Originally posted by schunacher
I have been appreciating your posts especially because I have an HLN567W with 309 FW. I have the green push problem where all blonds look like they have been swimming in a chlorinated pool especially in darker scenes. I have had some success with the user menu by pushing up the blue using the color temp (COOL2) and removing green using tint (63/37). Of course this only works for the S-Video inputs and I have the problem on all inputs.
I would like to apply your tweaks but instead of just copying them in, I would like to understand them. I liked your explanation in your post about the effect of the gamma correction; very concise and at the right technical level so that I can comprehend what is actually happening. Would you be so kind as to give a similar level of explanation about the effect of R/G/&B offset inputs and their relationship to the R/G/&B gain controls? I'm sure that it would be appreciated by many members of the forum.
OK, I'll give it a shot:
We don't know the physical response of the mirrors in the DLP chip, but we do know that they are preceded by a Gamma lookup table that makes them APPEAR to have the control characteristics of the 3 grids of an RGB CRT, in digitized format. Once we have selected Gamma, the video vs. light response curve (Gamma 0 is most like a CRT on the HLNs), what remains is to process each of our three Primaries to apply to these "grids".
All of the controls have close CRT counterparts:
SubContrast is like a Master Screen Grid control, acting as a gain multiplier on all three colors;
SubBrightness is like a Master Bias, setting the cutoff voltage source from which the individual Biases will be derived, and hence moving all three together.
So how will each of the colors be processed to drive these digital pseudo-grids:
Well, you take the input and separate it into its RGB components, and you have a signal that in analog space would look like a graph of one color's level vs. time. Each Primary is now MULTIPLIED by its Gain value, where 128 can be inferred to be "1", and other levels are proportionally above or below. This is effectively a multiplier on the Y axis of our analog video vs. time graph. The next task is to decide where zero is -- at what level should our "CRT" reach cutoff? The offset controls provide this function -- they move the baseline to where we think zero should be; think of the signals as coming in, being multiplied by Gain, and then all three get 128 subtracted from their adjusted instantaneous value to normalize them, then each of the three get its own Offset value added. Zero for each channel, after this sum, is Cutoff for that Grid, where positive numbers are above Cutoff and negative below. We now have a final value for each of the three primaries, which will now drive the three "Grids" that have been processed by the Sub and Gamma selections to emulate a CRT with appropriate transfer characteristics. The numbers we enter are mostly arbitrary, so long as we stay in the range where the math will not over or underflow, and we don't run out of our weakest Primary:
Like a CRT one color will run out of steam first at the bright end -- this is a fundamental limit, unless by accident you could make a good, bright source that was exactly 6500K, so all three primaries would "hit the wall" together. But our DLP's are again like a CRT -- Red is shortest in supply. So if we fix the Red processing curve, say by arbitrarily assigning Gain and Offset values of 128 to Red, then there will come a point where there is simply no more Red to be had if SubContrast is further advanced, attempting to make the picture brighter. I have suggested we identify that point, and set SubContrast just below it, so that Red will be able to trace the Gamma curve we have selected, without truncating the top end (which would cause a sudden shift toward Blue/Green at the top end, when Red hit the wall). Since there is some small interaction with SubBrightness, we want to recheck this max Red condition when we are able to make a final setting of correct Black level.
Color Temperature adjustment will now consist of adjusting the Gain and Offset of the other two Primaries to track with Red, and produce the same White temp at all IRE levels. But note that each is non-linearly processed by Gamma, and each has one additional multiplier: the color output of the lamp at that Primary. Green, for instance is stronger that Red, Blue stronger yet (compared to what is required for 6500K). So when all three grid-like curves are set for the same Gain and Offset, they would start together, but Green would reach a higher level than Red, and Blue the highest. Since the curves are non-linear, as soon as we adjust the Green and Blue Gain to place the max levels at the right amount -- the curves mis-track at lower levels. So we must iteratively adjust both Gain and Offset for Green and Blue to stay as close to Red over the WHOLE range as the quirks of their curvature allow. And there is the subjective factor that variations in Blue are less critical, because they change Temperature more than Tint, which is less noticed, whereas variations in Red-Green balance are readily perceived as Tint shift.
After all of these machinations, I wound up the these settings on my HLN467W, with 309 Firmware:
Red Gain: 128
Green Gain: 113
Blue Gain: 97
Red Offset: 128
Green Offset: 130
Blue Offset: 134