Low-Power PPG System Design for Wearables: LED Drive, Sampling Strategies & Power Budgets

Power consumption is the single most constraining factor in wearable PPG system design, and the difference between a naive and optimized implementation can span two orders of magnitude. A continuous green PPG measurement at the wrist can draw 10-20 mW with poor design choices or under 0.3 mW with careful LED drive optimization, duty cycling, and adaptive sampling. For devices expected to run days or weeks on batteries under 100 mAh, this difference determines whether the product is viable.

This article provides a rigorous treatment of power-aware PPG system design, covering the complete signal chain from LED selection and drive circuitry through analog front-end (AFE) design, sampling strategies, and system-level power management. For foundational PPG concepts, see our introduction to PPG technology. For how wavelength selection affects power budgets, see our wavelength comparison guide.

The PPG Power Budget Breakdown

Understanding where power is consumed in a PPG subsystem is essential before attempting optimization. In a typical wrist-worn reflectance PPG system, the power budget breaks down as follows:

LED drive: 60-85% of total PPG power. The LED is by far the dominant consumer. Green LEDs (525 nm) require higher drive currents than infrared LEDs due to higher melanin absorption and tissue scattering at shorter wavelengths. A typical green LED driven at 10 mA with a forward voltage of 3.0 V consumes 30 mW of instantaneous power. Even with duty cycling at 0.5%, average LED power is 150 uW per LED channel.

Analog front-end (AFE): 10-25%. The transimpedance amplifier (TIA), ADC, ambient light rejection circuits, and LED driver circuitry collectively consume 100-500 uW depending on the AFE architecture. Modern integrated AFEs like the Texas Instruments AFE4404, Maxim MAX86141, and Analog Devices ADPD4101 have optimized this significantly, with some achieving under 100 uW in low-power modes.

Digital processing: 5-15%. The microcontroller performing filtering, peak detection, and heart rate or SpO2 computation. With modern ARM Cortex-M0+ processors at 1-16 MHz, the processing power for basic PPG algorithms is typically 10-100 uW. More complex algorithms involving motion artifact removal add 50-200 uW (Ram et al., 2012).

Data communication: variable. BLE transmission of processed results adds 5-30 mW during active transmission but can be duty-cycled aggressively since results update at 1 Hz or slower.

Marefat et al. (2014) provided one of the first comprehensive power breakdowns for wearable PPG, demonstrating that LED power reduction yields the highest return on optimization effort. Their analysis showed that reducing LED drive current from 20 mA to 5 mA through optical path optimization reduced total system power by 62% while maintaining heart rate accuracy within 2 BPM (DOI: 10.1109/BSN.2014.6855527).

LED Selection and Drive Optimization

Wavelength-Dependent Power Requirements

Different LED wavelengths have fundamentally different power requirements for achieving equivalent signal-to-noise ratio (SNR) in PPG applications. This is driven by the wavelength-dependent optical properties of skin tissue.

Green LEDs (520-530 nm) produce the highest pulsatile signal amplitude at the wrist but also face the highest tissue attenuation. Wall-plug efficiency for green InGaN LEDs is typically 15-25%, meaning 75-85% of electrical power is lost as heat. Typical required optical power at the tissue surface is 1-5 mW for adequate SNR in reflectance mode (Tamura, 2019).

Red LEDs (660 nm) have better wall-plug efficiency (25-40%) and face less melanin absorption, but produce lower pulsatile signal at the wrist. The required optical power is similar to green (1-4 mW) because the lower absorption coefficient of hemoglobin at red wavelengths partially offsets the melanin advantage.

Infrared LEDs (940 nm) offer the best wall-plug efficiency (30-50%) and the lowest melanin sensitivity, but the AC/DC pulsatile ratio is 3-8x lower than green at the wrist (Maeda et al., 2011). This means more averaging or higher SNR is needed for equivalent heart rate accuracy. However, for SpO2 applications where infrared is required, the power advantage of higher LED efficiency is significant.

