#exp30

I ordered some boards for my Mac SE accelerator. I'm quite pleased with how they look.

Assembly wasn't too bad. The 0805-size resistors and capacitors can be a bit fiddly to work by hand, but they're still manageable. In all, it just took a couple hours to put together the first test board.

It fits in my SE with plenty of clearance between it and the floppy drive cages.

So the big question is — does it work?

Of course not! What fun would that be?

First run attempt was just a garbled screen, and not even the infamous checkerboard pattern. I didn't expect it to be fully functional at first attempt, but that pattern is a new one.

Time for the logic analyzer.

I can see the hand-off working. The accelerator asserts the Bus Request signal to take the SE 68000 off the bus, and waits for the Bus Grant signal in response. Once the 68000 is off the bus, the 68030 is brought out of reset and allowed to take over. That much worked as expected.

Next thing to look at is whether the 68030 is starting up properly. The first thing it's going to do is load the initial startup vector and stack pointer from ROM. These are 32-bit values, so each will take two access cycles over the 16-bit SE bus. It was getting stuck on these initial read cycles, waiting for a bus termination signal

The 68k processor family has a great asynchronous bus that relies on peripherals providing an appropriate bus cycle termination signal. Macintosh computers generate the termination signal synchronous to the system clock. What I found was the faster 68030 (running at 12MHz for testing) was sometimes ending one cycle and starting the next within a single clock pulse of the main 8MHz SE clock. The SE motherboard logic never saw the new CPU cycle begin.

My first draft of logic was largely asynchronous, and paid little attention to the system clock. But since the SE motherboard logic is synchronous and expecting 68000 bus cycles, I need my logic to better emulate 68000 bus cycles.

So I rewrote nearly all of the logic for the CPLD to roughly emulate 68000 bus cycles.

Never thought I'd be happy to see a Sad Mac error. That error means the 68030 is running enough code for the initial ROM checksum to fail. That's quite an improvement over not even loading the initial vectors.