Making an STM32 last at 150°C?

[StefanH]

Are there any guides or tips on making an STM32 MCU last in a high temperature (~150°C) environment? Specific voltage or clock speed tweaks?

The MCU would sleep most of the time, but now and then take and store some measurements.

Would certain families be better (for example because of different semiconductor nodes)? We're looking at the G4 and L4 MCUs (G4 for the embedded OPAMPs) but the application would work with pretty much any part number.

Similarly, would certain package types be more temperature resilient?

[Imen.D]

Hello @StefanH ,

When you run an MCU at high temperature (especially > 125°C), you have to take into account several factors and thermal parameters (thermal runaway, lifeTime usage estimation, TJ ...)

I recommend you this AN5036 Thermal management guidelines for STM32 applications, which describes the thermal management guidelines.

Note that several application notes are available, which present lifetime usage estimates by product series:

• AN5337 STM32H7 Series lifetime estimates
• AN5738: STM32G4 Series lifetime estimates
• AN5438: STM32MP1 series lifetime estimates

Hope this helped you!

Imen

[StefanH]

Thank you! That's great information.

The G4 vs H4 comparison is interesting, the G4 graph is an order of magnitude better than the H7 if I interpret correctly? But the H7 is listed in the table with higher junction temperatures than the G4. Am I correct to assume those are just given points and a G4 would likely fare a lot better at 140 or 150C than the H7?

[Pavel A.]

Generally the bigger and more complex the chip (lot of memories...) - the lower is its temperature limit.

Perhaps you're looking for automobile grade STM8. these little critters can stand 150 C ambient.

[waclawek.jan]

This is just like overclocking. It "works" (probably), but there's no guarantee.

And what's worse, some the effects of higher temperature are not immediate. Some of them may be, as even individual transistors behave differently at other temperatures than they were designed to (i.e. you may experience "not work"); but many are not. Higher electromigration rate, higher leakages especially in FLASH reduce lifetime exponentially.

Btw. are you talking about 150deg.C "junction temperature", or 150deg.C ambient? Note, that that's a huge difference.

You may consider using microcontrollers (other than STM32) which are designed for high temperature operation, or a solution which in some way avoids the elevated temperature.

JW

[StefanH]

Thanks for your insights so far, Pavel & JW.

The STM8 would be interesting, but don't fit the performance needs. A G4 at 26MHz would probably be the minimum. CPU active times at this performance level would be very short, far less than a second.

The environment temperature varies but can be up to 150C. Can't do much to avoid that. There is no expectation that the parts should last outside of their operating temperature range, but I'm trying to maximize life.

The issues with flash at high temperature I was aware of, it's another research topic :)

I found ST MCUs rated for 150C, but only in stock as far larger packages than the 1cm2 we have for the MCU. I know, not easy, but if it was I wouldn't be asking here.

[STOne-32]

Dear @StefanH ,

Indeed we have few customers going beyond our products specification from time to time and may work properly . They rely on their own application qualification at the required environment.  You can see that our STM32G4 can sustain up to 150 degree in AMR ( Absolute Maximum Rating) it is a good choice for your application. 

Cheers,

STOne-32

[waclawek.jan]

> The environment temperature varies but can be up to 150C.

Consider junction temperatures perhaps +15deg.C higher. Probably all effects increase exponentially with temperature.

If electromigration is part of the problem - and there are hints that it may be - then you want to avoid mcus built with small-feature technology. You may want to consider 'F3 rather than 'G4.

One strategy to mitigate FLASH leakage is to reprogram it periodically (there's a tradeoff in this, you have to balance it against wear).

You may want to characterize the problems by performing a mockup HAST. 150deg.C is not an extremely high temperature and its effects won't be obvious immediately, so you may want to accelerate things by heating them up way higher, perhaps at a couple of temperature points, and then try to estimate probablility of failure in time.

And, most importantly, don't forget, that you are definitively outside what's guaranteed by ST. Be prepared for surprises, have plan B.

JW

[StefanH]

After Imen's reply I indeed went Googling to see if I could find a similar document on the F4. Now you mention it, the mixed signal F3's might be interesting indeed. We'll indeed be testing, but first it's key to understand what to test and have some ideas of where the problems may be so we can measure / observe accordingly.

[Nikita91]

TI have a range of high temperature devices (-HT) up to 210°C developed for oil drilling.

https://www.ti.com/tool/TIDA-00002

[Joe WILLIAMS]

Dear StefanH,

As an alternate solution, if temperature is your primary concern, ST manufactures automotive grade MCUs. The SP5 has full operation guaranteed from -40 to 165 °C (TJ). Take a look at our SPC5 32-bit Automotive MCUs page to see if one of these automotive grade micros might work in your application. A SPC5 product selector tool is on the adjacent tab:

https://www.st.com/en/automotive-microcontrollers/spc5-32-bit-automotive-mcus/products.html

https://www.st.com/en/automotive-microcontrollers/spc5-32-bit-automotive-mcus.html

Several versions are available through the ST Authorized Distributor's inventory. If you need help locating alternate devices, please submit a request directly to the Online Support Team at:

 

https://my.st.com/ols.

Kind Regards

Joe WILLIAMS

STMicro Support

[Garnett.Robert]

Hi,

Does the chip have to operate continually at 150 C or is that just a transient?

Does the board have electrolytic caps, these degrade very quickly at high temperature? The better normal grade are rated at 12,000 hrs at 105 deg C. Their life reduces by half for every ten degrees above their rating. At 150 C these wouldn't last more than 700 hrs, probably much less. (100 hrs?)  You can get auto grade that are rated at 150 C (Kemet), but I'm not sure of the capacitance and voltages available.

Is it possible to insulate the board and cool it with air? You can get lightweight ceramic insulation material that might do the job.

 

Regards

Rob

 

[Personwithhat]

Hey just testing the waters here - Have you had any luck finding a small package MCU that will work up to 150C?

What solution did you end up going with?