Pulsed LED Drive and Duty Cycling

The most impactful power optimization in PPG system design is pulsed LED operation with synchronous detection. Instead of driving the LED continuously and sampling the photodetector, the LED is pulsed on for a very short duration (the "integration window") and the AFE samples only during this window.

The mathematics are straightforward. For a continuous LED at drive current I_LED and forward voltage V_f, the average power is:

P_continuous = I_LED * V_f

For a pulsed LED with pulse width t_pulse at sampling frequency f_s:

P_pulsed = I_LED * V_f * t_pulse * f_s

With t_pulse = 50 us and f_s = 25 Hz, the duty cycle is 0.125%, yielding an 800x reduction in average LED power. The key constraint is that the pulsed current must be high enough to generate sufficient photocurrent during the short integration window. Photodetector shot noise scales with the square root of the integration time, so shorter pulses require proportionally higher peak currents to maintain SNR.

Patterson et al. (2009) demonstrated that pulsed operation at 100 us pulse width with 50 mA peak current achieved equivalent SNR to continuous operation at 1 mA, while reducing average LED power by 50x (DOI: 10.1109/TBME.2009.2027338). This established pulsed drive as the standard approach for all battery-powered PPG devices.

Automatic LED Current Control

Static LED drive current wastes power because the optimal current varies with skin tone, sensor placement, contact pressure, and ambient light conditions. Modern AFEs implement automatic gain control (AGC) or automatic LED current adjustment loops.

The MAX86141 AFE, for example, implements a digital feedback loop that monitors the DC photodetector level and adjusts LED drive current in steps of approximately 0.2 mA to maintain the signal within the ADC's optimal input range. This prevents both signal clipping (drive too high) and poor SNR (drive too low). Texas Instruments' AFE4404 provides similar functionality through its programmable LED current DACs and automatic ambient light cancellation.

Wong et al. (2020) showed that adaptive LED current control reduced average power consumption by 34% compared to fixed-current operation across a diverse population with Fitzpatrick skin types I-VI, while maintaining SpO2 accuracy within 1.5% (DOI: 10.1109/JBHI.2020.2991643). The power savings were largest for lighter skin tones, where the default high-current setting was unnecessarily aggressive.

Analog Front-End Architecture

Integrated AFE Solutions

The AFE is the critical interface between the optical components and the digital domain. Modern integrated PPG AFEs combine LED drivers, transimpedance amplifiers, ADCs, and ambient light rejection circuitry in a single chip, dramatically reducing board area, component count, and power consumption compared to discrete designs.

Key specifications that affect power consumption include:

Transimpedance amplifier (TIA) bandwidth. Higher bandwidth allows shorter LED pulse widths (enabling lower duty cycles) but consumes more power. The optimal bandwidth is matched to the LED pulse width, typically 10-100 kHz for PPG applications. The ADPD4101 offers configurable TIA bandwidth from 12.5 kHz to 200 kHz, allowing designers to trade bandwidth for power.

ADC resolution and sample rate. Most PPG applications require 16-20 bits of effective resolution to capture the small AC pulsatile component (0.5-2% of the total signal) without quantization noise limiting performance. Higher resolution ADCs consume more power, but oversampling a lower-resolution ADC and digitally filtering can achieve equivalent effective resolution at lower power in some architectures.

Ambient light rejection (ALR). Sunlight and indoor lighting create large interfering signals that can saturate the photodetector. Active ALR using correlated double sampling or modulated/demodulated detection subtracts the ambient component in hardware, preventing saturation without requiring reduced TIA gain. This is critical for outdoor wearable use and avoids the power cost of running higher-gain LED currents to overcome ambient light.

For a comparison of how different AFE choices affect signal processing pipelines, see our PPG signal processing algorithms overview.

