1603  Power Consumption Analysis

1603.1 Learning Objectives

By the end of this chapter, you will be able to:

• Analyze power consumption across different device operating states
• Calculate average current for duty-cycled IoT devices
• Identify and quantify power consumption components
• Understand radio power states and their impact on energy consumption
• Create accurate power budgets for IoT system design

1603.2 Power Consumption Analysis

Understanding power consumption requires analyzing all device states and their duration. Most IoT devices cycle through multiple power modes:

%%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#2C3E50', 'primaryTextColor': '#fff', 'primaryBorderColor': '#2C3E50', 'lineColor': '#16A085', 'secondaryColor': '#E67E22', 'tertiaryColor': '#7F8C8D'}}}%%
stateDiagram-v2
    [*] --> DeepSleep
    DeepSleep --> LightSleep: Timer/Interrupt
    LightSleep --> Active: Wake Complete
    Active --> Sensing: Start Measurement
    Sensing --> Processing: Data Ready
    Processing --> Transmitting: Send Data
    Transmitting --> LightSleep: TX Complete
    LightSleep --> DeepSleep: Extended Idle

    note right of DeepSleep: 1-100 µA
    note right of LightSleep: 0.1-1 mA
    note right of Active: 10-50 mA
    note right of Sensing: 5-50 mA
    note right of Processing: 20-80 mA
    note right of Transmitting: 100-300 mA

Figure 1603.1: IoT device power state machine showing typical current consumption for each state

1603.2.1 Power State Definitions

State	Current Range	Wake Time	What’s Active
Deep Sleep	1-100 µA	100-500 µs	RTC only, RAM retained
Light Sleep	0.1-1 mA	10-100 µs	RTC + limited peripherals
Idle	1-10 mA	Immediate	CPU idle, peripherals on
Active	10-50 mA	N/A	CPU running, no radio
Sensing	5-50 mA	N/A	Sensors powered and sampling
Processing	20-80 mA	N/A	CPU compute-intensive
TX/RX	100-300 mA	N/A	Radio transmitting/receiving

1603.2.2 The Idle Power Myth: Radio Idle ≈ Radio Receive

WarningCritical Insight: “Idle” Radio Still Consumes Significant Power

Many developers assume a radio in “idle” mode consumes minimal power. This is FALSE for most wireless protocols:

Protocol	TX Power	RX Power	Idle Power	Deep Sleep
Wi-Fi	160-300 mA	80-120 mA	50-80 mA	10-20 µA
BLE	10-20 mA	10-15 mA	1-5 mA	<1 µA
LoRa	40-120 mA	10-15 mA	1-5 mA	0.2-1 µA
Zigbee	15-30 mA	15-25 mA	1-5 mA	<1 µA

Key insight: For Wi-Fi, idle mode (50-80 mA) is nearly as expensive as receiving (80-120 mA). The only way to save power is complete radio shutdown with deep sleep (10-20 µA).

For BLE and LoRa, idle is better than active but still consumes 1,000× more than deep sleep. Always completely disable radios when not in use.

1603.2.3 Average Current Calculation

For duty-cycled devices, average current determines battery life:

\[I_{avg} = \sum_{i=1}^{n} \frac{I_i \times T_i}{T_{total}}\]

Where:

• \(I_i\) = Current in state i
• \(T_i\) = Time in state i
• \(T_{total}\) = Total cycle time

Example: Environmental Sensor (1-hour cycle)

State	Current	Duration	Energy (mAs)
Deep Sleep	10 µA	3595 s	35.95
Wake + Init	20 mA	0.5 s	10
Sensor Read	5 mA	2 s	10
Process Data	30 mA	0.5 s	15
Wi-Fi Connect	160 mA	1.5 s	240
Wi-Fi TX	200 mA	0.3 s	60
Wi-Fi Shutdown	80 mA	0.2 s	16
Total		3600 s	386.95

\[I_{avg} = \frac{386.95 \text{ mAs}}{3600 \text{ s}} = 0.107 \text{ mA} = 107 \text{ µA}\]

Battery life with 2000 mAh capacity:

\[Life = \frac{2000 \text{ mAh}}{0.107 \text{ mA}} = 18,692 \text{ hours} = 2.13 \text{ years}\]

1603.2.4 Radio Power States as Energy Multipliers

%%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#2C3E50', 'primaryTextColor': '#fff', 'primaryBorderColor': '#2C3E50', 'lineColor': '#16A085', 'secondaryColor': '#E67E22', 'tertiaryColor': '#7F8C8D'}}}%%
graph TB
    subgraph Radio["Radio Power States (ESP32 Wi-Fi)"]
        R1["TX (200 mA)<br/>20,000× deep sleep"]
        R2["RX (100 mA)<br/>10,000× deep sleep"]
        R3["Idle (70 mA)<br/>7,000× deep sleep"]
        R4["Modem Sleep (20 mA)<br/>2,000× deep sleep"]
        R5["Light Sleep (0.8 mA)<br/>80× deep sleep"]
        R6["Deep Sleep (10 µA)<br/>Baseline"]
    end

