How to design a better pulse oximeter: Implementation -

How to design a better pulse oximeter: Implementation


It is more important than ever to design medical devices that are more convenient to use and consume less power. This article discusses design and implementation of a more effective pulse oximeter.

In the first article in this two-part series, we covered the technical specifications of a pulse oximeter. In this article, we’ll cover design considerations such as transmissive vs reflective, sensor positioning, perfusion index, motion artifacts, and specifics of designing with an optical AFE.

Transmissive vs. Reflective

A PPG signal can be obtained using a transmissive or reflective LED and PD configuration. A transmissive configuration measures the nonabsorbed light passed through a part of the body. This configuration is best suited to areas such as the finger and earlobe where measurement benefits from the capillary density of these body locations, which make the measurements more stable, repetitive, and less sensitive to variations in placement. Transmissive configurations achieve a 40 dB to 60 dB increase in the perfusion index.

Reflective PPG configurations are chosen when the PD and LED must be placed next to each other for practicality, such as with wrist- or chest-worn devices.

Figure 1. LED-PD configuration. (Source: Analog Devices)

Sensor Positioning and Perfusion Index

Positioning on the wrist and chest require greater dynamic range in the PPG AFE as the DC signal is greatly increased due to the depth of the arteries below static reflective components such as skin, fat, and bone.

Greater resolution in the PPG measurements will reduce the uncertainty in the SpO2 algorithm. With a typical PI of 1% to 2% for wrist-worn SpO2 sensors, the goal of pulse oximeter design is to increase the PI through mechanical design or to increase the dynamic range.

The spacing of the LED to PD will have a major effect on the PI. Too little spacing will increase LED to PD crosstalk or backscatter. This will appear as a DC signal and saturate the AFE.

Increasing this spacing reduces the effect of both backscatter and crosstalk but also reduces the current transformer ratio (CTR), which is the LED output to PD return current. This will affect the efficiency of the PPG system and require greater LED power to maximize the AFE dynamic range.

Rapidly pulsing one or multiple LEDs has the benefit of reducing the 1/f noise contribution to the overall signal. Pulsing the LEDs also makes it possible to use synchronized modulation at the receive side to cancel out ambient light interferers. Integrating multiple pulses increases the PD signal amplitude and lowers the average current consumption. Increasing the total PD area also increases CTR as more of the reflective light is captured.

For heart rate PPG measurement, a combination of a single large PD and multiple power efficient green LEDs has been adopted by many HR device manufacturers to be used on places where there is limited blood flow. Green LEDs are chosen due to their high rejection of motion artifacts. However, this comes at the cost of power. Green LEDs have a higher forward voltage than red and IR and high absorbance in human tissue, meaning a higher LED power is required to return meaningful cardiac information.

As SpO2 requires multiple wavelengths and most systems still incorporate high efficiency green LEDs for the HR PPG, the most common configuration for HR and SpO2 PPG systems is a single green, red, and IR LED array surrounded by multiple PDs, as seen on the ADI VSM watch in Figure 2. PD to LED spacing has been optimized to reduce backscatter and the baffle design reduces LED to PD crosstalk.

Figure 2. ADI VSM watch V4, baffle, and LED DP array. (Source: Analog Devices)

Multiple prototypes of the ADI VSM watch were trialed to verify the most efficient PD to LED spacing for our HR PPG and SpO2 measurement.

Motion Artifacts

Motion artifacts provide one of the greatest design challenges to a PPG measurement system. When motion is present, the width of the arteries and veins change due to pressure. The amount of light absorbed by the photodiode changes and this is present on the PPG signal because photons are absorbed or reflected differently than when a body is at rest.

For an infinitely wide photodiode area covering an infinitely long deep tissue sample, all photons will eventually be reflected to the photodiode. In this case, no artifact due to motion will be detected. This, however, cannot be achieved; the solution is to increase the photodiode area while taking capacitance into account—lowering AFE and providing filtering for motion artifacts.

The normal frequency for a PPG signal is between 0.5 Hz to 5 Hz while motion artefacts are typically between 0.01 Hz to 10 Hz. Simple band-pass filtering techniques cannot be used to remove motion artifacts from the PPG signal. To achieve high accuracy motion cancellation, an adaptive filter needs to be supplied with highly accurate motion data. For this purpose, Analog Devices has developed the ADXL362 3-axis accelerometer. This accelerometer provides 1 mg resolution with up to 8 g of range while consuming only 3.6 μW at 100 Hz and is available in a 3 mm × 3 mm package.

