I'm working on an application in which we are using a custom board with the NRF52382, which is powered by a battery. We sample our ADC every 6ms, record these values for about twenty seconds, and then transmit them to a phone using the NUS_uart service.
We've found that there is some noise introduced by what I assume is the link layer in the soft device sending acknowledgement packets to see if the connection with the paired phone is still present. If I disable the softdevice and don't advertise on the disconnect, when observing our signal coming in to the ADC with an oscilloscope, this noise seems to not be present. I believe that radio transmissions are dropping the reference voltage of our power supply, and thus affecting the information coming off our sensors.
I'm wondering what the optimal approach to preventing the radio from drawing power during our sampling / recording period would be. If it would be possible to turn off the link layer for a portion of time during the reading, I think this would be sufficient. Otherwise perhaps tuning the connection parameters in combination with the softdevice scheduler looks like another viable approach, but I'm unsure of the specifics to go about doing this.
