue and bright that way, instead of D65, so too I only intend to bring up the fact that the standards do NOT describe display gamma as 2.2, and it's at best misleading or obfuscating to state that, and at worst flat wrong.

How did you get 1.96,

Math and the HP/MS info. 1/0.51 Again, see the data Tom (HP/MS) and Erik Posted. Rec.709 encode curve fits closer to 0.51 NOT 0.45. Because of the linear tail. See above in your own post where Poynton calls Rec.709 ~0.5 NOT 0.45. If you're expressing one rounded to hundredths, you should express the other that way too.

From the HP/MS info Tom Posted: "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."


0.51*2.2= 1.12. Inside his recommend range for dim environments (1.1-1.2)
0.51*2.5=1.28. Outside his recommend range for dim environments (1.1-1.2)

He didn't say 1.25 was desired. He said it was acceptable. If you round to tenths, 1.28 (or 1.25) would be expressed as 1.3. Desired overall seems to be 1.1-1.2 as he and others reference many times. Elsewhere, in 'PICTURE RENDERING, IMAGE STATE, AND REC. 709' (2007) he states: "The electro-optical conversion function (EOCF) of a reference video display has never been adequately standardized. The de facto EOCF is a power function having an exponent between 2.32 and 2.4."

What about where he says: “If you view your monitor in a dim surround, use a lower value of about 2.2.” ? He seems pretty clear there too. You say that is 'flat wrong' and have started threads that proclaim just that (http://www.avsforum.com/avs-vb/showthread.php?p=13393930#post13393930). Maybe ISF can read a Poynton book and even make some degamma measurements?

You’ve been saying for months how 0.45*2.2=1.0 and how ‘wrong’ that is and how ISF is wack because of it. 0.45 is not the correct curve fit for Rec.709 as the data Tom, Eric and Poynton has shown. Despite you saying the ‘curve remains effectively similar’, it’s been shown that it isn’t effectively 0.45. Seems 0.51*2.2 is just about where you think 0.45*2.5 should be.


Again, what does your CRT measure native? Can you post your measurement data that shows your native 2.5 degamma? You do have measurement data? (not just verbally expressed as a single # where we have to trust your word on it) ;)
Dave

This EBU document is very clear. It recommends a monitor gamma of 2.35 as I had mentioned earlier.

This EBU document is very clear. It recommends a monitor gamma of 2.35 as I had mentioned earlier.FWIW Tom:

In practice, the Gamma Curve (Transfer Function) for essentially all displays is so far off from the theoretical inverse of the encoding function that it is totally irrelevant. The perspective that I have taken in the entire Shoot-Out article series is that the goal is for the viewer to see an image that is identical to that seen in a studio by the creators of professionally produced content – generally the director and cinematographer or videographer. So my gold standard has been the Professional Sony Trinitron CRT monitors that are found in just about every production studio. Whenever I have measured the Transfer Function of these monitors it has always been a virtually perfect power-law (straight line on a log-log graph) as far down as I measured with accurate photometers, which is below 1 percent of the luminance for Reference White

Sony PVM-20L5 (CRT) measured gamma 2.20

thomas

EBU-TECH 3320 and 3321 might be of interest to people here:
http://tech.ebu.ch/docs/tech/tech3320.pdf
http://tech.ebu.ch/docs/tech/tech3321.pdf

EBU is fairly clear about the 2.35 Gamma recommendation for viewing in a dim environment.

I believe ARIB TR-B28 (http://www.arib.or.jp/english/html/overview/doc/4-TR-B28v1_1.pdf) specifies 2.2 however. http://www.fe-tech.co.jp/en/master/master_top.html

EBU-TECH 3320 and 3321 might be of interest to people here:
http://tech.ebu.ch/docs/tech/tech3320.pdfBtw this validates what I've already posted above:

"Whilst the camera may have a nominal opto-electrical transfer function according to ITU-R BT. Rec.709, this is in practice modified by the intention of the director in camera control or in grading."

FWIW Tom:
Sony PVM-20L5 (CRT) measured gamma 2.20In terms of my own measurements and viewing preferences I absolutely agree. I actually prefer it just a LITTLE higher at about 2.25.

However, it is interesting that EBU has clearly settled on 2.35--which assuming a camera gamma of 0.51--provides a system gamma of 1.199, which is JUST within the upper end of the recommended range of 1.1-1.2. 2.2 comes in at 1.122, also within that range. 2.25 gives 1.148, which is smack dab in the middle of the range.

However, 2.5 gives a system gamma of 1.275 which is considerably outside of that range.

Given these numbers, I would argue that--assuming a dim viewing environment--any display gamma between 2.16-2.35 is within acceptable tolerances and dependent on source material and user preference, but that lower or higher is out of spec.

Thanks for the reply Tom:)

I'm going to quote what I posted a few pages above which was validated by the EBU paper...the "encoded" gamma or tonal pre-emphasis if you will is NOT usually a simple (Rec. 709) power function. So to suggest a given display gamma (2.2) should approximate the reciprocal of the encode is not really reality for most DVD motion picture releases.My point is that I've heard it said many times here on this forum, that the ultimate display cal would emulate what the director and cinematographer saw in the post production process.

Also I think one can ignore a given particular production viewing ambient and still be confident that the production reference monitor cal 2.2 is spot on.. a fact confirmed by (eng. design) the director of engineering from Ikegami as well Dr. Soneira's Sony measurements. A Sony/Ikegami $25K plus CRT reference monitor has custom electronics that make an ideal (perceptual HVS) 2.20 response possible. (Remember the encoded content may have a logarithmic / exponential / film-like "lazy S" pre-emphasis.)

Btw Tom I don't agree with that EBU paper when it suggest a CRTs gamma is 2.35 or "immoveable"
It has been found from the measurement techniques, progressively refined over several decades, that a correctly designed CRT display has an OETF gamma of ~ 2.35. This is part of the “immovable legacy effect” of CRT.So as you can see what Dr. Soneira says here is relevant:

"In practice, the Gamma Curve (Transfer Function) for essentially all displays is so far off from the theoretical inverse of the encoding function that it is totally irrelevant"

"So my gold standard has been the Professional Sony Trinitron CRT monitors that are found in just about every production studio."

"Whenever I have measured the Transfer Function of these monitors it has always been a virtually perfect (2.20) power-law (straight line on a log-log graph) as far down as I measured with accurate photometers, which is below 1 percent of the luminance for Reference White"

thomas

Just got a chance to actually start reading through (e.g. copying & pasting loads of text into a translation tool…) ARIB TR-B28 (http://www.arib.or.jp/english/html/overview/doc/4-TR-B28v1_1.pdf) and I think it might be of interest to some people here as they measure the response of a Sony BVM-D24E1WJ and an Ikegami HTM-1990R.

