Helligkeiten der C64 PAL-Palette - zum 100. Mal

Es gibt 322 Antworten in diesem Thema, welches 56.672 mal aufgerufen wurde. Der letzte Beitrag (23. November 2024 um 18:01) ist von Zibri_.

  • That's interesting. Thanks for the measurements.
    So the amount of the interference checkboard pattern could be a hint to the sine-like-quality of the signal?

    You can clearly see that the 8565R2 has the most evenly color allocation. We know that also from previous measurements. Esp. the lead color :D brown has no difference here for odd/even.

    The term under the chip rev. is the modulator "revision"?
    So from left to right ist is board-1, board-2..., right?

    The VIC-II chip does not output true sine waves. The newer modulators have a low-pass filter in the chroma input that resonates at the color carrier frequency. That will make the color signal look more like a sine wave. As far as I can tell the original modulator did not have this filter.

    One theory could be that the resonance in the filter also increases the output voltage.

    Could you capture the sine signals both in shape and absolute value for the board-1 and board-5? Each color burst and lets say green color amplitude. Each odd/even. If you want, each composite and chroma output. ;)

  • Those are indeed board1,2,3,4,5.

    The text refers to the modulator version. Sorry, I should have made that clear.

    There are raw exports of scope data Bitte melde dich an, um diesen Link zu sehen.. Each csv covers two lines (so, one is even, one is odd, but you need to look at the color data to see which is which). Channel 1 is luma/composite, channel 2 is chroma.

    The input picture used was the color bars that I pasted above. It displays colors 0 to 15 for 16 pixels on a grey background.

    You need to do your own processing. :)

  • There are raw exports of scope data Bitte melde dich an, um diesen Link zu sehen.. Each csv covers two lines (so, one is even, one is odd, but you need to look at the color data to see which is which). Channel 1 is luma/composite, channel 2 is chroma.

    The input picture used was the color bars that I pasted above. It displays colors 0 to 15 for 16 pixels on a grey background.

    You need to do your own processing. :)

    ...-hot-ac.csv has Luma/Chroma
    and

    ...-hot-ac-comp.csv has Composite?

    My collor allocation is correct? I assume you used a med. grey border?

  • The Chroma on the R1 board looks really bad. Zero is not at zero and changes and also the sine swing is not around zero. To me, it's a miracle that the TV can disyplay anything out of this signal.:whistling:

    So either

    a) your R1 board/modulator has some issue

    or

    b) that's one of the reasons why R1 was revised very quickly. Maybe the attemp 6569R2 also failed and that is why we only know the R3+.

    Without having analyzed your data in detail, I think we can say that the color amplitude somehow seems to be intended to have the color burst amplitude. Can you agree on that?

  • The 6561R1 actually looks quite nice here on my sony lcd tv with s-video. The sine does not swing about zero, but it doesn't have to. The color signal is multplied by the color carrier by the color decoder, that cancels out the average value. Otherwise composite video would not work at all.

  • The 6561R1 actually looks quite nice here on my sony lcd tv with s-video. The sine does not swing about zero, but it doesn't have to. The color signal is multplied by the color carrier by the color decoder, that cancels out the average value. Otherwise composite video would not work at all.

    Just wanted to say that the signal looks weird to me.:rolleyes:
    For my understanding, technically such non-symmetrical sine behaviour should at least be more critical to clipping, not?

  • The color amplitudes on the 6569R1 are

    Code
           amplitude angle(approx)
    burst-1    0.24  126
    burst-2    0.22 -126

    So the color burst angles here are 126° and 234°.
    According to PAL, it should be 135° and 225°.

    Any need to adjust the color angles accordingly?

  • Ideally the color burst on one line should be 90 degrees ahead or behind of the burst on the other. But apparently for the c64 signal this isn't exactly true. So the difference in burst phase in that particular example is more like 108 degrees. Hence the 126/-126 thing. Edit: or I am talking nonsense and it's just that my measurement is wrong. I did point out that the values were approximations. :)

  • I hacked your 6569R1 measurements of post Bitte melde dich an, um diesen Link zu sehen. and Bitte melde dich an, um diesen Link zu sehen. and compared them with your results of post Bitte melde dich an, um diesen Link zu sehen..

    The newer measurements show color angles with an offset to minus for the "negative" line (I think that's the even line).
    What was the difference of those measurements?

    The older results pretty much fit the silverdr's results for the 6569R5.

  • The angles at Bitte melde dich an, um diesen Link zu sehen. ware measured with a capture card, not with my crappy python code. So I would trust those over Bitte melde dich an, um diesen Link zu sehen./Bitte melde dich an, um diesen Link zu sehen., at least as far as angles are concerned. I believe/hope the amplitudes are more accurate.

  • The angles at Bitte melde dich an, um diesen Link zu sehen. ware measured with a capture card, not with my crappy python code. So I would trust those over Bitte melde dich an, um diesen Link zu sehen./Bitte melde dich an, um diesen Link zu sehen., at least as far as angles are concerned. I believe/hope the amplitudes are more accurate.

