Running a 3.3V STM32F0 on 6.5V

[inevitableavoidance]

I'm using an STM32F0 in a high volume low power project running on a LIR2032 that can be charged via USB using an MCP73831. The STM is specd at 2.4 - 3.6V, with 4.0V as the absolute maximum. When the battery is fully charged however the voltage will be 4.2V, 0.6V out of spec.

Now, if I power the device with 6.5V it actually still runs fine, except for a drawing scary high currents (81uA in standby compared to 3.12uA on 3.3V). What exactly happens when you apply a too high voltage to a device like this, and what are the actual risks?

Thanks in advance!
David

[technix]

You are lucky that the chip can withstand that high a voltage without letting out the smoke... It shortens the life of the chip significantly even if it works fine, and you have no guarantee that every chip would work at this extreme overvolting.

You should have added one of those battery buck-boost converters so your chip can operate at a stable 3.3V.

[Seekonk]

Use the voltage drop of a diode or two to supply power.  I was given a watt meter that didn't work. First I checked voltage on the chip and it was 11V on a 3V3 micro. Replaced the regulatoe and it has been working for years.  Who da thunk.

[hans]

I would be careful with a diode drop in your design. Under load it may drop too much that you want to use a Schottky diode, but in sleep a Schottky diode may only drop 100-200mV at 10uA (often not specified in datasheets), which then increases the supply voltage again.

An active regulator of somekind would do better.. Not sure if it needs to be buck-boost. You could perhaps also use a buck converter and run your whole design at e.g. 2.7V or another low voltage, such that the converter has enough headroom at the lowest battery voltage you need to support.

I wouldn't let a microcontroller run at elevated voltages all day.
Or in extreme cases for short periods. At a previous job I've seen a fault of a production test jig, that put 12V briefly (like 2s at most) on the internal 3.3V of the product during programming. Complaint was: the devices wouldn't program. I fixed the jig, but the devices used to diagnose the problem were scrapped. In that case a failure of such a device could lead to a road accident, so no chances taken.