Both displays were calibrated to 100cd/m² and the response was measured at 6500k/9300k with black set to 0.1/0.01cd/m² with (I believe) fullscreen and window patterns.
Page 90 in the PDF (80 in the footer on the page) onwards.

Black is set to 0.1cd/m2 brightness, input-output characteristic curve close to the 2.2 square. [both displays]

0.01cd/m2 black M1 monitor [the Sony] is set to brightness, low and middle level (20 percent) of the curve in the 2.6 square, Middle and high level (20%) of the curve in the square close to 2.4.

0.01cd/m2 black M2 monitor [the Ikegami] is set to brightness, low levels of 2.4 square in the middle of the square and 2.6, high-level Le curve in the square close to 2.4.

In a professional gradin g suite all the displays will be profiled and lutted with a color management system. The native characteristics of a given display are immaterial beyond its coarse contrast and gamut capabilities from the perspective of being correctable within desired tolerances.

I am currently using a variety of monitors in my workplace : LCDs and CRTs. I could not tell you what the native characteristics of the displays are : all I care about is that my color management system can correct them to whatever standard I happen to desire.

Print film , rec.709 at 2.2 and 2.5 (these are the standard presets however I can specify whatever target gamma I require and even tweak it interactively.

I don't know anyone who would unwrap a monitor and rely on its native characteristics as being in any sense accurate for professional color grading. You calibrate the whole chain not just the display.

You calibrate the whole chain not just the display.If a given display is calibrated to a defined color standard such as EBU, ITU, SMPTE or DCI then in turn accuracy and transparency now becomes the responsibility of your color management system.. not further display adjustment.

Mr. D are you saying it’s not possible to “calibrate” a Pro Sony or Ikegami CRT grade 1 monitor for a perfect 2.2 power law response and nominal color space or (Sony/Ikegami) should be designed differently?

Also I think one can ignore a given particular production viewing ambient and still be confident that the production reference monitor cal 2.2 is spot on..
...
So as you can see what Dr. Soneira says here is relevant:What happened to you using Dr. Soneira to try to convince people that the light on the screen would be linear, as part of what I think was about 3 years of you trying to find ways to back up your claim that these projectors could not do more than 255:1 on/off CR with 8 bit sources? You can't have it both ways. Is the light linear to support your 255:1 claim, or is a gamma like 2.2 correct?

--Darin

What happened to you using Dr. Soneira to try to convince people that the light on the screen would be linear..
Darin FWIW you are still conflating a nonlinear CRT response (voltage input vs light output) with how adjacent intensity codes for a digital payload actually correlate.

Fact is I asked Dr. Soneira a loaded hypothetical i.e. if a given display just happened to be linear, what might be the correlation between the encoded dynamic range per bit depth and the measured display light output?

Here is his word for word reply:

http://archive2.avsforum.com/avs-vb/showthread.php?t=639069&page=1&pp=30
If it were linear, then the luminance for signal level 255 would be 255 times the luminance for signal level 1, implying that the maximum Contrast Ratio possible was 255Note Dr. Soneira response did not imply (1/2.2) compression.


http://www.displaymate.com/ShootOut_Part_3.htm

The next question is how to distribute the brightness levels among the allowed digital values

1) Each step would also be a bit wider than the previous step, so the spacing between the steps would vary and be non-linear. Or,

2) In reality, the intensity steps are all separated by equal differences rather than equal ratios. That means that the brightness ratio between adjacent steps will then increase as the brightness decreases, so the granularity artifacts will show up first at the dim-end of the intensity scale.
I think was about 3 years of you trying to find ways to back up your claim that these projectors could not do more than 255:1 on/off CR with 8 bit sources? Not once did I ever suggest anything other than the limited DR that is possible per what is basically agreed upon in the industry as a (LDR) Low Dynamic Range medium.

Finally Dr. Soneira has said (same link as above) for a nominal (16-235) 8-bit digital payload it should be possible to represent 219^2.2 ~ 140,921 CR (or dynamic range). Here Darin Dr. Soneira is using the same logic you are.. which btw ties bit depth to dynamic range.. which IMO in itself is OK, just not in that manner, nor in the manner Greg or Guy used in the past.

Mr. D are you saying it’s not possible to “calibrate” a Pro Sony or Ikegami CRT grade 1 monitor for a perfect 2.2 power law response and nominal color space or (Sony/Ikegami) should be designed differently?

I'm saying its beside the point. In terms of a professional grading pipeline the end display is not what you rely on to define your working colorspace.

So the fact that a given manufacturer claims a set of criteria for their display device does not give weight to 2.2 being the solution to the question of gamma posed in this thread.

I agree with that..I was just defining what until just recently has been the industry benchmark, but now is not even manufactured any more that I'm aware of. Although I can assure you these calibrated "out of the box" ultra high end direct view Pro series reference CRTs are spot on SMPTE standard and needs no correction at all.

Their price tag suggest the color management (processing) is designed into the product;)

I agree with that..I was just defining what until just recently has been the industry benchmark, but now is not even manufactured any more that I'm aware of. Although I can assure you these calibrated "out of the box" ultra high end direct view Pro series reference CRTs are spot on SMPTE standard and needs no correction at all.

Their price tag suggest the color management (processing) is designed into the product;)

To be honest in my experience you are paying for the stability of the tube in voltage terms , not necessarily the CMS side of things ( you are pretty stuck with the mechanical properties of a given tube ) The color controls are really just for calibration than defining colorspace.

I digress:
http://cinespace.rsrhq.com/forum/viewtopic.php?p=279&sid=825e0cef0599f52dbc04d86068522940

Here is the opinion of someone I trust.

So the fact that a given manufacturer claims a set of criteria for their display device does not give weight to 2.2 being the solution to the question of gamma posed in this thread.Actually your linked reference validates my potition:

http://cinespace.rsrhq.com/forum/viewtopic.php?p=279&sid=825e0cef0599f52dbc04d86068522940
“Stu, I agree with your summary above. The response curve described in ITU-Rec.709 is in reference to a capture device; and that viewing devices do not behave in the same way.”

“Gamma 2.2 and 2.5 are the most common and it seems that over time gamma 2.2 is becoming the de facto standard as more material is produced for Rec.709 and not specifically PAL or NTSC.”

Michael Chenery

Actually your linked reference validates my potition:



I offer it in the interests of providing more information : not to win any arguments.

