How to Design a PPG Sensor: Complete Engineering Guide for Hardware Teams

Engineering guide to PPG sensor design covering LED selection, photodetector specs, optical geometry, AFE circuits, PCB layout, and signal conditioning for wearable devices.

ChatPPG Research Team·

Designing a photoplethysmographic (PPG) sensor that produces clean, reliable signals is a multidisciplinary engineering challenge spanning optics, analog electronics, mechanical design, and signal processing. A poorly designed sensor will generate data that no algorithm can salvage, while a well-designed sensor reduces the burden on downstream signal processing and improves the accuracy of every derived measurement, from heart rate to SpO2 to blood pressure estimation.

This guide covers the complete hardware design process for a PPG sensor, from LED and photodetector selection through optical geometry, analog front-end circuit design, PCB layout, mechanical packaging, and signal conditioning. It is written for hardware engineers and product designers building wearable or clinical PPG devices. For background on the physics of PPG measurement, see our introduction to photoplethysmography.

LED Selection and Optical Source Design

The LED is the active element in a PPG sensor, and its selection directly determines the signal quality, power consumption, measurement capabilities, and skin compatibility of the final device.

Wavelength Selection

Wavelength choice is driven by the target measurement. The three primary wavelengths used in commercial PPG devices are:

Green (520-530 nm): Maximizes the pulsatile signal amplitude (AC/DC ratio) in reflectance mode at the wrist. Hemoglobin absorption at green wavelengths is approximately 5-10 times higher than at infrared wavelengths, creating strong contrast between systolic and diastolic blood volumes. Tamura (2019; DOI: 10.3390/s19153420) showed that green PPG at the wrist achieves AC/DC ratios of 1-5%, compared to 0.1-0.5% for infrared. Green is the primary wavelength for heart rate monitoring.

Red (660 nm): Required for SpO2 measurement. At 660 nm, deoxyhemoglobin absorbs approximately 10 times more than oxyhemoglobin, providing the oxygen-sensitive component of the ratio-of-ratios SpO2 calculation. LED peak wavelength tolerance should be within 5 nm of the target (655-665 nm) because the steep hemoglobin absorption slope at red wavelengths means small wavelength shifts cause measurable SpO2 calibration errors (Tremper, 1989; DOI: 10.1378/chest.95.4.713).

Infrared (940 nm): The oxygen-insensitive reference wavelength for SpO2 (at 940 nm, oxyhemoglobin absorbs slightly more than deoxyhemoglobin, but the difference is smaller than at 660 nm). Infrared also penetrates deeper into tissue (5-10 mm), enabling measurement from thicker body sites. Infrared PPG is less affected by melanin absorption, providing more consistent signal quality across skin tones.

For a detailed analysis of wavelength trade-offs, see our green vs red vs infrared PPG guide.

LED Optical Specifications

Beyond wavelength, several LED parameters matter for PPG sensor performance:

Radiant intensity: Measured in mW/sr, determines how much optical power reaches the tissue and ultimately the photodetector. For wrist reflectance PPG, typical LED drive currents of 5-50 mA produce optical output of 2-30 mW. Higher power improves SNR but increases power consumption and may cause tissue heating. The IEC 62471 photobiological safety standard limits irradiance at the skin surface to prevent thermal or photochemical injury.

Spectral bandwidth (FWHM): Narrower bandwidth LEDs (FWHM of 20-30 nm) provide more wavelength-specific measurement, improving SpO2 accuracy. Broader bandwidth LEDs (FWHM of 40-60 nm) average over a wider spectral range, which can reduce sensitivity to manufacturing wavelength variation but degrades SpO2 precision.

Viewing angle: LEDs with wider viewing angles (120-140 degrees half-angle) distribute light over a larger tissue volume, which can improve signal stability with sensor motion but reduces peak irradiance. Narrower viewing angles (60-90 degrees) concentrate light in a smaller tissue volume, increasing peak irradiance and potentially improving AC signal amplitude for a given drive power.

Forward voltage and efficiency: Green LEDs have higher forward voltage (approximately 3.0-3.5 V) than infrared LEDs (approximately 1.2-1.5 V), which affects power supply design and battery life. Wall-plug efficiency for high-quality green LEDs is typically 10-20%, while infrared LEDs achieve 20-40%.

Multi-LED Configurations

Modern PPG sensors use multiple LED dies arranged around one or more photodetectors. Common configurations include:

  • Dual LED (red + IR): Minimum for SpO2 measurement. Used in fingertip pulse oximeters.
  • Triple LED (green + red + IR): Standard for wrist-worn devices requiring both heart rate and SpO2. Used in Apple Watch Series 6+, Samsung Galaxy Watch 4+.
  • Quad/multi-LED arrays: Research and advanced clinical devices use 4-8 wavelengths for enhanced SpO2 accuracy, carboxyhemoglobin detection, or methemoglobin measurement. Masimo Rainbow SET technology uses 7+ wavelengths.

LED placement symmetry matters. Arranging multiple LEDs symmetrically around the photodetector improves spatial averaging and reduces sensitivity to sensor tilt or lateral displacement on the skin. A common layout places 2-4 green LEDs and 1-2 red/IR LED pairs at 90-degree or 120-degree intervals around a central photodetector.

Photodetector Selection

The photodetector converts optical power to electrical current, and its specifications establish the fundamental noise floor of the PPG measurement.

Photodiode vs Phototransistor

Silicon photodiodes are the standard choice for PPG sensors due to their linear response, low noise, fast response time, and stable temperature behavior. PIN photodiodes offer wider bandwidth (faster response) than standard PN photodiodes and are preferred for high-speed PPG sampling. Avalanche photodiodes (APDs) provide internal gain but add excess noise and are rarely used in commercial PPG sensors.

Phototransistors provide higher responsivity (built-in gain) but suffer from slower response time (microseconds vs. nanoseconds for photodiodes), higher noise, temperature sensitivity, and nonlinear response. They are occasionally used in low-cost, low-performance PPG sensors but are generally avoided in quality designs.

Key Photodetector Specifications

Active area: Larger active areas (1-5 mm^2) collect more light but also collect more ambient light and have higher junction capacitance (which limits bandwidth). Typical PPG photodetectors have active areas of 1-3 mm^2. The optimal area depends on the LED-detector spacing and the desired balance between signal level and ambient rejection.

Spectral responsivity: The photodetector must have adequate responsivity at all LED wavelengths used. Silicon photodiodes have peak responsivity near 800-900 nm and declining responsivity at shorter (green) wavelengths. At 525 nm, typical silicon photodiode responsivity is 0.25-0.35 A/W, while at 940 nm it is 0.55-0.65 A/W. This means green PPG requires more optical power to achieve the same photocurrent as infrared PPG.

Dark current: The leakage current of the photodiode with no light incident sets a noise floor. Quality PPG photodiodes have dark currents of 0.1-10 nA at room temperature. Dark current doubles approximately every 10 degrees Celsius increase in temperature, which matters for body-worn devices where the sensor temperature can reach 30-37 degrees Celsius.

NEP (noise equivalent power): The minimum detectable optical power, typically 10^-14 to 10^-12 W/sqrt(Hz) for silicon photodiodes. Lower NEP enables detection of smaller pulsatile signals, which is critical for measurements through highly attenuating tissue or at low perfusion.

Optical Geometry and Mechanical Design

The physical arrangement of LEDs, photodetector, and optical barriers within the sensor housing is as critical to signal quality as the component selection.

LED-Detector Spacing

In reflectance mode, the spacing between LED and photodetector determines the depth of tissue interrogated and the ratio of pulsatile to non-pulsatile signal. Monte Carlo simulations of photon transport in tissue (Chatterjee & Bhattacharya, 2022; DOI: 10.1038/s41598-022-15240-2) show that:

  • At 2 mm spacing, the median photon penetration depth for green light is approximately 0.5-1.0 mm, primarily sampling the epidermal and superficial dermal layers.
  • At 4-5 mm spacing, penetration depth increases to 1.5-2.5 mm, reaching the deeper dermal vascular plexus where pulsatile flow is stronger.
  • At 8-10 mm spacing, detected light has penetrated 3-5 mm, but total detected power is reduced by 1-2 orders of magnitude compared to 4 mm spacing.

The engineering trade-off is clear: wider spacing improves pulsatile signal quality (higher AC/DC ratio) at the cost of lower total signal power (lower SNR before amplification). Optimal spacing must be determined experimentally for each wavelength, tissue site, and target population. For wrist-worn green PPG, 3-5 mm spacing is typical. For fingertip transmission mode, spacing is irrelevant because the LED and detector are on opposite sides.

Optical Barriers and Light Piping Prevention

Direct optical coupling between LED and photodetector through the sensor housing or the skin surface (light piping) is a major source of signal contamination. Light that reaches the photodetector without passing through pulsatile tissue produces a large DC offset with no AC component, degrading the AC/DC ratio.

Effective optical isolation requires:

  • Opaque barriers between LED and detector cavities, extending from the PCB surface to the skin contact surface. The barrier material should be fully opaque at all LED wavelengths (carbon-filled PEEK, black silicone, or anodized aluminum are common choices).
  • Anti-reflective coatings or matte surface finishes on internal housing surfaces to prevent specular reflections from coupling LED light to the detector.
  • Compliant skin-contact materials (medical-grade silicone, TPE) that conform to the skin surface and minimize air gaps where light could channel between LED and detector windows.

Contact Pressure Management

The contact pressure between the PPG sensor and the skin significantly affects signal quality. Too little pressure results in air gaps and variable optical coupling. Too much pressure occludes blood flow in the underlying tissue, reducing or eliminating the pulsatile signal.

Teng and Zhang (2004; DOI: 10.1088/0967-3334/25/6/016) demonstrated that the PPG AC amplitude from the fingertip peaks at a contact pressure of approximately 60-80 mmHg (equal to mean arterial pressure, where transmural pressure approaches zero). For wrist-worn devices, the contact pressure from a watchband is typically 20-40 mmHg, which is below the optimal range but represents a practical compromise between signal quality and user comfort.

Design strategies for consistent contact pressure include: spring-loaded sensor modules that decouple sensor pressure from band tightness, dome-shaped sensor protrusions that concentrate contact force on a smaller area, and compliant gasket materials that accommodate wrist curvature variation.

Analog Front End (AFE) Design

The analog front end converts the photodetector current to a digitized signal while rejecting ambient light, maintaining dynamic range, and preserving the small pulsatile component.

Integrated AFE Solutions

For most commercial PPG applications, integrated AFE ICs provide the best combination of performance, size, and development speed. Key specifications to evaluate include:

ADC resolution: 16-bit ADCs provide 96 dB of dynamic range, which is marginal for PPG where the AC signal may be only 0.1% of the DC level (60 dB below DC). 20-22 bit ADCs provide adequate headroom. The TI AFE4404 offers 22-bit ADC resolution; the MAX86150 provides 19-bit effective resolution.

Ambient light cancellation (ALC): PPG sensors must operate in environments with varying ambient light (sunlight, fluorescent lighting, infrared remote controls). Integrated AFEs use a sample-and-subtract technique: they sample the photodetector output with LEDs off to measure ambient light, then subtract this baseline from the LED-on measurement. Effective ALC requires fast switching (LED pulse widths of 50-400 microseconds) and synchronized sampling. The MAX86178 achieves ambient light rejection of greater than 100 dB.

LED driver current range and resolution: The AFE must provide programmable LED current to accommodate different tissue types, skin tones, and sensor contact conditions. Typical ranges are 0-50 mA in steps of 0.1-1.0 mA. Automatic gain control (AGC) algorithms adjust LED current in real-time to maintain the photodetector signal within the ADC optimal range.

Power consumption: Critical for battery-powered wearables. Modern PPG AFEs consume 50-500 microamps in active measurement mode, depending on sampling rate and number of active LED channels. Duty cycling (pulsing LEDs briefly and sleeping between samples) reduces average power to 10-100 microamps at typical sampling rates of 25-100 Hz.

Discrete AFE Design

For research applications requiring maximum flexibility or specifications beyond what integrated AFEs offer, a discrete AFE consists of:

Transimpedance amplifier (TIA): Converts photodiode current to voltage. The TIA feedback resistor determines gain (typically 100 kohms to 10 Mohms for PPG) and, together with photodiode capacitance, determines bandwidth. A feedback capacitor stabilizes the loop but limits bandwidth. For PPG with signal bandwidth of 0-25 Hz, TIA bandwidth of 1-10 kHz is adequate and provides significant noise reduction compared to wider bandwidth designs. Low-noise op-amps with input current noise below 1 pA/sqrt(Hz) are preferred (e.g., OPA381, AD8615).

Programmable gain amplifier (PGA): Provides additional gain (10-100x) for the AC component while maintaining the DC level within ADC range. Some designs use a separate AC-coupled path with high gain for pulse detection and a DC path for baseline level measurement.

Anti-aliasing filter: A low-pass filter before the ADC prevents aliasing of high-frequency noise into the signal band. Cutoff frequency should be set to half the ADC sampling rate, typically 50-500 Hz for PPG.

PCB Layout Considerations

PCB layout for PPG sensors requires careful attention to electromagnetic interference, crosstalk, and thermal management.

LED driver traces: High-current pulsed signals from LED drivers can couple into sensitive analog traces. Route LED driver traces on separate layers from photodetector and TIA traces, with ground planes providing shielding. Keep LED driver traces short and wide to minimize inductance and resistive voltage drops.

Photodetector traces: The photodetector output is a high-impedance, low-level signal susceptible to capacitive coupling from digital signals, LED drivers, and power supply noise. Guard ring traces driven at the same potential as the photodetector output can reduce leakage current and capacitive pickup. Keep photodetector traces as short as possible, ideally routing directly from the photodetector pad to the TIA input with no vias.

Ground plane design: Use a solid ground plane directly beneath the analog signal path. Avoid splitting the ground plane under the photodetector and TIA. Separate analog and digital ground domains should connect at a single point near the ADC.

Component placement: Place the photodetector and TIA components as close together as physically possible to minimize trace length and parasitic capacitance. Decoupling capacitors for the TIA power supply should be placed within 2 mm of the power pins.

Signal Conditioning and Digital Processing

After digitization, the raw PPG signal requires preprocessing before feature extraction or algorithm input.

Baseline Wander Removal

The PPG DC baseline drifts due to respiration (0.1-0.5 Hz modulation), sensor motion, vasomotor activity, and temperature changes. A high-pass filter with cutoff frequency of 0.3-0.5 Hz removes baseline wander while preserving the cardiac pulse (fundamental frequency of 0.5 Hz at 30 BPM, the lowest expected resting heart rate). IIR filters (second-order Butterworth) are computationally efficient for real-time implementation. For research applications, zero-phase FIR filtering or polynomial detrending preserves waveform morphology better than IIR filters.

Powerline Interference Rejection

50/60 Hz powerline interference can couple into PPG through ambient light modulation (fluorescent lighting) or electromagnetic pickup in the analog chain. If hardware ambient light cancellation is effective, powerline interference in the digitized signal should be minimal. Residual interference can be removed with a narrow notch filter (Q factor of 30-50) at 50 or 60 Hz, or with adaptive notch filters that track the exact interference frequency.

Motion Artifact Preprocessing

Hardware-level motion artifact mitigation should be complemented by algorithmic approaches. Including a 3-axis accelerometer (sampled synchronously with PPG at 50-100 Hz) enables adaptive filtering, ICA, and deep learning methods for motion artifact removal. See our comprehensive motion artifact removal guide for algorithm selection and implementation details.

Sampling Rate and Resolution Requirements

For basic heart rate monitoring, sampling rates of 25-50 Hz and 16-bit resolution are adequate. For SpO2 measurement, 50-100 Hz sampling is recommended to accurately capture the pulse waveform shape needed for ratio-of-ratios calculation. For pulse wave analysis and cuffless blood pressure estimation, sampling rates of 100-500 Hz and 20+ bit resolution are needed to resolve fine waveform features like the dicrotic notch and second-derivative inflection points (Elgendi, 2012; DOI: 10.1016/j.cmpb.2012.09.005).

Power Management for Wearable Devices

Power consumption is a critical design constraint for battery-powered wearable PPG devices. The PPG sensor subsystem (LEDs + AFE + microcontroller) typically accounts for 30-60% of total device power consumption during active monitoring.

Duty Cycling Strategies

LEDs are the dominant power consumer. Pulsing the LEDs for 50-400 microseconds per sample and sleeping between samples reduces average LED power by 100-1000x compared to continuous illumination. At 25 Hz sampling with 100-microsecond LED pulses, the duty cycle is 0.25%, reducing a 20 mA continuous current to 50 microamps average.

Adaptive Sampling Rate

Reducing the PPG sampling rate during periods of low activity (sleep, sedentary time) saves significant power. Many commercial devices use 25 Hz sampling during rest and increase to 50-100 Hz during detected activity. Activity detection via the accelerometer triggers sampling rate changes with minimal latency.

Multi-LED Power Optimization

When multiple LED wavelengths are used, they need not all operate at the same sampling rate. Green LEDs for heart rate can sample at 25-50 Hz, while red and infrared LEDs for SpO2 can sample at lower rates (1-10 Hz) when SpO2 is not actively being measured, activating full-rate sampling only on demand or when abnormal oxygen saturation is suspected.

Validation and Testing

Before deploying a PPG sensor design, systematic validation is essential.

Optical bench testing: Verify LED output power, spectral characteristics, and photodetector responsivity using calibrated optical equipment. Confirm that the optical path (LED emission pattern, barrier effectiveness, detector field of view) meets design specifications using tissue-simulating optical phantoms with known optical properties.

