Is this a practical problem, or is it homework? Ideas involving the 8255 and other devices of its ilk are generally suggested by inept instructors who get their ideas from a book published 20-30 years earlier. The pace of the industry is much faster than the pace of academia.

In college people should learn, not just solutions, but how to arrive at them. Exercises of this sort don't teach that at all unless they force the student to go all the way back to the begining and select an original approach.

BTW, Erik, LPC doesn't tell us much. Google turned up lots of info on linear predictive coding, which I encountered 20 years ago while working a problem with secure voice communication, and low-bit-rate voice encoding, but I still don't know what an LPC is. Perhaps you'd like to elaborate.

I'd be interested in what, exactly, you recommend, if I were in the O/P's position. The general mention of IIC or SPI doesn't do as much as the mention of specific devices would do. That way there's a datasheet to read, rather than a 2K-page block of white papers.

BTW, the first i805x's I got in the late '70's were 12 MHz 12-clockers. If there were slower parts before that, I wasn't made aware of it. Early 8048-family parts were 6 or 8 MHz parts, but within a year of original production, even those were provided in 11 MHz speed. Of course, they didn't have a serial port, but that favorite crystal of yours would have worked with 'em. The old 8255's and even the 8155's of the time would work with an 11 MHz 805x. The "fast" 8155's and 8255's were 250 ns and, much later, 200 ns (-5) parts, targeted at the 5 MHz 8085 and 8086 family CPU's.

There's been so much discussion of why an 8255 is bad for new designs, and economic and supply-chain issues certainly would be one consideration, but I'd like to see the specific devices that would be useable in its place.

I never liked the 8255 myself, though it did work for many things, because it wouldn't drive a transistor hard enough to do anything. A darlington, well, maybe, but a "normal" 5-cent transistor, e.g. 2N3904, well, not really. This meant that I had to buffer the device's output before it would do any useful work, e.g. drive a relay or LED. Once I put in a 35-cent buffer, I could just as well have used a 35-cent octal flipflop, e.g. 74374, or even 74C374, of which the latter could,typically, and in a pinch, source 24 mA and sink 12 or more. My own position was that, once a configuration was determined for a given application, the "family" logic (74xxNNN) worked out better and lower in cost. Today, of course, that stuff is seldom used, with a US$2-3 CPLD replacing it.

The difference, of course, is that one can slap down a schematic in 10 minutes that contains the MCU, a couple of 8255's and a few other pieces of hardware that provide, at least in theory, a lot of I/O. That sort of thing is for people who don't know how to analyze their own system requirements, however. Yes, the 8255 is far more flexible than what one can jam into a 44 pin CPLD of comparable cost, BUT, if one knows what the hardware requirements are, that fact becomes insignificant relative to the cost and availability of the hardware, because the flexibility of the hardware is no longer helpful.

Aside from programmable logic, however, I'd like to know what people here on this forum are using in place of devices of the 8255 class. It would be interesting to know what the performance of such devices is.

Note that I mentioned the TI 74BCT2424 registered multiplexer in a previous post, in this thread, I believe. It has some really hefty drivers, and demultiplexes a 16-bit port into two bidirectional registered ports, or, essentially a 32-bit port with very hefty output drivers. Does anybody know of something that could replace that device? I've had to build 'em myself, since the real McCoy is no longer listed.

RE