Choosing Optimal LED Wavelengths for PPG Sensing: A Design Engineer's Guide

The choice of LED wavelength is the single most consequential hardware decision in PPG sensor design, directly determining which physiological parameters can be measured, the achievable signal-to-noise ratio, the sensitivity to skin pigmentation, and the power consumption of the optical front-end. While consumer wearable users see only a green or red light flashing on their wrist, the underlying wavelength selection is grounded in the complex interaction between photon energy, hemoglobin absorption spectra, tissue scattering coefficients, melanin absorption, water absorption, and photodetector quantum efficiency.

This guide provides a rigorous treatment of wavelength selection for PPG sensing, covering the fundamental spectroscopy, practical engineering tradeoffs, and emerging multi-wavelength approaches. For a comparative overview of the three primary PPG wavelengths, see our green vs red vs infrared PPG guide.

Hemoglobin Absorption Spectroscopy Fundamentals

The entire foundation of PPG wavelength selection rests on the absorption spectra of the two primary hemoglobin species in blood: oxygenated hemoglobin (HbO2) and deoxygenated hemoglobin (Hb, also called reduced hemoglobin).

The Absorption Spectra in Detail

The molar extinction coefficients of HbO2 and Hb vary by orders of magnitude across the wavelength range used in PPG (400-1100 nm). Several spectral regions are particularly important for sensor design.

In the Soret band (400-450 nm), both hemoglobin species exhibit their highest absorption, with molar extinction coefficients exceeding 100,000 L/(mol*cm). While the absorption contrast is high, extreme tissue scattering at these short wavelengths limits penetration depth to less than 0.5 mm, and typical LED and photodetector performance is poor in this region.

In the green region (500-575 nm), HbO2 has two absorption peaks at approximately 542 nm and 577 nm, while Hb has a single broad peak near 555 nm. The extinction coefficients are 10,000-40,000 L/(mol*cm), providing strong absorption with sufficient tissue penetration (1-2 mm) for reflectance-mode PPG. Critically, the absorption difference between HbO2 and Hb in this region is relatively small, meaning the green PPG signal is largely independent of oxygen saturation and reflects total blood volume changes. This is why green light excels at heart rate measurement but cannot provide SpO2 information.

In the red region (600-700 nm), a dramatic separation between HbO2 and Hb absorption occurs. At 660 nm, the extinction coefficient of Hb is approximately 3,300 L/(molcm) while HbO2 is approximately 300 L/(molcm), a ratio exceeding 10:1. This large differential is the foundation of pulse oximetry. Tissue penetration at red wavelengths is 3-5 mm, enabling both reflectance and transmission mode operation.

In the near-infrared region (700-1000 nm), the relationship between HbO2 and Hb absorption inverts. At 940 nm, HbO2 has an extinction coefficient of approximately 1,200 L/(molcm) while Hb is approximately 700 L/(molcm). The absolute absorption values are lower than at shorter wavelengths, and tissue penetration is deepest (5-10 mm), enabling transmission through thick tissue sites. The isosbestic point at approximately 805 nm is where HbO2 and Hb absorption coefficients are exactly equal, making the PPG signal at this wavelength independent of oxygen saturation. Prahl (1998) compiled the comprehensive reference dataset for hemoglobin extinction coefficients that remains the standard reference for PPG system design (doi: 10.1117/12.154627, related compilation).

Beyond HbO2 and Hb: Dyshemoglobins

Clinical multi-wavelength pulse oximeters (co-oximeters) must also account for carboxyhemoglobin (COHb, from carbon monoxide exposure) and methemoglobin (MetHb, an oxidized form). COHb absorbs similarly to HbO2 at 660 nm, causing conventional two-wavelength oximeters to overestimate SpO2 in carbon monoxide poisoning. Detecting COHb requires additional wavelengths, typically near 630 nm where COHb has a distinct absorption peak.

MetHb absorbs strongly across the visible spectrum with a characteristic peak near 630 nm. At 660 nm and 940 nm, MetHb absorption is similar, causing the R-ratio to converge toward 1.0 (corresponding to approximately 85% SpO2) regardless of true oxygen saturation. Multi-wavelength pulse oximeters like the Masimo Rad-57 use 7-8 wavelengths to resolve all four hemoglobin species simultaneously. Barker and Shah (2017) provided a comprehensive review of multi-wavelength pulse oximetry approaches for dyshemoglobin detection (doi: 10.1213/ANE.0000000000002056).