Discrete vs. Integrated Tradeoffs

Discrete AFE designs using separate TIA, ADC, and LED driver components offer maximum flexibility but typically consume 2-5x more power than integrated solutions. The main advantages of discrete designs are:

• Higher TIA performance (lower noise floor, wider bandwidth)
• Ability to optimize each component independently
• Availability of higher-resolution ADCs (24-bit sigma-delta)

For research prototypes and clinical-grade systems where power is less constrained, discrete designs remain common. Bent et al. (2020) used a discrete AFE with a 24-bit ADC to achieve noise floors below 0.1 pA/sqrt(Hz), enabling measurement of extremely weak PPG signals from peripheral body sites (DOI: 10.1038/s41746-020-0234-6). However, for consumer wearables, integrated AFEs are nearly universal.

Sampling Strategy Optimization

Continuous vs. Intermittent Monitoring

The simplest power reduction strategy is to not measure continuously. Many physiological parameters derived from PPG do not require continuous data:

Resting heart rate can be measured in 10-30 second windows every 5-15 minutes, reducing average PPG subsystem power by 95-99%. The Apple Watch uses this approach for background heart rate monitoring, sampling periodically rather than continuously.

SpO2 spot checks require only 10-30 seconds of data, drawing full power only during the measurement window. Background SpO2 monitoring, as implemented in the Apple Watch Series 6+ and Garmin Venu, uses intermittent measurements every 15-60 minutes during sleep.

Heart rate variability (HRV) requires continuous measurement during the analysis window (typically 5 minutes for frequency-domain analysis) but does not need 24/7 monitoring. Most consumer devices measure HRV only during sleep or dedicated sessions. For more on HRV measurement, see our HRV chart by age guide.

Continuous heart rate during exercise is the most power-demanding use case, requiring uninterrupted sampling at 25-100 Hz for the duration of the activity. However, since exercise sessions are time-limited (typically 30-120 minutes), this high-power mode does not dominate total daily energy consumption.

Adaptive Sampling Rate

Within a measurement window, the sampling rate itself can be adapted based on signal quality and physiological state. During rest with clean signals, 25 Hz is sufficient for heart rate determination (Nyquist frequency of 12.5 Hz covers the maximum physiological heart rate of ~220 BPM = 3.67 Hz with substantial margin). During motion, higher rates (50-100 Hz) may improve motion artifact removal performance but at proportionally higher power cost.

Jarchi and Casson (2017) demonstrated an adaptive sampling scheme that monitored signal quality metrics in real-time and switched between 25 Hz (rest) and 100 Hz (motion) modes. This achieved equivalent heart rate accuracy to fixed 100 Hz sampling while reducing average power consumption by 41% during mixed rest/activity protocols (DOI: 10.1109/JBHI.2016.2636219).

Event-Triggered Sampling

Accelerometer-driven event triggering represents the most aggressive power reduction strategy. The approach uses a low-power accelerometer (consuming 5-15 uW) to continuously monitor motion state. The PPG subsystem is activated only when specific conditions are met:

• Periodic timer expiration (e.g., every 10 minutes for resting HR)
• Transition from rest to activity (accelerometer threshold crossing)
• External trigger (user-initiated measurement)
• Detected anomaly in previously measured parameters

This approach can reduce average PPG power to under 20 uW while still capturing all significant physiological events. The Oura Ring employs a variant of this strategy, combining accelerometer-driven activity detection with periodic measurement windows to achieve multi-day battery life on a battery under 20 mAh.

System-Level Power Management

Microcontroller Sleep Modes and Wake Strategies

The microcontroller managing the PPG subsystem spends the vast majority of its time idle between measurements. Effective use of low-power sleep modes is critical:

Sleep mode power: Modern ultra-low-power MCUs (Nordic nRF52832, STM32L4, Silicon Labs EFR32) achieve 1-3 uA in deep sleep with RTC running. This translates to 3-10 uW, which is negligible compared to LED power during measurement.

