By the end of this chapter, you should be able to:
Cellular radios are power-hungry, but special modes like PSM (Power Saving Mode) and eDRX (Extended Discontinuous Reception) let IoT devices sleep for hours or days between transmissions. This chapter covers techniques to stretch a single battery to last 10 years or more, making cellular IoT practical for remote, unattended sensors.
“Cellular radios love to eat battery power,” Bella the Battery began, “but PSM is my best friend! With Power Saving Mode, the radio turns completely off when there is nothing to send. I go from draining 15 milliamps down to just 5 microamps – that is three thousand times less power! It is like the difference between running a marathon and taking a nap.”
“PSM is controlled by two timers,” Max the Microcontroller explained. “T3412 says how long I can sleep before checking in with the network, and T3324 says how long I stay awake after sending data. For a sensor that reports once a day, I set T3412 to 24 hours and T3324 to just 30 seconds. Sleep almost all day, wake up briefly, send data, and go back to sleep!”
“eDRX is the middle ground,” Sammy the Sensor added. “With PSM, nobody can reach me while I am sleeping. But with eDRX, I set an alarm to wake up briefly every few minutes to check if anyone has a message for me. It uses more power than PSM but less than staying fully awake, and the network can still send me commands.”
“Here is a fun way to think about it,” Lila the LED suggested. “PSM is like putting your phone on airplane mode for a whole day – totally unreachable but amazing battery life. eDRX is like checking your phone every hour instead of every second – you might miss a message for a bit, but your battery lasts way longer. Choosing the right mode depends on whether anyone needs to reach your device!”
Before diving into this chapter, you should be familiar with:
In one sentence: Battery life in cellular IoT is dominated by idle power consumption, not transmission power, so PSM and eDRX enable 10+ year battery life by reducing sleep current from 15 mA to 5-10 µA.
Remember this: PSM = unreachable deep sleep (5 µA); eDRX = light sleep with periodic wake (15 µA to 1 mA).
Battery life is critical for cellular IoT. Power Save Mode (PSM) and extended Discontinuous Reception (eDRX) are key technologies that enable 10+ year battery life.
PSM enables devices to enter deep sleep while remaining registered to the network, achieving battery life of 10+ years for IoT applications.
Two timers control PSM behavior:
| Timer | Name | Range | Purpose |
|---|---|---|---|
| T3412 | TAU (Tracking Area Update) | 1 hour - 413 days | How long device can sleep before re-registering |
| T3324 | Active Timer | 0 - 186 minutes | How long device stays awake after transmission |
# Enable PSM with specific timers
AT+CPSMS=1,"","","01000011","00000101"
| | |
| | +-- T3324 (Active Timer)
| +----------- T3412 (TAU Timer)
+--------------------- Enable PSM
T3412 encoding (Extended T3412):
Bits 5-7: Timer unit
000 = 10 minutes
001 = 1 hour
010 = 10 hours
011 = 2 seconds
100 = 30 seconds
101 = 1 minute
110 = 320 hours
111 = Deactivated
Bits 0-4: Timer value (0-31)
Example: "01000011" = 001 (1 hour) + 00011 (3) = 3 hours
T3324 encoding:
Bits 5-7: Timer unit
000 = 2 seconds
001 = 1 minute
010 = 6 minutes
Bits 0-4: Timer value (0-31)
Example: "00000101" = 000 (2 seconds) + 00101 (5) = 10 secondsThe Mistake: Developers enable PSM with AT+CPSMS=1 but leave T3324 at the network default (often 0 or very short). The device enters PSM immediately after transmission, and firmware updates or configuration commands never arrive.
Why It Happens: T3324 (Active Timer) defines how long the device stays in RRC_IDLE state (listening for downlink) before entering PSM deep sleep. If T3324=0, the device goes directly from transmission to PSM with zero downlink window. The AT command format AT+CPSMS=1,"","","TAU","ACTIVE" requires explicit T3324 configuration.
The Fix: Always set T3324 to at least 10-60 seconds to allow downlink reception:
AT+CPSMS=1,"","","01000011","00000101"
^^^^^^^^ ^^^^^^^^
T3412=3h T3324=10sT3324 encoding (rightmost 5 bits): 00001=2s, 00010=4s, 00101=10s, 01010=20s, 11111=62s. T3324 multiplier (bits 5-7): 000=2s units, 001=1min units. Example: 00000101 = 5 x 2s = 10 seconds active window. Power cost: 10s @ 10mA idle = 0.028 mAh per wake – negligible compared to attach cost (0.3 mAh) but enables reliable downlink.
eDRX provides a middle ground between always-on connectivity and PSM deep sleep:
| Mode | Sleep Current | Reachability | Latency | Best For |
|---|---|---|---|---|
| Always Connected | 15 mA | Instant | 10-100 ms | Real-time applications |
| eDRX | 15 µA - 1 mA | Periodic | 2.56s - 43 min | Moderate latency tolerance |
| PSM | 5-10 µA | On wake only | Hours - days | Delay-tolerant sensors |
The Mistake: Developers copy eDRX configuration from LTE-M documentation to NB-IoT devices (or vice versa), not realizing the eDRX cycle ranges differ significantly. Result: either the command fails silently, or the device uses an unexpected cycle length.