Tissue Optical Properties and Wavelength Dependence

LED wavelength selection must account for the optical properties of all tissue layers between the LED and photodetector, not just blood absorption.

Scattering Coefficients

Tissue scattering generally decreases with increasing wavelength, following an approximate power law: mu_s' ~ wavelength^(-b), where b is typically 0.5-2.0 depending on the tissue type. At 525 nm (green), the reduced scattering coefficient of skin is approximately 25-35 cm^(-1). At 660 nm (red), it decreases to 15-25 cm^(-1). At 940 nm (infrared), it falls further to 8-15 cm^(-1).

Higher scattering at shorter wavelengths has two consequences. It limits penetration depth (beneficial for concentrating the measurement in superficial capillaries) and it increases the probability of photon detection in reflectance mode (because more photons are scattered back toward the surface detector). Jacques (2013) published the definitive review of tissue optical properties across wavelengths, providing the quantitative data essential for PPG sensor modeling (doi: 10.1088/0031-9155/58/11/R37).

Melanin Absorption

Melanin in the epidermis absorbs light with a broad, monotonically decreasing spectrum from UV through near-infrared. The absorption coefficient of melanin is approximately proportional to wavelength^(-3.33), meaning melanin absorption at 525 nm is roughly 4-5 times higher than at 940 nm. This wavelength-dependent melanin absorption has significant implications for PPG performance across skin tones.

For individuals with high melanin content (Fitzpatrick skin types V-VI), green PPG signal amplitude can be reduced by 50-70% compared to lightly pigmented skin (Fitzpatrick type I-II). Red PPG is reduced by 30-50%, and infrared PPG by 10-25%. Fallow et al. (2013) quantified these differences across skin types and found that infrared wavelengths produced the most consistent PPG signal amplitude across the Fitzpatrick scale, with a coefficient of variation of signal amplitude across skin types of 15% for infrared versus 45% for green.

This has important equity implications for PPG-based health monitoring. Devices relying solely on green PPG may underperform for users with darker skin tones, particularly under challenging conditions (motion, poor coupling). Infrared wavelengths provide more equitable performance but have lower baseline signal amplitude at the wrist. A multi-wavelength approach that adapts wavelength weighting based on detected signal quality offers the best path toward equitable PPG measurement.

Water Absorption

Water absorption is negligible below 900 nm but increases substantially above 970 nm, with a strong absorption peak near 1450 nm. At 940 nm (the standard infrared PPG wavelength), water absorption is beginning to increase but remains manageable. At wavelengths above 1000 nm, water absorption becomes a dominant factor limiting tissue penetration and signal quality. This is one reason why the near-infrared window for PPG is effectively limited to 700-1000 nm using conventional photodetector technologies. For short-wave infrared (SWIR) applications exploring glucose detection, water absorption must be carefully modeled and compensated.

Wavelength Selection for Specific Applications

Heart Rate Monitoring

For heart rate measurement in reflectance mode at the wrist, green (520-530 nm) is the optimal choice. The combination of high hemoglobin absorption, shallow penetration concentrating the measurement in the superficial capillary plexus, and high back-scattering probability produces the maximum AC/DC modulation ratio.

Maeda et al. (2011) compared PPG signal quality at 525 nm, 590 nm, 660 nm, and 940 nm at the wrist in 20 subjects and found that 525 nm consistently produced the highest SNR for heart rate detection, with an average perfusion index (AC/DC ratio) of 1.8% versus 0.6% for red and 0.3% for infrared. The absolute advantage of green depends on the body site: at the fingertip (transmission mode), infrared produces excellent signals and is often preferred due to lower power consumption.

For continuous 24/7 heart rate monitoring where power consumption is critical, many wearable devices use a hybrid approach: infrared for resting heart rate (adequate signal quality at lower LED power) and green for active heart rate during exercise (higher signal quality to compete with motion artifacts). For detailed analysis of how wavelength choice impacts heart rate monitoring performance, see our PPG technology overview.

Pulse Oximetry (SpO2)

SpO2 measurement requires exactly two wavelengths with contrasting HbO2/Hb absorption ratios. The standard pair is red (660 nm) and infrared (940 nm).

The wavelength tolerance for clinical-grade SpO2 is tighter than many engineers realize. The SpO2 calibration curve (mapping R-ratio to SpO2) is wavelength-specific. A shift of 10 nm in the red LED peak wavelength can introduce a systematic SpO2 error of 1-2%. For this reason, clinical pulse oximeters use tightly wavelength-binned LEDs (typically within 2-3 nm of the target wavelength) and may include individual device calibration. Consumer wearables use broader LED bins (5-10 nm tolerance), accepting slightly lower accuracy.

The 660 nm red wavelength sits on a steep portion of the Hb extinction curve, which is both beneficial (high sensitivity to saturation changes) and problematic (high sensitivity to wavelength errors). Some researchers have proposed alternative wavelength pairs. Kim and Liu (2015) evaluated red wavelengths from 630-680 nm paired with infrared wavelengths from 880-960 nm and found that 660/940 nm remained optimal for the 70-100% SpO2 clinical range, but 630/940 nm provided superior sensitivity for detecting low saturations below 80%.

Blood Pressure Estimation

Emerging PPG-based cuffless blood pressure estimation uses pulse wave analysis features extracted from the PPG waveform morphology. The optimal wavelength for blood pressure estimation is debated, but green light is generally preferred for wrist-based measurement because the higher SNR and sharper waveform morphology (due to shallow penetration into a well-defined vascular bed) provide more reliable extraction of features like pulse transit time, systolic rise time, dicrotic notch position, and waveform area ratios.

Infrared wavelengths, with their deeper penetration, sample a more diffuse vascular bed that smooths out the fine morphological features most relevant to blood pressure estimation. Elgendi et al. (2019) analyzed PPG waveform features at multiple wavelengths and found that green PPG at the wrist provided the most reproducible and informative feature set for blood pressure regression models (doi: 10.3390/s19143271).

Multi-Spectral Analysis and Advanced Biomarkers

Beyond standard heart rate and SpO2, researchers are exploring multi-wavelength PPG for expanded physiological measurement. These applications require careful wavelength selection based on the absorption spectra of the target analyte.

Total hemoglobin concentration: Wavelengths at or near isosbestic points (505 nm, 522 nm, 548 nm, 570 nm, 805 nm) provide hemoglobin-sensitive but saturation-independent measurement. Combining isosbestic and non-isosbestic wavelengths enables simultaneous measurement of total hemoglobin and oxygen saturation.

Bilirubin: Bilirubin has a strong absorption peak near 460 nm (blue). Transcutaneous bilirubin measurement for neonatal jaundice screening uses blue and green wavelengths, though this is typically not a PPG (pulsatile) measurement but rather a static reflectance measurement.

Glucose: Non-invasive glucose estimation via PPG remains one of the most pursued and most challenging goals. Glucose has weak absorption features in the near-infrared (NIR) around 1550-1650 nm and in the mid-infrared. SWIR LED/detector pairs at these wavelengths are being explored, but the glucose absorption signal is orders of magnitude smaller than hemoglobin absorption, making reliable detection extremely difficult. Caduff et al. (2009) reviewed multi-sensor approaches to non-invasive glucose monitoring and concluded that no single optical wavelength provides sufficient specificity for glucose detection (doi: 10.1007/s13534-011-0007-x).

LED Technology and Practical Selection Criteria

Beyond wavelength, several practical LED parameters affect PPG sensor performance.

Spectral Width and Binning

LED emission is not monochromatic. Typical LED spectral widths (FWHM) are 20-30 nm for green LEDs, 15-25 nm for red LEDs, and 30-50 nm for infrared LEDs. Broader spectral width means the "effective" absorption coefficient is averaged over a range of wavelengths, smoothing out narrow spectral features. For pulse oximetry, broader LED spectra can introduce systematic errors because the R-ratio calibration assumes a single wavelength. Webster (1997) analyzed the impact of LED spectral width on SpO2 accuracy and found that green LEDs with FWHM exceeding 30 nm could introduce errors of 1-3% SpO2.

Manufacturing variation in LED peak wavelength (binning) is another concern. LED manufacturers sort production output into wavelength bins, typically 5-10 nm wide. For consumer wearables, using tighter bins increases component cost. For clinical devices, tight binning (2-3 nm) is standard practice.

Radiant Efficiency and Power Consumption

LED wall-plug efficiency (optical output power divided by electrical input power) varies substantially with wavelength. Green LEDs (InGaN technology) have typical efficiencies of 10-25%. Red LEDs (AlGaInP) achieve 20-40%. Infrared LEDs (GaAs/AlGaAs) reach 15-30%. These efficiencies directly impact battery life in wearable applications.