    R1 --> R2
    R2 --> R3
    R3 --> R4
    R4 --> R5
    R5 --> R6

    style R1 fill:#E74C3C,stroke:#2C3E50
    style R2 fill:#E67E22,stroke:#2C3E50
    style R3 fill:#F1C40F,stroke:#2C3E50
    style R4 fill:#27AE60,stroke:#2C3E50
    style R5 fill:#16A085,stroke:#2C3E50
    style R6 fill:#2C3E50,stroke:#2C3E50

Figure 1603.2: Radio power states showing the dramatic difference between active and sleep modes

1603.2.5 Protocol Energy Comparison

Different protocols have vastly different energy costs per transmission:

Protocol	Typical TX Time	TX Current	Energy per TX
BLE (1 packet)	1-3 ms	15 mA	15-45 µJ
Zigbee	5-10 ms	20 mA	100-200 µJ
LoRa SF7	50 ms	40 mA	2,000 µJ
LoRa SF12	1,500 ms	40 mA	60,000 µJ
Wi-Fi (short)	100 ms	180 mA	18,000 µJ
Wi-Fi (incl. connect)	3,000 ms	160 mA	480,000 µJ
LTE-M	500 ms	220 mA	110,000 µJ
NB-IoT	1,000 ms	200 mA	200,000 µJ

Key observation: BLE is 100-1,000× more efficient than Wi-Fi per transmission. Choose protocols carefully based on range and data requirements.

1603.3 Power Budget Worked Examples

1603.3.1 Worked Example: Optimizing Power Budget for a Wildlife Tracking Collar

Scenario: You are designing a GPS tracking collar for deer monitoring in a wildlife research project. The collar must last 2 years on a single battery, transmit location every 4 hours, and weigh under 100g (including battery).

Given:

• GPS module: L96 (acquisition 26mA × 30s cold start, 22mA × 1s hot start)
• Radio: LoRa SX1276 (40mA TX @ +14dBm)
• MCU: STM32L4 (8mA active, 1µA deep sleep)
• Battery constraint: <50g → 2× CR2450 = 1,200 mAh total
• Transmission: 50-byte location packet every 4 hours

Steps:

1. Calculate GPS energy per acquisition:

   Cold start (first fix after power cycle):
   E_cold = 26mA × 30s = 780 mAs = 0.217 mAh
   
   Hot start (subsequent fixes with ephemeris):
   E_hot = 22mA × 1s = 22 mAs = 0.006 mAh
   
   Average (assuming 1 cold start per day, 5 hot starts):
   E_gps_day = 780 + (5 × 22) = 890 mAs/day = 0.247 mAh/day

2. Calculate LoRa transmission energy:

   TX time for 50 bytes @ SF7: 70ms
   E_tx = 40mA × 0.07s = 2.8 mAs per transmission
   6 transmissions/day = 16.8 mAs/day = 0.0047 mAh/day

3. Calculate MCU energy:

   Active time per cycle: 2s (wake, process, initiate GPS/radio)
   Active energy: 8mA × 2s × 6 = 96 mAs/day = 0.027 mAh/day
   
   Sleep time: 24h - (6 × 32s) = 86,208s
   Sleep energy: 1µA × 86,208s = 86.2 mAs/day = 0.024 mAh/day

4. Total daily consumption:

   GPS:     0.247 mAh/day
   LoRa:    0.005 mAh/day
   MCU Act: 0.027 mAh/day
   MCU Slp: 0.024 mAh/day
   Total:   0.303 mAh/day

5. Calculate battery life:

   Theoretical: 1,200 mAh / 0.303 mAh/day = 3,960 days = 10.8 years
   With 70% efficiency factor: 3,960 × 0.7 = 2,772 days = 7.6 years

Result: The design significantly exceeds the 2-year requirement. This margin can be used to:

• Increase transmission frequency (every 2 hours → still 3.8 years)
• Add activity detection with accelerometer
• Use smaller/lighter battery (600 mAh still gives 3.8 years)

Key Insight: GPS acquisition dominates the power budget (82%). Optimizations should focus on keeping GPS in hot-start mode (preserve ephemeris data) or using A-GPS for faster acquisition.

