The true pixel resolution of the Model I was 512x192, comprised of 64 columns by 16 rows of 8x12 pixel character cells. But of course the processor only had acc

The true pixel resolution of the Model I was 512x192, comprised of 64 columns by 16 rows of 8x12 pixel character cells. But of course the processor only had access to the character cell array, not the individual pixels. The pixels were created with a dozen (or 2?) 'MSI' chips hard wired to create that one mode. It would take the values stored in that character array by the CPU, pass them through a character generator ROM, into a shift register and thence into analog to the display. At 8x12, each character had ample room for lower case descenders, but to shave a couple bucks off the price tag, the character array (at 3c00-3cff) had only 7 bits per character, and either the character ROM didn't even have lower case, or there simply was no way to index them with only 6 bits. Yes, 6, on account of all the symbols with the MSB set were dedicated to 'semigraphic' glyphs; the 2x3 chunks of giant lo-res pseudo pixels you could elicit with poke statements. 64 of those, 64 for letters numbers and symbols, and that was that. The 512x192 was probably about all the cheesy tunerless television display could handle, which is why we had 64 columns and not 80. Still, many of the other home computers only had 40 columns, so point for TRS-80. Otoh, those other computers often had 24 or 25 rows, but we only had the 16. Had the character cell been 8x8 instead of 8x12, we'd have had 24 rows, but that would have required 1536 memory locations for the screen. The 1024 locations at 7 bits each was seven 1024x1 static RAM chips, but 1536 would probably have doubled that, not to mention complicated the address decoding. So SET and RESET graphics were limited to 128x48, but those were good times.