For a target photodetector signal level, the required LED drive power depends on the LED efficiency, tissue absorption and scattering at the operating wavelength, the LED-detector separation, and the photodetector responsivity at the operating wavelength. When all factors are combined, infrared LEDs typically require the lowest drive power for a given SNR at the wrist, despite the lower pulsatile signal amplitude, because the combination of higher LED efficiency, lower tissue scattering, and higher silicon photodiode responsivity at 940 nm compensates for the lower modulation index. This analysis is critical for analog front-end design optimization.

Thermal Considerations

LED forward voltage and thermal characteristics affect drive circuit design. Green LEDs have forward voltages of 2.8-3.5 V, requiring boost converter architectures from typical 1.8-3.3 V battery voltages. Red and infrared LEDs have lower forward voltages (1.7-2.2 V), enabling simpler direct-drive circuits from low battery voltages. Thermal management is important because LED peak wavelength shifts with temperature (approximately +0.1 nm/C for red and infrared LEDs, +0.04 nm/C for green LEDs), which can affect SpO2 calibration accuracy over the operating temperature range.

Multi-Wavelength System Architecture

Modern PPG sensors increasingly incorporate multiple LED wavelengths driven in a time-multiplexed sequence. The Apple Watch Series 7+ uses green, red, and infrared LEDs with a multi-channel photodetector. The Samsung Galaxy Watch 5+ adds a bioelectrical impedance sensor to the optical system.

Time-Division Multiplexing

Multiple LED wavelengths share a single photodetector by operating in a rapid time-multiplexed sequence. Each LED illuminates for a brief period (10-100 microseconds), and the photodetector output is sampled during each LED's active period and during LED-off intervals for ambient light subtraction. A complete measurement cycle including three LED wavelengths and ambient sampling can be completed in under 1 millisecond, enabling effective sampling rates exceeding 100 Hz per wavelength channel.

The timing architecture must ensure complete LED turn-off between channels to prevent optical crosstalk, and the photodetector and amplifier must settle fully between channels. Modern AFE chips like the TI AFE4404 and Analog Devices ADPD4101 provide hardware support for multi-wavelength time-division multiplexing with programmable timing and integrated LED drivers.

Wavelength Selection for Robustness

Beyond measuring specific analytes, multi-wavelength PPG can improve measurement robustness. Different wavelengths have different motion artifact characteristics (due to different penetration depths and tissue interactions), different ambient light susceptibilities, and different skin tone dependencies. Fusing information from multiple wavelengths can improve heart rate accuracy during motion, enhance SNR through wavelength diversity, and provide more equitable performance across skin types.

Lee et al. (2021) demonstrated that a three-wavelength PPG system (green, red, infrared) with adaptive wavelength weighting reduced heart rate estimation error by 35% compared to single-wavelength green PPG during high-intensity exercise, by dynamically selecting the wavelength channel with the best signal quality for each time segment.

Summary of Wavelength Selection Guidelines

The following table summarizes optimal wavelength choices for common PPG applications.

Heart rate (wrist reflectance): Green 520-530 nm. Highest AC/DC ratio, best SNR.

Heart rate (finger/ear transmission): Infrared 940 nm or Red 660 nm. Deep penetration, low power.

SpO2 measurement: Red 660 nm + Infrared 940 nm. Clinical standard, required pair.

Multi-parameter (HR + SpO2): Green 525 nm + Red 660 nm + Infrared 940 nm. Full capability.

Skin tone equity: Infrared 940 nm primary. Lowest melanin sensitivity.

Blood pressure features: Green 520-530 nm. Sharpest waveform morphology.

Lowest power consumption: Infrared 940 nm. Best overall photon efficiency.

Wavelength selection is ultimately an optimization problem that balances measurement goals, hardware constraints, target population, and power budget. For a deeper understanding of how wavelength interacts with photodetector selection and signal conditioning, see our companion hardware design guides.

Frequently Asked Questions

What is the best LED wavelength for PPG heart rate monitoring?

For reflectance-mode PPG at the wrist, green LEDs in the 520-530 nm range produce the highest pulsatile signal amplitude and best signal-to-noise ratio. Tamura (2019) documented that the AC/DC modulation ratio for green PPG at the wrist is 2-5 times higher than red and 3-8 times higher than infrared under equivalent conditions. For transmission-mode sensors at the finger or earlobe, infrared (940 nm) or red (660 nm) wavelengths are preferred because their deeper tissue penetration enables the light to traverse the tissue successfully.

