cheap, low-power ARM better than STM32F030?

[ralphd]

I've been looking at ARM MCUs for something with more horsepower and functionality than the 8-bit AVRs I've been using for things like battery-powered sensor nodes.
The 22mA power consumption @48Mhz is too high to run off a coin cell.  I could run it at 8Mhz, but then I might as well stick with an AVR.
I looked at some of the Cypress PSoC 4 chips, and while there are some cheap (<$1) ones, it looks like you have to use a proprietary windows-only IDE to configure and program the chips.  I want to be able to use open-source tools like arm-none-eabi-gcc and a serial uart bootloader or a cheap swd USB dongle.

Any suggestions for an MCU that needs < 10mA @24Mzh, or even better, at 48?

[HackedFridgeMagnet]

http://www.silabs.com/products/mcu/32-bit/efm32-zero-gecko/pages/efm32-zero-gecko.aspx

Energy micro says this one has 114uA/MHz

@48MHz is 5.47mA.

[6thimage]

I think the real question is how much processing the microcontroller needs to do. If you are recording some data every second or so, the microcontroller doesn't need to be in its run mode all the time, it can be in one of its low power states - stop is 19 uA, standby is 2 uA for the STM32F030.

ST also do a line of low-powered microcontrollers (STM32L), they aren't as powerful as the STM32Fs, but the STM32L0 has a run mode with a minimum of 87 uA/MHz and low power sleep modes of 440 nA and 250 nA (depending on if you want to keep the RAM on). Here's a link to them http://www.st.com/web/en/catalog/mmc/SC1169/SS1817.

[dannyf]

You can try the L family of chips.

CM0+ typically has lower current consumption - try Freescale for example.

Otherwise, Energy Micro is good here as well.

Having said that, low current consumption is more about programming than hardware.

[andyturk]

I'm not a fan of IDEs in general, but Cypress' effort is pretty good. You *want* to use their environment because it's the only way to configure PSoC's analog stuff and the programmable logic blocks. As with many IDEs, Cypress' GUI is proprietary, but it sits on top of a standard GCC toolchain. I do the minimum necessary in the IDE, and then edit the code with emacs and build with Makefiles. Segger's J-Link works fine with PSoCs, so you're not locked into proprietary debuggers either.

10mA @ 24MHz equates to 417uA/MHz. Pretty much every Cortex-M ever made will be more efficient than that.

[nctnico]

When looking for low power the key thing to look for is how long it takes the controller to wake-up and do something useful. The time needed to start the oscillator is one of the cullprits. Some microcontrollers (especially from the cheaper brands) take very long to wake-up so they waste a lot of energy doing nothing but waiting until the clock is stable.

If a device needs to be extremely lower power I look at TI's MSP430 range first.

[ralphd]

Thanks everyone for the suggestions.

Both the STM32L and the EFM32 look quite good as far as power usage.
The STM32L is almost double the cost of the 32F though.  Q10 pricing at mouser for the STM32L051K6U6 is 247c vs 125c for the STM32F030K6T6.  The 32L don't seem to be available in a qfp-32 or TSOP-20 like the 32F, so I'd either have to go with the bigger (and more expensive) qfp48 package or see if I could manage to solder a qfn.
The EFM32 is nice and cheap - EFM32ZG222F32-qfp48 146c/q1, but like the STM32L there doesn't seem to be many package options that are friendly for hand-soldering.  I've done QFP-32, so qfp-48 should be doable.  I don't need the high pin count though - a dozen GPIO would be plenty.

As for freescale, I don't want to learn a chip that might get discontinued after the Nxp merger.

I guess for now I'll stick with the 8-bit AVRs.  I was just looking at the 168pa datasheet and noticed it only consumes 2mA running at 8Mhz.  With a 32kHz crystal for the async timer in power-down mode it's under 1 uA.

[dannyf]

You may also consider STM8S/L if you are comfortable with 8-bit mcus.