21MX proms (per request)
jwest at classiccmp.org
Mon Sep 7 17:20:48 CDT 2015
But I do think there's merit in Alexandre's general argument that one could theoretically emulate the older bipolar PROMs with some mix of modern components ... and the programmable device would be (a) more easily obtainable than the old bipolar PROMs and (b) programmable with inexpensive, contemporary USB device programmers which are definitely big positives.
I agree... and have suggestions. Prior art exists, I have seen the arcade crew use adapters to modify the pins of newer devices to be compatible.
For loader roms, there are about 10 official images from HP. Plus, most users (myself for sure) would want around 2 images that are user-created. The system board supports 4 roms, selectable by front panel. The issue with the solution the arcade crew has done - the adapters make the new memory devices too tall to allow clearance above. In chassis after a certain date code there is a cutout to access the rom sockets without removing the system board. In those, the clearance issue is the bottom of the memory controller. In prior revs, the issue is the metal work not providing clearance. I'm not saying a solution can't be found, but whoever develops such a device needs to be aware of the clearance issues particularly on the older rev machines. The only thing I can think of that is likely to have proper clearance is a board that has pins going to each 4 rom sockets, and the replacement memories are placed under the board - not on top of it - carefully avoiding existing chips on the system board. Regardless... whoever designs such a board needs to have a 21MX in front of them to assure clearances. Also, I'd have to check, but the precise spacing between the rom sockets may well be different between the M and E versions. All that being said, it would be nice if the board supported say 12 images, any 4 of which could be selected as active.
Now, here's a better take on the problem.... design a card that goes into a memory slot. Most all machines have many empty memory slots even when full of memory. Then a ribbon cable could go down to a paddle board that plugs into the 4 rom sockets. The board in the memory slot could get power straight from the backplane. The ribbon/paddleboard relieves the clearance problem. Extra points if this board also houses 1mw of ram and connects to the memory controller as usual. Even better points if the board replaces the memory controller and provides both 1mw ram plus loader roms :>
For microcode, you'd sure have to find a way to get past the speed issue. Since the 21MX's take microcode on the cpu board, or a FAB board, or a FEM board (and the "fab" is very different on the 21MX/M)... you've got a lot of choices. My bet - the most functional approach is to ignore trying to deal with microcode prom replacement on the cpu and fab boards. Instead, provide a single "new" "FEM" board that presents 24 bit wide memory modules of appropriate speed. Basically, this becomes a WCS board - with modern memory devices. There's a pretty large number of microcode options that could get stuffed into such a device. The only down side - FEM boards take an I/O slot, and one HP OS in particular will have an issue where I/O slots are pre-allocated and not configurable at sysgen time. Extra points if the board can also include devices that everyone always puts in - namely a TBG and HSterm/BACI. Not sure how one could handle the select code logic in one slot, would probably need paddleboards. But my advice - solve the speed issue first ;)
More information about the cctalk