OK so the more I look into this, the more confused I get (and I have to make a presentation to my class a week on Monday) so please feel free to chip in if you feel you can shed some light D) on this. It's actually a really interesting subject and for those of us who are involved or specialise in the graphic arts, I think it's important to grapple with the fundamentals. So far I've gleaned that colour doesn't actually exist in nature, or at least not the way we see it; it's a perceptual matter of our brains interpreting blue-green and yellow-red wavelengths and translating them into the many different colours we can perceive in a daylight environment: a gamut of approx 2.38 million, according to the CIE 1931 XYZ colour model. It gets confusing, however, when we dissect our perception of colour based on the structure of our eyes and how the cells in our eyes differentiate between chromacity (colour) and luminance (brightness). The maximum number of perceivable colours can therefore be extended to in excess of 10 million, but not everybody agrees that chromacity and luminance should be combined. Moving on to digital reproduction of colour, it gets even more complicated: the 16.7 million "colours" of the 24bpp (24 bits per pixel) RGB colour model are not in fact colours but the maximum number of addressable RGB values (256³). So how many colours can an LCD display actually reproduce? According to Steve Upton, president of chromix.com, computer displays have a true colour gamut considerably lower than the advertised range of 16.7 million. Granted, he posted this back in 2005, but I'm curious how he gets these gamut values for the hardware - I've drawn a blank on that one. This is the part where I get really confused, because it's hard to find actual figures for how many colours electronic displays can actually produce as opposed to the number of RGB values they can theoretically address. What I understand from Steve's post is that the bit-depth of a display and the resultant number of addressable RGB values is not equal to the actual gamut of the display, and that the two are often confused. So has technology advanced in the last 10 years in such a way that we have been able to develop more saturated colourants in LCD panels and therefore display more actual colours? I know that newer LED backlit IPS panels are able to achieve deeper blacks and a wider gamut (eg. 99% of Adobe RGB), but what does that say about actual gamut? I'm confused! Edit: so I just un-confused myself (to some extent, at least) when I read an excellent explanation of gamut vs bit depth. The sRGB gamut does not have 16.7 million colours, but rather a 24-bit RGB colour model has 16.7 million increments across an existing colour space, whether it be sRGB or something else that is interpreted digitally. That explains why the sRGB and Adobe RGB gamuts can be so small in relation to human vision. It's the marketing garble that confuses us. Always! "Hey look everybody - this monitor can display 1.07 billion colours!" No, it can't.
ok. These figures on how many colors the human eye can see are all bogus. If you read the research done on how they measured, you will see flaws. If you don't think that the human eye can't see more than 16.7 million colors. Do the test your self. Make a picture, of the size of your screen (bigger the better), and make it s gradient from black to dark gray. You'll see stepping. Do you see stepping in real life when you look at objects? No, of course not. This highlight the limitation of monitors today. We can see several billions, if not trillions of colors. The monitor most people use are able to produce 6-bit per channel (red, green and blue are channels), so 2^6 * 2^6 * 2^6 = 262,144 colors, and uses frame rate control to emulate the missing colors to reach 16.7 million colors. What is that? When a color the monitor can't reproduce is detected by the monitor, it looks at a table or follows an algorithm to take 2 colors that it knows that, if it switches to continuously up to the speed of the monitor refresh rate, it will trick your eyes in seeing the correct color. The visual tricks works really well, as you can see. The colors accuracy isn't great, and difficult to adjust, but it works. Then you have true 8-bit panels, which those are 2^8 * 2^8 * 2^ 8 = 256x256x256 = ~16.7 million colors (or 16,777,216 to be exact). Those are more costly monitors, so they are hugely popular, but popular. Basically high-end consumer grade IPS panels, and select high-end TN panels, and many *VA panels. It doesn't determine the color accuracy of the monitor, but if you hae the equipment you'll have a much easier time to calibrate it. But usually, as you are investing in a high-end consumer monitor especially in IPS side of panels, manufacture do efforts in giving you nice accurate colors in some fashion. Most of true 8-bit panels are marketed to be able to produce 10-bit colors per channel, or 1.07 billion colors. They use the same system as 6-bit panels to achieve 8-bit colors, but here it is to achieve 10-bit colors per channels. For true 10-bit you need to go in the professional grade monitor territory. The big problem is that 1.07 billion color monitor is that, well first you need to use DisplayPort, but beside that (as that is easy to fix these days), but you need a compatible graphics card. All GeForce and Radeon cards can't output so many colors. Only select Quadro card and FirePro cards. And to top things over, you need the content to be so many colors, and supported software. PhotoShop is, if you wonder.
