Which avr is the lowest power?
I have an application where I need to run off a watch battery and before I start scouring all the data sheets, I thought I would ask here first.
official AVR Consultant
Not sure but logic says a "small die" PV probably?
(unfortunately a lot of the datasheets say "power consumption: TBD")
Most of the answer will be "it depends". We could recommend a Tiny10, say, but you need more I/O or other resources.
When you get down to "lowest", much will depend on the rest of your layout--the AVR's difference of say 1uA average from one model to the other could well be dwarfed by other draws and then be essentially n/a.
"Watch battery" covers a lot of ground. What cell have you chosen, and how many mAH does it have? Then, how long do you want it to last? That gives you your raw budget of average draw that you want to meet and hopefully exceed.
Then you go from there. What proportion of the time is in deep sleep, and what wakeup source is to be used? That gives you the "overhead"; the rest of the capacity can be used for "real work".
One size doen't fit all...
You can put lipstick on a pig, but it is still a pig.
I've never met a pig I didn't like, as long as you have some salt and pepper.
I understand "it depends." I'm looking for the part that is documented to have the lowest sleep current. I have also run into the "TBD" spec and it is frustrating.
Final battery selection will be made to meet the customer's battery life requirements once I get the application optimized for lowest power.
But Lee's right - you need to pick an AVR that offers the facilities required for the solution then out of a number of devices you've narrowed it down to you pick the one that can achieve the lost power. My bet, also, was that the new Tiny10 model might be the "lowest" as it's almost certainly the smallest chip of silicon that Atmel put an AVR label on, but suppose that 6 pins aren't enough for what you want to achieve - then what?
Ahhh, I love this scenario. :) It goes like this:
Engineer: I've optimized it for the lowest power!
Customer: Works great! Now how do we get it to last twice as long?
Engineer: But it already goes 20% longer than the spec you gave me.
Customer: Really? I wonder when we wrote that. It needs to last about twice as long as what it's doing now. So when we make those changes, can we reduce the cost on the board?
documented to have the lowest sleep current
So you take a crack at it. Pick a cheap common battery like CR2032, or whatever space/budget allow. Guess at the lifetime desired. Calculate the average draw.
Feasible? If not, back to square 1. Otherwise continue...
It looks like the front page of most P/PA datasheets has either 100nA or 0.1uA for sleep current.
A CR2032 is roughly 250mAH. So at a 100nA draw, it could power that AVR in power-down for 2.5 million hours: 100,000 days; 285 years. As the shelf life/leakage is about 1%/year, this tiny draw becomes immaterial. (In fact once you continue this analysis with most AVRs spending most of their time in deep sleep, you'll find that the total sleeping current usage becomes a non-factor. Hence my harping on it. when I get the total average app draw below the battery leakage current, I quit.)
I appreciate all the design input, I really do. But I simply asked which pico power device is the winner at this point.
I worked for a guy who complained constantly that when he asked me what time it was I told him how to make a watch. He was right.
I need to take a party of school kids on a camping trip but I want to keep fuel costs to a minimum - wonder if I'm going to be able to pack them all into a Fiat 500?
I simply asked which pico power device is the winner at this point.
As I said, the critical number to me is with the wakeup source enabled. I often use the watchdog timer for a periodic wakeup source. Now we are at a few uA draw, so the actual sleep current is a very small percentage of that.
But if you are using pin-change (say) then it is a different matter.
I guess you are just going to have to spend $20 for an assortment of P-chips ...
But I simply asked which pico power device is the winner at this point.
But it's the wrong question (as already stated). The answer is going to be applicatione specific. Any answer we provide is meaningless without the context of your application.
First thing you need to do is nail down your minimum application requirement in terms of AVR hardware resources. At that point you can ask, which AVR (having X, Y, & Z) has the lowest power consumption. [in the mode(s) you require]
Otherwise the answer is going to be the tiny10, it is the smallest AVR die (and possibly newest), with the fewest gates, as such it is going to consume the lowest active current (in theory). [tiny10 boasts <100nA power-down sleep current]
[getting back to your "what time is it?" example. Certainly if someone asked you that in person, it's an easy answer, as you are both located in the same time-zone, so that detail can be assumed. If you were, however, to ask this over the internet, the question lacks enough information to provide the answer you want... we would need to know what timezone you want to know the time in.
Writing code is like having sex.... make one little mistake, and you're supporting it for life.
Iluvatar is the better part of Valar.
Do they have good balance?
[tiny10 boasts <100nA power-down sleep current]
Now compare to the same chart for Tiny24--that one looks to be well >>under<< 100nA at room temperature.
[It looks like they are all higher at higher temperatures--another factor that is of no importance to 'dad?!? [["Have you stopped beating your wife yet? Answer 'yes' or 'no'"]] ]
It looks like the Mega48PA will run at near 0 current if you freeze it. The winner in the datasheet race so far. [~100nA at room temperature]
If one looks at the Mega164 chart, the power consumption is >>lower<< at room temperature than frozen, and definitely lower than 100nA. The room-temperature leader now...
A Tiny48 is a hair higher than a Mega48. ;)
Whoa--Tiny25 looks good at low voltage. Both room temperature and frozen.
Small-die fans: Want a surprise? Look at the Mega640 chart. Next to nothing.
Tiny261 is right in there as well.
But let's back up the 100nA truck...beep-beep-beep...
All the above is at the 1.8V level. What supplyV are you going to run at? So how does the "yes-no" answer apply?
A CR2032 is nominal 3.0V. Let's see about power-down current (interpolated from the charts) for 3.0V at room temperature:
So if you run off a 2V cell the answer is different than off a 3V cell.
But we are only supposed to give one answer...
theusch - Thanks for scouring the data sheets. You went out of your way on this one and I appreciate it.
The ATXMega16A4 is ~100nA given the same criteria. That should transport my camping school kids and some.
Yes, I still beat my wife - she's no good at canasta.
You all are surely aware of what "typical" means.
(though I'd blow the extra 15nA and get access to all those lovely mega48PA goodies myself)
So the smallest die (Tiny10) has the lowest value in reality? Wink
The ATXMega16A4 is ~100nA given the same criteria.
One of the reasons I asked the question is that the ATMega48/88(P)(PA) has been my device of choice for several projects now, and this new project was different enough that I was thinking there is something else out there I haven't discovered because of my mega48 blinders. Turns out to be a very viable choice once again.
Also, the PA is going to draw ~1/3 less when active as well.
I haven't done any low-power apps with any other PA chips. The Mega164 (family) has good curves and the PA migration app note also claims a 1/3 improvement in active draw.
© 2020 Microchip Technology Inc.