6

On the NES, the system I'm most familiar with, the NTSC models produce 240 lines of video image, but the top and bottom 8 pixels are usually considered to be "hidden" due to overscan. (The actual number of hidden pixels varies along TVs from 4 to 12 it seems)

Most other systems at the time either have a similar issue or reduce the vertical resolution in order to show a consistent border colour in the overscan area.

I wonder if, by affecting the quirks of NTSC video timing, a 240 pixel tall image but where all lines are showing on most TVs is technically feasible. For example by making line rate as high as possible within the tolerance, and frame rate as low as possible, so that more lines can be squeezed in less time.

Does this even affect the overal heigth of the image, reducing the probability of lines being hidden by overscan?

Note: I do not count having two interlaced frames of 120 lines to be a valid solution.

Improve this question
1
  • 1
    The actual number of visible lines on an NTSC TV varies considerably. Only 9 lines of each 262 or 263 line field are fundamentally not visible on NTSC TVs. If you've ever seen those closed captioning lines flashing at the top of the screen you've seen the supposedly always hidden line 21. The first line of a 240 line NES display would start on line 22 or 23. On a old CRT you can adjust it so it displays all 240 pixels, on most modern flat panel TVs you can turn off the overscan clipping.
    – user722
    Apr 22, 2017 at 4:45

4 Answers 4

Reset to default
4

No, you can't somehow compress the lines on the TV to make 240 lines visible. The TV format is fairly rigid (no matter whether NTSC or PAL), and there's no way to, say, make it 280 lines total by just altering the timing in order to have 240 lines shown. Most TVs won't show a picture at all if you'd try that.

Note that this is different from (multisync) computer monitors, where you can vary the timing to a wide degree, and affect the image this way.

Improve this answer
8
  • Your answer is in clear contradiction with Tommy's. He says that tweaking the timing is possible, but that to show 240 lines I'd have to deviate too much and it probably won't be accepted. That means it's possible to add more shown lines, but not up to 240. You say tweaking the timing is impossible. How can I know who's right and who's wrong ? By the way I never intended to have 280 lines, just 240 but lessening the risk of them being overscan'd.
    – Bregalad
    Apr 21, 2017 at 14:53
  • It's the Internet, you can't know who's right and who's wrong :-) The ultimate proof is to get an old TV, get a graphics card that has TV out and allows you to modify the timing, and try it yourself. But the thing is that for TVs, the horizontal and vertical sawtooth form is fixed. The beam starts at the top in the occluded area, and the next line starts at defined distance below the first, and so on. So you can change the timing all you want, there's no way to change the image on a TV.
    – dirkt
    Apr 21, 2017 at 16:09
  • 1
    I think they're not really contradictory; dirkt states, correctly, that a TV is built for a certain synchronisation frequency — its raster gun always scans at the same speed and it looks for syncs a factory-configured amount of time apart. I point out that analogue electronics aren't exact so there's a tolerance in the timing, but not enough to get even as far as 248 lines instead of 240. I'll bet almost any television can display 241 lines in the space of 240 as that's only a ~0.4% divergence. If your TV works with any 8-bit Atari then it clearly has no issue with a ~0.2% divergence.
    – Tommy
    Apr 21, 2017 at 17:55
  • @Tommy Ok, the contradiction lies in the fact that dirks claims (and possibly rightfully so), that even when we deviate the timing, this will not change the vertical spacing, while you say it would change the vertical spacing but that tolerance is so small that it's not useful. Although the end result is the same, the answer is no, this can't be done, yours is because of the narrow tolerance, while dirkt answer is because of the way the standard work. Well I guess it doesn't matter, I have my answer anyway^^
    – Bregalad
    Apr 21, 2017 at 18:24
  • On a classic TV, horizontal and vertical actions are completely disconnected — horizontal scans and retraces, verticals scans and retraces, neither is linked to the other. So horizontal timing has no effect on vertical and vice versa. Therefore if you squeeze horizontal timing a little this way or that you affect vertical spacing. Just not a lot.
    – Tommy
    Apr 21, 2017 at 18:57
4

Yes and no. Yes, you can play with the signal timing, but if you do it won't be NTSC video anymore, it will be something kind of similar to NTSC but with different timing. This means you might not be able to display your signal on an NTSC monitor (or you may damage your monitor attempting to do so).

There were some NTSC monitors that had a switch setting to shrink the picture enough so that no part of the picture would be covered by the monitor bezel. This feature doesn't turn off overscan, it just ensures that the overscan area remains visible.

