Richard Russell wrote:
paulb wrote:Would the BBC really have cared about the Teletext mode requiring 20K or even 15K, especially if the motivation is merely to show Teletext pages, not specifically to offer a low memory mostly-text screen mode?
The BBC would have cared indirectly about requiring a lot of extra memory, yes, because of the cost implications!
Right, but this connects with BigEd's comment about memory pricing. Had RAM been more abundant (an aside: I believe that RAM production was what Inmos was originally doing, and so there was a recognised problem at a level intersecting with national strategy), the BBC could have gone with a larger amount purely because it would have offered more capabilities, future-proofing, and so on. If that then got used to provide Teletext in another way, the mechanism involved could have been a distant secondary consideration.
Richard Russell wrote:But even if cost was no object, there are two other important considerations. One is the quality of the displayed text; as has already been pointed out the teletext font is highly legible and the crude 8x8 character set available in the graphics modes would never have been acceptable as a substitute. Secondly, accurate implementation of the full Videotex specification was a requirement, and as I mentioned in another post that would have been difficult, if not impossible, to achieve in software with the CPUs of the day.
I can't comment on Videotex specification issues, and it isn't for me to rule as to whether the 8x8 font would have been acceptable for the BBC's purposes, but I disagree on whether a multicolour mode 4 would have been suitable. No-one really complains about the default font on the Acorn machines.
Richard Russell wrote:It's also instructive to work out just how much memory would have been required for a bit-mapped display capable of reproducing MODE 7 with the same quality and capabilities as the hardware chip. The basic Videotex character cell is typically 8 pixels wide by 10 pixels high, before character rounding; after rounding it is equivalent to 16 x 20 pixels. So with 25 rows of 40 characters that corresponds to 320,000 pixels. If we assume 4 bits per pixel (at least 3 are needed for RGB) that's 160 kbytes, substantially more than the "20K or even 15K" that you mention. Can you imagine the cost and complexity of providing that amount of bit-mapped graphics memory with 1980s technology?
Well, I did a quick calculation for a mode 2 screen of 24 lines which works out at 15K. Of course, that doesn't offer a nice resolution, so you'd have to consider doubling it to offer a mode 1 or 4 screen resolution instead, working out at 30K. Make it 16K or 20K (or 32K or 40K) if you want a slightly nicer font. There would be bandwidth issues, certainly, since the chipset is only delivering 80 bytes per pixel line at most.
Then again, it all depends on how important Teletext capabilities are. I don't recall any of the competing machines offering the same level of Teletext support with the base hardware, and so relatively primitive software emulations (like the Jafa software solution for mode 7 emulation) are deemed acceptable for certain purposes as a result. I guess the BBC wanted to leverage their existing technologies a lot more than the rest of the industry eventually did.Edit:
In the context of the memory needed to provide a Teletext display, it's also worth considering how it was done on the Archimedes, which I seem to remember involved a 40K screen mode (320x256x4bpp). That provided a fairly disappointing visual result - not the same look as on the Beeb - but then the choice of implementation is connected with the vertical resolution supported by the base monitor type. It's worth reviewing those Cub monitors and noting that the 1431 and 1451 support higher vertical resolutions than the 3000 model (released for the A3000), which seems bizarre until one considers that it was probably done to support the interlaced display of the Teletext hardware of the Beeb. So, the method of supporting Teletext appears to have affected the specifications of RGB monitors, possibly deferring the cost of displaying it to that purchase instead.