Capture card readout attached
Hallo Besucher, der Thread wurde 52k mal aufgerufen und enthält 322 Antworten
letzter Beitrag von Zibri_ am
Helligkeiten der C64 PAL-Palette - zum 100. Mal
- Tobias
- Erledigt
-
-
The color signals (including color burst) are created by mixing a combination of sin (0), cos (90), -sin (-90) and -cos (180) waveforms.
So, for example, to generate blue, you mix a little of -90 degrees with a lot of 0 degrees. This is done using a resistive divider. (See picture.)
The amplitude of the mixed signal ranges from the input amplitude divided by sqrt(2), if the resistors are equal, to about the input amplitude if one resistor is significantly smaller than the other.On the chip the resistors appear to be implemented by using transistors of some kind.
On a 6561 the color resistors are turned off when a color is not selected.
On a 8565 the color resistors are always on, and there are separate gates that pass a particular mixed signal, based on the effective color.
I don't know why the amplitudes in the 8565R2 are all equal. Looks like the color signal is amplified inside the chip, maybe some clipping occurs there.
-
Also from the 8565R2 palette, there seems to be a smooter color amplitude. The 240°-drop for the green is caused by the light green which has a lower amplitude (curves are averages of both dark and light colors).
The average color amplitude is around 0.213ish. Slightly lower than the 0.225ish of the 6569R5. (brown is around 0.160ish, same as odd/even-avg of 6569R5)
If one devides the color burst 0.300V peak-to-peak by sqrt(2), that would result in 0.212. So maybe that is the factor to calcualte from the color signal amplitude to the color vector length (=sqrt(U²+V²)?
I also checked the palette's color angles. Bascially, the angles of your previous post are matched by the calculation, but:
Brown in the palette does not do this "overshoot". In the angle readings, brown is about 147°. But if you analyze the palette, it is pretty close to the target color angle of ~140°.
That's a bit strange because I assume that both angle and palette results come from the capture card.
Any idea? -
The brown color is truncated in the ycbcr-to-rgb conversion (indicated by the X at the end of the line)
You can also see this in the RGB value (e.g. 5e3d00) where the blue component is exactly zero.
The angle is calculated from the cb and cr values, which come directly from the capture card (after some averaging)
So that would explain the 140 / 147 difference.
Something similar applies to light green but there the difference in angles is not so large.
-
May I add that the resulting saturation is not strictly equivalent to the color amplitude but relative to the burst amplitude and for a given luminance...and also the result of the subcarrier deviation of adjacent lines ?!
Visually, a composite vectorscope display including a burst measurement bar will deliver the information in the best understandable way, IMHO.
-
It means that brown (and light green) also might have the same color amplitude like the others, but it is out of (s)RGB gammut and therefore the capture card reduces the values until blue value gets zero?
Hypotetically, brown would need a blue value lower than 0, the 8565R2's light green a higher green value than 255?In this case, no chance to 100% transfer the C64-PAL-brown into sRGB, correct?
But as brown is quite dark and light green is quite light, we could live with that color amplitude conversion error for the palette, right?
To me as a layman, a very dark or very light color cannot be perceived very "saturated" anyway. -
The brown color is truncated in the ycbcr-to-rgb conversion (indicated by the X at the end of the line)
the ycbcr values are according to BT.601? So 16 "is" a value of 0, 235 a value of 1 for Y? I think Pb and Pr values are from -0.5..+0.5. Which 8-bit values correspond to that range? 16 is -0.5 and 235(240) is +0.5? To calculate backwards from YCbCr to analogue YPbPr.
Just asking because we could calcualte into (Y)UV from your (y)cbcr values to check the color amplitudes un-truncated by the (s)RGB conversion.What does it help for the (s)RGB palette? Nothing. Just for better understanding of the origin and peace of mind.
-
You can reduce the overall saturation until brown becomes a valid color. I did this when I measured my boards originally. The side effect then is that all colors look kind-of washed out. Like I said earlier I believe, this might be aesthetically pleasing but is probably not how it was designed.
In the capture posted above I had adjusted the saturation so that the vpp value for each color would be roughly equal to the color burst (0.3V).
-
In the capture posted above I had adjusted the saturation so that the vpp value for each color would be roughly equal to the color burst (0.3V).
Where do you adjust the saturation? At the inputs of the capture card?
To me as a layman, my understanding is: just hook up the C64 video out to the capture card in and see what the capture card does.
-
The card has a tunable saturation, contrast and brightness, among others. The brightness I leave alone (black is black
). I find that I always have to adjust the contrast because the picture is either too bright (8565R2) or too dark (6561R3). Nowadays I tend to adjust the contrast in such a way that grey (color 12) ends up about halfway between black and white. After the contrast adjustment I then tune the saturation. The card also has a chroma AGC flag (so that saturation is automatically set based on color burst amplitude), but I don't know if it really works. So I always tune things manually. -
Happy to see your 8565R2 measurements to have further data
There you are. Both measured on the same boards as before so far.
-
There you are. Both measured on the same boards as before so far.
Where is "there"? Can't see anything attached.
EDIT: Sorry, it's in the old link, I found it. Thanks! I will check it.
-
Just a basic question:
Separate video luminace level ranges from (0) 300-1000 mVand
Composite video luminance (part) level ranges from (-300) 0-700 mV
?
I know both from the web, but nowhere there is a clear definition.
Thanks.
-
-
Just a basic question:
Separate video luminace level ranges from (0) 300-1000 mVand
Composite video luminance (part) level ranges from (-300) 0-700 mV
?
I know both from the web, but nowhere there is a clear definition.
Thanks.I am referring also to that:
1st measurements of silverdr and latest measurements of silverdr
-
I found it
Good!
These should still be mostly self-explanatory even if a bit more complex than plain luma but if any explanation's needed - don't hesitate to ask. -
I found it
Good!
These should still be mostly self-explanatory even if a bit more complex than plain luma but if any explanation's needed - don't hesitate to ask.Just hacked the 6569R5 values into the table.
Even I can understand it. 
if you just could shed some light on the different chrominance levels of your 1st measurements and todays please.
See diagrams above. -
Just a basic question:
Separate video luminace level ranges from (0) 300-1000 mVand
Composite video luminance (part) level ranges from (-300) 0-700 mV
?
I know both from the web, but nowhere there is a clear definition.
Thanks.Erm... no. The 1000mV range is the so-called p-p voltage The 700mV is specific to video signal and refers to (roughly) 100 IRE units. IOW if you take the -300mV to 700mV you will get the same p-p (peak to peak). Now this might be a bit confusing as I named the luminance measurement files with p-p numbers. To have a correct relation between the two measurements, the luminance measurements should be used with displayed difference between the upper peak and the so-called blanking level, which is where the lower measurement cursor is located. Please also note that you may not and probably won't get the very exact same numbers when plain luminance (or chrominance) signal is measured and when the composite video is measured as the combining circuit (the modulator unit in case of the 64) will leave its mark on the outcome.
P. S. I just noticed that in case of the black colour I didn't put the upper cursor where it should be but it is clearly visible that there's not really any difference between the blanking level and the black level, which is correct in case of PAL signal. I'll try to update this file anyway soon.