Darin FWIW you are still conflating a nonlinear CRT response (voltage input vs light output) with how adjacent intensity codes for a digital payload actually correlate.Nope, I'm not conflating that at all. Are you still so confused that you think these digital projectors can't do more than about 255:1 on/off CR with 8 bit sources, as you have claimed before?
Fact is I asked Dr. Soneira a loaded hypothetical i.e. if a given display just happened to be linear, what might be the correlation between the encoded dynamic range per bit depth and the measured display light output?You asked him some stuff, you misrepresented some stuff he said, claimed that he was the only professional who confirmed your position, then he responded to you right here on this forum and told you that he definitely didn't confirm your position. He isn't so ignorant to think that these digital projectors or CRTs are limited to 255:1 on/off CR with 8 bit sources. Even if somebody didn't have a clue about the reasons for things it is simple to do measurements that quickly show how ridiculous a 255:1 claim is. Have you done any actual measurements of light coming off a CRT (like you said you would)?

And please answer clearly whether it is still your position that the digital projectors discussed here are limited to 255:1 on/off CR with 8 bit sources (like Blu-ray). BTW: If that is still your position, you have a long ways to do to get to even basic understanding of this subject matter and a simple lightmeter would help.

--Darin

Darin here is three years of confusion condensed into one post:)
I've tried mentioning gamma, the any positive number divided by zero is infinity, how black-plus one can be relevant, etc., but it doesn't keep tbrunet from trying to make people think that 8 bit video can't have more than 255:1 CR. It makes it difficult to have a more advanced discussion when we are stuck down at this stuff.
--Darin


http://archive2.avsforum.com/avs-vb/showthread.php?t=639069&page=1&pp=30
Post #72
If it were linear, then the luminance for signal level 255 would be 255 times the luminance for signal level 1, implying that the maximum Contrast Ratio possible was 255

Darin here is three years of confusion condensed into one post:)You've probably been confused for more than 3 years. I've just seen 3 plus years of you knowing so little about the subject matter that you think these devices are limited to 255:1 on/off CR with 8 bit sources (not even counting your apparent inability to answer simple questions like whether you still claim these devices are limited to 255:1 on/off CR with 8 bit sources, in your own words).

Is it your goal to show people you don't have a clue about the subject matter? If so, you are doing a bang up job.

If pigs could fly ...

--Darin

One thing that I can state with certainty is that Darin is not confused. From what I have read, I can't say the same about tbrunet.

Darin here is three years of confusion condensed into one post:)One more about that post from Raymond Soneira where you quoted part of it. When he said in post 71:

http://archive2.avsforum.com/avs-vb/showthread.php?p=7170328&&#post7170328
This is a good question… you are making the common assumption that the light output for a display varies linearly with the signal level.he was trying to be nice to you. He could have put it more bluntly and said that you are clueless and know so little that you are assuming something that isn't true, but he decided to word it nicely for you. That may have been a mistake given that in all this time you don't seem to get that he was telling you that you were wrong.

Even him saying:
It turns out that the light output is not linear with signal level, but instead varies more like the square of the signal level.wasn't enough for you to comprehend that you were wrong and to get you to not try to make people later believe that Raymond Soneira had supported your 255:1 claim.

--Darin

Darin here is three years of confusion condensed into one post:)

I've tried mentioning gamma, the any positive number divided by zero is infinity, how black-plus one can be relevant, etc., but it doesn't keep tbrunet from trying to make people think that 8 bit video can't have more than 255:1 CR. It makes it difficult to have a more advanced discussion when we are stuck down at this stuff.

If it were linear, then the luminance for signal level 255 would be 255 times the luminance for signal level 1, implying that the maximum Contrast Ratio possible was 255
Are you aware that there is a signal level 0 which (with a perfect projector) gives us exactly 0 lumens? Now please open calc.exe and divide 255 by 0 to calculate the maximum possible contrast ratio.

I know Alan suggested I ignore Darin’s personal attacks and report them instead..but Darin is the one who repeatedly returns to the subject again and again.

Nominal 35mm negative film stock is capable of capturing (maximum) ~11 f-stops, or around 2000:1 contrast or Dynamic Range..and finally only ~200:1 for a film print. According to Dr. Soneira, Greg Rogers and Darin plain old “video” using Rec. 709 gamma compression can have as much ~1000:1 or even ~141,000:1 dynamic range.

Maybe we will see Mr. D or anyone else for that matter in the industry that could suggest 8-bit "video" has as much or more latitude than state-of-the-art negative film stock?


http://archive2.avsforum.com/avs-vb/showthread.php?p=7170328&&

In this case, for an 8-bit signal the maximum possible Contrast Ratio is 255^2.2=196,965,
BTW, for Rec.709 it’s 219^2.2= 140,921
For properly gamma corrected signals, that's:

1 / (( 1/219) / 4.5) = 985.5:1
............................................................ ........
Are you aware that there is a signal level 0 which (with a perfect projector) gives us exactly 0 lumens? Now please open calc.exe and divide 255 by 0 to calculate the maximum possible contrast ratio.
Orginally posted by Raymond Soneira
So what happens when the display only has a Contrast Ratio of 2000? For digital signal levels restricted to 16 to 235 as in Rec.709, signals below level 24 will be indistinguishable from the signal level 16 for black

@tbrunet, your why of quoting confuses me. I'm not sure whether you actually replied to my comment or not?

Anyway. My point is this: In digital 8bit video there is one level which represents true black (exactly 0 lumens). It's 0 for PC levels and 16 for video levels. Do you agree with that?

Dividing any number by zero tells one nothing about nominal performance envelopes or the potential dynamic range of a given format.

Dividing any number by zero tells one nothing about nominal performance envelopes or the potential dynamic range of a given format.
Then how do you measure on/off contrast ratio of a projector?

Would you agree that any projector reviewer today would measure an LED based projector (which can turn off LED lighting completely when video level 16 is asked for) to have infinite on/off contrast ratio?

madshi, fwiw thank you for not attacking me!

The capabilities of a given display does not reverse engineer itself into the real world DR limitations of the (format) source. No source contains infinite dynamic range.

Not 48-bit digital audio..
Not 33-bit log video..
Nor 8-bit "video"

madshi, fwiw thank you for not attacking me!

The capabilities of a given display does not reverse engineer itself into the real world DR limitations of the (format) source. No source contains infinite dynamic range.

