POF warning question

[John]

Hello,

This is a quick question regarding the behavior of the Power Supervisor POF warning. We implemented a fairly naive use of the POF warning where we would log an POF event whenever we received a POF warning. We set the POF warning threshold to 2.7V. After seeing the event when we didn't expect it, we added some debug to log a very recent battery voltage measurement to our log when we logged the POF warning. It showed the battery voltage to be 3.4V.

This led to a number of questions regarding how the POF works. Most of which I could answer based on the reference manual and questions and responses on the forum. However, there were two questions I could not answer.

One, was how long a duration of the voltage below 2.7V is needed to trigger the POF warning. My guess based on what I see in the reference manual is any transient below 2.7V will trigger the warning (assuming of course that the voltage went back above the hysteresis level after each trigger). Is this correct, or does the voltage need to remain below 2.7V for some small amount of time before the warning is triggered?

Two, was is it possible to see false triggers of the POF warning?

Thanks, John

[MartinBL]

Q1. You just need a little dip below the threshold to trigger warning.

Q2. I haven't been able to find reports of false triggers like this.

1. Do I understand it correctly that you measure 3.4V right before and/or right after the POF warning?
2. Is it possible that current spikes caused by e.g. the radio turning on or by other circuitry on your device cause a voltage drop low enough to trigger the warning?
3. Is it possible for you to use an oscilloscope to monitor the supply voltage?
4. It is advisable to disable/enable the POF right before/after your device goes to sleep. It needs HFCLK to run and it is safer to gracefully turn it off while it is not in use.
5. If the problem persists you might consider using LPCOMP instead and see if that helps.

[John]

A1: That's what I thought since brownout looks to be controlled by the same logic.

A2:

1: With the data we have the 3.4V was measured within a minute of the event. Not close enough, but it is doubtful the battery truly discharged below 2.7. A transient dip seems much more likely.

2: Absolutely IMHO. I'm an embedded FW guy, but that seems very likely. We'll see what the HW guys say.

3: We have a meeting in an hour or so. I strongly suspect a HW guy will get that action in the meeting

4: That's good to know. Nothing I read before suggested this.

5: Thanks for the suggestion. Plus the LPCOMP uses less power which is always a plus for us.

[John]

Also note, we are making an addtional change to our implementation. We are adding a battery measurement to the event handler rather than using the last measurement made. We are also looking at checking the battery voltage via a timer some short time after the event before taking any action to shut down gracefully. 2.7V actually provides us with quite a bit of margin unless we have something like a short on the board.

[John]

I think I have what I need here so I am marking the question answered and closed. Thanks!

[MartinBL]

4: It might not be mentioned explicitly anywhere. It is one of those "better safe than sorry"-scenarios. It is generally recommended to shut down your peripherals when not in use.