Yes, it was pretty clear to me that the performance profile of the DS89C4x0 family members suggests that it is not a microcoded machine. Further, since I'd been told years ago that the reason the originally specified 50 MHz max execution rate of 50 MHz was not reliably attainable due to a shortcoming of the FLASH process, I'd concluded that the necessary clock multiplication couldn't take place within the technology in use there. Clearly, since they need a full cycle of the clock to make the memory work, a multi-phased clock arrangement wouldn't make a microcoded process operate at the necessary speed either. I'm told that the way the UBICOM Pic-like MCU's get around the flash process limitations is with an on-board shadow RAM that carries the code and is reloaded on each reset. It's too bad the Dallas guys didn't do that.

Nevertheless, these parts have lots of potential in the areas that interest me most, i.e. those that require really fast response to changes in external conditions, quick interrupt response, a fast synchronous serial port, etc. The fact that one can, under software control modify the workings of the memory map and the memory access timing is also a handy feature, though it won't find use very often.

However, I do have something on my desk at the moment that will use the XRAM at full speed while accessing external peripherals at a reduced rate, and which uses both the synchronous serial ports. Those features alone justify the selection of that device family. The fact that it has two data pointers is also a plus, as is the fact that the memory map itself can be modified under firmware control. In one mode, it can use the XRAM as a small but fast data buffer. In another, it can use the entire 64KB of external data memory as a buffer for the data acquired via the sync serial port. In a third, it can take that buffered data, fetch a portion of it into the XRAM, shuffle it into the USB FIFO, then move the data to be transferred next to another portion of the data memory, change the memory map again so the XRAM becomes a target again, and repeats the process. I've not yet seen a device that can do all that, yet lives in a convenient package.

I'm building this as a one-of, so it will be wire-wrapped. A half-millimeter-pitch TQFP would not lend itself to a one-of.

RE