Low power micro controller?

[SeungminLee]

I have used atmega series for MCU, however, I found that its power consumption is too high for my project so I decided to change MCU.
Also I tried MSP430 series but power reduction is not satisfied..

The functions I need are
: SPI comm.
: UART comm.
: Port control - high or low TTL (6~10 pins)

What do you think which MCU would be best with low power consumption for this use?

 

[keith1200rs]

I think it depends on whether you can power down or sleep for part of the time. You do not say what power consumption you need or the speeds of any of the interfaces.

Some of the Microchip PIC NanoWatt devices are quite low power even when running, not just in sleep modes. Devices such as the PIC18F13K22 or similar for example, seem to have a lot lower running power than similar, older devices.

Keith.

 

[millwood]

lower power consumption is more of a software issue: how careful you are at configuring your mcu and how you make it interface with the outside world: I have powered a 12F675 driving a lcd display from a cup of salted water.

in general, 430 has some of the lowest power consumption, and the faster arm chips (arm7) have the highest power consumption.

but it is up to you to get the chip to run at low current consumption.

 

[keith1200rs]

I don't agree that it is totally down to how you program the micro. Some of the newer chips using smaller geometry processes and with internal low voltage regulation for the core make a massive difference.

Keith

 

[bobsanjose]

The information you provide is insufficient. How fast do the peripherals need to go, what else does the CPU need to do, is it s one time thing and then go to power down, is it a recurring thing, how often....????
Depending on your answers the recommendation would be different with any of the parameters changing. In general there are some very low power MSP430 out there, some very low power Cortex-M3, e.g. Energy Micro,
**broken link removed**
some Cortex-M0 e.g. LPC1100...
**broken link removed**

A few things that can be relevant based on different needs.
Lowest standby power
Lowest power while running at the frequency and using the peripherals you need
Fastest oscillator startup time from sleep
and more of this kind.

Definitely more information in the answer than in the question

Bob

 

[millwood]

I don't agree that it is totally down to how you program the micro. Some of the newer chips using smaller geometry processes and with internal low voltage regulation for the core make a massive difference.

Keith

let's check with the data.

12F675, a 10(?) year old chip:

operating curent: 8.5ua @ 32Khz and 2v, or 100ua @ 1Mhz and 2v.
standby current: 1na @ 2v.
not rated for extra low-power.


16F193x, a 1-yr old chip:

opearting current: 7.2ua @ 32Khz and 1.8v, or 150ua @ 1Mhz and 1.8v.
standby current: 60na @ 1.8v.
rated for extra low-power.

so a few things we can learn from the above:

1) regardless of their "low power" rating or not, they consume very little power.
2) current consumption goes up with frequency.
3) not much difference between xlp rated chips and non-xlp rated chips.
4) you can significantly reduce the current consumption if you put the chip into standby, by a factor of hundreds to hundreds of thousands of times.

just for your interest, the rating for a lpc2106 is 2.5ma at 28Mhz and 1.8v. the if you could run 16F193x at 28Mhz (28Mhz*4), its current consumption would be roughly 16.8ma.

the other way to say the same thing is that if you ran a lpc2106 at 250Khz (core) frequency (equivalent to 1Mhz for the PIC), it would consume about 22ua at 1.8v, far less than the XLP-rated 16F193x.

again, this shows that what's "lower power" is relative and there are a lot of things a programmer can do to lower current / power consumption.

 

[keith1200rs]

Yes, if you compare two totally chips with completely different on board peripherals, memory etc.... you prove nothing. A comparison of devices with similar capabilities on chip (memory, peripherals, processing speed) would yield a different answer.

Leakage current can be a problem for smaller process geometries, but they make up for it if you need to actually do some processing.

Keith.