Optical AFE

The positioning of the pulse oximeter generates several challenges. Wrist-worn SpO2 devices provide additional design challenges as the AC signal of interest is only 1% to 2% of the total received light on the PD. To achieve medical grade certification and distinguish between slight variations in oxyhemoglobin levels, a higher dynamic range on the AC signal is required. This can be achieved by reducing ambient light interference and decreasing LED driver and AFE noise.

Increased dynamic range is essential to measure SpO2 under low perfusion scenarios, and next-generation optical AFEs like the Analog Devices ADPD4100 (and ADPD4101) achieve up to 100 dB SNR. This integrated optical AFE has eight onboard low noise current sources and eight separate PD inputs. The digital timing controller has 12 programmable timing slots that enable the user to define an array of PD and LED sequences with specific LED current, analog and digital filtering, integration options, and timing constraints.

Increased SNR/μW is an important parameter for battery-powered continuous monitoring. This key metric has been addressed by increasing the AFE dynamic range while also lowering the AFE current consumption. The ADPD4100, for example, has a total power consumption of only 30 μW for a 75 dB, 25 Hz continuous PPG measurement including the LED supply. Increasing the number of pulses per sample (n) will result in a (√n) increase in SNR while increasing the LED drive current will have a proportional increase in SNR. 1 μW total system consumption will return 93 dB SNR for a continuous PPG measurement using a 4 V LED supply.

Automatic ambient light rejection reduces the burden on the host microprocessor while achieving 60 dB of light rejection. This is achieved using LED pulses as fast as 1 μs in conjunction with a band-pass filter to reject interference. In certain operating modes, the ADPD4100 automatically calculates the photodiode dark current or LED off state. This result is subtracted from the LED on state before conversion in the ADC to remove ambient light as well as gain errors and drift within the photodiode.

Design is further simplified with application-specific development tools. For example, the ADPD4100 is supported with the EVAL-ADPD4100-4101 wearable evaluation kit and ADI Vital Signs Monitoring Study Watch. This hardware seamlessly connects to the ADI Wavetool application to enable bioimpedance, ECG, PPG heart rate, and multiwavelength PPG measurements for SpO2 development. Embedded in the study watch is an automatic gain control (AGC) algorithm that tunes the TIA gain and LED current to deliver optimum AC signal dynamic range for all LED wavelengths selected.

Finger- and earlobe-based SpO2 readings are the easiest to design for as the signal-to-noise ratio is higher than wrist- or chest-based positioning due to the reduction in bone and tissue, which also reduces the DC component contribution.

For such applications, an optical sensor module like the ADPD144RI and photometric front end like the ADPD1080 enable developers to quickly skip the design challenges associated with LED and PD placement and spacing to achieve optimal power to noise ratios. This is possible because the optical sensor has an integrated red 660 nm LED, 880 nm IR LED, and four PDs in a 2.8 mm × 5 mm package. The spacing between the LEDs and PD have been optimized to give the best signal-to-noise ratio for SpO2 high accuracy PPG measurements. The device has also been mechanically optimized to reduce optical crosstalk as much as possible, even when the sensor is placed under a single glass window.

The ADPD1080 is an integrated optical AFE with three LED drive channels and two PD current input channels in a 17-ball, 2.5 mm × 1.4 mm WLLCSP. This AFE works well for custom design low channel count PPG products where board space is critical.

click for full size image

Figure 3. ADPD410X block diagram. (Source: Analog Devices)

click for full size image

Figure 4. ADPD4100 simultaneous red (right) and IR (left) PPG measurement. (Source: Analog Devices)


  1. Toshiyo Tamura. “Current Progress of Photoplethysmography and SpO2 for Health Monitoring.” Biomedical Engineering Letters, February 2019.
  2. Jihyoung Lee, Kenta Matsumura, Ken-Ichi Yamakoshi, Peter Rolfe, Shinobu Tanaka, and Takehiro Yamakoshi. “Comparison Between Red Green and Blue Light Reflection Photoplethysmography for Heart Rate Monitoring During Motion.” 2013 35th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), July 2013.

Robert Finnerty is a systems applications engineer at Analog Devices where he works in the Digital Healthcare Group based in Limerick, Ireland. He works closely with the Vital Signs Monitoring Group, focusing on optical and impedance measurement solutions. Rob joined the precision converters group within ADI in 2012 and has focused on low bandwidth precision measurement. He holds a bachelor’s degree in electronic and electrical engineering (B.E.E.E) from National University of Ireland Galway (NUIG). He can be reached at

Related Contents:

For more Embedded, subscribe to Embedded’s weekly email newsletter.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.