How Do Smartwatches Measure Heart Rate? Reading Blood Flow
Most smartwatches shine light into your wrist from LEDs on the back. A light sensor then measures how much light bounces back. Each heartbeat adds a little blood under the skin, and that blood absorbs more light. From how often this light rises and falls, the watch works out your heart rate per minute. The heart rate it shows is a guide for looking after your health. On its own, it cannot tell you whether you have a disease. Even features approved as medical devices, such as the ECG, do not replace a doctor's diagnosis.
What you'll learn
- How light can read changes in blood flow, and why green light is used
- How a wave-like signal becomes "beats per minute," and how arm movement is removed
- Why readings can be off (fit, movement, cold, tattoos)
- How the ECG and blood oxygen features differ from measuring heart rate
- The chips inside, and how they run on a tiny battery
Prerequisites: None. Reading "How Do LEDs Work?" and "How Do Image Sensors Work?" first makes the parts about light and electricity easier to follow.
In everyday life
Take off a smartwatch and turn it over, and you may see a green light flickering. The heart rate on the screen climbs while you run. Your heart rate while you sleep shows up as a graph. Some models can also record an ECG, or measure the share of oxygen in your blood. All of this comes from a small sensor touching your wrist and the semiconductor chips that run it.
How it works 1: using light to see changes in blood
Each time the heart contracts, it pushes blood out to the whole body. With each push, blood also reaches the tiny blood vessels in the skin of your wrist. The amount of blood there rises slightly. Apple explains that blood looks red because it reflects red light and absorbs green light. So the more blood there is at a given moment, the more green light the skin absorbs.
On the back of the watch are an LED that gives off light, and a part that turns the light it receives into electricity. This part is called a photodiode. The LED's light scatters inside the skin. Some of it is absorbed by blood, and the rest bounces back to the photodiode. The light that arrives drops a little with each heartbeat, and recovers before the next one.
This way of reading the rise and fall of light is called photoplethysmography, or PPG. Most of the light that arrives is nearly steady light bouncing off skin, bone and other tissue. Allen (2007), a widely read paper explaining PPG, also splits the signal in two. One is a small part that changes with each beat. The other is a large part that changes only slowly. The key is to pick out a small change inside a lot of light, accurately.
Why green light?
Green light is strongly absorbed by the red pigment in blood, so the change from each heartbeat shows up clearly. This pigment is called hemoglobin. Tamura and colleagues reviewed past research in a 2014 paper. According to it, bright green LEDs becoming practical is what made wearable heart rate monitors widespread.
Research also suggests that green light is less affected by movement. A Japanese research team (Maeda, Sekine and Tamura, 2011) ran tests with jumping and arm swinging. The pulse measured with green light matched the ECG better than the pulse measured with infrared light. Still, invisible infrared light is also used. According to Apple, which light is used depends on the model, and on things like whether you are exercising or sleeping.
How it works 2: chips that turn light into numbers
A photodiode passes more current when more light reaches it ("What Is a Diode?" and "How Do Image Sensors Work?" explain how). But this current is very small, and it can't be used for math as it is. So it goes through a circuit that turns the small current into a voltage that is easier to handle, and makes it bigger. This circuit is called a transimpedance amplifier.
The signal, now a voltage, is turned into numbers tens to hundreds of times a second. The circuit that does this is called an AD converter, or ADC ("What Is an Analog Chip?" has more). Sunlight and room light also sneak in through gaps at the wrist. So the light is also measured at moments when the LED is off. That value is subtracted from the value when the LED is on.
Some chips combine everything from the circuit that drives the LED to the circuit that makes numbers. Such a chip is called an analog front end (AFE). It sits right behind the sensor and handles the first stage of processing. One example is the AFE4404 from Texas Instruments (TI). Its documents describe setting aside time to measure only the surrounding light, and subtracting that amount.
How it works 3: from a wave to "beats per minute"
The stream of numbers also contains wobbles other than the heartbeat. Breathing, and changes in how hard the watch presses on the skin, make slow, large wobbles. Fine noise from the circuit and the light makes fast, small wobbles. To keep only the heartbeat's rhythm, math is used to remove the unwanted wobbles.
The simplest way is to average the last few values. Fast noise cancels out when averaged, and a smooth wave remains. This calculation is called a moving-average filter. Real products combine more refined calculations, and Garmin calls its own method proprietary.
