A 30-Picowatt Processor For Sensors

Roland Piquepaille writes "University of Michigan (U-M) researchers have developed an ultra low power microchip which 'uses 30,000 times less power in sleep mode and 10 times less in active mode than comparable chips now on the market.' It only consumes 30 picowatts in sleep mode, which means that a simple watch battery could power the chip for more than 200 years. Of course, this is not a processor for your next computer. It is designed for sensor-based devices such as medical implants, environment monitors or surveillance equipment. However, the design is very clever." Roland's blog has some more information, including a die picture of the chip, known as the Phoenix.

Doc Oc...

[clang_jangle (975789)]

...might want a Beowu ---- oh, forgive me. I know where the door is...

Re:

[by Anonymous Coward]

I don't think you know well enough yet. Let me give you a foot-- er, hand.

Watch batteries don't last 263 years...

[Cliff Stoll (242915)]

The lithium CR1216 batteries on my shelf started corroding after 4 years. Several of the AG3/CX41 alkaline batteries began leaking after 5 years. Still untouched, in their wrappers.

Re:

[dean.collins (862044)]

interesting, I was at the East Coast DFJ venture capital competition a few weeks ago where one of the contestants was a 'radiation based' battery that lasted 20+ years, these two things together could drive a circuit forever. (sorry dont have links). Dean

Re:Watch batteries don't last 263 years...

[QuoteMstr (55051)]

Are you thinking of an RTG [wikipedia.org]? These things last a few decades before the thermocouples lose the ability to transform the heat to electricity.

Re:

[dean.collins (862044)]

to be honest most of it went over my head but yes it used an isotope to power it, they had sold some samples to lockheed for continuous ciruit running for 20+ years etc at $2000 a pop. one of the points they were making was they could power a pacemake for 20+ years now with this battery, my question was didn't radiation cause problems but apprently not. Cheers, Dean

Re:Watch batteries don't last 263 years...

[Detritus (11846)]

The typical RTG uses Pu-238, which is an alpha source and easy to shield.

Re:

[Sj0 (472011)]

I've always wondered if an RTG coupled to banks of supercapacitors could be the solution for an electric car you'd never have to fuel up during the effective life of the battery? If you ran out of juice, just rest the night and you're ready to rock and roll for another days travel.

Re:

[TheLink (130905)]

RTGs don't provide that much power for the weight (and generate quite a lot of heat). Petrol has about 34 megajoules per litre. 50 litre tank = 1.7 gigajoules.

From the wiki, a 39kg RTG can generate a max of 390W (electricity) and 7.2kW(heat). 390W of electricity after 8 hours will give you 11 Megajoules - the amount of energy in 330ml of petrol (think size of soda can), and that assumes 100% efficiency.

Now if you can also convert all of that 7.2 kilowatts of heat to energy - making a total of about 7.6kW (a

No but Lithiium pacemaker batteries last ...

[celtic_hackr (579828)]

A very long time. I have replaced lithium primary batteries in train equipment that have lasted more than 10 years and still retain most if not all of their power. The batteries are a backup so don;t get used a lot, but the ones used in pacemakers usually last ten or more years. The plain truth here is batteries last as long as the application that they are needed for. Which is why retailers don't sell Lithium chromate batteries to the general masses. Who would ever buy an Duracell when they could own a SAFT Lithium Chromate that lasts 10 time longer?

915x915um^2

[cyfer2000 (548592)]

Can some one explain to me how this chip is connected to the world?

Re:915x915um^2

[CTho9305 (264265)]

That die isn't particularly small - 1mm by 1mm. Plenty of existing chips have dies that size, but they're just packaged in larger packages to space out the pins. This picture [ctho.ath.cx] shows a die (about 3mm*3mm) and the remains of its package so you have an idea of how small a die can be, even in a large package.

Re:

[oldhack (1037484)]

You connect it with Denon's ethernet cable, to get all the most processing power possible.

The article sucks

[Facegarden (967477)]

There is some better info farther into the article, but the first thing they say about the chip is rediculous:

"So how did these scientists build this very efficient chip? The answer is extremely simple: they've reduced the battery size. 'Phoenix is the same size as its thin-film battery, marking a major achievement. In most cases, batteries are much larger than the processors they power, drastically expanding the size and cost of the entire system, said David Blaauw, a professor in the Department of Electrical Engineering and Computer Science. For instance, the battery in a laptop computer is about 5,000 times larger than the processor and it provides only a few hours of power.'"

