Can the MCU be accidentally damaged by overclocking?

[gbigden]

Posted on April 27, 2015 at 16:48

Specifically, the STM32F303. It is tricky to set up the clocks and sources, dividers and PLL multipliers etc. If I get it wrong can I damage the chip by overclocking it?

#clock #stm32f3

[Tesla DeLorean]

Posted on April 27, 2015 at 17:08

I think the complexity is overstated.

The damage would occur from overheating, not expressly from clocking too fast or selecting the wrong gears, it's not going to chew up the gears in a mechanical sense. The VCO can generate 400-500 MHz signals, but is likely to top out a some maximum in terms of what it and the underlying silicon characteristics dictate. The failure occurs when critical paths in the design can't supply an answer during the clock period, resulting in the state and progress of states to fall over, and for the behaviour to stop being predictable. ie numbers don't add up properly, cache/prefetch supply the wrong data.

[gbigden]

Posted on April 27, 2015 at 17:27

Thanks, but is there a thermal shutdown, or do I just keep my finger on the chip? 🙂

[pmoore]

Posted on April 27, 2015 at 20:04

Gary:

There is a linear temperature sensor in most of the STM32Fxxx parts as part of the ADC.  Many of the newer parts, the F303 included, store ''Temperature sensor calibration values'' for the sensor, so you don't have to calibrate each part yourself to get a reasonable idea of what the processor core temperature is. Parts clocked outside of temperature and/or speed ratings often do not operate within manufacture listed specifications or warranties.  You would be running off the edge of the known map into an area marked ''Warning!  Beyond here there be dragons.'' if you overclock the core. On the upside, in most cases (especially for short periods), if the clock configuration is very far outside of the specifications, segments of the STM32Fxxx just do not work as desired, so you quickly understand something must be wrong and try a new configuration. ST has pretty good documentation on how to configure the clocks...and now has utilities that will write the clock configuration code for you and keep you from accidentally running outside of the specifications. -Patrick

[Tesla DeLorean]

Posted on April 27, 2015 at 20:56

Depends if you think the temperature sensor is sufficiently localized. I'm not aware the processor implements any thermal based shutdown function.

There may be ways you can get the part in to thermal runaway. I'm not sure how you'd do it, but such an event will cause physical damage. Temperature alters the behaviour of the transistors. If you run the part too fast, where it still functions correctly, it is likely to generate more heat, CMOS draws the power at is transitions states, and will rise as clock increases. At some point you will clock the part faster than the critical paths permit (temperature, voltage, speed), at this point the part will malfunction, most likely crashing and ending in the Hard Fault Handler at some point. I don't think this will cause physical damage to the processor. What will cause physical damage is an uncontrolled motor running something through the end stops, or a pipe bender that's lost where the floor and ceiling are.

[gbigden]

Posted on April 28, 2015 at 10:06

Thanks. I am just worried about setting a timing parameter wrong and having the chip self destruct