From the cleaned-up wave, the watch measures the time from one trough to the next. If the gap is 1 second, that is 60 beats in a minute (60 seconds). If the gap is 0.75 seconds, divide 60 by 0.75 to get 80 beats per minute. To keep things simple, this example uses just one beat. In practice, the watch averages several beats before showing a number.
How it works 4: telling arm movement apart
When you run or swing your arms, the watch shifts slightly on your skin, and it presses with a different force. That changes how much light bounces back, and makes a wave that looks a lot like a heartbeat. This is the biggest headache for heart rate monitors that use light. Garmin, too, lists changes in sweat, vibration, temperature, posture and distance from the skin as things that disturb the signal.
So the watch uses its built-in accelerometer. This small part measures the rhythm of your arm's movement ("What Is MEMS?" shows how it works). Put simply, if the light wave has the same rhythm as your arm swing, the watch treats that rhythm as movement and removes it. It then follows the remaining rhythm as your heartbeat. Measuring the noise separately and subtracting it is also the idea behind "How Noise Cancelling Works."
When the rhythm of the arm swing is close to the heart rate, telling them apart gets hard. One study tracked the heart rate of 12 people running at up to 15 km/h (Zhang et al., 2015). It used a method that also uses acceleration. The average error was about 2.3 beats per minute. But this is a result under set study conditions, and it does not show how accurate any product is.
Why readings can be off
Putting the makers' official explanations together, there are four main causes.
- Fit: If the band is loose, light leaks in through gaps, or the watch shifts each time you move. Fitbit recommends wearing it two to three finger widths above the wrist bone during exercise, not too tight.
- Movement: According to Apple, regular motions like running or cycling give better results than tennis or boxing.
- Cold: In the cold, less blood flows through the skin, so the change is smaller. In cold places, Garmin recommends warming up indoors before you start.
- Tattoos: According to Apple, the ink, the pattern and how dark the tattoo is can block the sensor's light.
The effect of skin tone is still being debated. A 2020 study by Duke University in the US and others tested six devices on 53 people. Differences in error by skin tone were not big enough to rule out chance. In other words, they were not "statistically significant." On the other hand, error during activity was on average about 30% higher than at rest. There were also differences between devices.
A published response urged caution about this result. It pointed to the small number of people, and to problems with the scale that sorts skin tone by eye. The original researchers replied that one study cannot prove there is "no difference," and that more research is needed. In a 2022 document, Apple writes that skin pigment strongly absorbs the light the watch uses. So, Apple says, it adjusts the LED current and the amplification on the receiving side.
According to Apple, when the signal is weak, the watch makes the LEDs brighter and measures more often. If wrist readings are not steady, Apple also suggests a chest-strap heart rate monitor connected by Bluetooth. Apple, Fitbit and Garmin all list fit first among the ways to get accurate readings.
ECG: measuring electricity, not light
Heart muscle contracts in response to electrical signals. These signals reach the surface of the body as tiny voltages. A record of them is called an electrocardiogram (ECG). PPG looks at the blood that increases as a result of each heartbeat. An ECG looks at the electricity that causes the heartbeat itself.
On an Apple Watch, the back and the dial on the side (the Digital Crown) have metal parts that pick up electricity, called electrodes. Touch the dial with a finger of your other hand, and the watch can measure the voltage between your left and right arms, across your heart. It records like this for 30 seconds. Apple advises resting your arm on a table or your lap and keeping it still during the recording.
The voltage the electrodes pick up is very small, and electrical noise from the surroundings rides on top of it. So ECG circuits generally amplify only the difference between the voltages of the two electrodes. Noise that appears equally on both cancels out. This circuit is called a differential amplifier.
Hospitals use ECGs that look at the heart's electricity from 12 directions. The watch's recording is close to one of these directions, called Lead I. In 2018, the US Food and Drug Administration (FDA) authorized this app as ECG software that can be used without a doctor's prescription. In Japan, the Minister of Health, Labour and Welfare approved it in 2020 as a "home electrocardiograph program." It became available in Japan in January 2021.
Even so, the watch's ECG is not a diagnosis in itself. Apple says that this app cannot detect a heart attack, blood clots or a stroke. Apple also tells users to contact emergency services right away if they think they may be having a heart attack, for example because of chest pain. In Japan, the Ministry of Health, Labour and Welfare also sent a notice to the prefectures on January 27, 2021. Its title is "On the proper use" of these programs.