Not 48-bit digital audio..
Not 33-bit log video..
Nor 8-bit "video"

Why do you not answer madshi's simple question?

fwiw thank you for not attacking me!
:)

The capabilities of a given display does not reverse engineer itself into the real world DR limitations of the (format) source. No source contains infinite dynamic range.

Not 48-bit digital audio..
Not 33-bit log video..
Nor 8-bit "video"
The problem for me is this: If you say that a specific source format "XXX" has a potential max contrast ratio of "YYY", then that reads to me as if any projector can not display a better contrast than "YYY" when display a video encoded in format "XXX". Or in other words: Blu-Rays are encoded in 8bit YCbCr. Now since you're claiming that 8bit has a max contrast ratio of something between ~1000:1 and ~141,000:1 then that reads to me as if no projector could possible show a greater on/off contrast than ~141,000:1 when displaying a Blu-Ray disc.

But that's far from the truth, because in reality a good LED based projector could easily measure with infinite on/off contrast ratio even when displaying a movie encoded in black&white (= 1 bit format).

The only question I have attempted to answer is.. what IS the DR limitation for a standard DVD disc that is encode per standard Rec. 709 format.

No digital format (Audio/Video) can encode infinity..

Now since you're claiming that 8bit has a max contrast ratio of something between ~1000:1 and ~141,000:1 .No..the DR is closer to ~200:1 or even less considering adjacent codes should be below our HVS perception threshold.

Yes your displays measured light output must be much greater..but perceptually it is the source that will determine max image density and granularity.

No..the DR is closer to ~200:1 or even less considering adjacent codes should be below our HVS perception threshold.

Yes your displays measured light output must be much greater..but perceptually it is the source that will determine max image density and granularity.
There's no argument over that granularity is severely limited with 8bit. The problem is that you've often used the terms "max potential contrast ratio". I think that has caused some confusion over what your point is. Can I get one straight answer to get clarification about your position?

Do you claim that no projector can measure higher than ~200:1 contrast ratio when displaying a DVD?

I think many people read your posts that way. But I think that's not what you're really trying to say?

IMHO talking about "max potential contrast ratio of a source format" is highly misleading in itself because to an average consumer (like me) talking that way seems to imply that the source format limits the measured contrast ratio the projector/display could possibly achieve. I hope that this is not what you're saying?

..talking that way seems to imply that the source format limits the measured contrast ratio the projector/display could possibly achieve. I hope that this is not what you're saying?madshi thats fair and goes directly to the issue of perceptual coding itself. i.e.

It is my belief that perceptually beyond an ideal (grayscale) tracking of ~200:1, one is unable to neither quantify nor parse (process) dynamic range much greater than that i.e. temporally in real time.

I’m not here to argue the merits of sequential on/off..I’m more interested in the bulk of the image rather than if a projector has a 25 ft-L white point and has a measured black level of .002 ft-L ..rather I would be much more interested in the performance of my display that could do .07-14 ft-L track (luminance) grayscale with absolute transparent ability.

I'm still not sure what you really mean. Do you say that:

(1) ... the human eye can not see more than ~200:1 ANSI contrast?
(2) ... a projector can not possibly show a higher ANSI contrast than ~200:1 because it's limited by the 8bit video format?
(3) ... something else?

Much empirical data confirms the “real” instantaneous DR figure; I could with confidence say somewhere between 200-300:1 could easily be a maximum. I will digress to an expert in the field of electronic displays...I'm sure Darin is certain to suggest Mr. Putman does not understand reciprocal functions nor comprehend a CSF.

http://www.hdtvexpert.com/pages/shmontrast.htm
:rolleyes:
Empirical data suggests the human eye is limited to a dynamic range of 100:1 at any given instant. That means that if you look at a scene with objects of different luminance values, you won’t be able to discern more than a 100:1 difference between the darkest and lightest objects. Of course, the instant your eye moves, its built-in auto-iris function raises and lowers the grayscale boundaries. That’s what allows you to perceive shadow detail and also pick out a white cat scurrying along in a field of snow”

“The truth is, grayscale is the single most important attribute of any electronic display. Without shades of gray, we don't have contrast. Without shades of gray, we can't create wide color palettes. Grayscales are where it all begins when a projector or monitor first comes to life on the drawing board.

I can easily see the diffence between a 3000:1 and 30'000:1 on/off, likewise the difference between 300:1 ansi contrast for 500:1 is also pretty obvious (with the right material). I will say that I think 400:1 and higher ansi contrast isn't really ideal for 8bit video where the number of graduations are limited and will highlight banding and and compression issues.

If I understand you correctly, it seems that you are overlooking an obvious point, tbrunet.

Even if what you are saying is true and the eye is in fact limited to seeing an instaneous CR of 200 or 300:1, that says nothing about how well a display can represent performance in a low APL scene. If a display can't do a convincing black when the eye expects to see black, those scenes will look terrible, even if the instantaneous CR of points in that scene are at your maximum levels.

Having a projector with a higher on/off CR suggests (but does not guarantee) that that particular display will do a good job in low apl scenes.

This is quite easy for anyone to see which also makes me feel that I am misunderstanding what you are trying to say.

@tbrunet, I'm sorry, but this kind of discussion (not getting any straight answers to my questions, sudden topic changes) makes my head ache. I guess I'll have to give up... :(

..that says nothing about how well a display can represent performance in a low APL scene.This is what I meant by the adjective "tracking" ..specifically grayscale (luminance) tracking through the entire DR. I should not have included the term ANSI in my post above.

I can easily see the diffence between a 3000:1 and 30'000:1 on/off ...As can most people. The reality is that people here can see the differences between 2000:1 and multiple thousands to 1, including things like enabling a dynamic iris when the projector is already doing more than a couple thousand to 1 and goes higher, using 8 bit sources, yet Thomas wants to stick to his claim of over 3 years ago that these projectors cannot do more than 255:1 on/off CR. At one point he told me he would actually measure a display, but over 3 years later he still hasn't not reported on doing that and just came up with excuses. It may be relevant that he has now taken a position that he doesn't care how the on and off portions of the on/off CR test measure, he is going to stick with his position. Back then, he even told one person who had measured thousands to 1 on/off CR that something must have been wrong because these projectors cannot do more than 219:1, 235:1, or 255:1 (whatever his number of the day was). Now he does just about everything he can to keep people from realizing he was wrong as he was telling people like Greg Rogers and Guy Kuo that they didn't know what they were talking about. Guy explained things in terms that just about anybody could understand and yet tbrunet decided to persist as if his numbers like 255:1 for the most on/off CR possible is true.

