NTSC scan lines used by 8-bit computers

Question

Several computers of the late 70s used a TV set for their display. I'm not talking about the TRS-80, which came with a monitor that was basically a stripped-down black and white TV set, but computers that could plug into your TV, thereby saving the purchase price of a monitor. The Apple II (with third-party modulator), Atari 800, TI 99/4. All had vertical resolution 192.

Then the Commodore 64, released in 1982, had vertical resolution 200.

Why the difference? Why didn't the earlier machines also go for 200?

Conjecture: the limiting factor was the title safe area. Tolerances of TV sets actually sold in the US, tightened up over the years, so that in the 70s, only a 192-line area would reliably be visible, but by the early 80s, that extended to 200.

Alternate conjecture: the limiting factor was the known title safe area. 200 would in fact have been fine on 70s TV sets, but computer designers didn't have enough experience with different models to be sure of that yet, and rationally played it safe.

A possible line of evidence to distinguish between these: if the first conjecture is true, the 70s computers should all have used pretty much the same range of scan lines, and the C64 extended slightly in each direction.

Of the 240 scan lines nominally present in an NTSC field, which 192 are used by the Apple II, Atari 800, TI 99/4? And which 200 by the C64?

Answer 1

I found one document which might shed some light why 192 lines was used before 200 caught on.

The TI 99/4 used a TI TMS9918 VDP, which based on the document, was designed as a chip for simple low cost systems, i.e. a chip for early home computers. It was allegedly a chip with highest resolution full color capability of any single chip systems of it's time. It claims that the resolution is near the limit for ordinary color television, being the highest practical for home computers used with ordinary TVs.

It does acknowledge that Motorola MC6847 offers higher resolution but says it trades colour variety with higher resolution.

So the 192 lines offers 24 rows of text with font height of 8 pixels. Now for example C64 offered both 24 and 25 text rows, so the extra 8 scanlines to get up to 200 scanlines and 25 text rows may not sound like much more complex.

In reality, the compexity might grow. For example in graphics mode, the TI VDP splits the screen into three 64-line sections adding up to 192 lines, and each 256x64 section uses 256 tiles of 8x8 pixels, so three separate 256 tile maps are used.

As 64 is a power of 2, it likely made things much easier to detect overflow from 63 to 0, and jump to next section, than to compare with some other number. And a reason why all sections are identical and add only up to 192 scanlines.

How it all works with NTSC lines is a bit difficult to assess, as many chips tend to invent their own line numbering scheme.

NTSC starts line numbering from 1 after last non-blanked line, with the first blanked line in vertical blank area. There are 3 blank lines with equalizing pulses, 3 sync lines, 3 blank lines with equalizing pulses, and 11 standard blank lines. That's 20 total blank lines, with lines 21 through line 262.5 active, so 241.5 active lines. 21 is generally reserved for closed captioning and we can ignore the half line, so that's 240 lines of video sent. So generally, lines 21 and 262 would not be visible any more. Earlier on, the blanking was shorter and 486 interlaced lines was possible, or 243 per field.

As the TI VDP uses 262 total lines, not 262.5 which means progressive. Going through in same order, there are 3 blank lines, 3 sync lines, and 13 blank lines. So 19 total blank lines. Then there's 27 lines of top border, 192 lines of actual screen data, and 24 bottom border lines. So active video data sent on 243 lines.

If we simplify things a bit and make some assumptions, in NTSC the center of the screen is halfway between active image lines of 120 and 121, or roughly 120.5 lines before the end of frame.

With TI VDP, the halfway of the 192-line screen is between lines 96 and 97, so there is 96 lines plus 24 bottom border lines, or 120.5 lines. The center of the screen matches assumed center of standard NTSC transmission.

If I find timings for other platforms, I will add them. NTSC details for C64 are much harder to find than PAL.

Answer 2

Older television sets used round picture tubes with a bezel to mask off the edges of the scanned area. Picture tubes were often circular, and to maximize useful areas, the image was expanded to the point that the corners would get cut off. Because horizontal and vertical driving circuitry would be affected by changes in temperature and aging, it would have been difficult to ensure the image stayed centered, and the interaction of the rectangular scanning pattern with the circular tube shape would have made variations in centering noticeable. To obscure that, television sets used a vaguely rectangular bezel.

Computers, of course, were almost invariably designed to make use of a rectangular screen area. A friend of mine had a VIC-20 attached to a rather old console TV which slightly cut off the screen corners, despite the fact that the VIC-20's screen, at 184 lines high, is among the shortest frame used by personal computers. The scan lines were somewhat wider than some computers (154 chroma clocks wide, compared with 140 for the Apple II and Commodore 64). The Atari 400/800 display was taller and wider (160 chroma x 200 lines) but its text output defaulted to omitting the left and right 4 chroma clocks, yielding 152x200).

Answer 3

Similar to the TI VDP answer, have a look at the Apple text mode layout described in this answer. (And keep in mind that the graphics video layout is based on the text layout).

There are 40 characters in 24 lines. 24 is 3 times 8, and you can see how the 40 bytes are fitted three times, with 8 bytes of "holes". This construction is repeated 8 times, covering the address space from $400 to $7FF.

Now imagine you'd have to fit a 25th line in order to move from 192 scan lines to 200 scan lines. This would have messed up the whole scheme, and since the video controller is implemented completely with discrete integrated circuits, it would have increased the number of circuits, space used on the motherboards, and price quite a bit.

So it's a no-brainer to just use 24 text lines to keep cost and complexity down.

A 25th line starts to make sense if you have a dedicated video chip (like the C64) or CRT controller like the MC6545, where you already have counters and compare registers, and if you have enough RAM to fit the no-power-of-2 extra information.

Answer 4

Conjecture #1: NTSC title-safe area (unlikely)

An NTSC video frame has 525 lines. After subtracting the 21 lines of vertical blanking, and the 20% of the visible area that was considered non-title-safe before 2002, you're left with 403 usable lines. Or 201 lines after halving it to deal with interlacing. So the Commodore 64 was pushing the limits of NTSC vertical resolution with 200 lines, but it seems to have worked well enough.

Conjecture #2: Wanting square pixels (unlikely)

One conjecture I had is that computer manufacturers just wanted to have square pixels. Since the 4:3 screen aspect ratio was dominant at the time, you'd want the vertical resolution to be a multiple of 3. If you also want to have a text mode with the same pixel resolution (as suggested by @dirkt) using 8×8 character bitmaps (which work well with 8-bit bytes), then the resolution needs to be a multiple of 24×32. And 192×256 is the largest that fits in the "safe" area.

But while 192×256 resolution has been used (e.g., in the Nintendo DS), the Atari 8-bits defined “high-resolution” graphics as 192×320 pixels, giving pixels with a 4:5 aspect ratio. The Apple II equivalent was 192×280 pixels, with 32:35 pixels, closer to square than Atari's, but still not square.

Conjecture #3: Binary math (plausible)

While "200" is a nice round number to people used to thinking in base-ten, it's not so much in binary, which is what computers use:

• 200 = 1100 1000
• 192 = 1100 0000

Suppose that you need to multiply a number by the vertical screen resolution. How do you do so on a CPU like the MOS 6502, which didn't have a multiplication instruction. Using bitshifts and additions:

• 192 * x = (x << 7) + (x << 6), two left shifts and one addition.
• 200 * x = (x << 7) + (x << 6) + (x << 3), three left shifts and two additions.

So 192, having one less 1 in its binary representation, is easier to work with in binary. Sure, it seems trivial now, but circa 1980, saving two instructions in a ROM graphics routine may well have been worth the slight loss in resolution.