Blood oxygen (SpO2): comparing red and infrared
Hemoglobin carrying oxygen and hemoglobin without oxygen absorb red light and infrared light differently. So the watch shines both kinds of light and compares how much each one changes with the pulse. From this, it can estimate the share of hemoglobin that carries oxygen. This share is called SpO2.
Apple Watch models that can measure blood oxygen also have red and infrared LEDs besides the green ones, and they measure this way. Takuo Aoyagi of Nihon Kohden, a Japanese company, found the principle of using the pulse to measure the share of oxygen. He presented it at an academic meeting in 1974. Pulse oximeters that clip onto a fingertip use the same method.
However, Apple says its blood oxygen feature is not for medical use, and is for fitness and wellness. The FDA also explains that oxygen meters sold for general wellness are not reviewed by the FDA. The FDA lists skin color, poor circulation, skin temperature and nail polish, among others, as things that can affect readings.
The chips inside, and saving battery power
| Part | Job | Learn more |
|---|---|---|
| LEDs (green, red, infrared) | Shine light into the skin | LEDs |
| Photodiode | Turns light that bounces back into current | Diodes |
| Analog front end | Drives the LEDs and turns current into numbers | Analog chips |
| Accelerometer | Measures arm movement | MEMS |
| Microcontroller | Calculates heart rate from the wave | Microcontrollers |
| Bluetooth chip | Sends results to your phone | Bluetooth |
| Lithium-ion battery | Powers everything | Batteries |
A watch battery is small, so every part is designed to use as little electricity as possible. The LEDs do not stay on. They flash for very short moments and then go off. In the MAX86141, an AFE from Analog Devices, you can choose how long the LED stays on. A longer flash lets more light arrive, which makes measuring easier. A shorter one saves electricity. TI's documents for the AFE4404 put the average current while picking up the PPG signal at under 200 microamperes (0.2 milliamperes).
The microcontroller that does the math stops its circuits and sleeps while it has no work. This state is called sleep mode. To send data to your phone, the watch uses a power-saving form of Bluetooth that sends in short bursts and rests for long stretches. This kind of Bluetooth also has a set of rules for sending heart rate monitor readings.
Common myths
Myth: The green light takes a picture of your blood vessels
In fact, a photodiode only measures the amount of light that reaches it. A camera's image sensor captures how light is spread out. It uses millions to tens of millions of tiny light-sensing cells, called pixels. A heart rate sensor has just a few light-receiving parts, and it watches changes in overall brightness. Saying the watch "looks" at blood with light makes it easy to picture an image.
Myth: A watch with an ECG feature can tell you if you have heart disease
In fact, the watch's ECG is a recording from just one direction. The use approved in Japan, too, is to classify the recorded waveform and let you know the result. It tells whether the shape shows a normal rhythm, or suggests a kind of irregular heartbeat called atrial fibrillation. Diagnosis is made by a doctor. Because it shares the name "ECG" with the hospital test, people tend to think it can do the same things.
Myth: With a light sensor, a watch can measure blood sugar without a needle
In fact, in February 2024 the FDA warned people not to use such products to measure blood sugar (glucose). The FDA has not authorized a single smartwatch or smart ring that claims to measure blood sugar on its own, without piercing the skin. A different case is an app that shows readings on the watch from a glucose meter that pierces the skin. People tend to expect that if light can reveal blood flow, it can also reveal what is in the blood.
By the numbers
| Item | Number | Note |
|---|---|---|
| Heart rate range it can measure | 30–210 beats per minute | Apple Watch (Apple's support page) |
| ECG recording | 30 seconds | How long your finger touches the dial |
| Average current of the AFE chip (while measuring PPG) | Under 200 µA | Example: TI's AFE4404 |
| Example of error in a study | About 2.3 beats per minute | 12 people running (Zhang et al., 2015) |
At 70 beats per minute, a day comes to 70 × 60 × 24, or about 100,000 beats. For every one of them, the watch keeps picking up a tiny change in light, right there on your wrist.
Learn more
- See the differences in a table: FIR vs IIR filters (two kinds of math for cleaning up a wave) / Wi-Fi vs Bluetooth vs LPWA
- Try the numbers: ADC SNR and ENOB calculator (how finely a signal is turned into numbers)
- Terms for your next step: op-amp (the basic circuit behind turning current into voltage) / dark current (current that flows even without light) / FIR filter
- Related explainers: "What Is an Op-Amp?" / "What Chips Are Inside a Smartphone?"