Helligkeiten der C64 PAL-Palette - zum 100. Mal

Zitat von mathop

6561 color resistors: (white bits)

Thanks mate!

On the first view, coming from VIC-II, the color amplitudes look strange. In other words: you can clearly see that VIC-I and VIC-II had different designers.

Btw.: one color here seems to be "special" again. Either by purpose or by mistake. This time it's not orange, it's red. When I change the cos value to get the target color angle value, also the red's color amplitude looks more reasonable compared to the other colors (esp. cyan which should have the same color amp.).

Zitat von mathop

Amplitudes of bright colors should be smaller than those of non-bright variants and color burst.

The color amplitudes are not constant in the VIC-I, their amounts somehow follow the theory (see EBU color bars).
If I add the (compared to the referenced color burst weak) color amplitudes to the light colors' Y-signals, the max. signal levels are still within the tolerance (before clipping).
For the dark colors, you could double the color amplitudes and still are below the signal levels of the light colors.

Is there a hint on the die-shot, how and how much the color amplitudes are switched?

Having seen the luminance resistors of VIC-I, actually I expected separate color resistor pairs for dark and light colors. But maybe it makes sense to have just one set and amplifiy the signal afterwards. So a potential phase/amplitude error is same for both the dark/light color.

For the real amount of color amplitudes, I think we can only do much guesswork without having any measurement data.
Also there might be the addiitonal parasitic phase(/amplitude) shift, different for ODD/EVEN. Again, just guessing.

Zitat von mathop

I don't have a VIC-20, so I can't do measurements myself.

In case someone would send you a VIC-20, could you (do you want to) measure it?

I think I have the answer to the question why the Commodore color order is like it is (impossible to judge from the VIC-II colors I would say):

The sequence of pairs are (were) the ones with the same color amplitude in VIC-I, similar like for the EBU color bar. In the example attached, I corrected the cos value of red btw. (angle and amplitude "fit" this way).

The color amplitudes in the die-shot seems to be around 1/3rd of the 75% EBU color bars' colors. I used it for the light colors. For the dark colors, I used double the color amplitudes.

I think this might be the closest palette approach that we can deliver without further measurements. It is kind of die-shot-palette.

EDIT: I forgot to mention that by using the estimated luminance levels, color angles and amplitudes, none of the calculated RGB colors was out of gammut.

Zitat von mathop

I somehow managed to get a VC20 so I now have one of those as well, except with no power supply. So that will have to wait until I find a way to power it.

That sounds promising. Maybe someone can help you out by temporarily providing a VC20 power supply? Do you have a device to load a program to display the colors?

I am really excited to see how the color angles will be, if/how the offset between ODD/EVEN is and especially which value the two different color amplitudes will have for each color. The luminance might be almost as we predicted with some minor deviation. Maybe some colors might be on exactly the same level which we calculated as very slightly different.

We have to keep in mind that at least two revisions of the 6561 do exisit. 6561E and 6561-101. 1st (older?) one also as ceramic version.
The die-shot you analyzed was 6561E I think.

I will be on vacation for one week. So if there is no feedback from my side in this period, you know why.

Zitat von mathop

I analyzed the 6561-101 die shot;

Didn't you use the one from here?
http://visual6502.org/images/p…S_6561_VIC_die_shots.html

That one is 6561E.

Zitat von mathop

I'm getting a bench power supply for other projects, hopefully I can power from that. (There's a VC20 PSU selling on ebay for 119.95- that's insane.)

I just made an offer for one VC20 incl. PSU.
If seller accepts, I could send you both and you can use them during the measurements.

Zitat von mathop

I managed to power up the VIC-20. Had to adjust quite a bit to get the luma voltage to a decent level.

Some video captures:

I did two passes, one with a mostly white display, and one with a mostly black

(See attachments)

Alles anzeigen

Thanks for the measurements.

I analyzed about the colors and it seems that there is a VERY big phase shift between ODD and EVEN lines. Especially for the light colors. As I dont know for sure what is ODD and EVEN, I just called them A and B at the moment.

