NRF52 reset after voltage/load step

I have a design that uses the NRF52810 as well as NRF9160. 

For safety reasons, it must have a fuse that is about 1 ohm into the system.  I'm seeing an issue after an input load-step of e.g. 1A (used to power the nRF9160 transmit), the nRF52 seems to reboot.  The only thing I could find suspicious is that the voltage that the nRF52 sees goes from ~2V to about 3V in about 1ms when the load is removed, and this is what causes the issue.  Interestingly its not when the load turns on and goes from 3V to 2V.  

Does this sound surprising?

I might try a regulator to just prove this step-up is the issue or not, but unfortunately, it will be very difficult for me to make this change due to some other circuitry that relies on 3.3V (stable) while the cellular isn't running! 

For what its worth, the reset pin is not used in this design as this was the first thing I thought (e.g. a glitch on the pin)

  • I found a footer note in the datasheet that says POR will reset if there's a 300mV/300ms rise.  That sucks but I guess that's exactly my issue.  Is there any further detail on this?  Ironically I have the nRF52 controlling the boost circuit, and have now implemented a controlled *rising* rate of change that seems to fix it.  However, it is VERY slow, so if you guys can explain that circuit at all, maybe I can be a bit smarter (I understand it would be "off datasheet" and carry some risk

  • This is a common issue, and ideally the nRF52 should be isolated while other circuits (solenoids, relays, modems ..) draw significant current which generates a voltage dip and the inevitable following voltage recovery. In principle this is easily done by using a schottky diode from the main voltage supply to the nRF52 VDD capacitor. The nRF52 VDD capacitor should be large enough to allow (say) an advertising burst or other BLE burst without that local nRF52 VDD dipping below that 300mV or so figure; maybe use 2 x 47uF with a voltage rating of at least VDD x 2 (say 6.3 volts) to avoid the ceramic capacitor derating issue. In practice use an ideal diode instead of a schottky to save all that wasted power; tie the EN to VDD so it is always enabled.