On Thu, Jun 20, 2024 at 10:15 AM Eric Smith via cctalk < cctalk@classiccmp.org> wrote:
> > That's where I'm starting, too. I'm using a 16500A (rather than a B, C, or > 1670x) because it should be much simpler to figure out how the processor > accesses the registers and memory of the modules. Then it should be > possible to build something like a USB4 PCIe alternate mode interface to > the analyzer backplane, replacing the analyzer's CPU card. > > That was my thinking too, but I am less ambitious on the backplane. I haven't measured the bus speed there yet but i suspect it's rather slow and uses block read / DMA style operations for the buffers (there are not enough address bits for much else). I thought it might be supported by a cypress ez-usb so old HP cards could be used for much wider/faster captures than when they're used in streaming mode. > I don't intend to reverse-engineer details of how the modules actually > work. Instead, I want to run the real 16700 software in simulation or > binary translation on a PC. > I want to look first at what really simple cards like a 16510 get in order to start/stop/transfer. However these typically don't work with the 16700 UI so although running that under emulation is attractive I don't think you can use anything older than a 16555. There's also the division between 16505 software (which I think is related to early 16700) and the 16500C which it drives over SCSI. > > The tricky part is going to be figuring out how the module correlation > circuitry on the CPU board works, because my plan requires reproducing > that. > Yes. I'll be happy getting a single card working. Multicard setups can use a real 16700 but if you're working to accelerate that it's a nice complement to a single-card option.