The mixed colors of A and B have the same angle deviation for dark and light colors.

Reference angle and amplitude always is what we got thru the resistor snake readings.

Btw. which VIC-I revision and board assy are you analysing here? Maybe we should mention it not to mess up in the future.

Zitat von mathop

VIC chip is 6561-101 - I did not check the board revision, but it is the old 9VAC board

I also receiced my ordered VC20. It is the newer board with the C64 power supply. I cannot check its condition at the moment because I don't have a fitting video cable. I could send it to you if you want to confirm other VC20 outputs. It also has the 6561-101 chip, I also could not see a board rev.# mentionend.
But maybe I should check it before not to send you a box of junk.

Zitat von mathop

the old 9VAC board with 2 trimmer pots at R7 and R32.

Oh. Basically a good thing to allow adjustments, but one we have to keep an eye on when measuring the signals. You knows whoever turned the screws...

https://atariage.com/forums/to…pro-tip-adjust-your-pots/

http://sleepingelephant.com/de…=Video_signal_adjustments

My board has just one trimmer pot and the red C48 crystal thing.

Zitat von mathop

I realized just now that the color saturation in #228 is off because the color burst signal is not the proper 300mV p-p. Although if you auto-adjust most colors become too saturated. Unfortunately I don't have time to do more measurements until this weekend.

Don't worry, no rush. It wasn't solved for 40 years, so some days or weeks don't count.

I averaged the Vp-p values of your two measurements (there is quite some value difference between them, but "shape" is consitent), devided them by 2 (->amplitude) and referenced this to 0.700V (assuming color burst amplitude was 0.150V). So I get the dimensionless color amplitude values (vector length of U/V). That is what is show in my graphs, not voltage values (sorry, I am often lazy in defining the axis).
So when I reference the measured values to an adjusted value, I could allign the color burst signal as it would have been properly set to 0.300V p-p. The above calculated color amplitudes seem to be a bit too high, which means the actual color burst was a bit too low during the measurement. Assuming that the measured burst was around 0.247V p-p, the "chip" color amplitude curve (from the snakes) quite nicely fits the line B (ODD or EVEN?). Both for the dark and light colors.
Btw. I am using the corrected snake value for the red cosinus, because angle and amplitude makes so much more sense with that intervention. Also that way, the chip ("target") amplitude curve nicely fits the "reasonable" amplitude shape, based on the EBU color bars. I still think that Commodore/MOS based the VIC-I's color individual "saturations" on those PAL-RGB-YUV basics.

I am still somehow shocked that the ODD/EVEN color angle difference seem to be up to 50° for the light colors. This looks more like color guessing than color generating.

Until further data, I guessed the luminance levels from your oscilloscope screenshots.

Also I analyze the RGB palette values from your capture card measurements.

For the moment, both fit each other not so bad I would say.

On luminance level 5 and 6, the assumed colors have to change places I guess.

Based on die-shot analyze, measurements, arranged screenshots, feedback and PAL fundamentals, attached luminance curve shape and color order is the best iteration step I have to offer at the moment.

Cool. I will have a look on it hopefully during the week.

Last week a played a bit with your previous measurments and summarized the results as below. Basically all data that is needed to create the 3 specific VIC-I palettes (ODD/EVEN (or maybe vice versa) and the MIX) is there.
After checking your latest data, we can adjust if necessary.

After that, we could check the die-shots if there is some difference in the color section 6561-101 ("R0") vs. 6561E. You said lumiance sections look identical?

Some VIC20 brightness levels are quite close to each other and the composite output doesn't make it easier to define the exact luminance levels (I used the center of color peak-to-peak). Do you see a chance to measure luminances directly at chip output pin like you did at the very start of this thread?

Zitat von mathop

Measuring directly at the chip has no effect for this particular board; the color signal is still mixed in.

Oh ok. Because there is some S-Video mod for the VIC20, I was just assuming that also the VIC-I has Luma/Sync and Chroma outputs separated.