billd@fps.com (Bill Davidson) (07/26/90)
In article <139554@sun.Eng.Sun.COM> poynton@sun.com (Charles A. Poynton) writes: >The final row is the chromaticity of white, which represents the relative >contributions ("colour balance") among red, green and blue. Unfortunately >most computer displays are WAY too blue. Daylight has the same colour as a >chunk of platinum heated to 6500 kelvin, hence the CIE defined standard >illuminant D65 to have this "colour temperature". Modern blue phosphors are >about twice as efficient as red and green phosphors, with respect to the >sensitivity of human vision and driving all three electron guns in a colour >CRT with the same amount of beam current produces a picture that is about >twice as blue as daylight -- very noticeably blue. This is done to achieve >the maximum possible brightness, at the expense of colour reproduction. The >additional brightness over a more sensible choice of 6500 K is only about >5%, due to the eye's insensitivity to blue, but in a market that has >historically had little interest in accurate colour reproduction, a 5% >brightness increase was worth the penalty. This makes me worry about possible extra problems for gamma correction. My understanding of gamma problems is that linear changes in voltage do no give linear changes in brightness. A gamma curve plots the changes of voltage vs. brightness. If one color of phosphor is more efficient than another color, might this also change the gamma curve? Should we really have three separate gamma values (one each for red, green and blue)? This might help bring the color balance back a bit closer to the proper value. I'm sure that the differences are small enough that for most applications it doesn't matter but still, if there's an error in my calculations, even if it's acceptable, I like to know about it. I don't recall Roy Hall's book mentioning this. He does mention correcting for a given color monitor but nothing about separate gamma curves. Am I just out in left field? --Bill Davidson
mrd@ecs.soton.ac.uk (Mark Dobie) (07/26/90)
In <10273@celit.fps.com> billd@fps.com (Bill Davidson) writes: >This makes me worry about possible extra problems for gamma >correction. My understanding of gamma problems is that linear changes >in voltage do no give linear changes in brightness. A gamma curve >plots the changes of voltage vs. brightness. If one color of phosphor >is more efficient than another color, might this also change the gamma >curve? Should we really have three separate gamma values (one each for >red, green and blue)? This might help bring the color balance back a >bit closer to the proper value. This is my understanding also. >don't recall Roy Hall's book mentioning this. He does mention >correcting for a given color monitor but nothing about separate gamma >curves. Am I just out in left field? I think this is a reasonable conclusion (but I'm no expert). I have a program (posted here a while ago) called ppmgamma which gamma corrects a ppm file. It allows you to specify separate gamma values for r,g,b so I think you're right. Now a question: On a related topic, in Foley and van Dam (the old one, I'm afraid :-() they mention the logarithmic relationship between human's perception of brightness and actual brightness. My question is, is this really a hard and fast logarithmic relationship? or does this just fit the experimental data best. Are there a variety of models for this? or is there some (biological?) evidence pointing to the log relationship. Just curious. >--Bill Davidson -- Mark Dobie M.Dobie@uk.ac.soton.ecs (JANET) University of Southampton M.Dobie@ecs.soton.ac.uk (Bitnet)
billd@fps.com (Bill Davidson) (07/27/90)
In article <3312@ecs.soton.ac.uk> mrd@ecs.soton.ac.uk (Mark Dobie) writes: >In <10273@celit.fps.com> billd@fps.com (Bill Davidson) writes: >>This makes me worry about possible extra problems for gamma >>correction. My understanding of gamma problems is that linear changes >>in voltage do no give linear changes in brightness. A gamma curve >>plots the changes of voltage vs. brightness. If one color of phosphor >>is more efficient than another color, might this also change the gamma >>curve? Should we really have three separate gamma values (one each for >>red, green and blue)? This might help bring the color balance back a >>bit closer to the proper value. >>[more babbling deleted] >>Am I just out in left field? > >I think this is a reasonable conclusion (but I'm no expert). I have a >program (posted here a while ago) called ppmgamma which gamma corrects >a ppm file. It allows you to specify separate gamma values for r,g,b >so I think you're right. Well, that's a poor way to judge it because *I* wrote that program so any confusion I may have gets carried over into what I write. I'm glad to hear that somebody at least noticed it. I wasn't sure if no news was good news or if no one cared about gamma correction. I've wondered about this question of separate gamma values for some time but had never bothered to ask about it before. I figured I'd at least build the capability into the program. In any case, I received some mail from someone who at least sounds like he knows what he's talking about. Why he chose not to post I don't know. My confusion could cause damage if people start thinking along the same lines so I'm including it here (name deleted since I didn't ask permission but he works for a company which is well known for, among other things, working with many types of CRT's at a hardware level). light_output = K * volts**gamma. This power-law nonlinearity is in the grid-cathode-voltage-to-beam- current transfer characteristic, which in turn comes from the geometry of the CRT grids, cathodes, and Maxwell's equations. Beam current can be measured in electrons/sec. The phosphor efficiency is just a scaling constant (photons per electron) which happens after the nonlinearity (and so won't affect the exponent -- gamma -- of the power-law nonlinearity). The geometry of the guns and grids in the CRT might have asymmetries that could result in differences in gamma among the three channels. So, the need for three separate gamma curves can exist but it's apparently not caused by differences in phosphor efficiency. Ideally, a monitor should have a gamma value of 1.0. I'm a little suprised that I've never heard of gamma values being a marketing issue (as in Monitor Company A sez: "Our monitors have a lower gamma than Company B's monitors" and Company C sez: "...."). Certainly if I were in the market for a high res color monitor and all other things were equal, I'd buy the one with the lowest gamma value. Gamma correction may make up for losses in intensity but in doing so it sacrafices a number of intermediate intesities. --Bill Davidson