So... they made it more efficient by giving it a smaller battery? That is so obviously backwards... They can give it a smaller battery because it's more efficient, but not the other way around... Or did i miss something? The article certainly doesn't help explain anything more if that is really come clever something-something going on...
-Taylor

Re:

[rm999 (775449)]

You are referencing the blog, not the article; the article is more clear about it. They explain that the cool part of this processor is that it is so efficient, it only requires a tiny battery the same size as the processor. The main thing that is preventing portable electronics from being smaller is their battery. Case in point: the laptop which has a battery 5000 times larger than the part that is actually doing the work.

Kind of an obvious and not so ground breaking statement, but at least it makes sense.

Re:The article sucks

[lagfest (959022)]

That's how Roland Piquepaille understood it. Stuff like this is why I despise his blag.

From TFA:

The timer "isn't an atomic clock," Hanson said. "We keep time to 10 minutes plus or minus a few tenths of a second. For the applications this is designed for, that's okay. You don't need absolute accuracy in a sensor. We've traded that for enormous power savings."

...

Phoenix engineers used much narrower power gates that restrict the flow of electric current. That strategy, coupled with the deliberate use of an older process technology, cut down on energy leaks.

The art of compromise

[EmbeddedJanitor (597831)]

Processors such as the Microchip PIC10F are down to the nanoWatts level. That is more than fine for many/most purposes. These things cost well less than a buck each and include various peripherals. They can operate in a wide range of voltages and temperatures.

These 30pW sleep mode CPUs will allow things to go to the next level of minaturisation, but will need reduced cost and will need to prove that they are reliable.

There is a huge issue with power consumption vs stability. Basically, each bit in a CPU hol

Re:

[mustafap (452510)]

>These 30pW sleep mode CPUs will allow things to go to the next level of minaturisation, but will need reduced cost and will need to prove that they are reliable

And also, will require that the sensors reduce their power consumption too. A complete system consists of more than just a processor and a battery.

I'm a fan of the PIC10F myself. Used it in a few magic tricks for a guy in vegas. You'd be surprised how much processing power magicians pack these days :o)

Re:

[arielCo (995647)]

So... they made it more efficient by giving it a smaller battery? That is so obviously backwards... They can give it a smaller battery because it's more efficient, but not the other way around... Or did i miss something? The article certainly doesn't help explain anything more if that is really come clever something-something going on...

Yup, real dumb. Closer to the proverbial horse's mouth [umich.edu]:

There's nothing special about its size [...] But Phoenix is the same size as its thin-film battery, marking a major achievement. In most cases, batteries are much larger than the processors they power, drastically expanding the size and cost of the entire system [...] "Low power consumption allows us to reduce battery size and thereby overall system size. Our system, including the battery, is projected to be 1,000 times smaller than the smallest known sensing system today"

The article goes on to the potential new applications with really tiny sensors, mostly embedding hordes of tiny bugs into the target organism/structure for distributed, robust monitoring.

Re:

[Ceriel Nosforit (682174)]

Posted by Roland Piquepaille @ 9:38 am

How I would do it

[kipman725 (1248126)]

sounds like it just does automaticly what a micro programed for use with sensors does. Usualy you have a counter in the chip that is driven from a clock pulse derived from the main system clock that generates an interupt (waking the proccesor up) every time it overflows. As each overflow is the same time interval apart this can be used to run a subroutine that checks whether it's time to run the sensing program yet. If it is the sending program is run and the proccesor goes back to sleep only waking on e

Theoretically...

[seanadams.com (463190)]

What's with all these idiots who think "theoretically" is a synonym for "not really"? This gem in particular:

"Theoretically, the energy stored in a watch battery would be enough to run the Phoenix for 263 years."

Note that it's carefully worded to say "the energy stored in.." not to that a watch battery actually _could_ do this. Because it couldn't. The battery's internal resistance and chemical processes would cause it to drain itself long before you'd ever consume a meaningful portion of that energy.

Only in very specialized applications where you have extremely weak, but continuous sources of power, could you realize any benefit to a picowatt vs a nanowatt of consumption. For batteries or supercaps, the power source will self-discharge at a much higher rate anyway.

Re:

[bh_doc (930270)]

Why the hate on "theoretically"? In the example you gave, they could have removed "Theoretically", and that sentence of the article *would still* have been a correct statement. It's talking about the energy consumption of the processor, not the limitations of the power source. Your suggested substitution ("Not really") would've been wrong.

Re:

[seanadams.com (463190)]

Why the hate on "theoretically"? In the example you gave, they could have removed "Theoretically", and that sentence of the article *would still* have been a correct statement.

