Memory board rev 1.1

Posted on Mon 05 July 2021 in MUPS16

I finally got a bit of time to fix the bugs in the memory and MMU boards, and to solder up the memory board (waiting for some chips for the MMU still).

The rev 1.1 design of the memory board is alomst the same as the 1.0, except that I
  • tightened up the RAM memory decoding so that it didn't respond to writes to the high memory-mapped space;
  • replaced the IS61C5128AS-10TLI chips with the slightly slower IS61C5128AS-25TLI ones;
  • added a set of pin headers on the left of the board that expose the MISO lines from the SPI ROM chips

The first of these was just a bug (RAM decoding was responding to _any_ write outside the ROM address space, so a write to something like the LCD mapped at 0x900000 would have written to 0x100000 as well.

The main reason for replacing the RAM chips was that I had six sitting in a box, so I didn't need to order any more. In addition, the slower chips come in a TSOP-32 package that's much more friendly for debugging (you can actually get a probe onto each leg, since they're standard 1.27mm pitch). The access time is still ridiculously over-specced for this system anyway, at 25ns.

The final change to add the pin headers was to make programming the ROM chips easier. I can now plug in a cable to program the chips, and read them back. Previously I could program them OK, but reading the contents for verification meant clipping wires directly to the MISO legs of the ROM chips, since I forgot to bring them to the board edge anywhere.

Renderings and photos

Rendering, in glorious green solder mask:

Getting a little tight on the routing here. Not quite as bad as the MMU, but close:

Getting ready to solder everything, with the chips laid out one one pcb before being transferred to the lower one after it's had solder paste applied:

The new and old boards, side by side. I still prefer the look of the RAM chips on the old board, but the new ones are much easier to work with:

The setup for reading and writing the serial ROM chips. The extra breadboard is needed because the Arduino uses a fixed pin for the MISO line, so I needed a 2-1 mux to pick which ROM output was routed to that pin:

Testing

First, I tested the new setup for reading and writing the ROM chips in-place, and, somewhat to my amazement, both the hardware and software worked first time:

(this was writing an image containing some basic test code for the system; you can see some error messages starting at address 0x100). Since the ROM contents all looked correct, I put the reset board back in, and after a couple of seconds of the power-on-reset sequence, checked that the ROM contents had been correctly copied to RAM at address 0x800000, and hey presto:

Finally, the all-important unit testing to make sure the RAM worked, and this time everything worked perfectly!

$ python3 -m pytest mups16/test_board/mem_test.py -v -s
===========================================================================================================================
test session starts
===========================================================================================================================
platform linux -- Python 3.6.9, pytest-5.3.1, py-1.8.0, pluggy-0.13.1 -- /usr/bin/python3
cachedir: .pytest_cache
rootdir: /home/ali/git/cpu/python
collected 6 items

mups16/test_board/mem_test.py::test_ram_bytes_low_odd PASSED
mups16/test_board/mem_test.py::test_ram_bytes_low_even PASSED
mups16/test_board/mem_test.py::test_ram_bytes_high_odd PASSED
mups16/test_board/mem_test.py::test_ram_bytes_high_even PASSED
mups16/test_board/mem_test.py::test_ram_words_low PASSED
mups16/test_board/mem_test.py::test_ram_words_high PASSED

Very happy with how that all turned out. Fingers crossed the 1.1 revision of the MMU is as successful.

memory pcb