9

As I recall, on the Commodore 64's text mode, if you chose certain colour combinations, the letters would be badly obscured by some kind of interference which seemed to affect each character cell individually. If you put a yellow A atop of a blue background, it would put several yellow pixels inside that character cell beyond the ones which made up the A.

Other combinations looked well, some reasonable, some terrible. I had around three different Commodore 64s from different batches, revisions, which all exhibited this problem, all connected by RF to a small cheap TV.

Yellow on a blue background would be the worst affected, green on a blue background would be somewhat better, and light blue on a blue background would be splendid and clear. White or greys on a black background also would be clear. I'm sure I'm not misremembering the fact this happened, but I find VICE doesn't seem to emulate it.

Why did certain color combinations cause bad interference in the textmode on the C64?

Omar and Lorraine
  • 38,883
  • 14
  • 134
  • 274
  • 2
    "look terrible" on a bad TV, or on a decent monitor ? – Tommylee2k Sep 21 '18 at 11:32
  • @Tommylee2k Good point, I was using a cheap TV at that time, but I don't remember the same thing happening with my ZX Spectrum. – Omar and Lorraine Sep 21 '18 at 11:38
  • I might be biased, but I think on the C64 any colour combination looked awkward ;) The C64 had the worst choice of color of any home computer of its time, IMHO. – tofro Sep 21 '18 at 11:40
  • Matter of taste, obviously - But I like my reds to be red (rather than pink), my yellows to be yellow (rather than beige) and my blues to be distinctive (rather than baby blue). Realistic pictures were rare in 16-colour-time anyhow, and any attempt to look realistic just looked more ridiculous than good. But, taste is taste - But the color scheme has always been (then and now) a reason why I never wanted one. – tofro Sep 21 '18 at 11:51
  • Yellow on blue the worst? Try blue on pink! ;-) – Fabby Sep 21 '18 at 12:14
  • @tofro No joke, that palette is awful. Nothing but washed out pastels. But I have to say that The Bear Essentials has some of the best looking graphics I've seen on C64 in terms of making the best of the colors available. – db2 Sep 21 '18 at 12:52
  • @tofro There may not have been a lot of good combinations, but there were some. My favorite was dark gray on light gray. Some might consider that cheating, since they might not consider levels of gray to be colors. Nevertheless it did look really good. BTW, I had a color monitor, so I didn't have to fight a TV's limitations. I didn't mind the default white on blue either. IIRC, IBM had done a study which indicated white on blue was easy on the eyes. – RichF Sep 21 '18 at 20:08
  • Possible duplicate of Does every retrocomputer and console with NTSC composite output have 'artifact color' ability? – Thomas Sep 25 '18 at 22:50
  • 1
    @Thomas It's not a duplicate, since the interference I'm talking about is not a new colour but always the same as the foreground colour in that character cell. – Omar and Lorraine Sep 26 '18 at 07:48
  • I understand, but it is the same reason: because the pixel clock is faster than the ntsc color system, the different combinations will be reproduced with a different fidelity; this can cause some colors to be off, some to be more saturated, etc; which is why some combinations come out good and some don't – Thomas Sep 26 '18 at 11:57
  • @Thomas does the same thing go for PAL? – Omar and Lorraine Sep 26 '18 at 11:58
  • yes; PAL is similar to NTSC; just different carrier frequency and the phase alternated from one line to another; so you have a bit more color stability, but it's similar. While the effects may be slightly different, it's overall the same mechanic at play – Thomas Sep 26 '18 at 12:03
  • 1
    @Thomas It's not a duplicate because they're asking different questions and the questions have different answers. Just because the question text of one provides a partial answer to the other doesn't make them duplicates. – wizzwizz4 Sep 30 '18 at 10:46
  • I understand the questions are different in nature, but how should we handle the case where the answer is the same phenomenon; one question observes new colors, the other question observes nice vs. bad looking combinations, but both questions are answered with understanding the NTSC/PAL signal (additionally there is a ringing problem on some of the 8bits computers). Having a duplicate answer, or 2 similar answers with details tailored to the question makes me think that some information could be missed. What would be the best approach? – Thomas Sep 30 '18 at 19:13
  • @Thomas: Some systems were deliberately designed to maximize the usefulness of artifact colors (e.g. the Apple II), some were designed in ways that happened to make them useful (e.g. Atari 8-bit), some were designed to minimize them (e.g. TI 99/4a, VIC-20, and late model C64s), and some were designed in ways that neither minimized artifact colors nor made them useful (early-model C64s) – supercat Oct 04 '22 at 17:47

2 Answers2

15

This is just the well-known "AEC Glitch" of the C64's video output, exacerbated by the additional analog distortion of the RF output to a TV.

Anyone who used a C64 on any kind of display is familiar with the vertical stripes that appear. This is caused by interference leaking into the Luma (aka Luminance, or Intensity) portion of the video output. The interference can occur from several of the bus control or clock digital signals on the motherboard, but it is most directly associated with the changing state of AEC (Address Enable Control). This is why it is commonly called the "AEC Glitch".

Since the vertical stripes are due to regular variations in Luma as each horizontal line of the display is drawn, the distortion is minimized for text display by using colors that contrast on their Luma dimension. This maintains the necessary contrast between adjacent pixels to make the text more readable despite the Luma interference fluctuations. When you choose two colors that contrast on the Hue and Saturation, but not Luma, then the effect of the Luma interference is exacerbated. Adjacent pixels of similar Luma will tend to bleed together more dramatically with the interference pattern. This goes a long way to explaining the choice of light blue on a darker blue background for the default text screen. Also why contrasting intensity (luma) colors for foreground and background work best.

It should also be noted that this interference problem is frequently corrected by modders installing the so-called "Luma Fix".

Brian H
  • 60,767
  • 20
  • 200
  • 362
7

I don't know the C64's special glitch problems, but to me the effect you describe can be explained by going over the RF signal.

The analog TV signal that gets modulated on the RF encodes Luma and Chroma very differently (for compatibility reasons when color TV was introduced).

Luma gets a bandwidth of about 3 MHz, meaning that on the visible part of a scan line (50 µsec), you get maximum 150 highs and 150 lows, meaning 300 pixels. So the 320 pixels of HiRes graphics or 40-characters text can be more or less distinguished if they have a good chroma contrast.

Chroma, on the other hand, only gets about 1.5 MHz bandwidth, meaning 150 pixels. So, if foreground and background color differ mainly in hue or saturation, the TV signal can't transport 320 different pixels in a line, and you get a badly blurred picture.

To get a crisp picture (as far as you can call anything on an old-style TV "crisp"), you should choose a color combination that differs only in luminance, e.g. dark blue and light blue.

Even combinations that are typically regarded "good contrast" like white letters on red background will get blurred in the chrominance, e.g. the vertical bar of an uppercase "T" will become more light-red than white, only the horizontal bar will be white in the middle.

And if the RF circuit isn't perfect (and I've never seen high-quality ones in the consumer market), the chroma changes often were out of sync with the luma ones, maybe lagging behind by a few pixels.

Ralf Kleberhoff
  • 2,380
  • 10
  • 12
  • And if the RF circuit isn't perfect (and I've never seen high-quality ones in the consumer market) - the ones on the early Atari computers were pretty spectacular. I remember typing in a program for the Apple II that used color bleed to simulate more colors, and it did nothing on my 400. – Maury Markowitz Jan 05 '20 at 00:10