I should post this part first so that you understand my answer: I did the test, and the gradient is so freaking smooth it makes me want to cry. I had to zoom in to 100% and even then, any colour discrepancy was minimal, barely noticeable. My problem with your test is that it assumes that monitors can actually display 16.7 million distinguishable "colours", and as above I don't believe that is the case at all - as I said, 16.7 million refers to the maximum possible number of combinations of RGB values with 8 bits per channel. I'd love to measure (somehow) the actual number of colours that a computer monitor can display. Steve upton gave a figure of about 1 million, which seems more reasonable to me; it's just really annoying to read almost everywhere that LCD monitors display 16.7 (or more) million colours. Think about it... RGB values of (0,0,0), (1,0,0), (0,1,0), and (0,0,1) should be considered four distinguishable "colours" because they have unique identities. But they are not. They will all look exactly the same on a screen: black. Similarly, you could have (255, 0, 0), (255,1,0) and (255,0,1) as three distinctive reds. In practice, they will be totally identical. This is applicable to any of the 16.7 million RGB values, so you have to ask just how valid is that number in practice? My reference figure of 2.38 million for the human gamut is indeed out of date (in some ways at least) but the CIE 1931 XYZ colour model is still used today as a standard for demonstrating colour spaces in the context of human vision.
I can see the banding on that picture as clear as day, but that's because that's what the picture is - exactly 108 bands from black (0,0,0) to 42% grey (107,107,107) with no dithering. I doubt real-life gradients are made up of bands like that! Digitally speaking at least, smooth black-white gradients use far more than the 256 values available in strict greyscale. Here's my gradient from (0,0,0) to (107,107,107), identical to yours except that it makes use of a possible 756 RGB values and also uses dithering. I can see no banding in this gradient at 100% and it looks be-a-utiful, and it was achieved with only 756 colours. (Large file (>1MB) and must be veiwed 100%)
I do believe that real life shows it perfectly smooth, and doesn't use any tricks to achieve that. I can't find any flaw or limitation that comes close to the output of our monitors. As for that picture. Sorry, but I see banding. It's not as clear cut as mine, but I see it. It's particularly visible on motion (scrolling up and down on the picture), or pay close attention to it (all at 100% zoomed in). And yes, I am using a true 8-bit IPS panel. What it looks like the picture is ding is a dithering effect on the edge of the gradients. So yes, it's much less visible, but the limitation is still there.
I see one shade of grey on A. I see another shade of grey on B. Posting pictures is silly. To delineate the colours a monitor can display you need equipment that doesn't suffer from being the complete cluster **** that is the human body.
The human body is not the problem. The problem is our brain. Our brain is designed in a fashion to understand the real world, not visual tricks. When there is an error, it assumes that there is an error caused by something else, like eyes, and corrects it. This is also why we don't see a black dot where the optical nerve passes out of the eye to the brain, and is pretty darn good at hiding it, without feeling anything strange thing like an abstraction or something. A small dot at a specific location is when it fails (ie: you no longer see the dot), and that is when you only have 1 eye, (second eye is backup to fix the problem), and even then, it blends super super well the surroundings. Also, this is unrelated. The discussion is that we can see more colors than our displays can provide, and how these studies on how the human eye can only see x amount colors which is well under our monitors ability, so don't need more colors, is B.S (at least that is the point I am doing). I don't see stepping on shadows, I don't see a deterring effect anywhere, I transitions are above smooth. Take a walk on the park, sit down on a bench, and appropriate real life "graphics" for a moment, look at the shadows, the details and vividness of leafs or ground, and so on. I don't know how to explain this, but spend 10min, focusing on that. I know I have a few times. 'can't say I am not trying to enjoy nature at times and not looking at a screen all day and staying indoors.