    Ok, I understand. Maybe the script just caught the wrong reference for the EVEN lines. I added +6° here and with that amount, it looks quite ok.;)

  • Here is the result of post #150.
    The ratios of the luminances stay almost same for black and white border.


    But since the sync-black voltage also gets lower, I don't think this will affect the picture on a monitor.


    For example, if you adjust output voltages so that sync-black is at .3V you get something like this (vertical resolution is 0.004V)

  • Al little bit off-topic.

    Just found again that site:

    Bitte melde dich an, um diesen Link zu sehen.

    No source mentioned, but the UV values are totally out of PAL spec.

    U can be maximum +-0.436

    V can be maximum +-0.615

    Attached the C64 values.

    It i salso mentioned that the VIC-1 also had variants with 5 and 9 luminances. I can't imagine that as the 1st VIC-2 still had just 5 luminances.

    What do you think?

  • So absolute luminance informations is not helpful here

    Jein :wink: If we want to e. g. be more accurate with the palettes assigned to specific VIC chip variant then I see it rather important for example not to use #FFF for the one that generates darker "white" output. True, it would require broader sample base to fully confirm the findings but I wouldn't discard this part of the information so easily. And in the meantime I can possibly check about a dozen of different HMOSII boards to confirm or reject the thesis.

    A picture tells more than 1000 words.8)

    Just check the different "black levels" of the picture tubes. 1802 and a Sony 21"
    That's why I think that the absolute luminance levels of each individual C64 are not needed, just the relative levels.;)

  • Maybe the "manual" analyzing is not too bad. Time consuming and boring, but not bad concerning the results. ;)

    Attached the result from board Bitte melde dich an, um diesen Link zu sehen. separate video's oszi data.

    It seems that the color amps don't refer each line to the color burst, more kind of average of the color bursts.
    At least with this calcualtion, the result looks very similar to all previous color amlpitude sequels we obtained by different measurement methods like capture card or vectorscope.

  • I'm done with the color amps from the sep. video recordings. (don't feel to analyze the Copperdragon deviated signals atm :D)

    The NMOS chips' color amplitude swings around about 1.2x color burst.

    I will create one curve for all NMOS chips because

    a) the differences are not that big

    b) we don't know the root cause of the differences atm (measurement tolerance, chip lot, board, modulator)

    For HMOS-II color amplitude, I will use a constant value of 1.0x color burst.

    The swing shape is similar to the NMOS, but muuuuuch smaller.

    I averaged each color burst by 3 values, 1st values is always a bit overshooting. If I leave this 1st value out, the color burst as reference gets slightly smaller and so all colors will have a ratio of ~1.0

    Long story short: The 8565R2 palette should be a little less "vivid" than for the 6569Rxs. Can someone confirm this from his/her/its own experience?

  • It's been a while, I have since come up with a theory that might explain the phase differences.

    The VIC chip generates four sine-like signals internally, at 0, 90, 180 and 270 degrees. ('sin', 'cos', '-sin', '-cos') On even lines (for PAL), the cos and -cos signals are inverted. For each color sin or -sin is mixed with cos or -cos in some proportion. (See picture.) This is done using resistors. Due to parasitic capacitance the input signals will be attenuated and phase shifted depending on the size of the resistor: the larger the resistor, the more the signal will be attenuated and phase-shifted.

    Bitte melde dich an, um diesen Anhang zu sehen.

    Now for the 6569 there is a trick we can do. We can take the measured u and v vales on odd and even lines and use these to reconstruct the cos and sin signals before they are mixed. This is based on the observation that if you have two vectors, A and B, then define C=(A+B)/2 and D=(A-B)/2, then A=C+D and B=C-D. This does not work for the 8565 because there the color amplitudes are all the same, if the inputs were mixed properly the amplitudes would vary depending on the angle.

    Bitte melde dich an, um diesen Anhang zu sehen.

    See the graph above. Here black represents the 'sin' part of the color, and green/red the 'cos/-cos' part. Then the color itself is represented by the blue vector, and orange for the inverted phase. (Apologies to the colorblind.) If you take the color red, for example, the 'sin' input is very small, and also phase shifted to the right. This corresponds to the large resistor that is used for the 'sin' portion of the red color. On the other hand the 'cos' input is large and shifted to the left. This corresponds to the small 'cos' resistor.

    There is an additional subtlety here in that the color burst itself is also phase shifted - that's why the measured signal all appear phase-shifted to the left, since the corresponding resistors are smaller than the color burst resistor.

    Hope this makes some sense to someone. :)

  • Welcome back, Mr. Snake. ;)

    I understand in principle, but not fully understood everything.:D
    Can one say that the difference (unsymmetricy) is because only cos and -cos are inverted, but sin and -sin stay same? And sin and -sin are not "straight" on the x-axis.

    Do these findings fit to what the measurments say in quality and quantity?

    For the average color vector length (=sgrt(U²+V²) of NMOS VIC-IIs, you can assume around 0.258. (=1.2*color burst; = 1.2*0.215)

    In other treads, there are again discussions about "the" C64 colors. So maybe we should finish our stuff here soon.:saint: