Testing power use and how low device can continue to operate.
Using 2-AA batteries last measurement using below line returned 1388mv. voltage = T2Utils::readVoltage(T2_WPN_VBAT_VOLTAGE, T2_WPN_VBAT_CONTROL);
Using volt meter battery pack measured .96 volt.
This corresponds with the cut off of the power on the Talk2 BitBucket site.
"Although the board has been tested and able to run down to 0.75V, the recommended/start voltage must be between 0.9V and 3.3V."
Any idea of why VM and internal T2Util don't agree?
What might happen:
2xAA at 1.388V is the same to say that each cell was at 0.694V (if they drained at exactly the same rate). This is probably the lowest voltage a single cell can go before it's declared dead - in other words not able to deliver any energy.
Once there's no more energy for the step-up to rise the voltage, and, at the same time, deliver the required current, the step-up will basically "short" the battery trying... even no much energy will be dissipate (because there's almost none in the cells), this "short" would still bring the batteries voltage further down, in your case to 0.96V (or 0.48V per cell). That's my conclusion why you now read a lower voltage compared to the latest reading from the Whisper Node.
So the conclusion is:
Voltage is not all we need. The batteries also need to deliver some current to supply the demand required by the step-up circuit (or any other circuit). So, for Alkaline cells the voltage alert will depend if you're running with a single or two battery in series. Staying in the safe side, for single cell the alert voltage can be something between 0.95V and 1V, if you're using 2 batteries in series set your alert between 1.8V and 2V.
Other points to consider: Average Current
Do you know the current while the battery reading is being done? Is that a low-power sketch where you sleep frequently or are you just keeping everything running?
Cell voltage
Sometimes when using AA in series, one of the batteries might drain at different rate. This might result in one of the cells having very low voltage compare to the second cell. Additional it'll also have higher internal resistance, which could reduce the capacity to drain the energy from the "good cell".
Could please measure the cell voltage independently?
Voltage divider and ADC precision
The difference from the Arduino ADC and the VM is quite significant on your case, which really points to the initial explanation. Saying that the battery voltage measurement might have some small drift caused by the resistor tolerances, noise and the ADC precision itself.
Final Experiment
A suggestion is to plug the 2xAA at 0.96V to the Whisper Node, but power the board from the micro-USB. Now the board will be running from the external power supply, but you'll be able to check if the voltage reading from the Whisper Node is close to what your VM is saying.
Please let me know how it goes, this is a great topic to assist others!
I have since run other tests, logging results to a DB every 5s and was able to discharge a single battery into the .8v range before the whispernode dropped out.
The issue I found in 2 different tests when using 2 batteries, monitoring voltage left is not sufficient since batteries don't always discharge at the same rate and therefor one completely dies all of a sudden. In my tests and for my future projects I'll be running on a single battery and alerting at the 1V mark or so.
Using the USB power, the Whisper Node reports 996, within tolerance of the VM reading of .96!
So, maybe switch away from two batteries is best as xnih13 commented.
Means time to change battery holders!
This would allow node to alert base and then base to send alarm warning: email/text/sound/light.
People normally forget about it and jump into rechargeable ones + solar panels. We have supported a project where small motorized ball-valves where controlled by Whisper Node and everything was powered by D cells.
The calculations proved that implementation and running costs were much smaller with D cells. Also the battery expected life was close to 2 years, which was more than satisfying for the customer.
Have switched to a 18650 Lithium ion battery, which provides more than sufficient energy. We have radio nodes reporting to a base: RPI3 running Node-Red, Influxdb, and Grafana. This gives storage option on the RPI3 with cloud options. I did test with a TP4056 and 5 volt solar panel, but the 18650 (3.7v @2200mah) is plenty for my 8 month deployment.
