If you’ve ever tried to make a really low-power circuit — especially one that runs on harvested power — you have probably fallen into at least a few of the many traps that await the unwary in this particular realm of electronic design. Well, Dave Young has been there, seen the traps, and lived to tell about it. In these territories, even “simple” systems can exhibit very complex, and sometimes downright confusing behavior when all possible operating conditions are considered. In his 2019 Hackaday Superconference talk: Scrounging, Sipping, and Seeing Power — Techniques For Planning, Implementing, And Verifying Off-Grid Power Systems, Dave discusses a number of these issues, how they interplay with low-power designs, and tricks he’s collected over the years to design and, more importantly, test these deceptively simple systems.
Dave is an electrical engineer and his company, Young Circuit Designs, has worked in the test and measurement, energy, and low-power consumer industries. We were lucky to have him share some of his 15 years of experience on the Supercon stage this past November, specifically discussing devices powered from harvested energy, be it wave energy (think oceans not RF), thermal energy, or solar. The first lesson is that in these systems, architecture is key. Digging deeper, Dave considers three aspects of the architecture, as mentioned in the talk title: scrounging, sipping, and seeing power.
Scrounging Power
Dave cautions about overly optimistic estimates of available power. For example, a 6W-rated solar panel produces an average 4.11 watt-hours per day when averaged over the entire year at his location in Rochester, NY. This is, of course, much less than you might expect. Such a discrepancy isn’t unique to solar systems, either. So much so that Dave cautions us not to overlook another possibility: eschew scrounging altogether and use primary batteries with enough capacity to run low-power systems for their design lifetime. For ultra-low-power systems, this is certainly a possibility, even in environments where “free” power is available, and may represent an easier and more reliable design alternative.
There are other issues with solar systems, too. If you’ve done solar design, you may have seen maximum power point tracking (MPPT), because solar cells yield a variable amount of power depending on the load. To extract the maximum power from the cells for a given illumination, you may have to adjust the voltage that the cells run at by changing the current drawn. There are integrated circuits designed exactly for this purpose, and for large systems, this kind of architecture is a no-brainer. However, Dave points out that these MPPT ICs can consume 5-25 mA themselves, which can easily negate their benefit for small systems. Sometimes a linear charge controller is simply more efficient because it consumes less power itself.
Sipping Power
Besides the power source itself, a design must consider what to do with the power. In the second section of his talk, Dave starts with the premise that energy harvesting systems are always in either one of two states: they have power to burn, or are just about to run out of it entirely. While you can control the amount of time your system spends in either state with proper design, you must carefully consider how your system will behave in each. This feast-or-famine existence means you need to create a power breakdown for each state, and not just a single overall budget. This implies a careful examination of all the conditions the system may find itself in. For example, solar systems will have different power available during various times of day and year; coupled with different levels of battery charge, this creates a number of states that must be examined, and leads to a number of possible failure points.
Seeing Power
In the final segment of his talk, Dave discusses seeing power: in other words, determining where in your system the power is going. Here again we have the benefit of some hard-learned lessons. Fist, he suggests verifying your power budget(s) at the line-item level. By turning off chips one at a time with firmware, you can verify that their actual power consumption matches the design goals. For those parts that can’t be disabled in firmware, he recommends simply removing them from the PCB as you go. The value in such an exhaustive test is that you won’t be fooled by a system that meets the overall power budget but for the wrong reasons: such a system is likely to perform incorrectly under some combination of conditions.
There are many more lessons in Dave’s video than we can cover in this brief discussion. Definitely check out the full video of Dave’s excellent talk, and if you want to follow along at home, you can find the accompanying slides here.
Video is private
Still Private
My fault, updated now. Sorry about that.
Low power is a topic I enjoy reading about because there are always neat tricks that make you stop in your tracks and stare for a second.
Making sure that your GPIO is not secretly leaching power, or doing the math and finding out that always running and waiting for a change can be more wasteful than hibernating and waking up on a schedule.
There were a few Sparkfun articles I found that played with this and they were a good introduction. “Adventured in low-power land” or something…
It seems that they have buried the old blog posts and this is all I can find now. Still helpful, but more clinical and less story.
https://learn.sparkfun.com/tutorials/reducing-arduino-power-consumption
https://www.sparkfun.com/news/1842
Fix the video please, i’d love to watch it.
Thanks for fixing the vid.