Most people would be embarrassed if they asked a writer like Raymond Soneira if they were right about the 255:1 or 219:1 figure, the writer responded that they were wrong, and then this person later claimed that the writer had actually agreed with them and was the only professional to validate what they had said. But tbrunet soldiers on. Even quoting part of the response where Raymond Soneira said that only having 2000:1 on/off CR would be a problem as if it supports tbrunet's position that we can't get more than 255:1 or 219:1 on/off CR.

People can see tbrunet's question:
Not sure if my question maybe relative, but I would like to know i.e. given an 8 bit gamma encoded source (DVD) how can a display ever have more ANSI or simultaneous CR.. greater than 255:1, or in the case of Rec 709... 219:1? I find a displays accuracy/ability to produces "grayscale" probably this single most important attribute concerning CR?and Raymond Soneira's response here (in post #71):

http://archive2.avsforum.com/avs-vb/showthread.php?t=639069&page=3&pp=30
@tbrunet, I'm sorry, but this kind of discussion (not getting any straight answers to my questions, sudden topic changes) makes my head ache. I guess I'll have to give up... :(I think getting you to give up is part of tbrunet's goal. The truth obviously isn't part of his goal or he wouldn't have played some of the tricks he has over the past 3+ years on this subject matter. You've already seen that he will refuse to answer simple questions. Many of them would make it obvious that he is either trolling or doesn't know what he is talking about, so he tries to divert the subject to something else and refuses to answer.

For those not familiar with tbrunet's tactics, you can do simple measurements that show he is wrong, but he has also refused to answer questions like the following (which I've probably asked 5 plus times and he has refused to answer it every time):
If a projector/screen combination are doing about 10 ft-lamberts for reference white (100 IRE), how much should they be doing for the missing levels:

100 IRE: 10 ft-lamberts
75 IRE: ?? ft-lamberts
50 IRE: ?? ft-lamberts
25 IRE: ?? ft-lamberts
0 IRE: ?? ft-lambertsEven though he is unwilling or unable to answer questions like that, he claims that he understands the subject matter and people like Greg Rogers, Guy Kuo (the author of AVIA), Stacey Spears, etc. do not understand it. Anybody with a lightmeter could see that tbrunet is wrong about his 219:1 on/off CR limit with REC.709 sources, but even without that, if he understood the subject matter he could answer my question above by stating what algorithm/equation he used and giving the values.

I'm not sure what would drive a person to spend 3 plus years trying to convince people they were right about something that can be easily be shown to be wrong with measurements, but that is what I see here.

--Darin

As can most people. The reality is.. Read and learn Darin!

http://www.hdtvexpert.com/pages/shmontrast.htm

Have a nice day:)

Orginally posted by Raymond Soneira
So what happens when the display only has a Contrast Ratio of 2000? For digital signal levels restricted to 16 to 235 as in Rec.709, signals below level 24 will be indistinguishable from the signal level 16 for black Is your goal here to show that Raymond Soneira disagrees with you, or do you not understand what he said? He says 2000:1 on/off CR is a problem because it isn't enough, and you use this over and over to support your position about only 219:1 being possible. After these years it is probably time for you to understand that or stop posting it.

--Darin

Read and learn Darin!

http://www.hdtvexpert.com/pages/shmontrast.htm

Have a nice day:)For those who aren't familiar with tbrunet's tactics, what he did here is one of his main ones. In the post immediately preceding his response here I pointed out that he avoids simple questions like mine I posted and tries to divert to something else. This is a usual tactic for him. When he can't answer something that anybody who understood the subject matter would be able to answer, he posts a link or quote of something else. He used to quote Raymond Soneira all the time and then Raymond Soneira came here and told him in no uncertain terms to stop (and that Raymond had definitely not validated tbrunet's position). He stopped for a while, but then went back to quoting Raymond Soneira as if he agreed with tbrunet, when he clearly said he did not.

Don't expect tbrunet's behavior to change here. Given that he has been trying to mislead people for over 3 years, he is likely to continue that behavior, as he has done here.

--Darin

This will be my last post Darin

Based on the exact quote you've posted above, you suggested Dr. Soneira was NOT able to calibrate that display.. and those levels are indeed distinguishable?

Orginally posted by Raymond Soneira
So what happens when the display only has a Contrast Ratio of 2000? For digital signal levels restricted to 16 to 235 as in Rec.709, signals below level 24 will be indistinguishable from the signal level 16 for black

.

This will be my last post DarinThank you.

As we've been over before, tbrunet has refused to answer simple questions about the subject matter over and over, and usually comes back with some diversion, like maybe a question of his own. At one point I obliged and answered his questions, but he was never man enough to actually answer mine after I had. So, as I said before, I am not going to let him extend the conversation by answering his questions that he is using to avoid answering an outstanding question to him.

I think he is well aware that he is wrong on this subject matter. I don't believe anybody would be so clueless as to think they understood the subject matter when they can't answer even simple questions about it.

--Darin

I am not going to answer his questions that follow my questions he refuses to answer.The question was rhetorical, the Dr. knows how to calibrate displays..your knee jerk response was as I indicated above. WRONG as usual!

The question was rhetorical, the Dr. knows how to calibrate displays..Taking your position about him knowing what he is doing, he told you that you were wrong, he said that 2000:1 on/off CR was a problem because it wasn't enough, and yet you claim he supports your position about 219:1 or 255:1. I think people can see through that.

--Darin

Don't expect tbrunet's behavior to change here. Given that he has been trying to mislead people for over 3 years, he is likely to continue that behavior, as he has done here.

My ignore file would work a lot better if people quit quoting him. You know, after three years, and all. ;)

Back to the subject at hand...

One of the issues with some of the professional-grade LCDs is that they are stuck with 8-bit imaging, from some limited research I have done (e.g., Sony, JVC, Panasonic all use 8-bit per channel color). Given this, I would expect them to have issues if they are fed a >8-bit signal (e.g., 10-bit, 4:2:2 HD-SDI).

Bill

One of the issues with some of the professional-grade LCDs is that they are stuck with 8-bit imaging, from some limited research I have done (e.g., Sony, JVC, Panasonic all use 8-bit per channel color). Given this, I would expect them to have issues if they are fed a >8-bit signal (e.g., 10-bit, 4:2:2 HD-SDI).
I guess that could be worked around by using time based dithering, just as plasmas do. The amount of dithering needed to get from 8bit to 10bit would be fairly low, so I wouldn't expect any noticable disadvantages, if implemented well.

