[SOLVED] Searching for ultra low power microcontroller

[dani]

Hello all,
I am searching for ultra low power microcontroller for my upcoming project. Can any one please help me to find the best available solution comparable to MSP430.

Rom ~ 128
Ram ~ 64
Internal ADC ~ 10bit
etc

Regards
Dani

 

[millwood]

the answer depends on what you mean by "ultra low power".

 

[microlab]

hi
thare are many things when u want a real"ultra low power controller"
the main thing is that you have ti define the application in detail
like is it working on battery?
is the system is located at a remote place where will be difficult to replace the battery
what's the active time?
and so on
almost all controllers can be used according to our spec
i used the atmel 89c51 series PIC and AVR for the same application but with different techniques to increase the battery life(n adc's with 89 series)
u can use the MSP serious from TI or 8051 series from Silabs
they are designed for ultra power.

regards

ml

 

[dani]

As i mentioned it should be comparable to MSP430 means alternates to it. Yes it will run on a battery.

 

[millwood]

pretty much any mcu is "comparable" to msp430 at one level or another.

 

[bobsanjose]

I have done lots of comparing other devices to the MSP430 and there are NOT many that can in deed compete.
My favorite based on specs (have not used it) is the Gecko family from Energy Micro. The EFM32 families can match or sometimes even beat the MSP430 in all aspects. Power while running code, power in different stand by modes, wake up time, low power peripherals that can run without CPU.... you name it.
Energy Micro recently announced the EFM32GG (Giant Gecko)
**broken link removed**
family but that one will take a little longer until it is on the market. There are a couple families available already from EM.
**broken link removed**
STM32L might be a competitor as well. Most ultra low power MCUs do not offer the memory you are looking for because leakage in the memory kills most low power stand by behavior.

hth, Bob

 

[karesz]

Hi,
Renesas has a "ARM Core" newly too, as super low power type.
More commercially (of course) PIC micro...
K.

 

[millwood]

power consumption is, in most applications, programming driven. if you are careful, you can achieve low power consumption with a non-MSP430 chip than you would with it.

and in general, mcus don't take much power anyway.

 

[bobsanjose]

power consumption is, in most applications, programming driven. if you are careful, you can achieve low power consumption with a non-MSP430 chip than you would with it.

All things being equal, the programmer knowing what (s)he is doing, I have to strongly disagree. If an application goes to sleep and wakes up frequently, the internal oscillator and the wake-up time is of the essence. Only VERY FEW MCUs are good at this one, MSP430 and EFM32 being two of them.
Running on battery, the supply voltage range is very important. A range from 3.3 to 2.7V will not be able to make max. use of the battery's charge, while one from 1.8V to 3.6V uses up ever bit of the charge.
There is so much more relevant for low power / battery life and it seems your experience is somewhat limited within this field (no offense)

and in general, mcus don't take much power anyway.

If they are connected to mains or compared to an electric heater, this statement is surely true., If compared in an apple to apple fashion, depending on the MCU and assuming a smart programmer, this difference can easily be a factor >10 in battery life.
Recommended lecture: **broken link removed** that would be a comparison between PIC24 and MSP430. PIC24 is pretty good with low power, it is just a matter how you look at it. Compare this to an older architecture such as a 12-clock or a 6-clock 8051 and the difference will be much more. There is a good reason why so many semiconductor vendors are introducing new low power products, because there are major differences to the old ones as well as between the new ones.

 

[millwood]

All things being equal, the programmer knowing what (s)he is doing, I have to strongly disagree.

you are essentially saying the same thing: a good programmer is key to achieve low power consumption. I was making the programmer explicitly the factor, and your counter argument is assuming away that factor.

you certainly can find examples where the (software) envelope has been pushed to its limit and the hardware becomes the distinguishing factor. My point was that situations like that aren't every day occurrence.

 

[tenaja]

I have done lots of comparing other devices to the MSP430 and there are NOT many that can in deed compete.
My favorite based on specs (have not used it) is the Gecko family from Energy Micro. The EFM32 families can match or sometimes even beat the MSP430 in all aspects. ...

I was recently looking for a low power mcu, and was excited when I read EM's ad--the specs looked great. I wondered how they compared to the much more mature NXP LPC line... and the power consumptions are virtually identical. EM rates theirs at 180uA/MHz, whereas NXP rates theirs at a something like 9mA at 50MHz. If you do the math, they match.

Nifty marketing, anyway.

 

[bobsanjose]

I was recently looking for a low power mcu, and was excited when I read EM's ad--the specs looked great. I wondered how they compared to the much more mature NXP LPC line... and the power consumptions are virtually identical. EM rates theirs at 180uA/MHz, whereas NXP rates theirs at a something like 9mA at 50MHz. If you do the math, they match.

