How Smartwatch Heart Rate Sensors Actually Work (And Why They Still Miss Beats)

Millions of people strap glowing green lights to their wrists every day, paying hundreds of dollars for premium fitness platforms to track their sleep, strain, and heart rate. We trust these devices to guide our marathon pacing, monitor our recovery, and even detect silent arrhythmias.[5]

But beneath the sleek glass interfaces and marketing claims of clinical-grade accuracy, every smartwatch and fitness band is bound by the exact same physical constraints of optical physics. The technology powering them is called photoplethysmography (PPG).

The mechanism behind PPG is deceptively simple. A wearable device shines a light into the skin and uses a photodiode to measure how much of that light bounces back. With every heartbeat, a surge of blood expands the capillaries in your wrist.

Because blood absorbs light, the photodiode detects a slight drop in the reflected light during a heartbeat, and a slight increase as the blood flows away. By counting these microscopic fluctuations in light intensity, the watch calculates your pulse.

If you flip over an Apple Watch, Garmin, or Whoop, you will almost always see green LEDs. Manufacturers do not use green because it is the only option; they use it because hemoglobin—the oxygen-carrying protein in red blood cells—absorbs green light exceptionally well.

This strong absorption creates a massive contrast between the systolic (pumping) and diastolic (resting) phases of the cardiac cycle. In laboratory settings, green light provides a pulsatile signal magnitude that is significantly stronger than red or infrared light, giving the software a loud, clear rhythm to track.

However, green light has a critical physical limitation: it has a very shallow penetration depth. It barely makes it past the upper layers of the skin before scattering. This shallow optical path introduces the first major hurdle for wearable accuracy: skin tone bias.

Melanin, the pigment responsible for human skin color, is highly effective at absorbing short-wavelength light, including green. In individuals with darker skin tones—specifically Fitzpatrick scale types 4 through 6—the melanin absorbs a significant portion of the green light before it ever reaches the blood vessels.

This creates a degraded signal-to-noise ratio. A comprehensive review of wearable accuracy found that while devices perform adequately across all skin tones at rest, the error rate diverges sharply during exercise. Some smartwatch brands underestimated heart rates by 10 to 15 beats per minute in darker-skinned users during vigorous activity.[3]

The second, and arguably more disruptive, limitation of PPG is motion artifact. When a user runs, cycles, or lifts weights, the wearable physically shifts against the skin. The underlying tissue deforms, the optical path changes, and ambient light leaks into the sensor.

The physics of human movement creates a nightmare for signal processing. The frequency of typical human motion artifacts ranges from 0.01 to 10 Hz. Meanwhile, the frequency of a human heart rate ranges from 0.5 to 5 Hz.[2]

Because the noise of your arm swinging completely overlaps with the signal of your pulse, engineers cannot simply apply a basic low-pass filter to clean the data. If they filter out the motion, they filter out the heartbeat.

To solve this, manufacturers pair the optical sensor with a tri-axis accelerometer. The accelerometer records the exact frequency and intensity of the user's arm movement. Advanced algorithms, such as adaptive spectral filtering, then mathematically subtract the motion frequency from the optical frequency to isolate the true heartbeat.[2]

Despite these algorithmic gymnastics, intense physical activity remains the Achilles heel of wrist-based wearables. Studies show that the absolute error rate of optical heart rate sensors is, on average, 30 percent higher during activity than during rest.[1]

This explains why many devices struggle during activities with erratic, non-rhythmic arm movements—like CrossFit, tennis, or weightlifting—compared to the steady, predictable motion of running. The algorithms are constantly playing catch-up, trying to guess which spikes are heartbeats and which are just the watch bouncing against the wrist bone.

To push past these physical limits, the industry is slowly shifting toward multi-wavelength sensors. By incorporating red and infrared LEDs—which penetrate deeper into the tissue and bypass some of the melanin absorption—devices can cross-reference multiple optical paths to verify the pulse.

As smartwatches transition from casual fitness trackers to FDA-cleared medical devices capable of detecting atrial fibrillation, this baseline accuracy becomes a matter of public health equity. A sensor that drops data during a workout is an annoyance; a sensor that misses an arrhythmia due to skin tone is a liability.

Ultimately, a smartwatch is not an electrocardiogram. It is a highly sophisticated camera taking thousands of pictures of your capillaries, using machine learning to guess your heart rate through layers of moving tissue. As venture capital pivots heavily toward AI-driven medicine, the next leap in wearable accuracy won't come from brighter LEDs, but from better neural networks trained on vastly more diverse datasets.[6]