Posts by Tobias

    Meanwhile I have no idea what's going with this orange on 8565. I'm pretty sure that the square counts are correct. I guess the designers started off with something that more resembles the ratios on the 6569 but then had to adjust because the resulting colors were off?

    Somehow I wish you just say that it's an error.:) But I also had a look on the die-shot and you are correct. Maybe just an individual error of that run of production?
    In so many ways, the corrected cos value would make "sense". Color amplitude is adjusted to the other colors' common value and also the color angle will be the same as obviously the NMOS chips have.
    Also if I use the corrected color angle (~128°), the comparison between "target" angles and measured values (vectorscope) looks so much more straight forward than when using the uncorrected orange value (~122°). Please see attachment.

    There is in fact a patent from CBM (US4489284) that describes this circuit, more or less. (Could not add a link for some reason.)

    Cool. Do you have an idea what it is for? Maybe to compensate the angle dependant ODD/EVEN color deviation (see 6569R5 diagram) and make the mix color much more fit the target angle? Also the snakes' color/burst ratio in HMOS-II seems similar like NMOS (~1.3), but the measured color amplitude of HMOS-II looks more like 1.0 of burst. So it's the aim of that magic part to do all this stabilizing?
    You also had mentioned that the difference in HMOS-II colors is that they are always switched on. I don't understand what's the difference/effect for this. I have no clue about electronics basically. Always switched on would eliminate that parasitic phase errors of the sin/cos parts?


    Could it be that the magic "stabilizing" circuit does something to the orange (because color angle somewhere near color burst angle?) so that they had to make this cos-snake-adjustment to finally get the target color angle and amplitude for orange? I just dont understand why just for the orange.:oob:

    Just want to give you feedback from your VC-20 6561 luminance checks.


    When using corner ratio of about 0.3, the curve shape looks resonable compared to the screenshot. I also included 2 exsiting VC-20 palettes for comparison.
    On the 2nd pic, you can see the range of curve shapes depending on the corner ratio.


    Next step for VC-20 would be to have a look on the colors.:D Maybe we could gain also some insights for NMOS/HMOS-II VIC-II.

    Sodele.

    Dank mathops Analyse des Die-Shots sind wir hoffentlich etwas schlauer, was die VC-20 Helligkeiten betrifft.


    Anhand der Länge der Widerstandsmäander ("Schlagen") im Chip kann man über Umwege auf die Widerstände schließen. Wie gesagt gibt es beim 6561 für jede Farbe (14) einen Wert (bzw. jeweils zwei Varianten davon) plus schwarz.

    (Im 9-stufigen C64 gibt es schwarz und 3 Widerstände, die miteinander kombiniert werden können. Beim 5-stufigen R1 gibt es auch schwarz plus 3 Widerstände, diese können jedoch nicht miteinander kombiniert werden.)

    Weiß ist, wenn kein Widerstand aktiv ist (offen="unendlicher" Widerstand). Mit einem externen Widerstand auf dem Board/Modulator zusammen ergbit das einen Spannungsteiler, der dann entsprechend für das Helligkeitssignal verantwortlich ist.


    Je nachdem, wieviel von den Schlangen-"Ecken" man bei dieser Zählweise als Widerstand berücksichtigt, gibt es leichte Unterschiede. Aus den Zählungen von Farb-Schlangen hat sich ein Anteil von etwa 30% als nicht unvernünftig erwiesen. Somit ergbit sich der lilafarbene Helligkeitsverlauf für den VC-20.

    Zusätzlich habe ich nochmal die Werte für die Colodore- und Mike-Paletten als Vergleich mit angeführt. Die vierte Kurve zeigt die etwas mutige "Analyse" des Bildschirmfotos der per Augenmaß geordneten Stufen. Ihr Verlauf passt aber grundsätzlich zu dem Ergebnissen aus der Chip-Analyse.

    Das zweite Bild zeigt, in welchen Bereichen ja nach Ecken-Anteil der Helligkeitsverlauf variieren könnte.

    Ich fände es aber interessant zu sehen, wie Spiele wären, die die Extrafeatures ausnutzen würden. Z.B. könnte man ganze Spiele einfach in ein (oder auch mehrere!) .REU File(s) packen oder man könnte schon bei der Entwicklung den Turbo-Mode einbeziehen. Oder das alles kombinieren... ich sehe Doom, ich sehe Rebel Assault :D

    Das Thema mit Spiele explizit für Turbo-Modi zu entwicklen hatten wir schon:
    Super CPU

    I was referring to the chroma bits. See attachment. It appears to be a pair of buffers/amplifiers but I have no idea how these work.

    So that's the 8565R2, right?

    But even if there is some external magic, that does not explain the strange orange, right?
    Therefore, could you please check again if the square counting of the cos (left) part of the orange is really correct?


    From seeing the plain die shots, for me it is still a miracle that you can destingish the snakes at all. I think it is very easy to misjudge some of those almost non-existing lines, overlaped and crossed by other layers.


