PSM vs eDRX: Tracker Battery Life Math

Illustration

Tracker battery life comes from average current: sleep current, active-window current during T3324, and the charge of each report, divided into a derated battery capacity. Using Nordic's published 2.7 µA PSM floor and an example 1000 mAh cell, the floor alone gives only an upper bound, so measure report charge and granted timers on your board.

A battery claim for a cellular tracker needs a defensible number, and a datasheet floor current does not give one. This guide sets out the average-current method for PSM and eDRX, the published figures you can start from, and a spreadsheet layout you can rebuild in a few minutes. It contains no measured result and no Duobay price. The example inputs are marked, and every other input is yours to replace.

Which published figures can you start from?

Only a few numbers in this area come from sources we could read directly. Each row below is as of October 31, 2026. The Nordic figures come from a distributor-hosted copy of Nordic's own product brief, and the brief's version and test conditions should be re-checked on nordicsemi.com before you rely on them. The brief gives its figures at 3.7 V and 23 dBm TX.

This article does not quote the timer encodings for T3324 and T3412, so it does not decode raw timer values. Check the 3GPP NAS specification your modem vendor cites.

How do you compute average current and battery days?

Pick one wake-up cycle of length T. Add the charge spent in each state, divide by T, and convert the battery capacity into days at that current.

t_sleep   = T − T3324 − t_conn                          [s]
Q_cycle   = I_sleep × t_sleep + I_act × T3324 + Q_event  [µA·s]
I_avg     = Q_cycle / T                                  [µA]
days      = C × (1 − d) × 1000 / I_avg / 24              [days]

C in mAh. Multiplying by 1000 converts mAh to µAh. Dividing µAh by µA gives hours.

For eDRX, the radio listens only inside the Paging Time Window. If the device wakes inside the window and sleeps outside it, the average current from paging is:

I_edrx_avg = (I_ptw × L + I_sleep × (T_eDRX − L)) / T_eDRX      [µA]
I_total    = I_edrx_avg + Q_event × n / 86,400                  [µA]

L and T_eDRX in seconds. n = reports per day. 86,400 = seconds per day.

The second line adds the report charge, spread over a day. Because the window ends early when a page arrives, L is the longest monitoring time and the formula gives a conservative paging term.

Variables, and which ones are yours

Only the rows marked as published or example come with a value. Every other input is an assumption or a measurement for you to supply.

What does the published current alone give you?

A bound is the quickest check on the method. Take the example 1000 mAh cell with no derate and divide by each published current. Hours = 1000 mAh × 1000 / current in µA. Days = hours / 24. The result ignores self-discharge, every report and every active window, so read it as an upper bound only.

Inputs: the Nordic currents in the first table (source, as of October 31, 2026) and a 1000 mAh example cell. Results are rounded to whole numbers.

These day counts leave out the derate d, the report charge and the active window, which is why they cannot be a battery life. Put your own numbers into the sheet below and see which term is largest.

How do you build the spreadsheet?

One column of inputs and a few formulas are enough. Type the inputs into B1 to B8 and let the formulas fill B9 to B12.

Then add a sweep. Vary T3324 over the values your carrier might grant, and vary the report interval T, and watch B11 and B12. Compare the three terms in B10: the sleep term, the active-window term and the report charge. The largest term is where your effort should go. Each term has a different lever: the granted window, the report interval or the connection behavior.

What are the common pitfalls?

These follow from the figures and spec text above. They are engineering reasoning about how the numbers behave. None describes an incident at any company.

• The floor is not the average. The brief lists a 2.7 µA PSM floor and a 6 µA average at a 655 s eDRX interval for LTE-M. The average is more than double the floor. A battery estimate that uses only the floor leaves out every active term.
• The same interval gives different averages on the two radios. At 655 s the brief lists 6 µA for LTE-M and 9 µA for NB-IoT. Copying one radio's figure into a design that uses the other changes the result by half again.
• The eDRX range is not the same on both radios. TS 36.304 gives 1 to 256 hyper-frames for LTE-M and 2 to 1024 for NB-IoT. A long NB-IoT cycle may have no LTE-M equivalent.
• The window can end early. TS 36.304 says the UE monitors paging during the PTW or until a page for it arrives, whichever is earlier. Using the full L as the monitoring time gives a conservative paging estimate, but a received page also starts a connection. Its cost is not in any published figure here, so measure it.
• Requested is not granted. The network can grant timers that differ from the ones the device requests. A calculation built on the requested T3412 and T3324 describes a request. Read back the granted values and rerun the sheet.
• Vendor conditions do not travel. The brief figures are given at 3.7 V and 23 dBm TX for one module family. Another module, supply voltage or TX power needs its own datasheet row.
• A module figure is not a board figure. The published floor is for the module. A real tracker adds a SIM, power management parts and sensors, and the published numbers do not include them. Measure the floor on the assembled board.

What should you measure on your own board?

The vendor did not publish a per-report charge, an idle-window current or a carrier-granted timer for your case, so you need to measure them on your board. We have not measured them, and no result appears in this article. The pass line is set by you from your product requirements, and you should record the basis for it.

Fix the pass line before you run the test, and write down why you chose it. A line chosen after you see the result tends to land just under the result.

What should you think through before you choose?

Take this list to your firmware team and to any supplier who quotes a battery life for your tracker.

• Which radio, LTE-M or NB-IoT, does the design use, and does every current figure we use belong to that radio?
• What T3412 and T3324 does the firmware request, and what did the carrier grant on a real SIM?
• Which figures in the sheet are measured on our assembled board, and which are still vendor numbers?
• Does the sleep-floor measurement include the SIM, power management parts and sensors?
• What is the cell, what derate did we apply, and where is the basis written down?
• How does the calculation change if the carrier grants shorter timers than we requested?
• Which term is largest in the sheet today: the sleep floor, the active window or the report charge?
• What will we do if the measured result misses the promise on the product page?

If you want a power budget and battery recommendation for your tracker, tell us about your tracker. You can also read about our LTE and GNSS tracker development.

Primary sources

All pages below were accessed on 2026-10-31. Confirm figures and document versions on the live pages. This is an engineering write-up of public documents. It has no certification status.

• ETSI, 3GPP TS 36.304 V17.4.0, UE procedures in idle mode (sections on eDRX and the Paging Time Window): https://www.etsi.org/deliver/etsi_ts/136300_136399/136304/17.04.00_60/ts_136304v170400p.pdf (accessed 2026-10-31)
• Nordic Semiconductor, nRF9160 SiP product brief (distributor-hosted copy): https://cdn.sparkfun.com/assets/learn_tutorials/1/5/9/7/nRF9160_SiP_product_brief.pdf (accessed 2026-10-31)