killed sensor by touching Vcc with my soldering iron? :(

[exe]

Hi there!

I'm troubleshooting problems with our sensors. The problem is that they "forget" their configuration. Sensors are provided to us, so we are not the developers of it.

We suspected that issues with power could trigger erratic behavior. So, I decided to simulate problems with power by inserting a 20 Ohm resistor on Vcc. Sensor indeed went nuts. But I also noticed when I touch Vcc with my soldering iron, sensor reboots. The iron is grounded to mains, sensor is connected to computer via usb-to-serial adapter. The serial adapter is not halvanically isolated from usb, and is plugged to my laptop, the laptop is powered from the same power extender.

Now sensor seems to be half-dead: it blinks, it seems it responds to some commands (I can read back sensor id, and trigger things like calibration), but I can't read anything from it.

My main question is: what happened? Why soldering iron could make sensor to reboot? I checked with my DMM, there is not significant DC or AC voltage between the usb's Vcc and the iron tip. I only touched Vcc, no other pins or anything.

PS soldering iron is unmodified AIXUN T3A.

[fzabkar]

Measure the voltage between USB ground and mains ground. If it is sitting at half mains potential, then that could be the problem.

[Kleinstein]

Most PCs have a connection to ground. With desktops this is often a direct like to PE. With labtops this can be only capacitors to both the life and neutra wire, so that one has something like half the mains voltage to ground with capacitive coupling.

With a grounded soldering iron there can than be current flow, e.g. shoring the supply or injecting supply spikes. Even if not grounded it is usually not good idea to solder while under power, as ESD events can trigger a latch up and destroy or damage chips, even though many modern chips get relatively robust in this respect (also relatively low power supplies help to minimize damage).

[wraper]

We suspected that issues with power could trigger erratic behavior. So, I decided to simulate problems with power by inserting a 20 Ohm resistor on Vcc. Sensor indeed went nuts. But I also noticed when I touch Vcc with my soldering iron, sensor reboots. The iron is grounded to mains, sensor is connected to computer via usb-to-serial adapter. The serial adapter is not halvanically isolated from usb, and is plugged to my laptop, the laptop is powered from the same power extender.

Both solering iron and laptop are grounded. What else do you expect when shorting VCC to GND?

[exe]

Both solering iron and laptop are grounded. What else do you expect when shorting VCC to GND?

I measured again and now it's 4.5V DC between iron tip and VCC. I guess, I made a measurement mistake or something. I'm too tired today, will repeat my measurement tomorrow.

I was also told that it's possible to re-flash sensor. If sensor survives, then it's a success -- I reproduced problem that we have in the field: sensors "forget" their configuration, and it's probably due to quality of power.

The next question would be why this is happening. Sensors should not write anything to flash during operation. So, I'd assume if there are no writes to flash, then power supply issues should not corrupt memory. Either it's a wrong assumption, or sensor does some sort of recalibration or something.

[thm_w]

We don't know what a "sensor" is without some photos or other information.

If the sensor board contains a microcontroller, the brownout on Vcc may not be protected against, and then the micro goes wild running random instructions. Some of these instructions could be programming flash. I've seen this happen many times.

Other, less likely case, is that its crap flash thats corrupted.

If it is a micro you could check if it is read protected or not.

[james_s]

Regardless of exactly what happened, it's not advisable to solder components while they're live. Is there some reason you can't just unplug it before you mess with it?

[exe]

Regardless of exactly what happened, it's not advisable to solder components while they're live. Is there some reason you can't just unplug it before you mess with it?

My poor memory)

The idea is to reproduce realistic scenario on the field, I can't go to the vendor and say "hey, your sensors die when I try to solder when while they operate".

[exe]

We don't know what a "sensor" is without some photos or other information.

If the sensor board contains a microcontroller, the brownout on Vcc may not be protected against, and then the micro goes wild running random instructions. Some of these instructions could be programming flash. I've seen this happen many times.

Sorry, forgot to put these details. The sensor is based on ATSAMS70Q21 MCU, plus W9864G6KH (RAM, I presume), plus an UWB human presence sensor (which I'm not sure what it is, hiddent under a metal can). It basically detects if a person is present near-by.

If you've seen MCUs doing weird stuff with its own flash, then it's already good a datapoint. We also suspect that a power glitch in the grid caused a brown-out or something. What we have is three sensors got knocked out along with raspberry pi (though I'm not sure if rpi was completely dead, or we only had to replace its sd card). Two sensors forgot their configuration, one just reset itself.

And another topic how realistic it is to fix the problem in firmware, or another hardware revision. On the one hand, it's understandable that an MCU requires reasonably stable power rails, etc. On another hand, I personally would like sensors to be more robust to harsh environments, if possible.

If problem with power confirms, we'll probably install a UPS.

[Kleinstein]

Modern µCs usually have some protection against a brown out. This was more a thing in the 1990s (e.g. early AVRs tend to overwrite EEPROM position 0).  It may still need some care in the firmware to get a correct initialization / set-up.  A problem could be that under normal conditions the supply goes down relatively smooth and slow (e.g. ms range) while the soldering iron is more like a short and thus instant off and maybe a fast on / off sequence.  The short may in theory damage some moudules with internal voltage regulators and buffer capacitors from reverse current flow.

[thm_w]

Yes any modern micro will have brownout protection, if enabled. Many people either forget to enable it, until a problem surfaces, or purposefully disable it as it draws excess power. ATSAMS70Q21 looks like a beast though, so might not be designed around low power consumption.

If you are using a raspi I would consider a UPS. Especially if not using high grade industrial SD cards. Corruption used to be a common issue, not sure if anything has changed there.
https://hackaday.com/2022/03/09/raspberry-pi-and-the-story-of-sd-card-corruption/

You could also play with a power supply repeatedly switching it on and off and ramping it up and down to try to kill the sensor.

[exe]

In case you are curious what was the problem, we believe there are two issues:
1) firmware constantly frequently writes something to flash while it shouldn't, vendor promised to fix that
2) the usb hub was switching between external power brick and raspberry pi. Rpi couldn't supply all the power needed. We found that inside the hub there is a jumper that forces using external power only.

[GigaJoe]

btw, pi has current limiter over usb, as they assume a provided power supply has  a limited power

so i simply did bypass wire from 5v power socket to 5v usb pins
that may simplify your power design
add 6-8V tvs diode