Why It Happens: AT+CEDRXS uses the same command format but different AcT-type values and different eDRX cycle encodings: - LTE-M (AcT-type 4): eDRX cycles from 5.12s to 2621.44s (43.7 minutes) - NB-IoT (AcT-type 5): eDRX cycles from 20.48s to 10485.76s (2.91 hours) The same 4-bit eDRX value means different cycle lengths depending on technology!
The Fix: Use correct AcT-type and verify cycle encoding:
LTE-M: AT+CEDRXS=2,4,"0101" → 81.92s (1.36 min)
NB-IoT: AT+CEDRXS=2,5,"0101" → 327.68s (5.46 min)NB-IoT eDRX values: 0000=5.12s, 0001=10.24s, 0010=20.48s, 0011=40.96s, 0100=61.44s (1 min), 0101=81.92s, … 1111=10485.76s (2.91 hours). LTE-M values differ! Always verify with AT+CEDRXRDP to read the network-assigned eDRX parameters after configuration.
3G UMTS Complexity:
LTE Simplification:
Impact on IoT:
Discontinuous Reception (DRX) cycles determine how often devices wake to check for paging messages. Adaptive DRX adjusts cycles based on traffic patterns:
Signaling reduction: 30% fewer paging messages by matching DRX to actual device activity patterns
Operating system controls app wake-ups and radio usage:
Example: Android Doze mode, iOS Low Power Mode reduce signaling by 40-60% through aggressive batching
Application Resource Optimizer (ARO) analyzes cellular traffic patterns to identify inefficiencies:
Traffic Analysis:
Alignment Strategies:
Random Access Channel (RACH) handles connection setup. Massive IoT deployments can overwhelm RACH capacity:
Real-World Impact: Smart meter deployments with 100,000+ devices require RACH optimization to prevent “signaling storms” at peak hours (e.g., all meters reporting at midnight)
Time-Dependent Pricing (TDP) for cellular IoT explores how dynamic pricing can shape network demand, reduce costs, and optimize resource utilization.
How can pricing algorithms balance demand with network capacity?
Approach:
Results:
What pricing structures maximize revenue while maintaining user adoption?
Price Elasticity Findings:
Unexpected “Sales Day Effect”:
Revenue Impact:
| Metric | Before TDP | After TDP (1 year) | Change |
|---|---|---|---|
| Peak demand | 100% baseline | 70% baseline | -30% |
| Off-peak demand | 40% baseline | 52% baseline | +30% |
| Total data usage | 100% baseline | 108% baseline | +8% |
| User satisfaction | 3.2/5 | 3.8/5 | +19% |
| Network CapEx | $1M/year | $700K/year | -30% |
Time-Dependent Pricing connects to broader IoT monetization strategies. For business model frameworks and pricing psychology, see:
TDP demonstrates that technical optimization (signaling reduction) directly impacts business economics (30% lower data costs). Understanding both layers enables effective IoT solution design.
Calculating realistic battery life requires accounting for every current-consuming state, not just peak transmit and deep sleep. The following example walks through a complete calculation for a production NB-IoT water meter reporting daily readings.
Device specifications:
Current consumption by state (measured at 3.6 V):
| State | Current | Duration per cycle | Energy per cycle |
|---|---|---|---|
| PSM deep sleep | 5 uA | 23 hr 58 min (86,280 s) | 0.431 mAh |
| Wake + RRC setup | 80 mA | 1.5 s | 0.033 mAh |
| Transmit (23 dBm) | 230 mA | 0.8 s | 0.051 mAh |
| Receive (Rx window) | 48 mA | 2.0 s | 0.027 mAh |
| T3324 active timer | 15 mA | 30 s | 0.125 mAh |
| RRC release + PSM entry | 3 mA | 2.0 s | 0.002 mAh |
| Total per daily cycle | 0.669 mAh |
Annual consumption:
Daily energy: 0.669 mAh
Annual energy: 0.669 x 365 = 244.2 mAh
Self-discharge (3%/year for Li-SOCl2): 19,000 x 0.03 = 570 mAh/year
Year 1 total: 244.2 + 570 = 814.2 mAh
Usable capacity (80% of nominal, accounting for temperature and aging): 15,200 mAh
Projected battery life: 15,200 / 814.2 = 18.7 yearsWhy the real answer is lower (12–14 years):
The theoretical 18.7 years is optimistic. Three factors reduce it:
Coverage-limited retransmissions: In basements or underground meter pits, the NB-IoT signal may require 2–3 HARQ retransmissions. Each retransmission adds 0.8 s at 230 mA. With an average of 1.5 retransmissions per report, the transmit energy doubles from 0.051 to 0.102 mAh per cycle.
Network-imposed TAU: Even if T3412 is set to 24 hours, some operators force a TAU (Tracking Area Update) every 12 hours, adding a second wake cycle. This doubles the daily energy budget.