Nifty marketing, anyway.

They only do match if nothing but the frequency counts comparing apples to oranges, otherwise EM has significant advantages.
1. EM is using a Cortex-M3, which offers approx. 30% more performance than a Cortex-M0 the one that NXP is using. So, to achieve the same performance with an EFM32 running @ 39 MHz, you have to run the LPC1100 @ 50 MHz.
You can also check: **broken link removed** for the LPC1100 features
and **broken link removed** for a comparison of Cortex-M0 and Cortex-M3

2. The LPC1100 series, the one with Cortex-M0 is less mature than the EFM32, hit the market later and is using a newer CPU. Don't mix up the maturity of the LPC2000 or LPC1700 with that of the LPC1100.

3. EFM32 offer some really low power peripherals, such as a low power ADC and a low power UART, I haven't found another competitor with such good numbers on these peripherals. **broken link removed**

4. As previously mentioned, if an application runs from battery some aggressive power saving methods can be used such as putting the device into deep sleep mode as often as possible. To do this, there needs to be a very fast wakeup time. The EFM32 offers one of the fastest wake-up times in the industry.

All these features combined, running the EFM32 @ 39 MHz instead of 50 MHz for the LPC1100, low power peripherals and also very important the fast wake-up time, the EFM32 can achieve anywhere between 30% more battery life all the way to 100%+ more battery life although EM is using the CPU that needs more power.
Conclusion: EM must be doing something right!
p.s. don't get me wrong, NXP is next in line and better than all other Cortex-M options currently on the market. If you are looking for a device with 8-32 KB Flash, I would recommend the LPC1100 because the price is tough to beat. If memory should be more and the best possible low-power behavior is wanted, I would go with the EFM32.

As Millwood pointed out, a good programmer is key. A good programmer can achieve better results with a LPC1100 than a not so good programmer with the EFM32. Comparing apples to apples, good programmer to good programmer, the EFM32 will make a difference.

Cheers, Bob

 

[rca]

there is also some one at onsemi company, section Audio DSP, BelaSigna family.

 

[Malek Lamari]

I strongly recommend you to go for ST Microelectronics solutions; STM32L.

 

[spiralbrain]

NXP has a very bad support, they never provide samples and are not very well known in some parts of the world, Also the company has gone through a bit of trouble with selling of plants in Europe. Surprisingly the demand for their chips has suddenly gone up after making so many redundant in Europe. what is the use of getting something like LPC1110 series out when you cant sell them to the people who need them.

this is the most recent news: **broken link removed**
and NXP Job-Cuts: Due To 'Change In World Trade Patterns' Or Misjudgment? (Mannerisms)

Microchips Nano Watt is also worth considering https://www.microchip.com/en_us/technology/xlp/

Microchip has already done the comparison:

**broken link removed**

 

[bobsanjose]

NXP has a very bad support, they never provide samples and are not very well known in some parts of the world, Also the company has gone through a bit of trouble with selling of plants in Europe. Surprisingly the demand for their chips has suddenly gone up after making so many redundant in Europe. what is the use of getting something like LPC1110 series out when you cant sell them to the people who need them.

Your comment makes it rather obvious you are not a semiconductor person, rather a user that is upset for not getting samples, right? It is correct that NXP has been going through tough times lately, however the MCUs within NXP have done fairly well! I disagree with your support statement. Sure support experience is different for every single person and your experience might be bad. Another assumption here, your business is actually handled through a distributor and you are not very happy about this.
You want the LPC1100 devices when "people need them", go and order them from Digikey, pretty much all of them are available from stock, what is your problem?

Microchip has already done the comparison:

Comparison MSP430 and XLP
Sure Microchip has done a comparison. You always try to compare yourself in the most favorable way to the best in class.
Strange enough, TI has done a comparison too and imagine that, suddenly the MSP430 compared to the same Microchip device looks soooo much better.
It always depends who does the comparison and selects the power mode that suits them best.
Check out the much more comprehensive White Paper that TI did versus PIC. Read it and judge for yourself who did their homework better.
**broken link removed**
A quote from the TI White Paper:
" MSP430 Summary
• World’s lowest-power MCU family by design
– 100nA storage mode, <500nA standby, <1μA RTC mode
– 0–16MIPS <1μs, agile instant-on clock system
– Intelligent peripherals reduce CPU load, improve performance
– MSP430F2xx is lower power compared to the PIC24F XLP in all modes of operation, by the datasheet and on the bench
– At a measured 700nA in a true 3V standby mode with self-wakeup, PIC24F XLP is 75% higher power compared to MSP430F20xx at 400nA
– When evaluating ULP MCUs be cautious of incredibly low marketing numbers such as “20nA deep sleep”. These are often under unrealistic conditions such as 1.8V, no BOR, no RAM retention and no self wakeup – this is not a very useful state for a microcontroller."
--- end of quote ---

No offense!
Bob

 

[millwood]

NXP has a very bad support, they never provide samples

your support experience may depend on your coverage person(s), and your approach with them. Mine has been fairly helpful and visit us regularly. whenever she came, we make sure that she saw design wins we had with their competitors products and that made her more competitive in terms of sampling new dev boards - we don't buy chips ourselves.