One more on tbrunet trying to make people believe that Raymond Soneira supports his position. People can see what happened a couple of years ago when tbrunet tried to do something similar (starting at post #56):

http://archive2.avsforum.com/avs-vb/showthread.php?p=8121257&&#post8121257
Contrast Ratio CR NOT Correlated with Bit Depth
I want to publicly thank you Dr. Soneira! You have validated my position on the direct correlation between CR ratio and Bit Depth. FWIW you are the only professional on this forum to do so!First of all, I have not previously participated in this discussion, I certainly have not validated the above, and I strongly object to anyone using or invoking my name without my explicit permission.

Second of all, the above statement that the Contrast Ratio, CR, depends on bit depth is categorically wrong. CR is the ratio of Peak-Level Luminance to Black-Level Luminance. Those two special levels are absolutely independent of the Bit Depth, the Gamma Function, and the Transfer Function. Note, however, that all of the levels that lie between the Peak-Level and the Black-Level (17 to 234 for digital video or 1 to 254 for full range PC images and video) do explicitly depend on the Bit Depth, the Gamma Function, and the Transfer Function. Also, the digital levels near the Black-Level will be affected by the Contrast Ratio. I hope this clarifies my position. I consider the matter closed and will not be drawn into Ringer discussions of this type.

My complete 9-Part Display Technology Shoot-Out series is now available on the DisplayMate website. See http://www.displaymate.com/shootout.html. Please refer directly to the articles.--Darin

That's a good quote, Darin. It clears all the muddy water. Let me put it in bold just to stress the significance:

Raymond Soneira:
"CR is the ratio of Peak-Level Luminance to Black-Level Luminance. Those two special levels are absolutely independent of the Bit Depth, the Gamma Function, and the Transfer Function."

In this case, for an 8-bit signal the maximum possible Contrast Ratio is 255^2.2=196,965,
BTW, for Rec.709 it’s 219^2.2= 140,921madshi,

According to Chris Wiggles the quote above from Dr. Soneria is based upon the misguided assumption that digital intensity is coded linearly.

http://archive2.avsforum.com/avs-vb/showthread.php?p=7170328&&
Originally posted by Chris Wiggles (Post #72)
"That's two radically different things put together incorrectly that has no real meaning. They're just one random number raised to the power of another. This would mean that other bit depths would have greatly increased CR capability when that's not the case at all: lets plug in for 10-bit, 1024^2.2 would be 4,194,304. That's not true at all.."

"This mistaken math is based on the incorrect assumption that if it were coded linearly that it even would be 255:1, which it would not be at all. Hence the confusion still even with gamma"

So what happens when the display only has a Contrast Ratio of 2000? For digital signal levels restricted to 16 to 235 as in Rec.709, signals below level 24 will be indistinguishable from the signal level 16 for black
Originally posted by Chris Wiggles
“I'm not sure where you're getting that perception? Those levels will still be imaged at the display unless someone mistakenly misaligns black. The entire range will be distributed over a smaller contrast range, i.e. the limited 2,000:1 range of the display.”


Madshi if you "think" you’ve had an epiphany or want to put something into bold.. try explain to anyone why Greg Rogers or Guy Kuo miscalculated the figure posted above from Dr. Soneria by ~ 140,000?

For properly gamma corrected signals, that's:

1 / (( 1/219) / 4.5) = 985.5:1

Thomas, I think you need to reread the posts in that link because they don't support your conclusions.

Lawguy,

Please note how Chris Wiggles who is considered this forums "display calibration expert" attempt to correct Dr. Soneira who btw has a PhD in physics and renowned in the field of display technology. FWIW Lawguy if you had read carefully my post above…indeed there is great confusion on the correlation and my intuition is really spot on. From day one people have attacked me every time I've post my perspective..I've never accused anyone of intentionally misleading me or anyone on this forum.

I tried to explain to Darin the implications of Dr. Soneira’s response to me and come to find out my own interpolation was validated by Chris Wiggles himself. I’m only saying that IMO theses are the facts..you can believe what you wish.

thomas

There is no limit on contrast ratio based on its bit-depth, all it has to do with is the brightness of white and black. You can achieve any contrast ratio with 1-bit if you wanted. (black & white, no shades in-between)

What you might be able to work out, however, is how far an 8-bit source can be stretched before posterisation starts becoming an issue once the contrast ratio gets to a certain point, as each step in light will become too far apart and won't provide smooth gradations any more. Moving up to 10-bit or greater should allow you to use higher contrast ratios without introducing posterisation.

You can achieve any contrast ratio with 1-bit if you wanted. (black & white, no shades in-between)Put another way, one can achieve infinite dynamic range (contrast ratio) with one bit.

That in my book is NOT the real world.

madshi,

According to Chris Wiggles the quote above from Dr. Soneria is based upon the misguided assumption that digital intensity is coded linearly.No, you either don't understand what Chris said or are acting like you don't. For a clue, when you see "if it were", it doesn't mean that it is or that the person (Dr. Soneira in this case) said that it is.
Madshi if you "think" you’ve had an epiphany or want to put something into bold.. try explain to anyone why Greg Rogers or Guy Kuo miscalculated the figure posted above from Dr. Soneria by ~ 140,000?Madshi,

Most of this stuff has been explained to tbrunet multiple times. I think he purposely ignores simple explanations because he wants to extend the conversation and keep from admitting that he was wrong. As Greg said, his numbers were based on if you decoded strictly with REC.709 with the linear tail. His number was correct for that case. Dr. Soneira's number was for decoding strictly with a 2.2 gamma (no linear tail). His number was correct for that case. At the lowest end there can be big differences in relative numbers between using the linear tail of REC.709 on decode and using a strict 2.2 gamma, even though the differences in absolute level are small (small differences in absolute values near video black can give large differences in CR to those levels from 100 IRE). As both explained, these were not for total CR (since there is no limit there), but for CR with black+1.
Please note how Chris Wiggles who is considered this forums "display calibration expert" attempt to correct Dr. Soneira who btw has a PhD in physics and renowned in the field of display technology.Again, going by your assumption about Dr. Soneira being so infallible, Dr. Soneira told you in no uncertain terms that you were wrong. And yet you persist as if he agreed with you.

