Hard to set the colour levels with the 470pF. Chroma is being attenuated by about 50% after the cap which means 100% cyan now needs a chroma peak to peak swing of about 1.77V instead of 0.855V and the DAC can only do 1V. I might have to try the other DACs settings for Chroma to have it drive a larger range but then the source impedance is increased as well...
@Ltech said:
I think the only part missing now is increase chroma, the U an V by 10%
U= R-Y
V= B-Y
Can we only change R an B +10%, no the green part decrease.
Do we have to decrease G 5% and increase R&B 5% ?
I have no time to tweak it now.
My experience was never touch Green while color grading, it change the gain/luma of the picture at the end.
Sorry, have been sidetracked on the capacitor thing. I think it is hard to boost the chroma because of the limited DAC headroom. We can only safely drive colours to 75 percent if we want proper 0.3V sync levels and correct peak white at 0.7V above blank. If we boost Chroma it will exceed this for some colours and the DAC underflows in the colour space converter and messes things up.
We have a few solutions (for PAL, not looked at NTSC yet, but it's likely similar):
1) stick to Chip's approach of compressing the range down and live with the mismatch of levels vs the standard
2) reduce the colour burst level by 1.33 (100/75) to mean that 100% saturated colours are now represented by what 75% colour level voltages normally are (Wuerfel_21 approach). This should not exceed the DAC but would mean to match the 75% reference standard to the tester you'd need to feed 100% colour bars and the colour on screen would be far more saturated. Other than that discrepancy that it might look okay, but Luma and chroma would be mismatching. Unsaturated colours might appear a bit brighter than normal perhaps?
3) capacitively couple Luma to Chroma so we can drive over 1V. Not a perfect output and a little soft but colours are vibrant. Still investigating.
4) use S-Video primarily - with totally correct levels for both Luma and Chroma. As someone with a S-video capable device this is my personal preference, but there are people who would only have composite available and would want a solution for that.
Oh, didn't notice this thread got posted to until now
@rogloh said:
We have a few solutions (for PAL, not looked at NTSC yet, but it's likely similar):
1) stick to Chip's approach of compressing the range down and live with the mismatch of levels vs the standard
2) reduce the colour burst level by 1.33 (100/75) to mean that 100% saturated colours are now represented by what 75% colour level voltages normally are (Wuerfel_21 approach). This should not exceed the DAC but would mean to match the 75% reference standard to the tester you'd need to feed 100% colour bars and the colour on screen would be far more saturated. Other than that discrepancy that it might look okay, but Luma and chroma would be mismatching. Unsaturated colours might appear a bit brighter than normal perhaps?
3) capacitively couple Luma to Chroma so we can drive over 1V. Not a perfect output and a little soft but colours are vibrant. Still investigating.
4) use S-Video primarily - with totally correct levels for both Luma and Chroma. As someone with a S-video capable device this is my personal preference, but there are people who would only have composite available and would want a solution for that.
Chip's current approach results in white-clipped picture on my capture card (moreso on the S-Video input than composite).
I think the lowered chroma level approach is much better. It works fine on everything I've tried. And if a display really doesn't like it, you can just turn up the saturation knob to compensate.
Externally coupling the chroma is great, too, but while you're at it you really want to put a chroma trap filter on there to get rid of artifact colors. The datasheet for the AD725 color encoder has a circuit that can alledgedly filter either 3,58 or 4.43 MHz, but I think an opamp is needed (two for simultaneous S-video out).
Yeah the Y/C is the way to go to avoid this problem. A week ago or so I tried to boost the colour output levels of the chroma to match 100% colour bar saturation levels but have had problems getting the DAC to go that high. Something is stopping it from reaching the correct levels. Cyan and Red I think are meant to be 0.885Vp-p when fully 100% saturated but I couldn't get the DAC output to go that high when terminated with the 75 ohms. I was not sure what what I was doing wrong at the time and have since moved to other stuff but it would be good to try to get it right. There's scaling upon scaling going on and there might be a sqrt(2) factor needed here too. I think perhaps only one chroma component axis is being sent not the summation of both Usin(wt) and Vcos(wt) which may result in some further scaling or something....
Comments
Hard to set the colour levels with the 470pF. Chroma is being attenuated by about 50% after the cap which means 100% cyan now needs a chroma peak to peak swing of about 1.77V instead of 0.855V and the DAC can only do 1V. I might have to try the other DACs settings for Chroma to have it drive a larger range but then the source impedance is increased as well...
I think the only part missing now is increase chroma, the U an V by 10%
U= R-Y
V= B-Y
I have no time to tweak it now.
My experience was never touch Green while color grading, it change the gain/luma of the picture at the end.
Sorry, have been sidetracked on the capacitor thing. I think it is hard to boost the chroma because of the limited DAC headroom. We can only safely drive colours to 75 percent if we want proper 0.3V sync levels and correct peak white at 0.7V above blank. If we boost Chroma it will exceed this for some colours and the DAC underflows in the colour space converter and messes things up.
We have a few solutions (for PAL, not looked at NTSC yet, but it's likely similar):
1) stick to Chip's approach of compressing the range down and live with the mismatch of levels vs the standard
2) reduce the colour burst level by 1.33 (100/75) to mean that 100% saturated colours are now represented by what 75% colour level voltages normally are (Wuerfel_21 approach). This should not exceed the DAC but would mean to match the 75% reference standard to the tester you'd need to feed 100% colour bars and the colour on screen would be far more saturated. Other than that discrepancy that it might look okay, but Luma and chroma would be mismatching. Unsaturated colours might appear a bit brighter than normal perhaps?
3) capacitively couple Luma to Chroma so we can drive over 1V. Not a perfect output and a little soft but colours are vibrant. Still investigating.
4) use S-Video primarily - with totally correct levels for both Luma and Chroma. As someone with a S-video capable device this is my personal preference, but there are people who would only have composite available and would want a solution for that.
Oh, didn't notice this thread got posted to until now
Chip's current approach results in white-clipped picture on my capture card (moreso on the S-Video input than composite).
I think the lowered chroma level approach is much better. It works fine on everything I've tried. And if a display really doesn't like it, you can just turn up the saturation knob to compensate.
Externally coupling the chroma is great, too, but while you're at it you really want to put a chroma trap filter on there to get rid of artifact colors. The datasheet for the AD725 color encoder has a circuit that can alledgedly filter either 3,58 or 4.43 MHz, but I think an opamp is needed (two for simultaneous S-video out).
Yeah the Y/C is the way to go to avoid this problem. A week ago or so I tried to boost the colour output levels of the chroma to match 100% colour bar saturation levels but have had problems getting the DAC to go that high. Something is stopping it from reaching the correct levels. Cyan and Red I think are meant to be 0.885Vp-p when fully 100% saturated but I couldn't get the DAC output to go that high when terminated with the 75 ohms. I was not sure what what I was doing wrong at the time and have since moved to other stuff but it would be good to try to get it right. There's scaling upon scaling going on and there might be a sqrt(2) factor needed here too. I think perhaps only one chroma component axis is being sent not the summation of both Usin(wt) and Vcos(wt) which may result in some further scaling or something....