• Please review our updated Terms and Rules here

IBM System/23 Datamaster, display and character generation

I should have posted this a week ago, but I had no time to redo the schematic until today. I haven't retraced the parts in my board since last week so take this as a grain of salt.S23-video.png
The Datamaster has two clock sources for the CPU and three for the video subsystem. The latter can work at 18.432MHz or 18.2MHz if the European switch is set - the third source hasn't been identified yet. Two components receive the selected signal: the video shift register and a 74LS174 configured as a shift register, which samples an unidentified signal and feeds it to the 8275 as a clock.

The CRTC has two outputs to signal attributes and the s/23 uses them to select the value to the character ROM or another one generated by a 74LS139 and a couple of NOT gates. The result is fed to the video shift register and NANDed with something.

There is still a lot of work to do. I also feel something is wrong as the 139 also controls the select signal of the 157s...
 
I have found part of the banking mechanism. Apparently bit 6 of the character code gates the page of the ROM. This means 64 characters are always fixed while the other 64 depend on the page they are in. This would explain why the 'l' and the '1' characters share representation. 64 characters aren't that many to put the alphabet (lower case and upper case) and the numbers.
 
Nice! Yeah, I would've thought that as well, although those fixed 64 characters include a few less-obvious choices like ½ or ³. I suppose it's all good as long as the results aren't ambiguous to the end user - it's probably telling that this duplicity of "1" and "l" isn't really noticeable until you look at the ROM. After all, in terms of storage it's all EBCDIC where the two characters clearly have distinct code points.

Speaking of the character set: when I look again at the CRT section of the manual (see first post), it makes me wonder why there was a "Germany only" variant with reduced glyph heights.
One might think that they wanted better accommodation for characters like "Ö", to distinguish it more clearly from "ö" etc. But most other locales use plenty of diacritics and accented characters as well...
 
@VileR I have been using your program to see the character set as an example for my experiments with floppy images of the System/23. I hope you don't mind...
 
RetroAND said:
@VileR I have been using your program to see the character set as an example for my experiments with floppy images of the System/23. I hope you don't mind...
Click to expand...

Not at all... even if it didn't work as intended for the charset, at least it's good for something!
 
A couple of days ago I found that the Datamaster's CRTC has a slight difference in the interrupt generation system. For this reason stock 8275 won't work directly as spare parts without patching the ROM. I am patching ROS 02h at this point.
 
I have acquired a text-editor card. It adds a 8275 and 32KB of DRAM. There isn't a character ROM associated with the CRTC, therefore I suspect the on-board RAM is to debine characters. Further analisys will confirm or deny this statement.

Oh, and by the way, the previous post of mine is wrong. A Datamaster will work on any 8275, even the Soviet one.
 
You must log in or register to reply here.
Back
Top