Why does he say “all the chips” that do MPPT require 5 to 25mA (around 7 min into the video)? I’m assuming he’s refering to the quiescent current. There are several chips on the market that require nowhere near that. For example the TI BQ25570
“Full Operating Quiescent Current of 488 nA (typical)”
Max 900nA (700nA at 25C)
http://www.ti.com/lit/ds/symlink/bq25570.pdf
Yeah, that’s quite a ways below 5mA!
Yes thank you, this chip is several years old by now so it’s surprising he didn’t even mention it and just asserted a power consumption 3 orders of magnitude too high… What I would’ve liked to see is a comparison of modern popular harvesting ICs, (and even a discussion of indoor solar) especially between the BQ25570 which uses a boost converter to charge the storage element and buck converter for the system, vs. the AEM10941 which has lower cold start voltage but only has a boost for charging and uses LDOs for 1.8V/3.3V
Wish I’d seen this a few weeks ago. My xmas project was to build an oven thermometer, and I didn’t want to mess with a wall-wart. It’s a thermocouple, arduino pro mini, and an always-on LCD that updates 1/sec. It all sucks 50 microamps, which would only give me a half year on a CR2032. The set of AA cells I jammed in there should last 5 years, but it would be nice to run off a smaller battery.
Seven years ago the city water meter reader showed up at my door, as they did every few months. My house is pre-WWII — meter is in the basement, and I routinely don’t let them in the house, and just phone in the reading. This time, though, he was carrying a replacement meter to install that could be wirelessly read from the road: I’d never be bothered to read a meter again.
Naturally, as soon as he was gone, the cover was off, and Lo! I see there’s no fancy energy harvesting, no water wheel, no apparent power source. But there is: buried in the permanent silicone encapsulant is a pair of lithium cells that are supposed to last 20 years. This on an a device that has an always-on RF receiver (to listen for the poll from the roadside reader), and a RF transmitter, as well as the usual encoder on the wobbling-disk water meter itself.
7 years later I’m still getting correct bills, so it must still be working, and I know they have not changed it out. But it’s still a mystery to me how they got the power consumption of a datalogger and an always-on radio receiver down to the self-discharge rate of a lithium primary cell. I assume it just wakes up every few seconds to listen if there’s a beacon present.
It does have the FCC ID on it: fccid.io/EWQ100WC but it doesn’t shed much light on the low power magic.
As you speculate (“I assume it just wakes up every few seconds to listen if there’s a beacon present”) it isn’t always on.
My electronic bathroom scale lasts 10 years on a pair of CR2032 batteries, and it has to wake up and check the sensor often enough to tell if I’ve touched it with my toe to turn it “on.”
Batteries have insanely large amounts of power compared to the needs of low power circuits. And very little power compared to the needs of a carelessly built circuit.
Decades ago I read something about metersbeing activated by a radio signal from the passing truck. Transmit on one frequency, and the reading sent back on a harmonic.
This was early on, but it was presented as the way things would go. But maybe there was a change.
This must be the most (in)famous device in that catergory.
https://en.wikipedia.org/wiki/The_Thing_%28listening_device%29
And that was in 1945.
Previously featured right here in these hallowed pages… https://hackaday.com/2015/12/08/theremins-bug/
Your AMR system is an Itron, and the design of that device is really cool. I’ve torn a few down and analyzed them, they are running at a very low single microamp load most of the time. The designs I looked at are on ceramic boards with printed resistors ( no SMT devices ), laser trimmed and the caps are similarly done. The batteries at the time were from the “purple battery company” and those are a unique design as well. They boast an almost zero self-discharge – as long as the battery never sees a load above a certain level. Otherwise, it damages the passivation layer in the battery and it begins to self discharge at a relatively high rate ( as compared to the device drain ). The product design is still ahead of the curve for electronic assemblies and deserves a full review in these halls.
I spent many years in the AMI world, this is one of the coolest designs out there.
” as long as the battery never sees a load above a certain level”…
Yep….As well as the two battery cells in parallel there is a huge electrolytic (well, it *looks* like an electrolytic — might be a solid electrolyte or something else long-lived), that I can only assume supports the current for the RF transmit pulse, which those lithium thionyl chloride cells wouldn’t be able to handle.
The design is visually beautiful and obviously very high quality. I’d love to tear it down more than I have, but it says it has anti-tamper features in it, and I’d rather not incur the wrath of city hall, who technically owns the thing.
Would love to see your review of this Itron design. I love low power tricks.
Talking about low (no) power: https://en.wikipedia.org/wiki/The_Thing_(listening_device)