Optical Heart Rate Sensor Technology: PPG Principles, LED Design, and Performance

An optical heart rate sensor measures the cardiac pulse by shining light into tissue and detecting tiny changes in absorption caused by blood volume fluctuations with each heartbeat. This is photoplethysmography (PPG) — and the design decisions made around LED wavelength, photodetector size, optical geometry, and signal processing determine whether a sensor performs well in the real world or fails the moment you start moving.

How an Optical Heart Rate Sensor Works

Each heartbeat pumps a fresh bolus of blood into peripheral blood vessels. As arteries and arterioles distend during systole and collapse during diastole, the volume of blood in the tissue beneath a PPG sensor oscillates rhythmically. Blood absorbs more light than surrounding tissue, so this oscillating blood volume creates a corresponding oscillation in the amount of light that reaches the photodetector.

The PPG signal has two components:

• AC component: The pulsatile part, synchronized with the heartbeat. Typically 1-5% of total signal amplitude.
• DC component: The steady background absorption from venous blood, tissue water, melanin, and bone. Makes up 95-99% of total signal.

Heart rate extraction involves isolating the AC component frequency — typically 0.5-3.5 Hz (30-210 BPM) — from the DC baseline and superimposed motion artifact.

LED Wavelength Selection: The Core Engineering Trade-off

Green Light (~530 nm): The Consumer Wearable Standard

Most consumer smartwatches and fitness bands use green LEDs for heart rate measurement. The rationale:

1. Hemoglobin absorption is high at 530 nm — both oxyhemoglobin and deoxyhemoglobin absorb green strongly, maximizing the pulsatile signal contrast
2. Green light penetrates 1-2 mm into tissue, reaching superficial capillary beds and small arterioles
3. Green photodetectors and LEDs are cheap, efficient, and compact

The pulsatile signal amplitude with green light at the wrist is typically 2-10x stronger than with red or infrared light, which is why green became the default for heart rate (though not SpO2, which requires two wavelengths).

The Melanin Problem with Green Light

Green light is heavily absorbed by melanin, the pigment in skin. People with darker skin tones see dramatically reduced signal amplitude with green PPG sensors, sometimes 3-5x less signal than lighter-skinned individuals. This is a well-documented source of bias in consumer wearables.

Research by Bent et al. (2020) in NPJ Digital Medicine (DOI: 10.1038/s41746-020-0261-1) found that wrist PPG accuracy for heart rate correlated with Fitzpatrick skin type, with darker skin types showing higher mean absolute errors. The effect was larger during exercise than at rest.

Red (~660 nm) and Near-Infrared (~940 nm): For SpO2

Red and NIR LEDs are required for SpO2 measurement, exploiting the differential absorption of oxygenated versus deoxygenated hemoglobin. These wavelengths penetrate deeper tissue (3-8 mm) and have lower melanin absorption, but produce weaker pulsatile signals at the wrist compared to green.

Devices that combine green (for heart rate) with red/NIR (for SpO2) must manage LED multiplexing: LEDs fire sequentially in microsecond pulses, and the photodetector integrates light from each pulse separately. This time-division multiplexing allows multi-parameter sensing without optical cross-talk.

Multi-Wavelength Approaches

Advanced sensors use 4-8 wavelengths to improve measurement robustness:

• Orange (~590 nm): Intermediate between green and red, better melanin tolerance than green
• Yellow (~570 nm): Strong hemoglobin absorption with lower melanin sensitivity than green
• 850 nm NIR: Better tissue penetration than 940 nm, used in some multi-site devices

Polar's OH1 optical heart rate sensor uses two green wavelengths and four photodetectors arranged around the LEDs, averaging spatial signal variations to reduce motion artifact and site-to-site variability.

Photodetector Design

Detector Area and Sensitivity

Larger photodetectors collect more photons, improving SNR — but increase sensor footprint and current consumption. Wearable sensors balance this by using:

• Silicon photodiodes: 3-20 mm² active area, matched to LED emission spectra
• Ambient light filtering: Optical bandpass filters or electrical filtering to reject sunlight and fluorescent light at non-LED wavelengths
• Dark current minimization: Low reverse-bias to reduce noise floor in low-light conditions

The transimpedance amplifier (TIA) that converts photodiode current to voltage is a critical noise source. Its noise floor sets the minimum detectable pulsatile signal — important for low-perfusion body sites like the wrist.

Multi-Photodetector Arrays

Rather than a single large detector, some sensors use multiple smaller detectors arranged around the LEDs. Averaging signals from multiple detectors:

• Reduces spatial noise from vessel heterogeneity beneath the sensor
• Provides some motion artifact cancellation (if motion shifts the signal on one detector, it partially appears on another with opposite sign)
• Enables gradient-based motion rejection algorithms

The Valencell Benchmark reference sensor (used in Polar Verity Sense validation studies) uses this multi-detector approach, achieving heart rate accuracies below 1 BPM mean absolute error in controlled exercise conditions.

Optical Geometry: LED-Detector Spacing

The physical arrangement of LEDs and photodetectors determines the tissue depth sampled. This is the source-detector separation (SDS), a fundamental parameter in diffuse reflectance PPG.