Improve this answer
7
  • 1
    Right; even systems that deliberately flouted the rules did so in consciously negligible ways: the Ataris slightly extend their line time to fit exactly 228 colour cycles because it makes for easier clocking and nicer dot crawl, but that's only 0.22% off spec. Well within tolerances.
    – Tommy
    Apr 20, 2017 at 13:56
  • So some fine tuning is definitely possible, and it does indeed increase the likehood of more lines falling outside of the overscan area? What about if both the VBlank rate is diminished to the bottom of the acceptable tolerance and the HBlank rate to the maximum? How many lines will this show in the typical TV?
    – Bregalad
    Apr 20, 2017 at 14:13
  • Vertical blank doesn't matter here; it will have happened at some time, the gun will have gone to the top of the screen, now it'll be proceeding downwards at the normal rate. Horizontal line length matters because it determines how far the gun will have moved down between the start of each horizontal line. A classic TV will accept some tolerance because it won't expect its analogue components that imply timing to be rigorously on spec now and forever. If you are in the low end of the tolerance you'll get lines that are closer together. But there's no spec for the tolerance.
    – Tommy
    Apr 20, 2017 at 20:39
  • @Bregalad You can adjust the timing a fair bit to adjust when vertical and horizontal retraces happen (and VCRs were notoriously variable in this regard) but you can't change the speed the electron gun scans the CRT left to right, top to bottom. The number of lines actually visible is function of how the CRT has been calibrated, basically how far above and below the phosphors it starts and ends its scan.
    – user722
    Apr 22, 2017 at 4:59
  • @Tommy: On systems which output 263 lines of 227.5 clocks, rather than 262 of 228, each frame will produce a "checkerboard" pattern of chroma artifacts, and alternate frames will largely cancel each other out. Using 228 clocks makes it easier to deliberately exploit chroma artifacts, but using 227.5 will improve the quality of color with images that aren't designed to do so.
    – supercat
    Oct 5, 2017 at 19:54
4

It'd depend on your TV. Some may just show 240 lines, of course, but assuming yours is chopping off only eight in total, that leaves 232 visible. It looks like the NES scans at 63.5us per line, which is exactly to spec per NTSC, to fit 240 lines, each would need to be slightly less than 61.4us — 232/240 as long as 63.5, or a 3.33% digression from the standard. Which is actually quite a lot — it's close to 7.5 colour cycles — do I'd expect it to be rare that a TV would accept that sort of overclocking. But there's no hard and fast rule.

As to why Nintendo painted 240; you also probably wouldn't expect the lines a TV shows necessarily to be in the centre of the nominally visible region. Some might show the top lines but not the bottom, some the bottom but not the top, most something in between.

Improve this answer
2
  • Your answer is in clear contradiction with dirkt's. You say that tweaking the timing is possible, but that to show 240 lines I'd have to deviate too much and it probably won't be accepted. That means it's possible to add more shown lines, but not up to 240. He says tweaking the timing is impossible. How can I know who's right and who's wrong ?
    – Bregalad
    Apr 21, 2017 at 14:54
  • Oh, well, I've gone an answered in dirkt's comment area. But the short version: I'm arguing there's enough tolerance for some variation but not enough for drastic variation, he's asserting that TVs are built for NTSC timing only so there's no way you're going to be able to get drastic variations.
    – Tommy
    Apr 21, 2017 at 17:57
-1

Depends how forgiving your TV is about the line rate. If we consider that the actual sweep rate of both the horizontal and vertical sections are fixed, then by upping the line sync frequency we can physically fit more scanlines in the same space.

A good estimate for the normally visible number of lines is around 216. Thus we need to increase the line rate by 240/216 = 1.111, or about 1/9th. That's actually quite a bit when we're talking about adjusting the scan rate of a more or less fixed-frequency display. For a regular colour NTSC television, that'd increase from 15.734kHz to 17.482kHz, which is most of the way towards MDA scanrate, and certainly beyond even the 16.5kHz of "enhanced resolution" arcade monitors (which are likely just normal TVs tweaked to their limits to fit as many visible scanlines as possible within a roughly 50Hz refresh rate).

I'm not saying you'll not find a TV that's happy playing along with that, but I think it'd be difficult. You're more likely to blow quite a few. Might have better luck with a proper computer monitor, as they're apparently somewhat flexible (there are a bunch of "extreme" Amiga modes which tweak almost down to 15.0kHz, and well up into the 16s, all allegedly 15.7kHz monitor compatible), but it feels a bit too much even so; a multisync would be better, but that kind of takes the fun out of it because you can just set whatever mode you like. And even with the regular monitor... you could just set a normal 240-line mode and then play with the picture controls.

The other consideration is that you'd have to shorten each scanline by 10%. That's maybe not so much of a problem, if with a typical machine we may have 455 pixels times per line, the adjusted version is still about 410, but it's eaten into our blanking quite a bit; with a 320 pixel line, we had 135 spare clocks per line, now it's only 90, and that could actually be kinda minimal for getting active area, border, blanking, sync, blanking, border and repeat. The deepest workable overscan tends to be barely any tighter. Or in other words, a line was about 63.5us before, now it's 57.2, and the usual full width (albeit with overscan) active period is about 53.3 (and visible, around 48us tops). The penny's been considerably shaved, and we may well find ourselves hitting flyback before the beam's off the right hand size (if we assume it always flies back to the same place ... which itself could be an issue for generating a biased AC waveform in the flyback/sweep circuitry and thus making considerable DC current), maybe even displacing the image to the left...

So in conclusion, whilst it's a decent idea, and could possibly be workable with a more modest goal (and, heck, I have a feeling it might even be employed to produce fake interlace on some recent Atari ST demos by hacking the length of several early lines on every other frame - as this is a system with NO native hardware interlacing ability, but the demowriters claim a true 720x440i colour resolution on a system only normally capable of 640x200p colour or 640x400p mono), to achieve what you want would almost certainly demand just too much of an everyday TV and cause it to go up in smoke.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .