How well did you sleep last night? Did you wake up feeling truly rested, or somewhat groggy? Quality sleep is essential for our health and well-being, but with as many as one-third of American adults not getting enough sleep regularly, it’s a problem that’s desperate for solutions.

Sleep deprivation can affect everything from one’s mood and ability to concentrate to risk levels for various chronic conditions. On a larger scale, sleep disorders can impact workplace productivity, trigger accidents, and lead to other costly outcomes. Research published in the Rand Health Quarterly examined the topic in five industrialized countries: the U.S., the U.K., Japan, Germany, and Canada. The 2017 study, “Why Sleep Matters — The Economic Cost of Insufficient Sleep,” found that up to $680 billion a year is lost across these five countries because of insufficient sleep. 1

How Is Sleep Quality Measured?

Actually, both sleep quantity and sleep quality are important. The National Sleep Foundation recommends that most adults get seven to nine hours of sleep each night. As for the quality side of the equation, the Foundation notes that good quality sleep in adults “means that you typically fall asleep in 30 minutes or less, sleep soundly through the night with no more than one awakening, and drift back to sleep within 20 minutes if you do wake up.” The opposite conditions denote poor quality sleep. 2

Typically, several times a night, we cycle through rapid eye movement (REM) sleep and non-REM sleep stages. Longer and deeper REM periods generally occur as we move toward the morning hours. The following is an overview of the stages. 3

  • Stage 1: Shortly after falling asleep, the brain produces alpha and theta waves and eye movements slow down. This light stage of sleep generally lasts up to seven minutes.

  • Stage 2: This is also a fairly light stage of sleep, when the brain produces sleep spindles, or sudden increases in brain wave frequency. The brain waves then slow down.

  • Stages 3 and 4: This is when deep sleep begins, with the brain producing slower delta waves. No eye movement happens here. It’s at these stages when the body restores itself, repairing muscles and tissues, stimulating growth and development, boosting its immunity, and building up energy for the following day.

  • REM sleep: This period begins roughly an hour and a half after falling asleep, with each REM stage lasting up to an hour. On average, adults have five to six REM cycles each night. It’s at this phase when dreaming happens and when the brain goes through the process of storing information from the day in long-term memory.

Figure 1. The principle of reflective optical-pulse measurements. (Courtesy of Maxim Integrated)

What Is Obstructive Sleep Apnea?

There are a variety of common sleep disorders, including insomnia, narcolepsy, sleep paralysis, and restless legs syndrome. In addition, there is sleep apnea, which comes in various forms. Central sleep apnea happens when the brain doesn’t send proper signals to the muscles that control breathing. Obstructive sleep apnea (OSA) is characterized by brief, but repeated, interruptions of breathing during sleep.

OSA occurs when muscles in the back of the throat do not keep the airway open, even though the afflicted person is trying to breathe. According to the National Sleep Foundation, more than 18 million American adults have sleep apnea, which can cause fragmented sleep and low blood-oxygen levels. 4 Undiagnosed OSA can be considered a “hidden health crisis.” A Frost & Sullivan study commissioned by the American Academy of Sleep Medicine highlighted the following costs associated with undiagnosed OSA. 5

  • Workplace accidents: $6.5 billion

  • Motor vehicle accidents: $26.2 billion

  • Lost productivity: $86.9 billion

  • Comorbid diseases: $30 billion

Monitoring Sleep Quality Through Polysomnography

A traditional approach to assessing sleep quality involves spending the night at a sleep lab for a polysomnography (PSG) sleep study. A PSG study consists of the following:

  • An electroencephalogram (EEG) to detect the brain’s electrical activity

  • Electro-oculography (EOG) to measure eye movements

  • Electrocardiography (EKG) to measure heart rhythms

  • Electromyography (EMG) to measure movement of body muscles

In addition, the PSG also utilizes a pulse oximeter to continuously measure blood-oxygen levels, flow meters to measure air flow through the nose, and audio and video recordings to assess snoring, movements, and general sleep quality. The test can confirm an OSA diagnosis and also evaluate other sleep-related disorders and conditions.

While PSG is considered the gold standard for sleep monitoring, it also comes with high cost in resources and is not accessible to everyone who may need it. Given the preponderance of sleep insufficiency in the general population, not to mention the associated costs of these disorders, it would be invaluable to find an accurate yet more efficient method to monitor sleep. This is where wearable technology can help.

Vitals to Track for Assessing Sleep Quality

Over the last several years, a variety of sleep monitoring solutions have hit the market, from smartphone-based sleep apps that track movement to sensors integrated into mattresses and bed-side monitors that track movement, respiration, and heart rate. Then there are the wearable options, which provide a convenient, unobtrusive, and potentially more accessible method.

The first generation of wrist wearables attempted to monitor sleep by tracking motion using an accelerometer, which was not entirely accurate. The next generation began assessing heart-rate variability (HRV) on the wrist. Analog front-end devices have come a long way in terms of lowering the overall system power, making it practical to monitor heart rate and HRV continuously during sleep. The combination of HRV plus motion detection (actigraphy) provides a relatively accurate picture of sleep quality. HRV is also tied to stress and overall wellness, with more variation a typical indication of good health. With sleep apnea, HRV can be altered. OSA can result in increased absolute high-frequency power in some cases, while central sleep apnea can result in reduced very low frequency in some cases. This may be confusing as higher HRV is generally associated with good health. However, these methods alone aren’t enough to detect sleep apnea.

An assessment of blood-oxygen saturation level (SpO2) in combination with HRV measurements has, on the other hand, demonstrated an improvement in sleep monitoring accuracy. Pulse oximetry, the tool used to measure blood oxygenation, hasn’t proven to be a valid apnea detector on its own, but knowing the patient’s oxygen saturation and data about RR intervals (the intervals between successive heart beats) is much more telling. An SpO2 value for most healthy people is more than 95%. However, for people with sleep apnea, the SpO2 values can drop to dangerously low levels and, if not treated in time, can lead to chronic health conditions.

Designing Sleep-Monitoring Wearables with Greater Accuracy

To accurately monitor SpO2 on wrist wearables, designers need to take into account clinical standards, system requirements, a proper optical design, and smart algorithms to meet the clinical standards. However, before we delve into each of these building blocks, it would be helpful to understand the technology behind measuring pulse oximetry.

Wrist-based pulse oximetry uses the reflective approach, where a photodiode and an LED are placed adjacent to each other and the photodiode collects light reflected from various depths under the skin.

Pulse oximetry is based on two key principles:

  • The modulation of transmitted light caused by absorption of pulsatile arterial blood. Pulsatile arterial blood absorbs and modulates incident light that passes through tissue, forming the photoplethysmographic (PPG) signal.

  • Different absorption characteristics of oxygenated hemoglobin (HbO2) and deoxygenated hemoglobin (RHb) for different wavelengths of light.

A PPG signal is comprised of an AC component, representing the light absorbed by the pulsatile arterial blood, and a DC level, which captures the effects of light absorbed by other blood and tissue components. The perfusion index (PI) is the ratio of the AC signal to the DC level. The AC and DC components of the received PPG signals differ for different LED wavelengths.

Clinical Standards

While wearables equipped with sophisticated algorithms and artificial intelligence can uncover useful insights, medical professionals won’t truly trust that information unless it is from a device considered to be of clinical-grade quality. In other words, the data collected must be of the same trusted quality as that which is collected in a rigorously established clinical setting.

For SpO2 monitoring, the U.S. Food and Drug Administration has specified the average RMS errors (ARMS), based on sensor type, as shown in Table 1.

Table 1. Typical ARMS specifications by sensor type.

ARMS is calculated as follows:

System Requirements

When implementing SpO2 in wearables, it is prudent to understand the system-level requirements both from signal quality and from power budget perspectives. One of the key parameters in enabling SpO2 on any location of the body is the signal-to-noise ratio (SNR). The SNR requirements to service SpO2 are noted in Figure 2. At a 3.5% error, the perfusion index (PI) and corresponding dynamic range should be as follows to meet clinical-grade quality specifications.

  • PI of 0.02, dynamic range of 110 dB

  • PI of 0.05, dynamic range of 103 dB

  • PI of 0.1, dynamic range of 97.5 dB

PI is the ratio of the pulsatile blood flow to the static blood in peripheral tissue and indicates the strength of a person’s pulse. The higher the value, the stronger the pulse. Blood perfusion varies from individual to individual and is also dependent on ambient considerations. The ability to meet the SNR dynamic, is affected by the sensor device’s optical architecture as well as parameters like skin tone, body location, ambient temperature, and the presence of tattoos or sweat. The lower the PI, the higher the dynamic range required to capture an accurate signal.

The analog front end (AFE) SNR requirements need to take into account the PPG SNR requirements for algorithms, as well as SNR for the PI. The AFE SNR versus PI shown in Figure 2 assumes that SpO2 algorithms require a PPG signal of 35 dB. For instance, for a subject with a PI of 0.05% (66 dB), the AFE is expected to support an SNR of 101 dB (66 dB+35 dB) to calculate SpO2.

Figure 2. SpO2 on the wrist SNR requirements. (Courtesy of Maxim Integrated)

Optical Design

Since it’s the first stage in the receiving path of a wearable system, an optimal photodiode is one that provides high responsivity in detecting small heart-rate signals at key operating wavelengths. Two LEDs with different wavelengths are needed to measure SpO2. 6 For the best PPG signal, the LED illumination wavelength should be near the absorption peaks of HbO 2: 540 nm and 570 nm. However, most commercial PPG sensors utilize green LEDs emitting about 530 nm.

The spacing between the LED components and the photodiode plays an important role in enabling detection of PPG signals, as does the LED driving current. Using a large LED driving current increases the effective penetration depth of the incident light because of the higher light intensity. LED driving current is, however, typically limited by the manufacturer for a specified maximum power dissipation.

Figure 3. Impact of optical spacing to PI (AC/DC), based on Maxim’s optical simulation models. (Courtesy of Maxim Integrated)

Another method is to place the photodiode close to the LEDs. Though when placed too close, the photodiode becomes saturated by the large non-pulsatile components that come from the multiple scattering of incident photons by certain layers of the skin. Increasing the distance between the photodetector and the LEDs can generate a plethysmogram with a larger pulsate signal component, but this approach requires a higher LED driving current. 6

The graph depicted in Figure 3 shows the impact of optical spacing on PI (AC/DC), based on Maxim’s optical simulation models. The PI for SpO2 is highest around 5 – 9 mm.


Algorithms take the raw PPG signal from the AFE in a wearable and convert it into an SpO2 number. Developers have been able to accurately measure SpO2 from the fingertip and the earlobe for quite some time, but accurate wrist-based measurements have been more challenging to attain due to low blood perfusion in the area. AFEs with a high SNR, however, can provide a good enough signal quality for accurate, noninvasive SpO2 measurement.

Per FDA accuracy requirements for wrist-based SpO2 measurements, the root mean square error (RMSE) must be below 3.0% for transmissive pulse oximetry and below 3.5% for reflective pulse oximetry. The proprietary optical architecture as well as algorithms developed by Maxim for this application meets these clinical requirements, with an average RMSE of 2.92% and an overall RMSE of 3.11% (validated by Hypoxia Lab tests on 20 subjects). 6


Wearables designers can now take advantage of sensor interfaces with advanced algorithms that provide accurate optical pulse oximetry and heart-rate sensing for their applications. For example, with Maxim’s complete reference design, MAXREFDES103, customers can validate SpO2 across a wider spectrum of population. MAXREFDES103 includes the MAX86141 AFE sensor, which integrates high-current LED drivers, a low-noise signal conditioning AFE, an ambient light cancellation circuit, and a “picket-fence” detect-and-replace algorithm. The reference design also integrates Maxim’s optical design capabilities and the MAX32664 sensor hub MCU, helping to reduce design time by half. MAXREFDES103 meets FDA requirements for wrist-based SpO2 measurements.

Indeed, given the proper optical design, clinical-grade technologies, and sophisticated algorithms, wearable designers have the tools needed to develop accurate sleep-monitoring products. These products, in turn, can be a key in helping us get a better night’s sleep.



This article was written by Sudhir Mulpuru, Director of Product Management, Healthcare Business Unit, Maxim Integrated (San Jose, CA). For more information, contact Mr. Mulpuru at This email address is being protected from spambots. You need JavaScript enabled to view it. or visit here .

Sensor Technology Magazine

This article first appeared in the June, 2020 issue of Sensor Technology Magazine.

Read more articles from this issue here.

Read more articles from the archives here.