My Two Bits About 8 Bits

OK, time to own up: With 32-bit MCUs becoming so inexpensive as well as capable, I figured 8-bit MCUs were hanging in there for extended legacy designs. I mean, who would base a new design on a 35-year old 8051 architecture? As it turns out, plenty of people. 

First a clarification: I doubt that anyone has produced a straight 8051 MCU in 20 years; the newer designs surround a modified 8051 core with plenty of high-speed analog and digital peripherals that can often interoperate without having to wait on the processor core. If your proposed TurboFramistan doesn’t require a lot of high-speed data processing, then a slower, simpler processor makes a lot of sense. 

A case in point is Microchip’s new PIC16(L)F170X and PIC17(L)F17F 8-bit MCUs, which they rolled out recently at EE Live. Actually, Microchip has a large investment in its 8-bit PIC16 family, which it continues to expand. If you look at a block diagram for any of them it takes a moment to find the CPU core amidst all the peripherals. A simple core at legacy (read: low-leakage) line widths—plus, with all due respect, some slick engineering—enables what Microchip calls “eXtreme Low Power,” with active current down to 35 μA/MHz and sleep current as low as 20 nA. Microchip is hardly the only MCU manufacturer to post impressive power numbers, but the 8-bit architecture makes them a lot easier to achieve. 

Despite a heavy investment in low-power 32-bit ARM cores—both its Precision32™ family and EFM32 MCUs it acquired via Energy Micro—Silicon Labs is still building out its extensive 8-bit C8051 family. Their new C8051F85x/6x MCUs include a lot of analog and digital peripherals to target motor control applications. This is nothing you can’t do with a 32-bit MCU but try doing it for 30 cents. 

The bottom line is just about everyone makes 8-bit MCUs: Atmel, Analog Devices, Cypress, Freescale, Maxim Integrated, Infineon, Intersil, STMicroelectronics, NXP, ON Semiconductor—the list goes on. Some of these companies are more committed to 8 bits than others, and some like Texas Instruments are working hard to lure designers to “upgrade” to 16 bits. 

But legacy considerations aside—and cost, which is still a big factor—there are good reasons to stick with 8 bits. If your design will be compute intensive, then 32 bits is the way to go. But if it will mostly just sit there—like a sensor node—and only wake up briefly to take the occasional data sample, then an 8-bit MCU is a logical choice. 

We all love the latest technology, but sometimes an updated version of an old standard is still the way to go.