    It's hard to believe that Commodore selected a different orange for the HMOS-II, especially with higher amplitude than all other colors.
    The fact that the color angle automatically "fits" to the NMOS orange when I correct the cos value to have an equal color amplitude is a strong hint that there is something wrong here.:search:

    There is also some analog magic going on to the right of the color resistors, that would also have an effect.

    An effect to what? Which part are you referring to? The chrominance snakes of 8565R2 VIC-II? Or the luminance snakes of 6561 VIC-I?

    No idea how it works though, and I really have been staring at it for a while. Maybe someone who actually knows what they are doing can chime in here. :)

    If you can't shine light on it, I guess chances are quite low.:cry:

    If I just slightly change the +cos snake of orange, I can reduce its ampluitde to an expected level and at the same time, the angle gets very similar to the one of the 6569R1 orange snake.
    Can you please check that area again please? Maybe it was hard to figure out on the die shot picture.:freak Another possibiity is that Commodore implemented an error by miscalculating that snake length.:guenni:

    I also added the color burst snake length that I actually would expect for the 8565R2 keeping the signal measurements in mind.

    I also checked your color snake counts of 8565R2.
    Orange here looks "strange" regarding angle and amplitude.
    Also I am wondering that the color burst has not the same vector length as the colors. Measurement results for 8565R2 show aound color burst "saturation" for all colors (ODD/EVEN same here).
    Commodore's purple target seems to be mis-set by default to 53° instead of 60.7°.

    mathop

    So this "Shade" switch to give a 2nd brightness (add some amount to an exisiting amount) to the base colors actually does not really exisit, right?

    Yes it does exist, I should have made that clear, sorry.

    Apart from black, which is special, there are seven 'snakes', one for each color. Each of the snakes is tapped either at the end for the bright version or somewhere in the middle for the dark variant, based on whether the 'bright' line is set. I have tried to illustrate this in the photo.

    It means bright total square counts are dark square counts plus additional bright square counts?:oob:

    Maybey the do similar thing in the color section to generate those two stages of color amplitude that they mention in the VIC-I documentation.

    Square counts below

    The resistance is ultimately determined by the number of squares minus the number of corners times some factor.

    I figured they designed the chip so that the burst l and burst r resistances would be the same. That means a corner counts as about .7 of a square.

    I checked it and for my understanding, corner squares only counts 0.3. (total counts-0.7*corners) Please check.
    Pictures are my and your square countings. Not much difference.

    Computed Y values, assuming black is Y=0 (0.3 V) and white = Y=1 (1 V), once with counting corner as half a square, and once with counting corner as a whole square (so the actual value is something in between.)

    How did you calculate it? Does the VC-20 also has an "external" resistor in the modulator, so the VIC resistors and the external resistor act as voltage dividers?

    mathop

    You are legend.:D Thanks.
    I will have a closer look on it. Due to the fact the the corner squares are actually round, a factor of ~0.78 would be reasonable for them. So reduce the total square counts by the number of corner squares * (1-0.7854).
    Maybe red and blue are on the same level. Darker red than blue would be strange.


    if we round the corners, that would give each corner a factor of 0.7854.

    I am wondering how the color snakes' section (color angles and amplitudes) will look like if the VIC-I luminance sections alrewdy is that different.

    The doc says that the VIC-I has two color amplitudes. From the schematic, it lookls like to be around 1.4 / 0.9 of color burst. I assume that the light colors have lower color amplitudes not to clip the composite signal.


    Btw. how do you identify which snake stands for which color?

    Haaaaalt. STOPPPP!
    Alles auf Null. :D

    Der Blick von mathop auf den Chip hat hervorgebracht, was ich fast "befürchtet" habe:

    Der VIC-II war laut Wiki eine Neuentwicklung, weil die damaligen Entwickler den VIC-I nicht gut genug kannten, um ihn zu verbessern. Da gab es wohl schon zwei gescheiterte Versuche.
    Daher fraglich, in wieweit man das Wissen über den VIC-II auf den VIC-I übertragen kann. Vielleicht hat hier tatsächlich fast jede Farbe eine eigene Helligkeit.:gruebel:nixwiss:

    Jede Farbe hat hier tatsächlich eine eigene Helligkeit(sschlange). Werde ich zeitnah analyiseren.
    Jetzt haben wir uns soviele schöne Gedanken zu den möglichen Helligkeitsstufen gemacht und dann das....:whistling::D
    Folglich sind die Trial-Paletten erstmal alle für die Tonne, so schön manche auch ausgesehen haben. Die Stufen waren zwar schön, aber künstlich generiert. War dieser komische Bildschirmhelligkeitsverlauf doch nicht nur so wegen unsauberem Signal o.ä.

    Nach dieser Erkenntnis wage ich auch keine Farbtonschätzung (auf Basis des VIC-II) mehr ohne Messung oder Blick auf den Chip.:search: