8 NB-IoT Power Saving (PSM/eDRX)

8.1 Read the Timer Gate From Wake to Sleep

Start with Figure 8.1. Follow the device from active communication into its reachable idle windows, then into power-saving mode. eDRX stretches the interval between paging checks while the device remains registered and periodically reachable. PSM turns the radio off more deeply; the network cannot page the device until it wakes for its own activity.

Figure 8.2 adds the operational decision. Request timer values, capture what the network grants, observe registration and paging behavior, measure current across the full cycle, and compare results with the reachability contract. Feed the measured evidence back into configuration instead of assuming the modem accepted every request.

Imagine a meter that reports every six hours and allows downlink work for two minutes after each report. A command arriving just after that window may wait almost six hours. That can suit a tariff update but not an urgent valve closure. Put urgent local protection in the device or choose a more reachable mode; no cloud retry can wake a modem in PSM.

Energy arithmetic needs the whole cycle. Suppose active transfer draws 120 mA for 20 s, reachable idle averages 8 mA for 100 s, and PSM draws 0.02 mA for the remaining 21,480 s of a six-hour cycle. Charge per cycle is ((120\times20+8\times100+0.02\times21{,}480)/3{,}600=1.01\ \mathrm{mAh}). Four cycles per day use about 4.04 mAh before battery ageing, retransmissions, temperature effects, and regulator losses.

Predict the trace. After an uplink, expect the modem to remain reachable for the granted active interval, then show the eDRX or PSM transition. Send a downlink inside the window and expect delivery. Send another after PSM begins and expect it to wait until the next wake. Compare requested and granted timers in modem or network evidence, and reject any battery estimate built only from requested values.

Networks and modem firmware can grant, round, or reject timers differently. Measure reachability and current with the target operator, radio conditions, firmware, and battery system before making a lifetime promise.

Queue downlinks with an expiry and purpose. A configuration update may wait for the next wake, but an old open-valve command may no longer be valid then. When the meter checks in, the service should deliver only commands whose time and device state still satisfy policy.

Repeat the cycle under weak signal. Extra access attempts and longer radio-on time can dominate charge even when PSM current stays unchanged. Report the distribution of full-cycle use rather than one quiet laboratory trace.

Expose the next expected wake and last granted configuration to support staff. When a command waits, those values distinguish normal PSM delay from a lost meter. Avoid a countdown that assumes the device clock and network timer remain exact after failed attachment attempts.

8.2 Overview: Power Saving Is a Reachability Contract

Write When the Device Can Hear You

Picture a water meter that reports once each day and then sleeps. Firmware is the software stored on the meter. The service team may want to send a new setting, but the meter cannot save energy and listen all the time. The product promise must say when a reply can arrive and how long the team may wait.

Write one daily cycle. Name the report event, listen window, requested timer, granted timer, next wake, queued command, age limit, weak-coverage rule, energy budget, maintenance window, and owner.

Test a command just inside and outside the listen window, a changed network grant, a missed report, weak coverage, a retry, a restart, low energy, and a planned update. Measure the whole device current and the real command result. A low sleep current or requested timer is not field proof.

Keep urgent water safety local while the device sleeps. The wider service can wait for a planned contact, but it should not promise an immediate command when the radio is unreachable.

This opening does not choose one timer or power mode. Practitioner builds the reachability and energy gate. Under the Hood examines timer state, network grants, paging windows, retries, phase energy, firmware changes, and the evidence needed at representative sites.

NB-IoT power saving is not a battery-life button. Power Saving Mode (PSM) and extended discontinuous reception (eDRX) decide when a device can be reached after it sends or receives traffic. A useful design starts by writing the product's downlink requirement, then chooses the least reachable mode that still lets operations do the work they actually need.

PSM keeps the device registered from the network's point of view, but the sleeping device is not listening for arbitrary downlink. eDRX keeps a device periodically reachable by waking it for paging opportunities. Both modes are negotiated with the network, so requested timer values are only proposals until the field logs show what was granted.

Write the reachability promise before picking timers. A daily meter may only need a short post-uplink window for queued settings. A leak alarm may need a bounded acknowledgement path but not arbitrary cloud commands. A maintenance campaign may need a temporary window where devices listen more often, followed by proof that they return to the normal low-power policy. Those promises decide whether PSM, eDRX, a scheduled hybrid, or another technology is acceptable.

The approval evidence should include the granted timer values, the whole-board current trace, server receipts, missed-command behavior, and recovery from weak coverage. If any of those are missing, the power mode is still an assumption rather than a release decision. Record the approved timer policy with firmware version and operator profile, because both can change the reachable state later.

Use Figure 8.1 to relate every low-current interval to reachability. T3324 begins after uplink and bounds the active period; eDRX then offers scheduled paging opportunities before PSM removes ordinary downlink reachability until the next update. The timeline connects current measurements to application delay, preventing “long battery life” from being approved without a command-delivery contract.

Read Figure 8.1 along the timeline: an uplink starts the active timer (T3324), eDRX paging windows follow while the device is still reachable, then PSM sleep makes it unreachable until the periodic update. Do not assume the device can be reached just because it sent an uplink; reachability ends when T3324 expires. Power saving is a reachability contract, and each gate on the timeline is where the evidence for that contract is collected.

8.3 Practitioner: Build the Gate From Granted Timers and Traces

The practical workflow is a ledger. Record the requirement, request candidate PSM or eDRX behavior, read the granted values, and test the same run through modem logs, server receipts, and whole-board current measurements. The battery model should use measured phase durations rather than a single sleep-current number.

Do not measure the modem alone. Host processors, sensors, regulators, debug interfaces, flash writes, and retry logic can dominate the budget after the radio appears to be asleep. The proof is a wake-to-sleep trace from the final firmware on representative hardware.

8.4 Under the Hood: Timers Shape the State Machine

The underlying state machine has three reviewable boundaries. After a device-originated transaction, an active timer can keep the device reachable for pending downlink. eDRX can add periodic paging opportunities while reducing continuous listening. PSM can remove paging reachability until the device wakes by application event or periodic update. The exact values and transitions depend on the network grant and the modem implementation, so the state machine must be observed rather than assumed.

The state machine also includes the host application. The host decides when sensors run, when flash is written, when the modem is allowed to wake, how retries are bounded, and when queued commands are safe to apply. A modem may enter PSM correctly while a regulator, debug interface, sensor heater, or retry loop keeps the board above the expected current. Under-the-hood review therefore pairs modem state with board-level measurement.

Use fault cases to test the transitions. Failed attach, refused timer grant, missed acknowledgement, cloud outage, profile error, and weak coverage can all keep the device active longer than the clean trace. The approved timer model should state the retry limit, backoff behavior, event log, and support action for each case. That model is what lets a reviewer distinguish expected delayed reachability from a defect.

Use Figure 8.2 to ground under the hood: timers shape the state machine in observable evidence. It presents evidence loop for NB-IoT power saving from requirements to requested timers, granted timers, measured current, field retries, and rollout decision rather than an unsupported feature promise. Inspect Request beside Grant before the next claim.

Follow Figure 8.2 from the requirement: the device requests timers, the network grants what it will allow, the full-board current is measured with those grants in place, weak-coverage retries are repeated in the field, and only then does the rollout decision follow. A requested timer is not a granted one, so the loop turns on the grant step. Under the hood, the granted timers are what shape the device’s state machine.