QSIC update

Noel Chiappa jnc at mercury.lcs.mit.edu
Tue Mar 6 10:43:18 CST 2018

Hey, all, a quick update on recent progress.

I now have a working prototype to match Dave's (although he's still doing all
the real work), and we've made a minor improvement in them (re-wired things so
we can use shorter cables to the FPGA daughter-card).

I got my indicator panel working, it looks quite nice:


The inlay is an original from a TS08 that I happened to have lying around; Rod
S has made us a large batch of new blanks, but silk-screening of the captions
on the front of those is yet to happen. The bezel is also an original. (I have
a large stock of those, so it'll be a while before we need to work out how to
make new ones - probably resin casting, although 3D printing will be an option
too.) However, the guts are all new, and as you can see, the result looks just
like the real originals.

Dave has also worked out how to connect up the RKV12 (our name for the 22-bit
RK controller, by analogy with the RLV11/RLV12) to the internal 'block' RAM in
the FPGA, and then did the stuff to connect it to both the uSD card and the
internal RAM at the same time, with one drive connected to the RAM, so things
like swapping, etc don't 'waste' uSD writes.

Both of these are working quite reliably; the exciser/tester runs until we get
tired of the noise, and turn the machines off! :-)

His current project is to work out how to talk to the larger external RAM on
the FPGA daughter-board (the internal RAM isn't large enough for even a single
complete RK pack). I'm so looking forward to putting swapping, /tmp, pipes,
etc all on different platters, so as the system (Unix V6, natch :-) runs I can
watch the lights and see _exactly_ what's going on!

After that: turning the RK into an 'RPV12' (which should be pretty easy, the
RK11 and RP11 are very similar), and adding a mux so that the two controllers
can share the storage devices, etc. Those should both be done soon after the
external RAM (and maybe before, if Dave needs a break from that :-).

We also intend to do an 'extended RP11' (name not yet chosen, although I like
'RPV-12D - DEC's last was the RP11-C), which extends all the disk address
fields in the register to use the unused bits, giving us pretty massive
storage capability. With 16 bits of cylinder (up 7), 8 bits of surface (up 3),
4 bits of sector, that gives 28 bits of block number per drive; and with 8
drives per extended RP11, that's a total of 31 bits of block number per
controller. Convert the blocks to bytes, that's 9 bits more, so 2^40 bytes per
extended RP11, or 1TB!

Very shortly now we'll need to turn to starting on the design of the
'production' hardware.


More information about the cctalk mailing list