Boasting better than 60,000 customers, Microchip claims to be the king of the 8-bit microcontroller hill. But it's the 32-bit mountain that Microchip sees in its future. To that end, the company has developed its first 32-bit microcontroller unit, dubbed the PIC32. The device incorporates a MIPS32 M4K processor core, surrounded by all the peripherals you would expect in a PIC-based microcontroller.

"We have customers that continually ask for higher performance and more memory," said Patrick Johnson, director of Microchip's high-performance MCU division. "It was the natural evolution for us to move up to support our customers who wanted to be on that path. We're giving our 60,000 customers a family that's compatible among 8, 16 and 32 bits, and a way to migrate up or down."

My first question when I heard about the MIPS-based products was, "why not develop your own core?" And I really liked the answer. An obvious reason is that you can get to market much more quickly by employing an existing core. But more important, it gives Microchip and its customers access to the existing ecosystem that surrounds the M4K.

"We could have gone in lots of different directions with this offering," said Johnson. "We could have chosen a 500-MHz core, or a superlow-power version. But that didn't make sense for the majority of our customers." The MIPS core was chosen after Microchip evaluated a host of other alternatives, such as those offered by ARM, ARC and Tensilica.

It took about two years from the time the decision was made until samples were ready. While that seems like a long time, Microchip wanted to be sure that every last detail had been addressed, including having development kits ready for shipment. The company said it had working silicon for more than six months, and it has already gone through one revision of the product.

Microchip also wanted to ensure that all the external modules developed for the 16-bit MCU would work transparently with the 32-bitter. And the crop of analog components that work with the 8-bit family can all be used for 32 bits. Those analog parts generally operate at 3 volts, and the MCU can operate between 2.5 and 3.6 V. A standard voltage regulator can be employed.

An interesting feature of the M4K core is that it has an extensible instruction set, so Microchip can add instructions to optimize a microcontroller for specific applications.

The family is being manufactured by Taiwan Semiconductor Manufacturing Co. It's in a 0.18-micron process, beyond the current capabilities of the Microchip fabs (which go down to 0.25 micron). The plan is to continue with the TSMC fab for the foreseeable future.

While any silicon vendor can license the MIPS core, Microchip believes it offers some key advantages, starting with the migration path from 8 to 16 to 32 bits. And the MCU offers features that Microchip's existing customers have embraced, such as reset and code detection, brownout reset detection, the ability to push the voltage range beyond ±10 percent, two-wire debug modes, low pin counts, and development with tools already in their arsenals.

"We've found that our customers are almost immediately productive because the PIC32 environment looks like what they're already familiar with," said Johnson.

In addition to being a MIPS licensee, Microchip had some input into the core itself. For example, the core's trace capability borrows some of Microchip's technology, as it consumes fewer pins and gates than a traditional MIPS product. Another example is the complex breakpoints, as the architecture now contains a sequential breakpoint capability. And the number of shadow register sets has been extended to increase performance.