Wake-up latency: Transitioning from deep sleep to active mode takes 1-10 us on most MCUs, which is fast enough to not affect PPG measurement timing. The AFE typically requires a longer startup time (50-500 us for TIA settling and reference voltage stabilization), which should be accounted for in the power budget.

DMA-driven sampling: Using direct memory access (DMA) to transfer AFE data to memory without CPU intervention allows the processor to remain in sleep mode during data acquisition. The CPU wakes only when a complete buffer of samples is available for processing. On the nRF52832, this reduces active processing time by approximately 60% for continuous 25 Hz PPG acquisition.

Multi-Wavelength Power Considerations

SpO2 measurement requires sequential red and infrared LED pulsing within each sample period, effectively doubling the LED power compared to single-wavelength heart rate monitoring. Additionally, the reduced pulsatile signal amplitude at red and especially infrared wavelengths at the wrist often necessitates higher drive currents or longer integration windows.

The power impact is significant. A green-only heart rate measurement at 25 Hz might consume 300 uW average, while adding red/infrared SpO2 capability increases this to 1.0-2.5 mW due to the additional LED channels and higher drive requirements. For wearables that primarily monitor heart rate with occasional SpO2 checks, maintaining SpO2 as a spot-check feature rather than continuous monitoring is the standard power management approach.

For how multi-wavelength measurements relate to PPG blood pressure estimation methods, the power implications are even more significant, as blood pressure algorithms often require higher sampling rates and more sophisticated signal processing.

Power Budget Example: Wrist-Worn HR Monitor

To illustrate practical power budgeting, consider a wrist-worn device targeting 7-day battery life on a 40 mAh, 3.7 V battery (148 mWh total energy).

Total power budget: 148 mWh / 168 hours = 0.88 mW average.

Allocated to PPG subsystem: 0.35 mW (40% of total, remainder for MCU, BLE, display, accelerometer).

Design choices to meet this budget:

1. Green LED at 525 nm, pulsed at 20 mA peak, 50 us pulse width, 25 Hz sample rate. Average LED power: 20 mA * 3.0 V * 50e-6 * 25 = 75 uW.

2. Intermittent measurement: 15-second windows every 5 minutes during rest (5% duty cycle), continuous during detected activity (estimated 2 hours/day = 8.3% of time). Effective measurement duty cycle: approximately 12%.

3. AFE (MAX86141 in low-power mode): 170 uW during measurement, ~1 uW in shutdown.

4. Total average PPG power: 0.12 * (75 + 170) + 0.88 * 1 = 30.3 uW.

This is well under the 350 uW budget, leaving substantial margin for SpO2 spot checks, higher sampling rates during exercise, and component aging. Real-world devices typically target 50% margin to account for variability across users and environmental conditions.

Emerging Low-Power Techniques

OLED-Based PPG

Organic LEDs (OLEDs) offer the potential for conformal, large-area light sources that could reduce the required drive current by spreading illumination over a larger tissue area. Yokota et al. (2016) demonstrated an ultra-flexible OLED-based PPG sensor that achieved functional heart rate monitoring at LED drive powers below 100 uW by using a large-area (1 cm^2) emitter (DOI: 10.1126/sciadv.1501856). However, OLED lifetime, efficiency, and cost remain challenges for commercial adoption.

Ambient Light PPG

Several research groups have explored using ambient light as the illumination source, eliminating LED power entirely. Lee et al. (2012) demonstrated PPG measurement using only ambient indoor lighting captured by a high-sensitivity photodetector with a narrow bandpass filter. While this approach reduces power consumption to under 10 uW for the optical subsystem, it is unreliable in varying lighting conditions and is not viable for continuous wearable monitoring.

