Sleep Modes

Learning Goals 5 min

A Nano running at full pace draws ~30 mA — a 2000 mAh battery lasts 66 hours. The SAME Nano in deep sleep draws ~5 µA — same battery lasts 50 years. The trick is to sleep most of the time and wake up only when needed. By the end of this lesson you will:

1. Compare the 3 power states: active, idle / standby, deep sleep / power-down.
2. Wake the chip from sleep using the watchdog timer (periodic) and external interrupts (event-driven).
3. Measure your sketch's actual current with a multimeter (µA range) and pick the lowest viable power mode.

Warm-Up 10 min

Hardware:

• Arduino board (Nano / Pro Mini works best — minimal extra circuitry; UNO has USB chip burning extra mA).
• Multimeter with µA range.
• LED + 220 Ω + battery + power switch.

Why sleep matters

The same chip, the same code, vastly different battery lives. Cluster G is about choosing the right sleep mode and waking only when needed.

New Concept · AVR sleep + wake sources 25 min

The 5 sleep modes (AVR / ATmega328)

Power-down is the workhorse for battery projects. Wake on a button press (INT0/INT1) or a periodic watchdog tick.

Library: LowPower by Rocket Scream

#include <LowPower.h>

void setup() {
  Serial.begin(9600);
  pinMode(LED_BUILTIN, OUTPUT);
}

void loop() {
  // Wake up briefly, do work
  digitalWrite(LED_BUILTIN, HIGH);
  delay(50);
  digitalWrite(LED_BUILTIN, LOW);

  // Sleep 8 seconds (watchdog wake)
  LowPower.powerDown(SLEEP_8S, ADC_OFF, BOD_OFF);
}

Every 8 seconds: chip wakes, blinks LED for 50 ms, sleeps again. Average current: ~50 mA × 50 ms + 5 µA × 8 s ≈ 0.3 mA average. Vs continuously running (~30 mA), that's 100× longer battery.

Sleep durations: SLEEP_15Ms, 30Ms, 60Ms, ..., 1S, 2S, 4S, 8S. For longer, call powerDown in a loop.

Wake on external interrupt

const int WAKE_PIN = 2;

void onWake() { /* dummy; ISR must exist */ }

void setup() {
  pinMode(WAKE_PIN, INPUT_PULLUP);
}

void loop() {
  attachInterrupt(digitalPinToInterrupt(WAKE_PIN), onWake, LOW);
  LowPower.powerDown(SLEEP_FOREVER, ADC_OFF, BOD_OFF);
  detachInterrupt(digitalPinToInterrupt(WAKE_PIN));

  // After waking — do work
  digitalWrite(LED_BUILTIN, HIGH);
  delay(200);
  digitalWrite(LED_BUILTIN, LOW);
}

Button press → INT0 fires → chip wakes from SLEEP_FOREVER → LED blinks → goes back to sleep. Average current: essentially zero between presses. A doorbell on a coin cell that lasts 5 years.

BOD (Brown-Out Detector)

The brown-out detector wakes the chip if V_CC dips below a threshold. Useful for reliability — without ~25 µA. For maximum battery life, disable with BOD_OFF as shown.

What you can't do in deep sleep

Most peripherals power down. You can't do delay() (Timer 0 is off), Serial.print() (USART off until you call Serial.begin again), millis() (counter frozen). Wake up first, do work, sleep again.

Worked Example · Measure your actual current 25 min

Setup

1. Pro Mini + battery (AAA pack, 3.3 V version).
2. Multimeter set to µA range.
3. Probe in series with the battery + lead.

Baseline

Upload an empty sketch (only setup + loop). Read current. Pro Mini at 3.3 V: typically ~5 mA. UNO: 30+ mA. Note this is the "active doing nothing" baseline.

Add LowPower

#include <LowPower.h>
void setup() {}
void loop() {
  LowPower.powerDown(SLEEP_8S, ADC_OFF, BOD_OFF);
}

Upload, switch to µA. Should drop to ~5–10 µA on a Pro Mini. ~3000× lower than running flat out.

Add a wake-on-press

The §3 second sketch. Press the button → meter spikes to ~5 mA briefly → returns to µA after.

Profile your real project

For any battery build, measure current in two cases:

1. Active doing work (e.g. reading IMU + publishing): how long does it last?
2. Asleep waiting: how long is the chip asleep per minute?

Average current = (active mA × active duration + sleep mA × sleep duration) / total duration. Compare to battery mAh to estimate runtime.

Try It Yourself 15 min

Mini-Challenge · Compute your build's runtime 10 min

For one battery project (plant monitor, doorbell, etc.):

1. Measure active mA + active duration per cycle.
2. Measure sleep mA + sleep duration per cycle.
3. Compute weighted average.
4. Divide battery mAh by average → runtime in hours.
5. Compare to your previous estimate without sleep.

You'll usually find 10×–100× improvements with proper sleep.

Recap 5 min

Sleep modes trade responsiveness for power. Power-down at ~5 µA is the standard for battery projects. Wake via watchdog (periodic) or external interrupt (event). The LowPower library hides the AVR register tweaking. Tomorrow: pick the right battery for the project.

Sleep mode

A reduced-power state where some peripherals are turned off. AVR has 5 levels.

Power-down

The deepest sleep — only the watchdog and external interrupt logic stay alive. ~5 µA on AVR.

Watchdog timer

An independent oscillator + counter. Can be configured to wake the chip after up to 8 s. ±10% accurate.

External interrupt

A wake source triggered by a pin change. INT0 (D2) and INT1 (D3) on UNO. Pin-change interrupts on more pins.

Brown-out detector (BOD)

Voltage-sag reset circuit. Useful for reliability; consumes ~25 µA. Disable for maximum battery life.

LowPower library

Rocket Scream's wrapper for AVR sleep. One-line API: LowPower.powerDown(SLEEP_8S, ADC_OFF, BOD_OFF);

Active vs sleep duty cycle

The fraction of time the chip is active. 1% active + 99% sleep = ~99× battery savings.

Homework 5 min

1. Measure your Pro Mini / Nano in deep sleep with a µA meter.
2. Build the wake-on-button doorbell. Confirm µA between presses.
3. Read ahead to ARD-L04-38 (Battery Selection).