Why might an N MOSFET switch output flip with no change in gate voltage (and then stick)?

In lots of places, we use a simple N-FET switch to turn on/off 24V outputs. We use a 30V rated part and stay well under the power/current limits.

We've started to see recently however a couple of instances of LEDs turning on of their own accord when wired as below:

simulate this circuit – Schematic created using CircuitLab

In a non-failed state, if the MCU pin goes high, the FET is turned on, the voltage at the Drain (LED's cathode) goes to 0V and the LED lights. In the failed state, even when the MCU output is 0V, the LED lights and the Drain voltage is ~0.15V. A high output still pulls the Drain to 0V, but it returns to 0.15V when the FET is switched back off.

I realise that when the switch is open the drain of the FET is floating rather than in a defined state, but how is the LED turning on when there is nowhere for current to flow to?

The datasheet for the part we are using does stipulate a "Zero Gate Voltage Drain Current" (no FET is an ideal switch) but it's measured in micro-amps which is orders of magnitude too small to light the LED. Using the datasheet we can also see that the part is being used within spec as we don't come close to the maximum voltages/currents.

Can anyone suggest what might be going on?

• Is there a facility to disconnect the MCU from the MOSFET gate? Where does the 24 volts come from? What interference might be on the 24 volt power line? What is the physical (cabling) distance between MOSFET and LED? Does the MCU and MOSFET driver share a common 0 volt supply? Jun 15, 2020 at 13:07
• The MCU and FET are both on the same PCB so not easy to disconnect. R2 in the above schematic could be removed if there's something to be ganied from doing so? Jun 15, 2020 at 13:25
• That's 40% answered. Jun 15, 2020 at 13:28
• Sorry, fell victim to the 'enter to post' rather than 'new line'. Then spent more time editing than the allowable window... 24V is the shared supply for the whole system. There's nothing isolated so the derived MCU voltage (3V3) shares the same 0V. The supply itself comes from an industrial supply when in situ, but the PCB in question also shows the same behaviour when powered with a bench supply. The LED is in an external button, but this is only down 5cm of cable so physically quite close. Jun 15, 2020 at 13:45
• Fairchild don't make MOSFETs anymore - they are made by ON semi (unless that part of the fairchild portfolio has been sold off by them) so, maybe someone somewhere decided to pick an alternative supply ans screwed up in some way. Jun 15, 2020 at 13:51