Sub-Threshold Analog Processing

Ultra-low-power analog circuits operating in the sub-threshold region can perform basic PPG signal conditioning (amplification, filtering, peak detection) at power levels below 1 uW. Teng and Bhatt (2021) designed a complete PPG analog front-end operating at 0.8 uW total power using sub-threshold CMOS design, enabling always-on heart rate monitoring with minimal digital processing overhead (DOI: 10.1109/TBCAS.2021.3055781). This approach trades bandwidth and dynamic range for extreme power efficiency, suitable for applications where only basic heart rate detection is needed.

Conclusion

Low-power PPG system design requires optimization at every level of the signal chain, from LED wavelength selection and pulsed drive strategies through AFE architecture and system-level sampling strategies. The dominant lever is LED power management through duty cycling and adaptive current control, which together can reduce average LED power by 1000x or more compared to continuous operation. Combined with intermittent measurement scheduling and aggressive microcontroller sleep mode usage, modern PPG systems can achieve average power consumption well below 100 uW while maintaining clinical-grade heart rate accuracy. For researchers and engineers working on next-generation wearable devices, understanding these power tradeoffs is as important as understanding the signal processing algorithms that extract physiological information from the PPG waveform.

References

• Marefat et al. (2014) provided one of the first comprehensive power breakdowns for wearable PPG, demonstrating that LED power reduction yields the highest return on optimization effort. Their analysis showed that reducing LED drive current from 20 mA to 5 mA through optical path optimization reduced total system power by 62% while maintaining heart rate accuracy within 2 BPM (DOI: 10.1109/BSN.2014.6855527).
• Patterson et al. (2009) demonstrated that pulsed operation at 100 us pulse width with 50 mA peak current achieved equivalent SNR to continuous operation at 1 mA, while reducing average LED power by 50x (DOI: 10.1109/TBME.2009.2027338). This established pulsed drive as the standard approach for all battery-powered PPG devices.
• Wong et al. (2020) showed that adaptive LED current control reduced average power consumption by 34% compared to fixed-current operation across a diverse population with Fitzpatrick skin types I-VI, while maintaining SpO2 accuracy within 1.5% (DOI: 10.1109/JBHI.2020.2991643). The power savings were largest for lighter skin tones, where the default high-current setting was unnecessarily aggressive.
• For research prototypes and clinical-grade systems where power is less constrained, discrete designs remain common. Bent et al. (2020) used a discrete AFE with a 24-bit ADC to achieve noise floors below 0.1 pA/sqrt(Hz), enabling measurement of extremely weak PPG signals from peripheral body sites (DOI: 10.1038/s41746-020-0234-6). However, for consumer wearables, integrated AFEs are nearly universal.
• Jarchi and Casson (2017) demonstrated an adaptive sampling scheme that monitored signal quality metrics in real-time and switched between 25 Hz (rest) and 100 Hz (motion) modes. This achieved equivalent heart rate accuracy to fixed 100 Hz sampling while reducing average power consumption by 41% during mixed rest/activity protocols (DOI: 10.1109/JBHI.2016.2636219).
• Organic LEDs (OLEDs) offer the potential for conformal, large-area light sources that could reduce the required drive current by spreading illumination over a larger tissue area. Yokota et al. (2016) demonstrated an ultra-flexible OLED-based PPG sensor that achieved functional heart rate monitoring at LED drive powers below 100 uW by using a large-area (1 cm^2) emitter (DOI: 10.1126/sciadv.1501856). However, OLED lifetime, efficiency, and cost remain challenges for commercial adoption.
• Ultra-low-power analog circuits operating in the sub-threshold region can perform basic PPG signal conditioning (amplification, filtering, peak detection) at power levels below 1 uW. Teng and Bhatt (2021) designed a complete PPG analog front-end operating at 0.8 uW total power using sub-threshold CMOS design, enabling always-on heart rate monitoring with minimal digital processing overhead (DOI: 10.1109/TBCAS.2021.3055781). This approach trades bandwidth and dynamic range for extreme power efficiency, suitable for applications where only basic heart rate detection is needed.