Signal quality assessment: Measure the AC/DC ratio, SNR, and ambient light rejection ratio on human subjects across a range of skin tones (Fitzpatrick types I-VI), body compositions, and environmental conditions. The AC/DC ratio for green PPG at the wrist should typically be at least 0.5% under resting conditions for adequate algorithm performance.

Reference comparison: Compare PPG-derived heart rate and SpO2 against gold-standard references (ECG for heart rate, arterial blood gas for SpO2) following the protocols described in our clinical validation guide. For heart rate, target a mean absolute error below 3 BPM at rest and below 5 BPM during moderate activity. For SpO2, target root-mean-square error below 3% (Arms) per FDA guidance for pulse oximeters.

Environmental testing: Evaluate sensor performance across the operating temperature range (-10 to 50 degrees Celsius for consumer devices), under various ambient light conditions (0 to 100,000 lux for sunlight exposure), and during water exposure if the device is rated for water resistance.

For comprehensive guidance on designing and executing clinical validation studies for PPG devices, see our clinical validation study design guide and for regulatory considerations, our FDA regulatory pathway guide.

Conclusion

PPG sensor design is a systems engineering problem where optical, electronic, mechanical, and algorithmic elements must be co-optimized. Component selection (LEDs, photodetector, AFE) establishes the fundamental signal quality, optical geometry and mechanical design determine how effectively that signal is captured from tissue, and analog circuit design preserves signal integrity through digitization. No amount of signal processing can compensate for a fundamentally flawed hardware design, which is why getting the sensor right is the essential first step in building any PPG-based medical or wellness device. The design choices outlined here, validated against the clinical and regulatory standards described in our companion guides, form the engineering foundation for reliable, accurate PPG measurement across the full spectrum of applications from consumer wearables to clinical monitors.

Frequently Asked Questions

What LED wavelength should I use for a wrist-worn PPG sensor?
For heart rate monitoring at the wrist, green LEDs (520-530 nm) provide the highest AC signal amplitude and best signal-to-noise ratio in reflectance mode. If you also need SpO2 measurement, you must add red (660 nm) and infrared (940 nm) LEDs. Most commercial wrist-worn devices use at least green plus infrared, with many adding red for three-wavelength operation. Select LEDs with narrow spectral bandwidth (FWHM less than 30 nm) and radiant intensity sufficient to achieve adequate photodetector current after tissue attenuation (typically 5-20 mW optical output per LED).
How far apart should the LED and photodetector be in a reflectance PPG sensor?
For reflectance-mode wrist PPG, the optimal LED-to-photodetector spacing is 3-7 mm. Closer spacing (less than 3 mm) captures too much light from superficial non-pulsatile tissue layers, reducing the AC/DC ratio. Wider spacing (greater than 10 mm) reduces total detected light to the point where shot noise dominates. The optimal distance depends on wavelength, tissue type, and desired penetration depth. Green light sensors typically use 3-5 mm spacing, while infrared sensors may use 5-8 mm to take advantage of deeper penetration. Empirical optimization with tissue phantoms and human subjects is essential.
What analog front end (AFE) IC should I use for PPG?
The most widely used integrated PPG AFEs are the Texas Instruments AFE4404 and AFE4900, Maxim (Analog Devices) MAX86150 and MAX86178, and ams OSRAM AS7058. These ICs integrate LED drivers, transimpedance amplifier, ADC, and ambient light cancellation in a single package. Selection criteria include number of LED channels (2-6), ADC resolution (16-22 bits), sampling rate capability, ambient light rejection ratio, power consumption, and package size. For prototyping, the MAX86150 offers a good balance of features and ease of use. For high-performance multi-wavelength designs, the AFE4900 or MAX86178 provide more flexibility.
How do I minimize motion artifacts in PPG sensor hardware design?
Hardware approaches to reduce motion artifacts include: (1) Minimizing sensor-skin gap with compliant materials and spring-loaded sensor modules that maintain consistent contact pressure; (2) Using optical barriers between LED and photodetector to prevent direct optical coupling and light piping through the housing; (3) Incorporating a 3-axis accelerometer as close to the PPG sensor as possible for algorithmic motion compensation; (4) Using multiple photodetectors at different spacings to enable differential measurement that cancels common-mode motion artifacts; (5) Designing the wristband or housing to maintain consistent sensor-skin coupling during wrist movement. No hardware design eliminates motion artifacts entirely; algorithmic processing remains essential.