In theory, you're absolutely right. :)

Re:

[hey! (33014)]

Well, is it helpful to know that the amount of energy needed to run this in sleep mode or thirty years is sufficient to raise a one pound weight by about 2 1/4 inches [google.com], or roughly 5.6cm?

It may make you happy that the math doesn't seem to right in any case [google.com], based on a CR2032 battery, which is rated at 3V, 220mah. 263 years sounds more like what you'd get [google.com] running off a large electrolytic capacitor [digikey.com], but that is an even more annoying notion.

Re:

[seanadams.com (463190)]

no, I think you're missing the point. They aren't really saying you can use the same battery and have it last 200 years. They're saying you can use a MUCH SMALLER battery and have it last just as long.

Actually, I got that point exactly, which is why I said the watch battery metric is stupid. The problem is that marketroids actually think that more people will understand it if they rate it in "watch battery years" instead of watts, just like storage capacities are easier for us to understand in terms of "lib

Re:

[hvm2hvm (1208954)]

And they are right. I don't work with watts every day so I don't understand exactly what 30pW mean. This is how the human brain works: comparisons with other things in life. You know that a Hummer is big because most other vehicles are smaller. A sky scraper is big only when compared to a 10 story building (which in turn you must compare to your own height). Stop picking on these details, it's pointless. And BTW, your condescending way of saying "I am smart because I can understand the real way of measuring

Re:

[seanadams.com (463190)]

And they are right. I don't work with watts every day so I don't understand exactly what 30pW mean. This is how the human brain works: comparisons with other things in life. You know that a Hummer is big because most other vehicles are smaller.

OK.. so distances are better explained in terms of Hummers placed end-to-end, instead of miles? The watch battery comparison is just as meaningless, because people don't need to power anything for 200+ years, nor do they have an intuitive feel for how much energy a wa

Re:

[hvm2hvm (1208954)]

No, distances are not better explained in terms of Hummer lengths rather than miles. That's because people use miles all day and can relate to them (actually kilometers in my case but that's another story). And at least me, when someone tells me that there are 20km from point A to B I still need to compare that to some other distance I know (like from my home to my school, or from my city to my cousin's). Also, they do give technical information about the power usage. They just put an additional information

Chuck Moore is doing things like that quite often

[K. S. Kyosuke (729550)]

"Measurements show that Phoenix consumes 29.6pW in sleep mode and 2.8pJ/cycle in active mode.â

Maybe I am miscalculating something, but did not Chuck Moore [slashdot.org]'s x18 perform at about 8.3 pJ/cycle in 2001 already, seven years ago? That was at 1.8 V. I guess things would get much better at 0.5 V. But their idle power is something that seems truly interesting to me. Too bad there is not much info on the design right now. Finally something interesting!

bs

[serbanp (139486)]

So, what's so special about this IC? Sure, it sits idle most of the time (except that they need to have a RTC to wake up every 10 minutes). That's what any digital CMOS IC would do if the clock would be cut off.

A turned-off switch in series with the system would drop the consumption to even lower levels while in sleep mode.

Somebody is desperately trying to justify grant money spent on pizza and beer...

Re:

[SeekerDarksteel (896422)]

Yes, because there's no such thing as leakage current. And no one uses DRAM or SRAM....

You can't just stop the clock then start it back up and expect that power consumption will be negligible during that time or that the circuit will work properly when the clock is restarted.

Re:

[imgod2u (812837)]

If I'm reading the article right, they're not just "turning the chip off". They're putting it to sleep with the important states retained. That is, the chip will wake up in the state it was put to sleep. This reduces the need to have a boot-up sequence and state restore, which, for chips that target the pico-watt region, eats up a lot of available energy.

10 Times Less? How is that possible?

[b4dc0d3r (1268512)]

I have one item at 10 units of electricity. I have another item using 10 times less electricity. So it uses (10 units * 10 = 100) 100 less units of electricity, for a total of -90 units.

Does that make any kind of sense to any of you?

Wouldn't you want to say 1/10th and 1/30,000th? Or even be cool and say "one order of magnitude" or even "5 orders of magnitude and a third applied to the result".

(please disregard the less/fewer issue here, one thing at a time)

Sleep for 200 years?

[Chris_Jefferson (581445)]

Wow, I can power a chip for 200 years in sleep mode! Wait. I can power my computer for even longer if I just turn it off.

