Low-Power Sensor Node

Learning Goals 5 min

Combine sleep modes (L04-37) and battery selection (L04-38) into the classic IoT field-device pattern: wake, sample, transmit, sleep. The complete blueprint for any battery-powered sensor that needs to last months or years. By the end of this lesson you will:

1. Implement the wake / sample / transmit / sleep cycle on a Pro Mini + sensor + radio.
2. Power down all peripherals between cycles (sensor + radio + chip) for true µA average draw.
3. Estimate runtime from measured current samples and pick the right wake interval for your use case.

Warm-Up 10 min

Hardware: Pro Mini 3.3 V (or any low-power AVR), DHT22 (or similar), nRF24 / RFM69 / LoRa radio, battery + monitor.

The pattern

while forever:
  wake from sleep
  power on the sensor (MOSFET switch)
  read sensor (~20 ms)
  power off sensor
  power on the radio
  publish (~50 ms)
  power off radio
  sleep for the configured interval

Each cycle takes ~100 ms of active time. Sleep takes 15 min – 1 hour depending on application. Active time is 0.005–0.01% of the cycle. Average current drops from ~30 mA to ~50 µA.

New Concept · The wake/sample/transmit cycle 25 min

Switching peripherals via MOSFET

Many sensors draw 1–5 mA even when idle. Cutting their power in sleep saves a lot. Use a P-channel MOSFET on the high side: GPIO HIGH = peripheral off; LOW = on.

Wiring:

• Battery + → P-MOSFET source.
• MOSFET drain → sensor V_CC.
• MOSFET gate → 10 kΩ pull-up to battery + + Arduino GPIO.

The structured loop

#include <LowPower.h>

const int SENSOR_POWER = 5;       // MOSFET gate (LOW = on)
const int RADIO_POWER  = 6;

const int WAKES_PER_SAMPLE = 7;   // 7 × 8 s ≈ 56 s; or 450 × 8 s ≈ 1 hour

int batteryThreshold = 3300;       // mV

void sleepInterval() {
  for (int i = 0; i < WAKES_PER_SAMPLE; i++) {
    LowPower.powerDown(SLEEP_8S, ADC_OFF, BOD_OFF);
  }
}

void wakeAndDoWork() {
  // 1. Sensor on
  digitalWrite(SENSOR_POWER, LOW);
  delay(50);                       // sensor wake-up time
  // 2. Sample
  float temp = readTemperatureSensor();
  // 3. Sensor off
  digitalWrite(SENSOR_POWER, HIGH);

  // 4. Battery check
  int vbat = readBatteryVoltage_mV();
  if (vbat < batteryThreshold) {
    // refuse to transmit if battery is low
    return;
  }

  // 5. Radio on
  digitalWrite(RADIO_POWER, LOW);
  delay(20);                       // radio warm-up
  radio.begin();
  radio.send(makePacket(temp, vbat));
  radio.end();
  digitalWrite(RADIO_POWER, HIGH);
}

void setup() {
  pinMode(SENSOR_POWER, OUTPUT);
  pinMode(RADIO_POWER,  OUTPUT);
  digitalWrite(SENSOR_POWER, HIGH);
  digitalWrite(RADIO_POWER,  HIGH);
}

void loop() {
  wakeAndDoWork();
  sleepInterval();
}

Pre-wake checks

Sometimes you want to do nothing if conditions aren't right. Common pre-wake gates:

• V_BAT too low → skip transmit, only beep low-batt warning.
• Sensor reading hasn't changed significantly → skip transmit (saves more battery, fewer packets to the gateway).
• Time of day (e.g. 9 PM – 7 AM "sleep mode" — only one reading per hour).

Reducing transmit cost

The radio is usually the biggest energy sink. Strategies:

• Higher SF on LoRa = longer per-packet time but longer range. Pick based on coverage.
• Compress payloads. Send a 4-byte struct, not a 30-byte JSON.
• Send only on change. Cache last sent value; transmit when delta exceeds threshold.
• Aggregate. Sample 12 times an hour, send hourly with 12 values.

Worked Example · Build a 2-year temperature logger 25 min

Bill of materials

• Pro Mini 3.3 V (~3 mA active, ~3 µA deep sleep).
• DHT22 (~1 mA active, draws ~0 µA off).
• RFM69CW LoRa-like radio (~30 mA tx peak, ~0.1 µA off).
• P-MOSFET (SI2333DDS or similar).
• 3 × AA NiMH (3.6 V, 2500 mAh).

Math

• Active per cycle: 100 ms × ~30 mA ≈ 0.8 µAh.
• Sleep per cycle: ~3 µA × 15 minutes = 0.75 µAh.
• Per cycle total: 1.5 µAh.
• Cycles per year: 365 × 24 × 4 = 35040.
• Per year: 35040 × 1.5 µAh = 53 mAh.

2500 mAh / 53 mAh per year = 47 years. Battery shelf life will limit you to ~10 years long before capacity is consumed.

Real-world results

Reality is messier: leak currents, parasitic loads, charge inefficiency on the boost regulator. Field deployments of this pattern typically last 2–5 years on AA cells. That's still phenomenal — battery becomes the device's effective lifetime.

Verify in practice

1. Measure µA in sleep.
2. Measure mA during active cycle + duration.
3. Compute weighted average.
4. Divide battery mAh by average → years of life.

Try It Yourself 15 min

Mini-Challenge · Estimate one of your project's runtime 10 min

1. Pick a battery project.
2. Measure active mA + duration per cycle.
3. Measure sleep mA + duration per cycle.
4. Compute weighted average.
5. Divide battery mAh by average → years.
6. Compare to the naive "always running" estimate.

Recap 5 min

Low-power sensor node = wake-sample-transmit-sleep with peripherals MOSFET-switched. Average current in µA, runtime in years. The pattern of every commercial battery-powered IoT device. Tomorrow we add solar and the device runs forever.

Wake/sample/transmit/sleep cycle

The four phases of a low-power node. Active 0.01–1% of the time; sleep the rest.

Peripheral power gating

Using a MOSFET to cut power to sensors / radios when not in use.

P-channel MOSFET (high-side switch)

Switches V_CC to a peripheral. Gate held HIGH = off; LOW = on. Cheap, low loss.

Sensor wake-up time

How long after power-on before a sensor gives valid readings. DHT22 ~50 ms; BMP280 ~10 ms.

Send-on-change

Transmit only when the value differs significantly from the last sent value. Reduces radio time.

Field-device pattern

The industry-standard low-power node template. Sleep most of the time; do work briefly; sleep again.

Homework 5 min

1. Sketch out the cycle for one of your projects on paper.
2. Estimate runtime with naive (always on) vs proper sleep cycle.
3. Read ahead to ARD-L04-40 (Solar-Powered Field Sensor).