This month, we take a look at the battle for the "world's lowest-power MCU" and what that really means to designers (hint: look closer).
Not surprisingly, we saw announcements from many microcontroller companies this month at Embedded World along the theme of “our new part achieves the lowest (blank) power …” followed by the supporting fine print. I honestly haven’t tried to compare the numbers until now.
The great news for designers is that MCU vendors are paying attention. One told me that customers who previously weren’t too concerned with power have now figured out that low-power devices usually cost less from a system perspective, with less board space (due to integrated power-saving features like DC-DC regulators) and smaller and less expensive power supplies. The upshot is that even if a device isn’t battery operated, it makes sense to use the least power necessary.
So why is there an intense battle for the lowest-power spec? If a part supplies the features you need at the cost you’ll accept and delivers the lowest power in your application, then I suppose it matters what the specific number is. But, as with all good specs, your mileage may vary.
Let’s boil the swamp and sample some recent announcements, looking only at new MCU families and where the stories were focused instead of getting too far into comparing numbers:
· Microchip went for an active current spec of 50 microamps/MHz on their new 8-bit PIC parts. In their briefing, they went beyond that and spun it to say that these new parts could operate faster on a given CR2032 battery because of this lower active power. Also, they are pushing the 20 nA sleep current of these families at the low end.
· TI’s MSP430 “value line” parts focused heavily on what you get for 25 cents, but during their briefing they pointed out that MCUs sleep more than 99 percent of the time, so their figures target standby ranges – 0.1 microamps RAM retention, 0.4 microamps standby, and 0.7 microamps Real-Time Clock (RTC) mode.
· Silicon Labs brought a sub-GHz radio to the party, so their active current figures are a bit higher for the integrated part but probably lower for the system at 160 microamps/MHz. During their briefing they also focused on sleep current, down to 315 nA in RTC modes and as little as 25 nA with RAM retention.
Wow. That was just a summary of what crossed my desk in the past few weeks, and you can see how all over the place the numbers are. I’ve done plenty of trade studies, and I’m sure you’re wondering the same thing – how to make sense out of these numbers and choose a part from the shelf.
There are a few things to consider when making a choice:
1. Invested? Chances are you’ve significantly invested in an MCU architecture and its coding and debug strategy. One fundamental guideline that never seems to change much is that for swapping to make sense, you should look for a 4x improvement minimum and ideally 10x. Chasing a particular spec for a percentage improvement doesn’t necessarily pay off. You are often better off pushing your vendor for roadmap improvements. (These announcements show vendors responding to exactly that.)
2. Stuck? Until the spec happens to hit your sweet spot with the right feature, that is. Let’s say you were looking for a part that could double the frequency you could operate using a CR2032 cell. Or maybe you were stuck at 10,000 ADC samples and need to go to 200,000 samples without increasing power. Or perhaps a sub-GHz radio requirement just popped up. Maybe there’s some other requirement that just happens to fit your application.
3. Overworked? We mentioned the sleep metric. One other point that was made during the briefings was that you can actually save power by waking up quickly, working harder and faster less often, and getting back to sleep. A faster, higher-power active core might win the overall power battle if it computes, samples, and sends data faster.
4. Measured? Maybe your application will get the performance quoted in these press releases. Maybe it won’t. This isn’t the PR department’s fault; it’s a reality – these figures have been measured under certain conditions, and your application might be and probably is different. Get a development kit and figure out exactly what you have in front of you; don’t blindly follow a data sheet until you have firsthand experience.
Now maybe you’re totally messed up on these specs, but I hope not. I’d like you to tell me what you see here. Whatever the exact power specs are, I think the trend in MCUs is outstanding, and even this limited sample of announcements shows the creativity and engineering that is going into cutting power consumption. MCUs are clearly powering down as technology gets better, and the result will be more efficient embedded devices.