PIC overvoltage failure mode

[evsource]

I have a PIC circuit that monitors a high voltage battery bank. There are two sides of the circuit - what I'll call the HV side and the LV side. The HV side measures the battery bank voltage, and transmits the reading through an optocoupler to PICs on the LV side. The HV pic gets its power from the 5V supply on the LV side, using a 5V to 12V DC-DC converter (the HV side of the DC-DC has a 5V regulator to regulate the 12V down to 5V for the PIC).

Recently, the DC-DC converter failed, and created what appears to have been a surge on the LV 5V supply. It took out all the microcontrollers on that 5V rail, and some other hardware running on the 5V. The 5V supply itself and a 74LS nand gate survived the carnage.

A couple questions about this type of situation:

1) Would it be possible if the PIC saw some voltage surge (I have no idea how high it was, high enough to kill the PICs), that it would output "high" on its I/O pins during the failure? I seem to have evidence that this happened, and after the PIC was dead, the I/O ports seem to then be stuck in the "low" state.

2) Any idea what sort of voltage surge and for what duration it would take to kill a PIC?

3) Any idea why the nand gate survived and not the PICs? It's got an absolute maximum supply voltage rating of 7V. Is it likely that PICs are more sensitive?

I'm dealing with an 18F2620 and 16F88. If anyone asks for my compiler version, I'm going to laugh.

[FvM]

1. Yes
2. Don't know. Somewhere beyond the bearable ESD energy.
3. Yes, uP has smaller structures and is most likely more sensitive.

Generally, a transsorb-diode (power Z-diode) at VCC can be expected to prevent from this kind of fatal damage.

[asmboy]

the actual voltage that it takes to destroy a pic depends a lot on what your port connect schematic looks like. In some cases - a fault in a component being driven by a port ( in or out TRSI - does not matter)
that causes the input protection didodes to switch on and shunt the over volts form a port line - to the Vdd rail .
that is a way to ruin the pic - EZ - and the 5v analog regulator can be popped by this effect too.
( making it appear that the regulator dies first - but not 4 sure true - )

By the same token - if the pic has no connections to any of its I/o - it could likely stand more Vdd over-volts yet - its the bad I/O path that will most likely destroy the part first.

i would suggest that you read about TVS diodes and fast clamp ZENERS

To get true protection for supply based Over volts -
put a fast 5.6V 200 mw zener in parallel with Vdd to ground and bypassed by 10 uf at 10v ( tantalum ok ?) - Now feed your 5V PSU source for the Vdd pin thru a 10 - 15 ohm resistor to the new connection just outlined

or just brute crowbar the darned thing directly using a 6V beefy TVS diode and darn the consequences.

i have had issues before with designs that i've been called on to troubleshoot - that used low quality isolation converter modules.

A COMMON PROBLEM IS THAT UNDER TOO LIGHT A LOAD - some China cheapies i encountered CAN GO OVER VOLT - just as may have happened to you.

this could occur if you are using the same supply to pull relays or have set up a micro control situation where there is BIG load variation - depending on machine state.

[evsource]

[Guest]

the 5.6v zener option is probably the one u want - the right TVS will have a sloppy knee or be too much $$

[evsource]

[asmboy]

some zeners make NOISE - especially when hanging very close to the knee.
( well BELOW - or SATURATED OVER the knee - not so much - its JUST in the knee zone that its dicey , so in the event you were running at 5.45v on a 5.6 you could start to generate some very broadband noise unexpectedly - and thats a real problem if you are using the Vdd ref option for A/d conversion - in particular .

AND it makes sure of a good low - Vdd source impedance - that will have been upset by the 10 ohm series resistor.

sorry i should have explained better - on some issues i'm liberal ,
but my design sense is pretty conservative, having been forged in the '60's.

AS a broader point - you won't go wrong by having some decent low Z capacitor for your Pic Vdd inlet - i tend to use a 47 @10V Tantalum B package SMT part on every design i author or modify.

AND BTW: i use a 100ohm series resistor and 5.6 v Inboard clamp zener
on ANY port output or input line that might be subject to a reverse voltage too, so long as ultra fast switching is not required.
That way you +5.6 protection AND reverse volt ( -.6v clamp) protection too.

on a signal path - the off state zener also exhibits some junction capacitance that forms an RC lag network for fast pulses. Use your slide rule and the zener datasheet to see if you can get away with THAT Bit of protection.

[Ttelmah]

The commonest solution, is to have a thyristor, with perhaps a 5.6v zener feeding it's gate, and a resistor from the gate to ground. A small capacitor in parallel with this, can be used to slightly slow the response, and limit noise at the transition. Key to remember with a TVR, or a crowbar like this, is that there must be a fuse or some form of resettable current limiter in the supply.
At the end of the day, you can obviously monitor the rail, with the PIC itself (A 'classic' way, is to have the PIC using it's comparator, to compare just under half the supply with - say - a 2.5V reference - if the comparator sees the half supply go above 2.5v, then it triggers, and this can be used for a warning, soft turn off, etc..
However you need to think carefully about the Di/Dt of your supply. If a transistor fails, it's resistance can drop to a tiny fraction of an ohm. You add the lead resistances, then look at the capacitances on the line. Just how fast can the supply rise, in it's worst case failure mode. In most cases, if you have something like a high voltage spike, rails feeding into the failing component, will already be well above their normal ratings, and significant energy will be present in capacitors etc.. The rate of rise will be well above anything that a 'software' solution can deal with.
If you look at well designed SMPS units, many will have a physical crowbar, limiting the rail to a figure, and also a 'warning' output, that comes on, when the supply is above or below it's rated level. They are aimed at different things, with the former being the 'last resort' saviour....

Best Wishes

[evsource]

[asmboy]

zener is BACKWARDS - will INSTA trip s drawn - as the zener is just an ordinary .6V FWD drop as you drew it
ALSO - i would put 1K to ground from zener anode insteda 100
and 2 to 5k in series with the zener to NOT over current the gate - if it was a 2n506x or whatever sensitive gate SCR - where justs a few uA will trip the sucker

[evsource]

[Ttelmah]

As a couple of further comments:
You don't detect the supply is falling, you detect it is rising. When the thyristor fires, the rail is killed in a few nSec, if you chose a fast thyristor. Far too fast for the processor to do anything. What the processor can do, is attempt to shutdown, if it sees an overvoltage _before the thyristor fires_.
Generally a gate resistor is not needed, since as soon as any current flows into the gate, the thyristor starts to pull the rail down, protecting it's own gate. One won't hurt, but will slow the response a tiny bit.
I'd be talking 100pF for the capacitor, rather than 100nF. The larger the capacitor the slower the response. You want speed.

Best Wishes

[evsource]

[asmboy]

i've blown sensitive gate 2n5060s by NOT having a small limit resistor in the crowbar - and the failure mode is Anode to cathode SHORT when it does fail - so it is cheap insurance to add that ONE extra bit of r - IF using sensitve gate SCRS for teh duty

[evsource]