Microchip has already done the comparison:

I suppose you know that that comparison is biased, right?

 

[spiralbrain]

Dear Bob,

first of all I must thank you for taking the time to answer my post in such detail. About myself now, yes I am a user, I am not a silicon man. The reason I mentioned samples is because it is NXP's policy to only send samples through the distributors. This keeps out students or first time users who wish to just explore the technology. you might argue why students? its because todays students are tomorrow's engineers. This is what TI just realized few months ago and released the Launchpad kit. why do you think there was a need for that? $4.30 delivered free worldwide. I am sure they made a loss doing that.

Maybe yes, as you and millwood pointed out, support may be regional or I am in a market that is considered less important by NXP. LPC1100 is a brilliant chip, its here to change the industry only if NXP can sustain itself first.

The TI's comparison looks more like frustration than a comparison. But yes TI may have a better product at 3v and they also have their own video YouTube - The MSP430F2xx Ultra-Low Power compared to the PIC24F XLP Challenge The best way to test this is to try them both. I have never worked on the MSP430 I would love to try it after I saw your die hard fan following for it.

I would still argue though that for some applications where ram retention is not an issue the PIC will out perform the MSP. what is the harm in saving the context in flash and restoring it after the device wakes up? its better than keeping the ram powered for 90% of the time the device is in sleep.

Now BOR, correct me if I am wrong, why would you require a BOR? unless you had a device other than the micro that was taking away too much juice from the battery (in this case you would not worry about a few microamps your micro saves for you). when talking of micro and nano amps the battery source would take a few years to get to the limit where there would be a BOR. And if there were a BOR this would mean that the micro is no longer able to use the power source and is not utilizing the battey to its utmost limit. Now the 1.8v factor, why 1.8? It is not usefull at 3v but a 3v battery is not 3v forever is it? it will drop and a micro that performs better at lower voltages will win the race in the end. And 1.8v is becoming the new standard in electronics technology.

In the end a smarter utilization of resources and some clever programming can save more power if deep sleep is used. Microchip announced the reduction of deep sleep current from 20nA to 9nA which takes the bar further down. Microchip Announces Industry s First 8-bit Microcontroller With 128 KB Flash Memory in 28-pin Package PIC18F47J13 Technology Inc announced the

No offence as well, arguments are educational!

 

[millwood]

what is the harm in saving the context in flash and restoring it after the device wakes up?

that lengthens wake-up time - where TI has a huge advantage over competition. short wake-up time can be critical for applications where the mcu is active for a very short period of time - typical of a low power consumption application.

its better than keeping the ram powered for 90% of the time the device is in sleep.

it depends. the only thing here is the leakage current through the sram. if you can make it small, then keep sram on during sleep isn't a bad choice.

if you cannot, then going to flash would be preferable.

marketing is always misleading, from both microchip or ti. the key for us to be able to understand limitations of those devices and act accordingly.

 

[bobsanjose]

Dear Bob,........ snip ............

Microchip announced the reduction of deep sleep current from 20nA to 9nA which takes the bar further down. Microchip Announces Industry s First 8-bit Microcontroller With 128 KB Flash Memory in 28-pin Package PIC18F47J13 Technology Inc announced the

No offence as well, arguments are educational!

First of all, absolutely no offense taken! I love arguments that are based primarily on facts and it is only natural that personal likes / dislikes become part of the discussion.
Here comes another parameter into the game that has not been mentioned yet. I am talking about self discharge of a battery. Depending on the type of the battery self discharge can be substantial and in fact much more than the sleep current of a super low power MCU. IIRC, the batteries I looked at in the past were much higher than the 20 nAs from Microchip. Just puts things in perspective.
As Millwood pointed out, saving parameters in Flash is a big disadvantage if the battery power system has to wake up frequently. Programming a flash takes lots of time, typically several milliseconds and lots of power. if you have a system that checks for an event every 10 ms (did the user push a button), then saving in flash is useless.
On the other hand for product storage on the shelf before first time use, string in flash is very useful.
So, I guess there is room for both options, the XLP option and the MSP430 option.

Cheers, Bob