Home Pictures (PAL) Pictures (NTSC) Pictures (NTSC CXA2025)

accurately reproducing the Video Output of a Commodore C64

Considerable Parameters


Video System


Video Timing

frame freq ~50 Hz ~60 Hz
gamma 2,8 2,5
line freq 15625 Hz 15734 Hz
video bandwidth 5.0 MHz 4.2 MHz
color bandwidth I: 1.3 MHz, Q: 0.4 MHz
color subcarrier 4.43361875 MHz3.57954545 MHz
VIC-II pixel clock7.882 MHz 8,18184 MHz
VIC-II color clock17.734472 MHz 14.31818 MHz
C64 CPU clock 0.985248 MHz 1.02273 MHz
C64 Cycles per line 63 65
VIC-II total lines 312 263

Color Spaces

  • Y - Luminance ("Brightness")
  • U - Color Difference
  • V - Color Difference
  • YCbCr

    RGB to YCbCr YCbCr to RGB
     Y  = 0.2989 * R + 0.5866 * G + 0.1145 * B
     Cb = B - Y
     Cr = R - Y
     G = Y - (0.1145 / 0.5866) * Cb - (0.2989 / 0.5866) * Cr
     B = Cb + Y
     R = Cr + Y
    YCbCr to YUV
      U = 0.493111*Cb
      V = 0.877283*Cr

    YUV (PAL)

    YUV (CCIR 601) is the PAL colorspace. It's a slightly modified YCbCr for better broadcasting.
    RGB to YUV YUV to RGB
     Y =  0.299 * R + 0.587 * G + 0.114 * B
     U = -0.147 * R - 0.289 * G + 0.436 * B   (B-Y;Cr)
     V =  0.615 * R - 0.515 * G - 0.100 * B   (R-Y;Cb)
     R = Y + 0.000 * U + 1.140 * V
     G = Y - 0.396 * U - 0.581 *  V
     B = Y + 2.029 * U + 0.000 * V

    YIQ (NTSC)

    NTSC uses the YIQ colorspace. Compared to YUV, YIQ is rotated by 33 degr clockwise, which results in different bandwidth for the color components (I: 1300 kHz, Q: 400 kHz)
    RGB to YIQ YIQ to RGB
     Y =  0.299 * R + 0.587 * G + 0.114 * B
     I =  0.596 * R - 0.274 * G + 0.322 * B
     Q =  0.212 * R - 0.523 * G - 0.311 * B
     R = Y + 0.956 * I + 0.621 * Q
     G = Y - 0.272 * I - 0.647 * Q
     B = Y - 1.105 * I + 1.702 * Q
    YIV -> RGB (Sony CXA2025AS US decoder matrix)
     R = Y + 1.630 * I + 0.317 * Q
     G = Y - 0.378 * I - 0.466 * Q
     B = Y - 1.089 * I + 1.677 * Q

    VIC-II Colors

    Commodore VIC-II Color Analysis


    especially NTSC users approached me a lot in the past, mentioning that the colors generated by VICE don't quite match what they know from the real thing. some possible causes which are addressed here are:


    NTSC (old lumas)

    SONY CXA2025

    SONY CXA2025, alternative color set


    generally for PAL the colors are more defined and stable, so the above should generally give you colors which are pretty close to what you expect from the real thing:


    PAL (old lumas)

    AEC Line Glitch

    Somehow the AEC line is coupled into the Luma Output, the result is that the current slope of the AEC line is visible (added to?) in the luma signal. Or in other words, a rising edge of the AEC signal produces a tiny bit more luma (for about one pixel) and a falling edge produces an even less tiny bit less luma (for about two pixels).

    This artefact is drastically exagerated here (obviously). On most real setups the brighter lines are barely visible (they become more visible if you decrease saturation). The darker lines are usually hardly visible at all, except on modded C64s that have their RF-Modulator removed for better picture quality.

    NMOS Output Characteristics

    The NMOS output drivers work by the following rules:
    • when going from high to low, the transition is relativly fast (about 0.5 pixels) because the line is activly pulled low.
    • when going from low to high, the transition is relativly slow (about 2 pixels) because the line is just "left alone" which causes it to go high.
    This has the consequence that in high resolution patterns with high contrast between the pixels (eg black and white) the actual hires pixels will never be able to reach their intended full brightnes and turn out darker, and infact smaller, than they should.


    svideo (default)


    PAL Decoding

    in PAL the color signal is always mixed with that of the previous scanline. this has the following consequences:
    • colors with the same luminance mix perfectly when drawing them in alternating lines
    • additionally, since the VIC (or the circuits in the C64) can/do not encode the color signal with a perfect phase offset of 180 degrees every other line, these mixed colors turn out differently depending on which colors are in odd/ even scanlines.
    another Artefact of PAL decoding is that Saturation is slightly different every other line (exagerated here).

    NTSC Decoding

    SONY CXA2025

    Artefact Color

    these are the most interesting bits for us from this usenet posting:
    FWIW, my "folk hypothesis" growing up was to consider the horizontal signal to be divided into [R][G][B] sub-pixels, and Apple screen bits to be approximately a half-pixel wide, e.g. [0][0] is the same width as [R][G][B]. The bit-7 shift would offset things so the same two bits [0][0] would overlap the neighboring [G][B][R] subpixels instead. A pattern of [1][1] would light up RGB (==white). A pattern of [0][1] would light up mostly [B]=blue and [1][0] would light up [R][G]=orange, or with the bit-7 shift [0][1] would light up [B][R] = violet and [1][0] would light up mostly [G] = greeen.

    Now I *know* that NTSC does not address individual pixels, but my "folk-hypothesis" was remarkably effective so it burnt into my brain.

    Looking at page 96, I see that it's strangely almost correct - there are 180 cycles across the screen of a color signal which goes through the spectrum MROYGCBV, and if the monochrome signal is "on" in a particular field of that you get those colors; Apple's pixels do address half of one of these cycles each. So [0][1] lights up GCB=green, [1][0] lights up MRO = purple. Shift a little bit over and you get. [0][1] lighting up CBV=blue and [1][0] lighting up ROY=orange.

    This also explains the "fringing" rather nicely. The pattern [0][0][1][1][0][0] which in (Apple Hi-Res)theory is perfect black-white-black will inevitably have the "white" signal "leaking" to either magenta or yellow on the edges depending on where it is within the cycle.

    Visible Area

    Different monitors show differently much of the VIC's output. The following figures give the maximum possible numbers:


    Overall, this makes a maximum visible area of 402 by 292 pixels.

    Aspect Ratio

    The VIC's pixels are not actually square. The actual aspect ratio can be calculated as follows:


    Another, but less accurate, way of calculating the pixel aspect ratio is the following:
    PAL Image, 1:1 Aspect Ratio PAL Image, PAL Aspect Ratio NTSC Image, PAL Aspect Ratio


    NTSC Image, 1:1 Aspect Ratio NTSC Image, NTSC Aspect Ratio PAL Image, NTSC Aspect Ratio






    general Theory




    other simulation efforts


    last updated 28.April 2012