Temperature derating: Water meters in cold climates (below -10 C) experience 20–40% capacity loss in Li-SOCl2 cells. The passivation layer that forms during long sleep also causes initial voltage sags on wake-up, potentially triggering brownout resets.
Adjusted realistic estimate:
Adjusted daily energy (retransmissions + forced TAU): 1.24 mAh
Annual energy: 1.24 x 365 = 452.6 mAh
Annual self-discharge: 570 mAh
Annual total: 1,022.6 mAh
Usable capacity (cold climate, -10 C): 12,160 mAh (64% of nominal)
Realistic battery life: 12,160 / 1,022.6 = 11.9 yearsKey insight: The T3324 active timer (30 seconds at 15 mA = 0.125 mAh) consumes more energy than the actual data transmission (0.051 mAh). Reducing T3324 from 30 seconds to 10 seconds saves 0.083 mAh per cycle – a 12% battery life improvement at the cost of a narrower downlink window. For meters that never receive downlink commands, T3324 = 2 seconds is optimal.
The battery life calculation shows how T3324 timer optimization impacts longevity. In this scenario, the network forces a TAU every 6 hours, so the device wakes 4 times per day instead of once:
Baseline configuration (T3324 = 30 seconds, 4 wake cycles/day due to forced TAU): \[ E_{\text{daily}} = 4 \times (0.051 + 0.033 + 0.027 + 0.125 + 0.002) + 0.431 = 1.383 \text{ mAh/day} \]
Optimized configuration (T3324 = 10 seconds): \[ E_{\text{active}} = 10 \text{ s} \times 15 \text{ mA} / 3600 = 0.042 \text{ mAh (vs 0.125 mAh)} \] \[ E_{\text{daily, opt}} = 4 \times (0.051 + 0.033 + 0.027 + 0.042 + 0.002) + 0.431 = 1.051 \text{ mAh/day} \]
Battery life improvement: \[ \text{Gain} = \frac{1.383 - 1.051}{1.383} = 24\% \text{ battery life extension} \]
Using the adjusted usable capacity from the worked example (12,160 mAh for cold climate):
The 10-second T3324 optimization adds 7.6 years of battery life – significant for deployments targeting 15+ year operation.
PSM and eDRX power modes connect to broader cellular IoT architecture: the T3412 (TAU timer) and T3324 (Active timer) are negotiated between device and MME during the attach procedure covered in Cellular IoT Overview. Control Plane vs User Plane optimization from NB-IoT Architecture impacts power consumption - Control Plane saves energy by avoiding bearer establishment overhead. Signaling reduction strategies build on LTE DRX fundamentals but extend cycles from seconds to minutes/hours.
Setting T3412 (TAU timer) to 24 hours maximizes battery life but means the device may not check in for 24 hours during error conditions, and any downlink command takes up to 24 hours to deliver. Map application requirements to timer settings: alert device (requires <1 min downlink latency) → T3324=60s, no long PSM; periodic reporting (latency tolerance = report interval) → T3412 = report interval, T3324 = 5s; monthly meter read → T3412 = 30 days, T3324 = 5s.
Calculating battery life from datasheet current values (PSM: 1.5 µA, TX: 200 mA) without measuring actual waveforms misses: startup oscillator current (2–5 mA for 5–50 ms), network registration current (10–50 mA for 2–30 s), IP data transmission bursts, and post-transmission dwell before PSM entry. Use a power analyzer (Nordic PPK2, Otii Arc) to capture the full current waveform from sleep → wakeup → registration → transmit → PSM, then calculate actual energy per cycle.
TCP keep-alive packets (sent every 60–120 s by default) prevent the device from entering PSM. A device configured with PSM T3412=1hour but TCP keep-alive=60s will never enter PSM, consuming 50–100 µA average instead of <5 µA. Disable TCP keep-alive for IoT applications, or set TCP keep-alive intervals to 2× the PSM cycle duration. Use application-level health checks on reconnection instead of TCP keep-alive.
Cellular modules have multiple power states beyond just PSM vs active: airplane mode (RF disabled), minimum functionality (AT+CFUN=0/1/4), and hardware sleep pins (PWRKEY, DTR). Not using hardware sleep pins means the MCU must keep the module powered even when not needed. For maximum power savings, use the module’s PWRKEY or SLEEP_IND pin to cut module power during long inter-transmission periods (>1 hour), accepting the 3–15 second power-up and re-registration penalty.
| Next Chapter | Description |
|---|---|
| Cellular IoT Deployment Planning | Coverage analysis, carrier selection, and real-world deployment considerations |
| eSIM and Global Deployment | Multi-carrier strategies for international IoT connectivity |
| LTE-M Interactive Lab | Hands-on simulation with PSM and eDRX configuration exercises |
| NB-IoT Power and Channel | Detailed NB-IoT power calculations and channel access mechanisms |
| NB-IoT PSM and eDRX | Deep dive into NB-IoT-specific power saving mode implementations |