 

[bobsanjose]

millwood,
given a must have set of features SPI (how fast) and UART (how fast), the 32 kHz comparison is probably useless. Latest Cortex-M0 and Cortex-M3 run in the neighborhood of 180-250 uA/MHz and that is from a 32-bit device, would compare to a PIC not just running 4 MHz but probably more like 8-10 MHz if you need to handle 16-bit SPI data.
On the other hand I totally agree with you in terms of "low power" naming, frequently bogus and once again, all relative.
My favorite comparison usually relates to cars because almost every engineer can relate to these.
The statement that a new car has better mileage than an old one is correct if you compare apples to apples but a 1980 VW Rabbit still offer much better mileage than a 2010 Hummer. And by the way, the difference is not just software

Keith, you are so right that an old device with 0.25 um or larger geometries will fare much better in leakage while a new comparable feature device will do better in active current while executing the same program from the flash.
My favorite overall in regards to low power is EFM32, very low stand by with RTC, low power SRAM retention, low power active and very fast oscillator start up from sleep. Energy Micro thought about all aspects of low power. They also have low power UART, ADC and more. A device with an on-chip oscillator that can not be used for the required UART will not be suitable for an application that goes to sleep frequently. ......
Not enough information to make a qualified recommendation.

Bob

btw. I do these kind of feasibility studies for a living, so if you need an in depth evaluation of a list of low power requirements, let me know.

 

[millwood]

Yes, if you compare two totally chips with completely different on board peripherals, memory etc.... you prove nothing.

that example is to show you that just because a chip is label'd "low power" doesn't mean it necessarily consumes less power than a chip that is NOT label'd "low power".

you cannot compare two identical devices because they don't exist. all you need to do is to understand two things:

1) cmos devices consume current only when they are in transition, from on to off and from off to on. so everything else being equal, lower frequencies -> lower power consumption.
2) a mcu consumes much less when it is in standby. so if you want to drastically reduce power consumption, program the chip so it stays in standby as much as it can.

A comparison of devices with similar capabilities on chip (memory, peripherals, processing speed) would yield a different answer.

maybe you can add your examples here?

 

[kender]

I have used atmega series for MCU, however, I found that its power consumption is too high for my project so I decided to change MCU.
Also I tried MSP430 series but power reduction is not satisfied.

For our reference, could you post power consumption in all 3 cases: original atmega (your experimental), MSP430 (your experimantal), desired (your spec).

Cheers,
- Nick

 

[keith1200rs]

The statement that a new car has better mileage than an old one is correct if you compare apples to apples but a 1980 VW Rabbit still offer much better mileage than a 2010 Hummer. And by the way, the difference is not just software

Bob,

The issue of cars is a different one and easily solved - weight! Cars used to be a lot lighter. As engines have improved in efficiency the cars have got so much heavier any improvements are lost. I had [edit - long since in the scrap yard] a 1980 VW Golf and it weighed 15CWT. Look at a similar sized car now and they weigh at least 25CWT. That is a lot of extra weight to haul around.

With processors, my requirements are usually low power while actually doing something and I have noticed some useful improvements recently in 16 bit ones I needed at fairly high speeds with minimal opportunities to send it to sleep. I am often stuck with having to use chips which can run at 3.3V or even 5V so some of the newer PIC ones which have a core regulator are useful even if their consumption is not so good while asleep. Also, some of them have a mode where the clock is kept running and it can wake from a 'sleep' mode in less than a clock cycle. Not really a proper low power sleep mode but useful at reducing power in systems which normally wouldn't be able to use sleep modes as they take too long to wake up.

The power improvement of smaller geometries is simple fact, as is the higher leakage. I design ICs, although not usually the smallest geometries.

Keith

Added after 3 minutes:

you cannot compare two identical devices because they don't exist. all you need to do is to understand two things:

1) cmos devices consume current only when they are in transition, from on to off and from off to on. so everything else being equal, lower frequencies -> lower power consumption.
2) a mcu consumes much less when it is in standby. so if you want to drastically reduce power consumption, program the chip so it stays in standby as much as it can.

Oh, yes, I realise that. I also know power consumption of digital logic depends on clock rate, voltage and gate capacitance.

Keith