1603.3.2 Worked Example: Comparing Protocol Energy Costs for a Smart Agriculture Sensor

Scenario: A soil moisture sensor transmits 100-byte readings every 30 minutes. Compare Wi-Fi, BLE, and LoRa for a 2-year battery life target.

Given:

• Sensor reading: 5mA × 0.5s = 2.5 mAs per reading
• MCU (nRF52840): 5.5mA active, 1.5µA deep sleep
• Processing: 10ms per reading
• Readings per day: 48

Protocol Comparison:

Protocol	Connection Time	TX Time	Total Energy/TX
Wi-Fi	3s @ 160mA = 480 mAs	100ms @ 200mA = 20 mAs	500 mAs
BLE	0ms (advertising)	5ms @ 15mA = 0.075 mAs	0.075 mAs
LoRa SF7	0ms	100ms @ 40mA = 4 mAs	4 mAs

Daily Energy Consumption:

Component	Wi-Fi	BLE	LoRa
Radio (48 TX)	24,000 mAs	3.6 mAs	192 mAs
Sensor	120 mAs	120 mAs	120 mAs
MCU Active	26.4 mAs	26.4 mAs	26.4 mAs
MCU Sleep	130 mAs	130 mAs	130 mAs
Daily Total	24,276 mAs	280 mAs	468 mAs

Battery Life (2000 mAh @ 70% efficiency):

Protocol	Daily (mAh)	Life (days)	Life (years)
Wi-Fi	6.74	208	0.57
BLE	0.078	17,949	49.2
LoRa	0.130	10,769	29.5

Conclusion: Wi-Fi gives only 7 months of battery life, making it unsuitable without energy harvesting. BLE achieves 49 years (theoretical), allowing for smaller batteries or higher transmission rates. LoRa provides excellent range with 29 years life. For this application, BLE is optimal for short-range or LoRa for long-range.

1603.4 Real-World Example: Smart Agriculture Deployment

A practical deployment of 100 soil moisture sensors across a vineyard:

Initial Design Choices:

• ESP8266 with Wi-Fi (familiar technology)
• 4× AA batteries (6V, 2000mAh effective)
• Transmit every 15 minutes

Measured Power Profile:

State	Current	Duration	Energy
Wi-Fi Connect	160 mA	5 s	800 mAs
Wi-Fi TX	200 mA	0.5 s	100 mAs
Idle (no sleep)	15 mA	894.5 s	13,418 mAs
Per cycle		900 s	14,318 mAs

Average current: 14,318 / 900 = 15.9 mA

Battery life: 2000 mAh / 15.9 mA = 126 hours = 5 days

Problem: Sensors died after one week. Replacement labor cost: 100 sensors × 30 min × $20/hr = $1,000 per cycle = $52,000/year

1603.4.1 Optimized Redesign

Changes Made:

1. ESP32 with deep sleep (10 µA vs 15 mA idle)
2. LoRa radio (40 mA TX, no connection overhead)
3. Transmit every 30 minutes (reduced from 15)
4. Static network configuration

Optimized Power Profile:

State	Current	Duration	Energy
Wake	20 mA	0.1 s	2 mAs
Sensor Read	5 mA	0.5 s	2.5 mAs
LoRa TX	40 mA	0.1 s	4 mAs
Deep Sleep	10 µA	1799.3 s	18 mAs
Per cycle		1800 s	26.5 mAs

Average current: 26.5 / 1800 = 0.0147 mA = 14.7 µA

Battery life: 2000 mAh / 0.0147 mA = 136,054 hours = 15.5 years

Result: Battery life improved from 5 days to 15+ years—a 1,000× improvement.

1603.5 Common Power Budget Mistakes

WarningAlarm Bells: Common Power Budget Mistakes

1. Ignoring Radio Idle Current

WRONG: "Radio off = 0 mA"
RIGHT: Check datasheet - many radios have 50-80 mA idle current!

2. Underestimating Wi-Fi Connection Time

WRONG: "Wi-Fi TX takes 100ms"
RIGHT: Connection + negotiation + TX can take 2-10 seconds

3. Forgetting Pull-up Resistors

10kΩ pull-up at 3.3V = 330 µA CONTINUOUS
This alone consumes 3× more than ESP32 deep sleep!

4. Assuming Datasheet Sleep Current

Datasheet: "10 µA deep sleep"
Reality: Add peripherals, pull-ups, LEDs = 100+ µA
ALWAYS MEASURE on actual hardware!

5. Ignoring Voltage Regulator Quiescent Current

LDO quiescent: 30-100 µA typical
Can exceed MCU deep sleep current!
Choose low-IQ regulators for battery designs.