As far as the claim about a display with 2000:1 being a problem because it will crush lots of levels, we've already been over how this actually goes against your claim, yet you keep bringing it up. In this case the proof is in the images. And anybody who gets a Sony Qualia, JVC HD2K, Sharp 12k, InFocus IN83, or one of the Samsung DLP from Joe Kane, puts them in a mode that is around 2000:1 on/off CR and calibrates them with SMPTE and/or DVE, then checks to see how many levels cannot be seen in any image (are crushed to black), they will see that it is not the case that lots of levels above video black are crushed. A couple or few may be, but if setup right it will not be 8 crushed levels. Of course, a person could choose to crush all sorts of levels no matter the on/off CR, but there is no need to do that in this case. As Dr. Soneira told me when I asked for clarification in email, it depends on how the lookup tables are setup. Of course it does, and for the front projectors I mentioned I don't believe any of them have what I would consider a design flaw of crushing that number of levels, even though they are limited to only around 2k:1 on/off CR in some modes. And as I said, all this goes against your argument anyway, since your argument was the opposite of Dr. Soneira's statement about 2000:1 on/off CR being a problem because it isn't enough. That is, more than 2000:1 on/off CR is a good thing according to Dr. Soneira, and your claim was the opposite.
I tried to explain to Darin the implications of Dr. Soneira’s response to me and come to find out my own interpolation was validated by Chris Wiggles himself.No, it definitely was not validated by Chris Wiggles. Please don't act like Chris saying, "if it were encoded linearly" means that it is encoded linearly. It is not in that space and Chris knows it isn't. He was pointing out a result of one way of thinking, but there was one thing in Dr. Soneira's post that some of us may have missed the first time reading it. That was:
Here we’re ignoring level 0, which is a special case that we’ll discuss at the end.Dr. Soneira was not saying that 141k:1 was the highest CR possible for video black at 16 and reference white is 235, he was saying that it was the highest if you ignore the video black level and use black+1 (level 17), because even a million to one would be lower than is possible for a display to do for on/off CR, which doesn't use the black+1 level, but uses the black level. Both Dr. Soneira and Greg Rogers said that there would ignore the total possible for the moment. I'm pretty sure I've explained this before. There is not limit for the total range, so looking at what happens to black+1 is interesting. But it leaves the door open for somebody to go around misleading others about what is possible.

I'm guessing the implication you tried to tell me about Dr. Soneira's response to you was that things are linear, when he clearly told you that you were wrong about that.
I’m only saying that IMO theses are the facts..you can believe what you wish.Again, you have your facts wrong and I think people who can understand a simple "if" doesn't mean something is the case, know that.

I would like to believe that everybody is after the truth, but you've had opportunity after opportunity to show that you were actually after the truth and shown otherwise. I made the mistake once of giving you the benefit of the doubt that you were actually looking for the truth and you proved me wrong. If you were after the truth you would answer the questions about what different levels would be and then keep on that until you could get them right, not try to divert to new things every time somebody asks a very simple question that you can't even figure out.

And here is a simple case where you have the opportunity again to show whether you are after the truth or not. It is clear that you've misinterpreted or misrepresented what Chris said. He definitely did not say that things are linear. Now that I've pointed out the "if" in the statement you can continue on acting as if it says something different that it actually says, like you have with some other things, or you can admit that it doesn't say that things are encoded linearly. And Dr. Soneira told you 2 and half years ago that things were not encoded linearly when he said:
It turns out that the light output is not linear with signal level, but instead varies more like the square of the signal level.yet you persisted with trying to make people believe that Dr. Soneira agreed with you that things are encoded linearly and the highest CR possible is 219:1 or 255:1.

As long as your behavior is a person trying to keep people from the truth I am not going to believe that the truth is something you are after. My perspective is that you are doing anything you can think of to keep from admitting that you were wrong with your claim that 219:1 or 255:1 was the highest on/off CR possible with 8 bit video, and that you owe Greg Rogers and Guy Kuo apologies.

BTW: About 7 posts above this one I already posted Dr. Soneira's post where he said that the bit depth does not determine the maximum CR (which agrees with what Chris said about that). It was the use of black+1 instead of black where the disagreement/confusion came in about that earlier.

--Darin

I tried to explain to Darin the implications of Dr. Soneira’s response to me and come to find out my own interpolation was validated by Chris Wiggles himself. I’m only saying that IMO theses are the facts..you can believe what you wish.

I don't think I have ever validated ANYTHING you've said. Certainly not as countrary to Darin's explanations to you. I have repeatedly asked you to cease quoting me since you have absolutely no intellectual integrity, or intellect to speak of at all. You distort, manipulate, and otherwise confuscate topics out of ignorance or trolling immaturity. If you think I disagree with Darin on this topic, you are SORELY mistaken.

Nearly every reputable video expert around these parts INCLUDING Dr. Soneira have explicitly rejected your absolutely ridiculous, stupid, ignorant, and stubborn nonsense in the most vehement ways possible, yet you still peddle the same nonsense.

I wonder if you still believe that no display can achieve an on/off cr greater than 2,000:1? Or whatever the limit is that you think today, which I am sure will change since you never answer any of the questions posed to you, you simply copy and paste meaningless and unrelated nonsense, or link to somebody else since you have no understanding whatsoever of what you're talking about.

Put another way, one can achieve infinite dynamic range (contrast ratio) with one bit.

That in my book is NOT the real world.

That is correct. It's called a light bulb. If you were smart enough to operate a simple light switch, which apparently is beyond your capabilities, then you'd realize that.

Quite obviously you don't live in the real world.

From day one people have attacked me every time I've post my perspective..I've never accused anyone of intentionally misleading me or anyone on this forum.


I recall that on day one you didn't even REALIZE that digital video even EXISTED. It had to be explained to you what digital even WAS, and you still are of the belief that there are volts on an optical disc.

Nearly every reputable video expert around these parts INCLUDING Dr. Soneira ...OK Mr. “digital video” guru,

Dr. Soneira suggested per Rec. 709 he calculates 219^ 2.2 = 140,921 CR

According to Chris’s limited mathematical prowess the Dr. just seems to be raising one number to the power of another? Your words not mine.

Also according to Chris’s comprehensive knowledge of display calibration, he suggested the Dr. appears unable to calibrate the black level for said hypothetical 2000:1 on/off ?

Darin, the linear tail is really there to de-emphasize real world artifacts of capture devices. It does this by reducing gain in this near black region. On the display side of things this (linear tail) function is not even practical since it would require near infinite gain from your displays response.. and it would also effectively undo its original purpose regarding noise suppression. In the production end the function is used for accurate image processing though.. and the target for a nominal Rec. 709 encoding is in fact a ruler flat 2.2 transfer function from blk-wht.

Darin's post above tries to make a distinction between 'black' or luma decimal code word 16 and 'black +1' (code 17).. and that one quantization level has the magical effect of providing infinite dynamic range for a given digital format.