For wrist reflectance PPG:

• Short SDS (1-3 mm): Samples superficial tissue only (epidermis, papillary dermis). High signal amplitude but sensitive to surface optical changes.
• Long SDS (4-8 mm): Deeper tissue sampling (subcutaneous fat, deeper arterioles). Lower amplitude but more stable signal from larger vessels.
• Multi-SDS designs: Some research sensors use multiple SDS to separate signal contributions from different tissue layers.

Most consumer wristbands use SDS of 2-4 mm, hitting capillary beds and small arterioles in the dermis. Clinical-grade photoplethysmography research systems (e.g., Biopac MP160) allow configurable SDS for controlled tissue depth sampling.

Signal Processing Chain

Ambient Light Cancellation

Even with optical filters, sunlight during outdoor exercise swamps the photodetector. Wearable sensors combat this with:

• LED modulation: Alternating on/off cycles at high frequency (100-1000 Hz), synchronously demodulating the detector signal to extract only light from the LED
• Differential detection: Subtracting a "dark" measurement (detector reading when LED is off) from the LED-on reading
• Optical isolation: Physical design that minimizes direct LED-to-detector paths and ambient light ingress

Motion Artifact Rejection

Motion artifacts are the dominant challenge for wrist optical heart rate sensors during exercise. The accelerometer data captured alongside PPG is used to:

• Estimate dominant motion frequencies (step rate, arm swing)
• Apply spectral subtraction or adaptive filtering to remove motion-frequency components from the PPG spectrum
• Gate heart rate outputs: suppress readings when motion artifact power exceeds a confidence threshold

Algorithms like TROIKA (Zhang et al., 2015, DOI: 10.1109/TBME.2014.2359372), JOSS, and WFPV use the accelerometer as a reference signal for adaptive noise cancellation, achieving mean absolute errors below 2 BPM during running at paces of 6-8 mph.

Performance Benchmarks Across Device Classes

Research-grade optical sensors (Masimo SET, Nonin, Nellcor): MAE <1 BPM in clinical conditions, validated across skin tones and perfusion states, FDA cleared Class II.

Dedicated wearable HR monitors (Polar OH1, Verity Sense, Wahoo TICKR FIT): MAE 1-2 BPM across exercise intensities including high-intensity intervals. Designed and validated specifically for exercise HR tracking. No SpO2.

Consumer fitness smartwatches (Apple Watch, Garmin, Samsung Galaxy Watch): MAE 2-5 BPM across exercise intensities, performance drops at high intensity (>150 BPM) or non-steady-state activities. Include SpO2 and multiple health sensors.

Budget fitness trackers (<$50): MAE 5-15 BPM during exercise, often worse. Lack robust motion rejection algorithms and use cheaper optical components.

Internal Resources

For related technical depth, see our articles on PPG sensor design fundamentals, LED wavelength selection for PPG, motion artifacts and HRV accuracy, and PPG photodetector technologies.

FAQ

What wavelength do optical heart rate sensors use? Most consumer wearables use green light (~530 nm) for heart rate measurement because hemoglobin absorbs green strongly, creating a clear pulsatile signal. Devices that also measure SpO2 add red (~660 nm) and near-infrared (~940 nm) LEDs, using the differential absorption between these wavelengths to calculate oxygen saturation.

Why do optical heart rate sensors fail during exercise? Motion artifacts are the primary cause. When your wrist moves, the optical coupling between the sensor and skin changes, and blood redistributes within tissue, creating signal variations that mimic or overwhelm the cardiac pulse. Additionally, peripheral vasoconstriction during intense exercise reduces pulsatile signal amplitude. Better sensors use accelerometer-based motion rejection algorithms to compensate.

Are arm-based optical sensors more accurate than wrist sensors? Upper arm optical sensors (like Polar OH1 worn on the upper arm or Scosche Rhythm+) often outperform wrist sensors during exercise because the upper arm has better optical coupling and experiences different motion patterns than the wrist. Several validation studies have shown arm-based Polar sensors achieve chest strap-level accuracy (<2 BPM MAE) during running and cycling.

What affects the accuracy of an optical heart rate sensor most? The three biggest accuracy drivers are: (1) motion artifact rejection capability — the quality of the signal processing algorithm, (2) sensor-skin optical coupling — how well the device contacts and stays against the skin, and (3) body site perfusion — how much blood volume is available for sensing. A technically excellent sensor worn loosely on a cold, poorly-perfused wrist will underperform a simpler sensor worn snugly in a well-perfused location.

Why is green light used instead of red for heart rate? Green light produces a stronger pulsatile signal at the wrist compared to red because hemoglobin absorption is higher at 530 nm and the short optical path length limits the background tissue absorption. The pulsatile AC component is typically 2-10x larger with green versus red at the wrist, giving a better signal-to-noise ratio for heart rate extraction. Red light is reserved for SpO2, where its differential behavior versus NIR carries the oxygen saturation information.

How do multi-LED optical heart rate sensors improve accuracy? Multi-LED sensors fire different wavelengths in sequence or pairs, then combine their readings. This allows: (1) cross-validation between channels to identify artifacts, (2) using different wavelengths' complementary noise patterns to improve SNR, and (3) spectral unmixing of hemoglobin components for SpO2. Some sensors also use multiple photodetectors to spatially average signal variations, reducing single-point noise.