nRF9151: Estimation of LTE energy consumption

[klaasg]

Hi,

We are trying to get an accurate estimation of energy consumption of LTE searches and data sessions.

We currently do this based on XMODEMSLEEP and CSCON notifications.

Example: if there is 2 seconds of RRC Connected, this will count for (2 seconds * ... mA) = ... mAh of energy spent.

The problem: it is hard to get a good estimation for these energy values. If we have the same data transmission on different networks, it can consume double the energy on one network in comparison to another network. This makes it very hard to estimate, because the times will be almost the same.

Our question: can we somehow retrieve more metrics of the cellular modem? We're thinking about 'TX time', the actual amount of (milli)seconds the modem was transmitting, or TX power, the output power the modem is set to. We see this power can differ depending on the network you're registered to.

Thank you.

[Achim Kraus]

Not sure, what's to goal you want to achieve.

In my experience, even statically located devices will face pretty different "radio conditions" over time. Including network searches and so on. With that, the runtime from battery is rather a range. An assumption, which worked quite well over some years, is to use the consumption on mid-term to calculate long-term, e.g. the consumption since the last restart and that of the last 2 weeks to calculate a forecast combining the two. For the npm1300 there is also an AI based function to determine, how much energy is left in the battery.

An other approach I use now for a year is testing with replacing the battery (none-linear relation of voltage and energy) by a super cap., which has a close to linear function of voltage and energy ( e.g. LIB1620Q4R0407, 4V, 400F others will obviously work as well). With that you may run the device for 2-3 weeks and you will get pretty realistic numbers for the energy consumption.