Durability of the PIC18F45K20

[envisionelec]

A couple years ago, I had a design built on contract around the PIC 18F45K20. This is a nominal 3.3V part so you can imagine my shock when I looked at the schematic and saw it running at 5V. I pointed it out to the engineer and he shrugged his shoulders and continued development. Thoroughly convinced it would blow up at some later date, I decided to do an experiment and let the test circuit run for a month straight on 5V.

It survived. It was not running hot and it passed every test I could throw at it, including running at full speed with internal PLL. Because the BOM price was already redlining, I simple dropped a diode inline with Vcc (to get it down within ABS Max spec) and made the decision to build hundreds. Every one of the field units is still working fine.

I don't think there is a lesson in here except that you *can* run this particular uP over "rated" voltage and get away with it. Note that the new version incorporates a PIC24F and a 3.3V regulator because while I may be foolish, I'm not a fool.   

[amyk]

The datasheet says absolute maximum of 4.5V, so maybe you were just lucky (or they have very wide margins on their process, which may narrow if they do a die shrink.) The 3.3V parts might just be identical die to the 5V version, that were selected and marked as such because they could run within spec at that voltage, and they give a lower absolute maximum to give them flexibility to move to a smaller lower-voltage process in the future.

I think running at full speed wouldn't be a conclusive test since if anything overvolting will make it able to run even faster.

[millerb]

I'd expect the life span in that condition to be dependent on how much you're sourcing/sinking through your IO, that's going to be the main cause of thermal issues and electromigration. In my experience PICs can take a beating but I sure wouldn't want to be rubbing against any absmax params on something I had to ship.

[envisionelec]

I'd expect the life span in that condition to be dependent on how much you're sourcing/sinking through your IO, that's going to be the main cause of thermal issues and electromigration. In my experience PICs can take a beating but I sure wouldn't want to be rubbing against any absmax params on something I had to ship.

Consumer electronics. I would not do it in anything that really mattered. 

[AndyC_772]

The 4.5V rating is the absolute max to avoid damage to the part; the upper limit for which functional operation is guaranteed is 3.6V, so anything above that is a grey area in which the device may or may not operate correctly. Have you tried EEPROM or Flash writes with the voltage above spec?

Personally I'd use the 18F45K22 instead, which has become one of my favourite chips - it effectively supersedes the 45K20 and will run at up to 5.5V.

[bxs]

I will not talk in 18F45K20 in specific, but in my history with 8bit PICs I have done really bad things, even 15V once in a 16F and so far, never killed one 

That 16F with 15V still worked at 5V, but consuming quite a bit of current...

About all those new PICs with new fabric process, I don't trust them, not even close, like I trusted in the good and old stuff but to be fair, I have also made some mistakes with those new PICs and like I said, so far no problems.

But in a product use a 3.3V part at 5V is really a bad thing and even after discovering it, keep it at 5V 

[itdontgo]

One of our old boards used to have a power supply glitch.  Basically the 4v rail used to go to as much as 18v!  Everything used to blow up apart from the PIC16F.  They're not voltage sensitive like that.

[amyk]

That's only parts of the die, I'd attribute the toughness more to being fabricated in an older process with bigger geometries and thicker gate oxide.

[AndyC_772]

Thinking about it some PICs are designed for high voltage programming so maybe they do make them tolerant.

Only on the MCLR/VPP pin, which is designed for 12V or thereabouts. Apply that voltage to any of the other pins and you're asking for smokey trouble.

[SeanB]

I had a box full of TTL, avionics computer, that had a PSU go unhappy and give the unregulated 14V rail to it. Ran for a few hours before it was changed, and aside from one failed IC, a 5406 30V open collector inverter ( Yes, I appreciated the irony) nothing else was faulty. Still have the board. Replaced every thing in the 5V regulator aside from the capacitors, and fixed the non functional crowbar ( hard to test that one, fuses are expensive for that unit, as you have to unsolder so much of the power supply block to get to them to unsolder and replace) on the 5V rail.

[brainwash]

I cannot attest to the durability of 18F series but I've stressed quite a lot of 16F and 10F series, intentionally. Overvoltage, undervoltage, back-emf, shorting all outputs, oscillating power supply. Except one which did not want to be programmed anymore (I have not looked into it, it might have been a connection issue) not a single part died. Granted, this is not testing, just anecdotal evidence.

Last week I put 9V through an 18F4520(?) for quite some time until I noticed the current draw. I was debugging a prototype board a friend had made and the regulators were not fitted properly. The board also had some shorts ('by design') and in the end a peripheral pin died. I'm not exactly sure when it died, but by the time we got to write code for it we had to fix about 20 something board issues. The offended pin was wired to a 74LVC4245A octal chip.

Actually, if someone has the chips and the inclination do to so, I would love to see a reliability compare between architectures: one more thing to add to the PIC vs AVR vs ARM vs ... topic. I would love to do it to TI valueline chips but I only have a handful of them.