Naturally - you're using a U2410 which doesn't support hardware calibration; my display has 12-bit processing, programmable 12-bit LUT and has recently been hardware calibrated. When it comes to displaying smooth gradients, there's utterly no contest. I anticipated this; my main reason for posting that picture was to show that it's possible to reduce banding with a slightly larger palette, and it's possible to eliminate banding if you use the right hardware and if it has been configured correctly. So I agree with you that screens have limitations (of course they do), but not all screens have the same limitations. The problem here is that you place the emphasis on what is shown, when in actual fact the emphasis should be placed on how we interpret what is shown, and that's when it gets very complicated. As per theshadow2001's example, our perception of colour is dependent on context, which is why dithering in gradients is able to achieve a much smoother transition and altogether eliminate banding. Nobody is saying that LCD monitors have a greater colour gamut than the human eye; in fact I said precisely the opposite (monitor 1 million vs human vision about 10 million). Once again, 16.7 million "colours" is not a gamut; it's bit-depth.
Just a wee update on one thing that was confusing me. I sent an email to Chromix (wanting to speak to Steve Upton) and one of the tech guys (Pat) got back to me and was very helpful in providing additional info about colour spaces and the actual volume of a given gamut. Pat corroborated what Steve said and gave similar figures, around 1 million colours (in terms of L*a*b values) for the sRGB gamut and a little over 1 million for AdobeRGB. This helps me (and should help others) to understand colour relative to the different areas in which it is used: perception by us, and generation by devices. Whilst LCD screens can in theory produce 16.7 million (or more) RGB values, they cannot produce that many discernable colours, which is an intrinsic gamut limitation.
When you say discernible. Discernible by what? Are the colours so close together people can't actually perceive the difference. Or is a pixel physically incapable of actually producing a different colour for each address combination? I'm still not sure where the limitation is. They person or the monitor itself.
It's both, e.g. do the colours really exist if we can't discern them? After all, that's fundamentally what colour is: what our brain interprets from what is seen. If you take three RGB values like (255,0,0) and (255,0,1) or (255,1,0), you have three pure reds that are totally impossible to distinguish from one another, but three supposedly disparate colours. Even as high as (255,20,20) the reds appear virtually the same; it's only by placing the two colours next to each other that you see any difference (and even then, I think you'd need a high quality display, since the only change I see is saturation...and in addition to that, it depends on the viewer, as everybody perceives colour differently. Perhaps my reds are more saturated than yours anyway!!). Certain things we can take as fact: nobody can perceive any difference between (255,1,0) and (255,0,0), so to all intents and purposes there is no difference other than numerically. But is there a red that is redder than (255,0,0)? Well, there are probably hundreds of them, but they can't be displayed electronically. There's a thought!
The confusion isn't so much what gamut monitors can cover; it's more a matter of how terms like "gamut", "bit-depth" and "16.7 million colours" are used interchangeably, which results in a lot of confusion. And yes, high end monitors can show more colours, but (again) it's got nothing to do with bit-depth, e.g. an 8-bit monitor might have a wider gamut than a 10-bit monitor, therefore it can display more colours, although not with as many steps or increments. We've had all this digital stuff drummed into us for so long that it's hard to move away from it and look at the facts in isolation from the numbers. In the strictest sense, more bits does not equal more colours! I realise that the human eye is able to discern many colours, but where things like gradients are concerned, it's not so much about the number of colours as it is about the context of the colours and how the eye interprets what it sees. My example of a very limited number of colours, actually fewer than 750, shows a perfectly smooth gradient from grey to black (well, it's very smooth on the right hardware, for sure!).