6. Not Accounting for Peak Currents

Battery voltage drops under load
Brown-out resets cause device failure
Solution: Add bulk capacitor or choose battery with low ESR

1603.6 Power Budget Calculator

Use this interactive calculator to estimate your device’s battery life:

viewof calc_voltage = Inputs.range([1.8, 12], {value: 3.3, step: 0.1, label: "Operating Voltage (V)"})
viewof calc_active_current = Inputs.range([0.1, 500], {value: 50, step: 0.1, label: "Active Current (mA)"})
viewof calc_sleep_current = Inputs.range([0.001, 10], {value: 0.01, step: 0.001, label: "Sleep Current (mA)"})
viewof calc_active_time = Inputs.range([0.1, 60], {value: 5, step: 0.1, label: "Active Time (sec/hour)"})
viewof calc_battery_mah = Inputs.range([100, 10000], {value: 2000, step: 100, label: "Battery Capacity (mAh)"})
viewof calc_efficiency = Inputs.range([0.5, 1.0], {value: 0.7, step: 0.05, label: "Efficiency Factor"})

calc_sleep_time = 3600 - calc_active_time
calc_avg_current = (calc_active_current * calc_active_time + calc_sleep_current * calc_sleep_time) / 3600
calc_battery_life_hours = (calc_battery_mah * calc_efficiency) / calc_avg_current
calc_battery_life_days = calc_battery_life_hours / 24
calc_battery_life_years = calc_battery_life_days / 365

md`### Power Budget Results

| Metric | Value |
|--------|-------|
| Sleep Time per Hour | **${calc_sleep_time.toFixed(1)} seconds** |
| Average Current | **${calc_avg_current.toFixed(4)} mA** (${(calc_avg_current * 1000).toFixed(2)} µA) |
| Power Consumption | **${(calc_voltage * calc_avg_current).toFixed(3)} mW** |
| Effective Capacity | **${(calc_battery_mah * calc_efficiency).toFixed(0)} mAh** |
| Battery Life | **${calc_battery_life_days.toFixed(1)} days** (${calc_battery_life_years.toFixed(2)} years) |

${calc_battery_life_years >= 5 ? "✅ **Excellent**: 5+ year battery life achieved!" :
  calc_battery_life_years >= 2 ? "✅ **Good**: 2+ year battery life - suitable for most IoT deployments." :
  calc_battery_life_years >= 1 ? "⚠️ **Moderate**: 1-2 year life - consider optimizations." :
  "❌ **Poor**: Less than 1 year - significant optimization needed."}`

1603.6.1 Understanding the Calculation

The calculator uses the duty cycle formula:

\[I_{avg} = \frac{I_{active} \times T_{active} + I_{sleep} \times T_{sleep}}{T_{total}}\]

And battery life formula with efficiency factor:

\[Life = \frac{Capacity \times Efficiency}{I_{avg}}\]

The efficiency factor (default 0.7) accounts for:

• Temperature effects on battery capacity
• Voltage cutoff (can’t use full capacity)
• Self-discharge over deployment lifetime
• Battery aging/degradation
• Non-ideal DC-DC converter efficiency

1603.7 Knowledge Check

Question 1: Your ESP32 IoT device shows these measured currents: Active (Wi-Fi TX) = 120mA for 5s every 15 minutes, Idle (no sleep) = 80mA rest of time. What optimization provides the GREATEST battery life improvement?

Current calculation: I_avg = (120×5 + 80×895)/900 = 80.2mA with idle. With deep sleep: I_avg = (120×5 + 0.01×895)/900 = 0.678mA. That’s a 118× reduction! The idle current at 80mA for 99.4% of the time completely dominates the power budget. Deep sleep reduces this by ~8,000×, making it by far the most impactful single change.

Question 2: A 10kΩ pull-up resistor at 3.3V on an unused GPIO pin continuously draws current. What is the current draw, and why is it a problem?

Using Ohm’s law: I = V/R = 3.3V / 10,000Ω = 330µA. For a device targeting 55µA average (5-year battery life), this single pull-up exceeds the entire power budget by 6×! This continuous draw never sleeps, negating all other optimizations. Fix: configure unused pins as outputs or use higher resistance (100kΩ) where timing permits.

1603.8 Summary

Key takeaways from power consumption analysis:

1. Sleep Dominates: For duty-cycled devices, sleep current × time often exceeds active consumption
2. Radio Idle is Expensive: Radio idle mode can consume 1,000-10,000× more than deep sleep
3. Calculate Average Current: Use the duty cycle formula to determine realistic battery life
4. Watch for Hidden Drains: Pull-ups, LEDs, debug interfaces, and regulators all consume power
5. Measure Real Hardware: Datasheet values are best-case; always measure actual consumption

1603.9 What’s Next

Continue to Energy Cost of Common Operations to understand how individual operations contribute to energy consumption.