Looking for Low current deep sleep MCUs

[JPortici]

Hi,
I have to redo a remote because the transceiver has gone EOL and no pin to pin replacement is there.
New transceiver has been selected, this would also be a good time to change the MCU.

In this application the remote is in sleep for most of the time, and exchanges data with another gadget at the press of a button or in "active" modes.

there are three modes:
- Active connecting, 4 second sleep, 60ms active
- Active connected, 32 second sleep, 60ms active (this is the mode the remote will be the most. average current from the transceiver is 6.024/21.61 uA best/worst case scenarios)
- Idle, 256 second sleep, a couple of ms active just to occasionally check battery then go back to sleep
Pressing a button will put the remote in "Active connecting" mode

During activity current is dominated by the transceiver, but during sleep current is dominated by MCU.

Current MCU is a PIC16LF1823, chosen because i know it quite well... and it has a very low deep sleep current (in this case it is called "power down" in the datasheet, whereas for other families/brands with power down RAM power is removed, deep power down only way to exit is pulling a pin high/low)

I'll Assume i can trust the datasheet.
1823 has a Power down current + WDT (to exit from sleep) of 530 nA, says the datasheet (3.0V, 25°C). I can confirm it's in this ball park for all tested units during developement and production.

However, working with this chip is a pain in the butt, very little Flash/RAM.
Solution #1 is to scale up to 1824 or 1825 (increase flash/ram), make the code a little less hand optimized but more readable and call it a day.
Solution #2 is to choose a different MCU with low cost, low pin count (max 20 for TSSOP, max 32 for QFN) and a power down current in the same ballpark

What i've looked at for now:
From Microchip:
- Other PIC16: LF15325, LF18324. Basically, updated cores/peripherals. A bit more flexible, also for power managed modes but datasheet is still "Preliminary" with incomplete data. More than one year ago i rejected them for the same reason. sigh.
- only other family that had all the requirements (low cost, low pin count, low power down current, SPI and ADC) was the PIC32MM. this boasts 520nA typical power down current + WDT. Other plus would be easier programming
From NXP:
- LPC1111FDH20 has a very high deep sleep current, 6 uA. 220nA for deep power down and 4 always on registers (enough to save all the variables needed). Problem, with DPD only way to wake up the MCU is to pull the wake pin. The new transceiver doesn't have a low power oscillator and automatic wake up like the old one so i can't use this mode. pity.
- LPC822M101JHI33 is probably the only good candidate i found: seep power down with wake up timer, 1.1uA. Still very high, but apparently less variations across temperature range

And that's it. I didn't find anything else, that's a very short list. I accept suggestions
- I checked out some ATTiny or SAM parts but power down currents were way higher. Same for STM32 L0
- I'd avoid having to learn another family for this (8051, MSP, ..) unless i have better results

Suggestions?

[danadak]

Possibly this -


http://www.cypress.com/file/45776/download


Regards, Dana.

[MosherIV]

I think the LPC1111FDH20 from NXP covers the ARM Cortex M0 series.

Maybe look at another chip vendor for something more suitable, ST Micro :
http://www.st.com/en/microcontrollers/stm32f0-series.html?querycriteria=productId=SS1574

Also, do not avoid 16bit micros. I believe the good old MSP430 from TI is exceptional for low power and because they are going out of fashion, TI are making them cheap now.

[JPortici]

Thanks dana, Only PSOCs in pricerange seems to be PSOC4s, datasheet says 2uA deep sleep current. not that good

Again, I looked at the STM32L0 family. sopposedly the lowest power they make. not so low power, unless i read the datasheets wrong, or they made it hard to understand (wouldn't be their first time  )
Remember i'm not looking at LOW POWER in running mode, but only at deep sleep and power down modes! it's being hard for me to find something as low power as pic16 in that aspect

[alm]

I second the MSP430 suggestion. Should be a nicer architecture and more flash than the 8-bit PICs. Something like this part should be able to handle everything in low power mode 3 (0.5 uA typical) and is available with up to 16 kB flash / 512 bytes RAM. They will probably be more expansive than the PIC16, however.

EDIT: An alternative would be to use an external watchdog timer (like this one) together with any MCU that can do a very low power deep sleep with wake up from pin change interrupt. For example, that MSP 430 would be down to 0.1 uA typical plus 35 nA typical for the watchdog timer.

[TK]

I second the MSP430 suggestion. Should be a nicer architecture and more flash than the 8-bit PICs. Something like this part should be able to handle everything in low power mode 3 (0.5 uA typical) and is available with up to 16 kB flash / 512 bytes RAM. They will probably be more expansive than the PIC16, however.

MSP430 is a nice option, and they are available in DIP package

[jwm_]

Msp432 is the msp low power design with an arm core. But msp430 is not bad at all, it's right in a sweet spot where 16 bits makes sense.