Bad article

[UK Boz (755972)]

The guy (who admits to not knowing his stuff so perhaps we can forgive him) really hasnt got a clue

The processor is designed specifically for sensors that wake up, do a few calculations and go back to sleep, these type of devices are genrally battery powered and off grid and generally make a decision whether to power up some other device eg to transmit the data. The device would probably be useless for anything involving serious processing, even the processor in an optical mouse would probably wipe the floor with it!

Barring that there are billions (yes billions not millions) of sensor devices out there currently using PIC/AMR/8051 derivatives that may benefit from this technology.

Interestingly we are getting to a level of power where even the most inneficient generator (or a low power radio signal) and a rather small capacitor could power it forever

Re:

[Muad'Dave (255648)]

I'm currently using these critters in that very application: http://www.digi.com/products/wireless/zigbee-mesh/xbee-series2-module.jsp [digi.com] A ZigBee module that can wake up, take a few sensor readings, transmit them to the 'mother ship', and go back to sleep. To me that's the cat's meow in distributed sensor technology - a fully meshed sensor array that can react to external interrupts and transmit data wirelessly and run on batteries for years. Their dev kit contains 5 modules, RS232 and USB dev boards, anten

If it's 30-picowatt in Sleep mode...

[Siriaan (615378)]

Surely it's a bit of a stretch to call it a "30-picowatt Processor"?

Yes

[dreamchaser (49529)]

It's a lot like my 0 watt lightbulb. It uses 100 watts when turned on, and zero when turned off.

Body Heat Power

[Doc Ruby (173196)]

300pW is 7.17017208E-11 calories per second, or 6.19502868 millionths of a calorie [google.com] per day. Over 160,000 of these sensors could be powered on a single calorie a day. Adults eat about 2500-3000 calories a day or more.

If these devices can be powered by a nanoscale heat engine, they could live indefinitely, as long as their host human is alive to measure.

Re:

[assassinator42 (844848)]

Adults eat about 2500000-3000000 calories per day.

Re:

[Doc Ruby (173196)]

True. Damn nutritionists use "calorie" and "kilocalorie" interchangeably. That really burns me up.

Re:

[Muad'Dave (255648)]

So I could embed these in my skin or lungs (wherever there's a temperature differential) and power them via body heat? Cool idea.

Watch battery for 200 years?

[nmg196 (184961)]

> a simple watch battery could power the chip for more than 200 years

Rubbish! Even if you draw ZERO power from a watch battery, it will be totally flat in less than a tenth of this time. They have a 15-20 year shelf life and obviously that will only get worse if you put it in a device that draws power from it. You would need at least 10 batteries to power the device for 200 years.

Re:

[kvezach (1199717)]

Use an RTG [wikipedia.org] with a very long-lived isotope, perhaps? Usually, long-lived isotopes put out too little power to be of use, but most anything should be able to provide enough for a 30 picowatt CPU... (Still, the efficiency would suck because of the Carnot limit; it might be better to find a beta emitter and make use of the electrons directly.)

Re:

[Muad'Dave (255648)]

Very interesting book excerpt on that very subject: http://www.nanomedicine.com/NMI/6.3.7.1.htm [nanomedicine.com]

This paper outlines using a Gd148 source to power medical implants. Fascinating.

A solid sphere of pure Gd148 (~7900 kg/m3) of radius r = 95 microns surrounded by a 5-micron thick platinum shield (total device radius R = 100 microns) and a thin polished silver coating of emissivity er = 0.02 suspended in vacuo would initially maintain a constant temperature ... [of 600K] ... with a 75-year half-life, initially generating 17 microwatts of thermal power which can be converted to 8 microwatts of mechanical power by a Stirling engine operating at ~50% efficiency.

My thought would be to skip the Stirling engine and go RTG.

Re:But...

[kipman725 (1248126)]

no

Re:But...

[mustafap (452510)]

No, it runs an age old Operating System called "Main Loop".

Re:

[MobileTatsu-NJG (946591)]

"We are not here to make you money with clickthroughs..."

Your rant about Roland served Slashdot some ads. Here's a thought: Don't post any comments in Roland's threads, then suddenly Slashdot will have an incentive to toss the tosser.

Re:

[jacquesm (154384)]

or better still, when you see RolandP take a weeks break from /.

(at the current RolandP posting frequency that would mean ./ will soon only consist of the editors and Roland...)

What really gets me is that I get suckered in all the time to just take the stories one by one without checking who posted them and after reading the summary I sort of get this sinking feeling, check back and sure enough... It's like being rickrolled only worse.