We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.
I am looking for the lwest power Cortex M-?
I don't want to spend years searching the internet for a low-power ARM Cortex controller.
Does anybody know of one?
(And yes, it is Keil related since I'll be using Keil to program it)
Thanks,
--Cpt. Vince Foster 2nd Cannon Place Fort Marcy Park, VA
Hello captain,
check out this link:
www.energymicro.com/index.php
Tamir
Thanks! (It would have taken a while before I stumbled across that)
You probably know, but it bears repeating: power consumption is very application dependent - so getting the application right may be far more important than (just) choosing the "lowest-power" (sic?) processor
Andy,
Yes I do know that, and right after I posted my question, I thought somebody would catch me on that one. I figured it would be one of the regulars.
My application will use all of the features available most of the time when turned on for several hours (and even more peripherals)... then go into sleep mode.
A low-power product for a high-power end result, possibly...
"high-power end result"
This project is to help people and not splatter them.
But yes, this project uses multiple micros and shall be battery operated.
The link supplied by you looks great, so I called them and left a message for 'the guy' to call back. I also emailed 'him' and he did respond: production units will be ready in November/December time frame. He shall provide the pin-out and packaging so I can design it in.
A typical usage needs to be on and last 24 hours before it is recharged.
So. much thanks to you Tamir!
What size is the battery?
Maybe you could give some indications what your power budget is for a single chip. At least allowed power/MIPS for the core. Possibly some indication of allowed power for all the extras.
Most (all?) Cortex-M3 are quite nice when it comes to power/MIPS since the manufacturer buys the processor core as a macro cell and the macro cell for a given technology should have a quite fixed power consumption.
ARM specifies figures down to 0.085 mW/MHz on the TSMC 0.13G process. I think the 0.18G process will consume 0.19mW/MHz. The ARM7TDMI on the 0.18G process is around 0.28mW/MHz.
Another thing is if you are going to use a common oscillator, or if each processor will have their own crystal. Processors will normally consume less power when receiving a nice square wave clock signal.
Unless you are forced to use Cortex-M3, you may also look at the ARM996HS which has asynchronous clocking and have a core specified to only 0.045 mW/MHz.
But in the end, the peripherials may be the big difference, since they can stand for a significant percentage of the total consumption and are what differs between the manufacturers.
I haven't had the time to read through all the information from the link Tamir posted, but did the guy you got in contact with inform you exactly what Energy Micro did better than other Cortex-M3 manufacturers?
Battery Size is TBD dependent upon the electronics, but so far the calculated budget allows for a 32 hours on a 2800mA/H LiIon: in what I think is 'worst case'. But of course simulation versus the real thing can be wrong.
Per is right about the external peripherals' power usage. Those have been addressed and mitigated.
A purpose behind knowing a low-power Cortex is an evaluation of the actual MIPS truly needed to accomplish the same goal as a 'fast' 8051.
Like cars, you'll find that they size the gas tank to ensure at least 300 miles between refueling to prevent the owner from thinking it is a gas hog, or get annoyed by constantly going to the gas station.
The same will happen on this project. I don't want the customer spending 'too much time' charging the unit.
Although I said before, that 24 hour mark is for the heavy user who uses it for 8+ hours at a time, while the typical user would use it for about an hour at a time.
Another aspect of the project is to convert the discrete CPUs (uCs) into an FPGA fabric. Thus the code-monkey-book shall be followed to make that "C" to VHDL transition... then I'll clock it down to reach for optimal battery life while still maintaining performance.
MCB1000 ?
No such thing listed at http://www.keil.com/boards/
"2800mA/H"
I expect you meant 2800mAh @ 3.6V since you need voltage*current*time to get energy, and that you are talking about a one-cell battery. Say 8W usable after losses from a DC/DC.
Counting 12 hours operation and a 90% power reduction for the idle hours, that would give about 0.6W when active and 0.06W in idle. Assuming 50% for the cores to process instructions, and 50% for internal modules (DMA, UART etc) and external logic, you would have 0.3W for instruction processing.
If the claim of 0.09mW/MHz is true for the 0.13G process, your 0.3W would then be enough for 300/0.09 [MHz] = 3.3 GHz or 33 processors running at 100MHz.
Processors using the 0.18G process would give you 300/0.19 [MHz] = 1.6 GHz.
I really must have computed something wrong, but the figures sounds quite large ;)
Just a footnote, but I normally add a *0.5 as safety margin when doing initial calculations, to cover ageing of batteries, unknown late-stage requirement changes etc and still have a bit of extra engineering margin. Better to be able to reduce the battery size after measuring on prototypes than to have to figure out how to fit a larger battery pack in a too small box after the tools for making the box have already been produced.
Absolutely!!!
Follows the general "make it work first; optimise later" approach.
Please check Cypress PSCO5 (available only on january)