I'm sorry but thats not true for audio nor video. Also Darin, perceptual encoding is not magic either..simply the camera video is gamma corrected for an end-to-end correlation that is perceptually ~1:1 and for digital payloads specifically to make better use of the limited bit depth range by not wasting codes on perceptually useless detail. Gamma has three purposes:

1) distribute quantization error (artifacts)
2) shift tonal values for perceptual intent
3) linearize your displays response

Also according to Chris’s comprehensive knowledge of display calibration, he suggested the Dr. appears unable to calibrate the black level for said hypothetical 2000:1 on/off ?Again, as others can probably see, tbrunet is claiming that Dr. Soneira is correct in order to claim that he (tbrunet) is correct, when Dr. Soneira's statements go in the opposite direction of tbrunet's stated position. Paraphrasing part of Dr. Soneira's response here:

http://archive2.avsforum.com/avs-vb/showthread.php?t=639069&page=3&pp=30

at post #71 to tbrunet's question:
Not sure if my question maybe relative, but I would like to know i.e. given an 8 bit gamma encoded source (DVD) how can a display ever have more ANSI or simultaneous CR.. greater than 255:1, or in the case of Rec 709... 219:1? I find a displays accuracy/ability to produces "grayscale" probably this single most important attribute concerning CR?Dr. Soneira said that 2000:1 on/off CR would be a problem because then levels 17-23 would be crushed. That is, that low an on/off CR would cause a problem in the grayscale (those levels of grayscale could not be seen). Yet tbrunet both wants to claim that Dr. Soneira's position there is gospel and that grayscale is the most important thing, with on/off CR above 255:1 not mattering. You can't have it both ways. If only having 2000:1 on/off CR screws up 7 levels of grayscale, then only having 255:1 on/off CR would screw up 17 levels of grayscale. But it is just another diversion from tbrunet to keep from admitting that his 219:1 or 255:1 on/off CR limit with 8 bit video was wrong. It doubt he will address this and will keep trying to extend the conversation by ignoring it and trying to confuse people. Not something an honest person would do.

tbrunet can complain about accusations of "intentionally misleading" people all he wants, but refusing to answer the simplest questions like:
If a projector/screen combination are doing about 10 ft-lamberts for reference white (100 IRE), how much should they be doing for the missing levels:

100 IRE: 10 ft-lamberts
75 IRE: ?? ft-lamberts
50 IRE: ?? ft-lamberts
25 IRE: ?? ft-lamberts
0 IRE: ?? ft-lambertsand what level 17 should be if level 235 is 10 ft-lamberts, while claiming he is the only one here to understand the subject matter, are intentional things that tbrunet does in hopes that people won't realize that he has been wrong in his crusade of multiple years to make people believe that 219:1 or 255:1 is the most on/off CR possible with the 8 bit video we get.

One time tbrunet answered the question about what level 17 should be if level 235 is 10 ft-lamberts, and guess what, it exactly coincided with what Greg Rogers had said. tbrunet didn't realize it since he gave it in ft-lamberts and Greg gave it in ratio, but they were the same answer and when I pointed this out, tbrunet claimed he had made a mistake, refused to give any new answer and to this day has refused to say whether the real answer was darker (higher CR) or lighter (lower CR) than what Greg had said. That is an intentional thing that tbrunet has done, so his complaining about people saying he is "intentionally" deceiving people fall on deaf ears here.

tbrunet,

I think you are proving me right when I said:
My perspective is that you are doing anything you can think of to keep from admitting that you were wrong with your claim that 219:1 or 255:1 was the highest on/off CR possible with 8 bit video, and that you owe Greg Rogers and Guy Kuo apologies.You probably still can't or won't answer whether you think Greg's answer understated how much CR 8 bit video provides or overstated it (which you accused him of over and over), but will continue playing the victim card and backing stuff from Dr. Soneira that actually contradicts your stated positions.

I bet other people could have figured out that Dr. Soneira was telling them in no uncertain terms that they were wrong if they had asked him your question and gotten the responses he gave.

I have little doubt that you will want to extend the conversation without ever admitting that your 219:1 or 255:1 on/off CR limit with 8 bit video position is wrong.

--Darin

Dr. Soneira said that 2000:1 on/off CR would be a problem because then levels 17-23 would be crushed. Chris says the Dr. simply misaligned the displays back level;)

Since gamma (1/2.2) compression is effectively tied to a CRT’s inverse gamma, results in their dynamic range being coupled.

It is mathematically impossible to code 11 f-stops, or around 2000:1 dynamic range into a standard "video" (color space) digital payload regardless of its bit depth. The Rec. ITU-709 transfer curve itself will clip highlights and crush content near black.

This is why specialized or custom gamma LUTs are implemented inside digital production workflows. Just as contouring artifacts are bit depth related so is dynamic range limited by a given format (color space) i.e. the standard ITU-Rec 709 video standard. Just by customizing the coding or gamma LUT,one could encode more dynamic range per given bit depth.

Chris says the Dr. simply misaligned the displays back level;)

Since gamma (1/2.2) compression is effectively tied to a CRT’s inverse gamma, results in their dynamic range being coupled.

It is mathematically impossible to code 11 f-stops, or around 2000:1 dynamic range into a standard "video" (color space) digital payload regardless of its bit depth. The Rec. ITU-709 transfer curve itself will clip highlights and crush content near black.

This is why specialized or custom gamma LUTs are implemented inside digital production workflows. Just as contouring artifacts are bit depth related so is dynamic range limited by a given format (color space) i.e. the standard ITU-Rec 709 video standard. Just by customizing the coding or gamma LUT,one could encode more dynamic range per given bit depth.As I said:

I think you are proving me right when I said:
My perspective is that you are doing anything you can think of to keep from admitting that you were wrong with your claim that 219:1 or 255:1 was the highest on/off CR possible with 8 bit video, and that you owe Greg Rogers and Guy Kuo apologies.Most people have enough self respect not to exhibit the behavior you exhibit here. As I said, it is definitely intentional that you refuse to answer simple questions that would show you to be wrong and try to extend the conversation away from places where you obviously don't (or didn't) have a clue what you were talking about and continued pushing things that simply were not and are not true.

BTW: This is a little bit like a person claiming they understand calculus and wanting to discuss it when they can't even figure out that 2 + 2 is 4. You are still at the stage sticking to your 2 + 2 is 5, but want to claim you are the only one who understands calculus and won't address your 2 + 2 = 5 claim. Someday you might graduate beyond remedial stuff with this subject matter, but sticking to your 219:1 or 255:1 claims shows that you are either dishonest or don't even have a beginning knowledge of the subject matter.

--Darin