MSP430F5232: Old device in a new PCB has started hanging after several thousand units deployed.

Latchup (triggering internal parasitic SCR) requires a power cycle to clear. Assuming the part wasn't damaged of course. Which is why you do not want to exceed the absolute maximum voltage ratings (or diode current) on device pins.

But I can't imagine why you would use an optocoupler that way.

Hi Javier,

Javier Pena said:

My suspicions are that the optocouplers are supplied with a +5V line and directly connected to MSP430F5232 pins thus providing TTL levels to a non TTL device but ,could this be hanging the device in such a way?

Once you apply a voltage outside the Absolute maximum almost any behavior is possible, we don't qualify the behavior of the device at all outside the Absolute Maximum conditions. Prolonged exposure or higher voltages are obviously worse. 

To David's point this does sound similar to latch up behavior. 

Best Regards,
Brandon Fisher

well, thinking about it certainly behaves as a latch up.

The origin of this problem is that previous version didn't have isolation and run directly to client`s  STM MCU which runs at 3.3V. New version works at HV and isolation was chosen, the problem is that the client company made this changes without consulting us  and supposing that input port was TTL compatible as some STMs mcu´s (theirs).

Obviously working at 5V was not desired, nor recommended

The "weird" thing is that measuring the pin is "only" at 3.8V I suppose that current output of optocoupler is not high enough to blow the diode in the GPIO input and it circulating towards 3.3V rail.

After all, can I take that this behaviour can be a direct consequence of this bad tension level? 

.

3.8V isn't weird as that is 3.3V plus a forward diode drop.

It is possible to apply more than the Vcc level to a pin but only if the current is less than the maximum diode current shown in the data sheet. For example, in order to get two way communication with 5V character LCDs, I include a series resistor on the data lines. Large enough to limit the fault current (I also have to worry about the series pass regulator on Vcc) but not so large as to degrade signal transition times.

The data sheet for the optocoupler shows an open collector output. Which means there must be a pullup resistor to 5V. Swap that for one with more resistance to limit current. Or remove it and use a pullup to 3.3V on your board.

analysing the clients company circuit it has a 2,2K pullup resistor in the output pin of the optocoupler., doing the math 5V-3.8V/2.2K = 0.54mA, we have a series resistor from that pin but right now its populated at 0 ohm. Diode current max is +- 2mA so it should be within range.

Changing the clients PCB is no longer an option as they have build +3000 (:facepalm:).

I am just not seeing the benefit of using an optocoupler for isolation and then breaking isolation by providing power to it. The point of using an optocoupler is usually to completely isolate the transmit and receive systems. The receive side should be powered only by the receiving system. I wouldn't even connect the grounds without a very good reason.

Or perhaps this was a really bad attempt at level shifting.

Sorry if I have not explained correctly, our device and the output part of the optocoupler is powered by a isolated DC/DC converter so (in this  case) isolation is well done.

this is the link to such DC/DC : www.mornsun-power.com/.../IB2405LS-1WR3.html

I thought of that but it was just too dumb of an idea to mention. Why waste precious PCB real estate (and money) on an isolated DC/DC converter when with just one extra wire, you can use power from the target system.

Because our device has a USB port and some operators chose to connect their laptop to a HV(+1-2 KVolts) connected device instead of disconnecting the device as the instructions marked, and doing this isolates the usb.

That's why.

after looking for information about Latch up (triggering internal parasitic SCR), I´ve found this paper about it: https://www.ti.com/lit/wp/scaa124/scaa124.pdf

If I understand the paper correctly this Latch up works as a short thus an increase in current should happen. If this is the case, that has no happened to me as current is the same in the blocked state as before.

Also if you review the thread , Voltage in the pin is 3,8 V with 0,54 mA in the diode so it should be within parameters.

kind regards, Javier.

Javier Pena said:

Voltage in the pin is 3,8 V with 0,54 mA in the diode so it should be within parameters.

So Vcc is 3.5V?

no, its 3,3V so max would be 3,6V.