[JPortici]

Reading through the MSP430 manual..

MSP430FR2111 - seems like LPM4,RTC,VLO is what i'm looking for. 590nA @25°, 3.0V. Basically on par with PIC16 XLP, not bad at all
still 2x the price of the pic in the quantities i'm looking forward to get, but it's a good excuse to get a launchpad devkit

Correct me if i'm wrong: with this mode i would
- Put data to keep in FRAM
- Go to sleep
- With RTC event MCU restarts
- Bit (saved in fram) tells if i have to do full initialization of transceiver or skip to program

is it correct?

[Yansi]

You could also use STM32L0 

(Supply current in Stop mode well under 1uA at normal temperatures)

[TK]

Reading through the MSP430 manual..

MSP430FR2111 - seems like LPM4,RTC,VLO is what i'm looking for. 590nA @25°, 3.0V. Basically on par with PIC16 XLP, not bad at all
still 2x the price of the pic in the quantities i'm looking forward to get, but it's a good excuse to get a launchpad devkit

Correct me if i'm wrong: with this mode i would
- Put data to keep in FRAM
- Go to sleep
- With RTC event MCU restarts
- Bit (saved in fram) tells if i have to do full initialization of transceiver or skip to program

is it correct?

I am sure there are non-FRAM MSP430 chips at lower prices

[mikeselectricstuff]

Take a look at the recent PIC16F15xxx chips - lots more memory and a lot cheaper.

[dgtl]

Silabs (former Energy Micro) has been targeting the low-power segment. If you are looking for ARM uc-s, then check for their stuff as well. In addition they advertized nice energy profiling tools, that measure system current correlated to program trace, so you can optimize the higher consumption parts. Didn't check the current consumption numbers, though.

[JPortici]

I checked again the L0 manual but it seems i can't exit stop mode with the watchdog/RTC event. Please, correct me if i'm wrong

I am sure there are non-FRAM MSP430 chips at lower prices

I sorted for minimum price, that was the first one that ticked all the boxes

@mike, 16f15xxx and 16f18xxx are on my list even though their datasheet is still incomplete for electrical characteristics. IIRC difference between the two for what i care about is in HEF vs EEPROM and thus slight difference in price (< 10 cents at low quantities, <5 cent in higher quantities)

RE: EMF chips.. well, i'm going to use a silabs transceiver. I did check the EMF arm chips (also because i wanted to use on with integrated transceiver at one point) but i found the documentation really hard to interpret/understand, harder than ST if you can believe it, i got a bit put off. Too bad because chips are really promising

[alm]

What about the external watchdog in some tiny package? Do not want the extra BOM item? Could turn out the be cheaper and lower power than getting a more expensive micro that has a watchdog timer that works at a sufficiently low power level.

[Yansi]

I checked again the L0 manual but it seems i can't exit stop mode with the watchdog/RTC event. Please, correct me if i'm wrong

Yes it of course can be waked up by RTC from stop mode. That is the purpose of the RTC Alarm function. Watchdog can also wake it up.

[MosherIV]

Yes it of course can be waked up by RTC from stop mode. That is the purpose of the RTC Alarm function. Watchdog can also wake it up.

Check again, there are 3 low power modes. I had to double check and I agree with JPortici.

The table says that only an external IRQ will wake the processor from STOP. Probably because in STOP mode the ARM core and all peripherals are no longer being clocked!

Another gotcha about stop mode is that the processor pretty much wakes up and has forgotten where it was when put to sleep, ie like waking up and booting up again, so the code has to check if it is waking from STOP mode
I have never had to use this STOP mode before only the first level of low power.

[Elf]

In addition to the MSP430 you might take a look at the Silabs EFM8 "Sleepy Bee." Up to 4k RAM, 64k flash. Minimum 50-80nA sleep mode consumption, depending on peripherals and supply voltage. Wakes up from sleep on:

• RTC events
• Port match event
• Comparator rising edge

It can also be stopped / reset for less power draw, with the GPIO retaining state. Data sheet: http://www.silabs.com/documents/public/data-sheets/efm8sb1-datasheet.pdf

[JPortici]

yep, since RAM is being powered off exept for the five registers in the always_on domain, only thing the mcu can do is do a reset.

checking for reset from stop could be done by putting data on the always_on registers hopefully they are cleared on a POR to start from a known state, but with STM and their manuals... who knows? do they actually write POR values now in the manuals?
ah, i should end the rant because i wouldn't be using it anyway, adding an external chip to a costlier chip, on a tight bom an board is nonsense.