Why are exactly 660 nm and 940 nm used for pulse oximetry?

These wavelengths were chosen because they maximize the differential absorption between oxygenated hemoglobin (HbO2) and deoxyhemoglobin (Hb). At 660 nm, Hb absorbs approximately 10 times more light than HbO2. At 940 nm, HbO2 absorbs approximately 2-3 times more than Hb. The ratio of these differential absorptions produces maximum sensitivity in the clinically relevant SpO2 range of 70-100%. Additionally, practical LED and photodetector technologies perform well at these wavelengths, and decades of clinical calibration data exist for this specific wavelength pair.

Can a single wavelength measure both heart rate and SpO2?

No. Heart rate can be measured at any wavelength where a detectable pulsatile signal exists (green, red, or infrared). SpO2 measurement fundamentally requires at least two wavelengths with different ratios of HbO2 to Hb absorption. This is because SpO2 is calculated from the ratio of pulsatile absorbance at two wavelengths, and a single wavelength provides no information about the relative concentrations of the two hemoglobin species. The minimum configuration for SpO2 is red (660 nm) plus infrared (940 nm).

What emerging LED wavelengths are being explored for PPG?

Several non-standard wavelengths are under active research. Yellow-orange LEDs (590-610 nm) are being investigated for improved venous oxygen saturation and total hemoglobin measurement. Multiple near-infrared wavelengths (730 nm, 810 nm, 850 nm) enable multi-spectral analysis for carboxyhemoglobin and methemoglobin detection. Short-wave infrared (SWIR) LEDs at 1300-1700 nm are being explored for non-invasive glucose estimation, though clinical validation remains preliminary. UV-A LEDs (365-405 nm) have shown promise for fluorescence-based skin perfusion assessment.

References

• In the near-infrared region (700-1000 nm), the relationship between HbO2 and Hb absorption inverts. At 940 nm, HbO2 has an extinction coefficient of approximately 1,200 L/(molcm) while Hb is approximately 700 L/(molcm). The absolute absorption values are lower than at shorter wavelengths, and tissue penetration is deepest (5-10 mm), enabling transmission through thick tissue sites. The isosbestic point at approximately 805 nm is where HbO2 and Hb absorption coefficients are exactly equal, making the PPG signal at this wavelength independent of oxygen saturation. Prahl (1998) compiled the comprehensive reference dataset for hemoglobin extinction coefficients that remains the standard reference for PPG system design (doi: 10.1117/12.154627, related compilation).
• MetHb absorbs strongly across the visible spectrum with a characteristic peak near 630 nm. At 660 nm and 940 nm, MetHb absorption is similar, causing the R-ratio to converge toward 1.0 (corresponding to approximately 85% SpO2) regardless of true oxygen saturation. Multi-wavelength pulse oximeters like the Masimo Rad-57 use 7-8 wavelengths to resolve all four hemoglobin species simultaneously. Barker and Shah (2017) provided a comprehensive review of multi-wavelength pulse oximetry approaches for dyshemoglobin detection (doi: 10.1213/ANE.0000000000002056).
• Higher scattering at shorter wavelengths has two consequences. It limits penetration depth (beneficial for concentrating the measurement in superficial capillaries) and it increases the probability of photon detection in reflectance mode (because more photons are scattered back toward the surface detector). Jacques (2013) published the definitive review of tissue optical properties across wavelengths, providing the quantitative data essential for PPG sensor modeling (doi: 10.1088/0031-9155/58/11/R37).
• Infrared wavelengths, with their deeper penetration, sample a more diffuse vascular bed that smooths out the fine morphological features most relevant to blood pressure estimation. Elgendi et al. (2019) analyzed PPG waveform features at multiple wavelengths and found that green PPG at the wrist provided the most reproducible and informative feature set for blood pressure regression models (doi: 10.3390/s19143271).
• Glucose:* Non-invasive glucose estimation via PPG remains one of the most pursued and most challenging goals. Glucose has weak absorption features in the near-infrared (NIR) around 1550-1650 nm and in the mid-infrared. SWIR LED/detector pairs at these wavelengths are being explored, but the glucose absorption signal is orders of magnitude smaller than hemoglobin absorption, making reliable detection extremely difficult. Caduff et al. (2009) reviewed multi-sensor approaches to non-invasive glucose monitoring and concluded that no single optical wavelength provides sufficient specificity for glucose detection (doi: 10.1007/s13534-011-0007-x).