Only two chips that seems to make it are PIC16, which already works, and msp430. i was looking if i could achieve same low power down / stop / whatever this manufacturer calss it power and eliminate some of the headaches.. not adding new headaches

the more i look into it the more i think i'll just go with a pic16 with more memory and more relaxed programming
just excecuting sleep instruction halts the processor and the internal clock, power goes down to some tens of nA. Then WDT is separate and can be enabled so LPWDT is a couple hundred nA with sufficient accuracy, at WDT event clock starts again and after it stabilyzes code resumes excecution from last instruction. ram retention voltage is 1.5V, the whole thing can't operate at less than 2.3V so no problems. as simple as that

[Yansi]

Yes it of course can be waked up by RTC from stop mode. That is the purpose of the RTC Alarm function. Watchdog can also wake it up.

Check again, there are 3 low power modes. I had to double check and I agree with JPortici.

The table says that only an external IRQ will wake the processor from STOP. Probably because in STOP mode the ARM core and all peripherals are no longer being clocked!

Another gotcha about stop mode is that the processor pretty much wakes up and has forgotten where it was when put to sleep, ie like waking up and booting up again, so the code has to check if it is waking from STOP mode
I have never had to use this STOP mode before only the first level of low power.

I know what I am saying.  You shall look again, not me.  (The RTC generates events for the EXTI controller. Reference manual RM0377, chapter 22.5).

[suicidaleggroll]

Only two chips that seems to make it are PIC16, which already works, and msp430. i was looking if i could achieve same low power down / stop / whatever this manufacturer calss it power and eliminate some of the headaches.. not adding new headaches

You don't have to go to LPM4...LPM3 should be more than sufficient.  It still keeps ACLK and the crystal active, which you should be able to use to drive a timer to wake back up.  The simple msp430g2553 consumes 500nA at 1.8v in LPM3.  That's 900nW...it would take 127 years for that to drain a single 1 Wh AAA.  How low do you need to go?

[MosherIV]

I know what I am saying.  You shall look again, not me.  (The RTC generates events for the EXTI controller. Reference manual RM0377, chapter 22.5).

Fair enough (and my apologies). Another case of the reference manual contradicting itself.
The table in the Power Control section clearly states that  Stop mode "HSI and
HSE
oscillators
OFF"

Which to me means that the RTC external clock is switched OFF, so the RTC should not be running.

Guess someone will have to try it to see which part of the ref manual is correct.

Like I said, I have not tried the STOP low power mode.

[JPortici]

I know what I am saying.  You shall look again, not me.  (The RTC generates events for the EXTI controller. Reference manual RM0377, chapter 22.5).

You are right! I don't know why i understood that i couldn't run the LSI oscillator in stop mode
(and when i wrote that i had JUST read both the datasheet and the reference manual)

[krho]

1823 has a Power down current + WDT (to exit from sleep) of 530 nA

I don't know what you were looking at but STM states the following for STM32L072KB   
0.29 ?A Standby mode (3 wakeup pins)
0.43 ?A Stop mode (16 wakeup lines)
0.86 ?A Stop mode + RTC + 20 KB RAM retention

[Yansi]

I know what I am saying.  You shall look again, not me.  (The RTC generates events for the EXTI controller. Reference manual RM0377, chapter 22.5).

Fair enough (and my apologies). Another case of the reference manual contradicting itself.
The table in the Power Control section clearly states that  Stop mode "HSI and
HSE
oscillators
OFF"

Which to me means that the RTC external clock is switched OFF, so the RTC should not be running.

Guess someone will have to try it to see which part of the ref manual is correct.

Like I said, I have not tried the STOP low power mode.

Why would you run RTC in your low power application from HSI/HSE in the first place?
RTC is kept running from LSI (internal RC) or LSE (crystal).

There is none contradiction in the RM in this case.

[MosherIV]

Why would you run RTC in your low power application from HSI/HSE in the first place?
RTC is kept running from LSI (internal RC) or LSE (crystal).

Yes, sorry my mistake.  Was not paying attention, busy doing the day job.

[wraper]

you might take a look at the Silabs EFM8 "Sleepy Bee." Up to 4k RAM, 64k flash. Minimum 50-80nA sleep mode consumption, depending on peripherals and supply voltage. Wakes up from sleep on:

• RTC events
• Port match event
• Comparator rising edge

It can also be stopped / reset for less power draw, with the GPIO retaining state. Data sheet: http://www.silabs.com/documents/public/data-sheets/efm8sb1-datasheet.pdf

+1, in addition it has decent free IDE and cheap dev board. http://uk.farnell.com/silicon-labs/slstk2011a/dev-board-efm8sb20f64g-sleepy/dp/2469341 which has built in J-link debugger, can be used for programming/debugging external device as well.
EDIT, it also has 2% precision internal oscillator which might be useful in this case. Also silabs MCUs are very cheap as bonus.

[danadak]

If you look at stop mode you can get down to tens of nA at room temp,
hot ~ 1 uA. Note startup time from stop mode is 2 mS. Startup time
of course impacts avg power consumption in LP designs,


http://www.cypress.com/file/121271/download


Regards, Dana.

[snarkysparky]

PIC24F04KL series.