How smartwatch sensors work: why orientation is irrelevant and skin contact is everything
Founder & CEO, Smartlet · CentraleSupélec engineer · Concours Lépine 2025, Awarded · CES 2026
The most common objection from people that have not tried out the Smartlet setup is that the Apple Watch heart rate sensor cannot “see” the other watch when the Apple Watch faces in. Understanding the heart rate sensor requires a small amount of knowledge about physics, but once understood, the objection is dispelled once and for all.
In this article we will discuss the physics and answer all of the questions such as: Will the heart rate monitor work? Will the step counter work? Will the sleep tracking work? The answer to all of the questions is yes and we will give the reason for each answer. All of the variables that have been found to be important to the heart rate sensor in peer reviewed studies are preserved by the Smartlet system.

Table of contents
- Photoplethysmography: an optical technique that does not know, which way is up
- Why reflectance PPG is inherently direction-agnostic
- Accelerometers and gyroscopes: three-dimensional by design
- What peer-reviewed research says about smartwatch accuracy
- The one variable that actually matters
- Smartlet and sensor performance in practice
- Explore compatible configurations
- Frequently asked questions
Wichtigste Erkenntnisse
| PPG measures from the microvasculature | Not from the radial artery. Not from the surface skin. The capillary network is distributed throughout the wrist tissue - on all sides. |
|---|---|
| Reflectance mode is direction-agnostic | LED and photodiode on the same side of the skin. No gravitational orientation required. Validated by Charlton and Marozas, NIH/PMC 2021. |
| Skin contact is the only variable that matters | Not orientation. Not wrist position. Sensor-to-skin contact quality determines signal accuracy. Scientific Data/Nature, 2025. |
| Mean bias: 0.27 bpm vs clinical ECG | npj Digital Medicine living review, 2026. 82 studies, 430,052 participants. The accuracy argument against inner wrist wearing does not hold. |
| Smartlet maintains consistent contact | The pod presses the smartwatch against the inner wrist. This is the exact condition Apple recommends for workout and continuous monitoring accuracy. |
Photoplethysmography: an optical technique that does not know, which way is up
Blood volume changes in the wrist capillaries are in the tissue of the whole wrist, and therefore are not just in the surface skin of the wrist. So, when measuring PPG signals on the inner wrist, a similar effect is observed as on the outer wrist.
Why reflectance PPG is inherently direction-agnostic
Depending on the mode in, which the light is transmitted or reflected, there are different Photoelectric Plethysmography (PPG) sensors. Transmission PPG is a method used in some wearable devices. In hospital environment finger clips and ear probes are also using the transmission PPG principle. The LED is placed on one side of the tissue, and the photodiode on the other side.
How optical sensors work
As the light travels through the tissue it interacts with the blood vessels, and the signal is generated. The transmission PPG requires a precise location between the LED and the photodiode in the tissue. This is the reason that the hospital clips have to be placed in a particular angle or position.
Why inner wrist positioning works
All wrist wear smartwatches use a reflectance PPG method. The LED and the photodiode are situated on the same side of the skin. The LED emits light that is absorbed by the skin, and the reflection is made by blood vessels within the capillary bed and is detected by the photodiode. There is no transmission of light passing through the tissue. No knowledge of the spatial positioning of the lights, and the photodetectors on the opposite side of the body is required. It is a direction less method.
MDPI Electronics review 2023 Reflectance
MDPI Electronics review 2023 Reflectance is confirmed as the technology used in wrist wearables as it can measure oxygen saturation from the skin surface without having to traverse tissue, which is what has to be done with clinical pulse oximeters. The fact that the wearable has to be held or positioned at some angle is irrelevant. The wrist is not the optimal site for measurement of pulse oximetry as it was at other sites, but this is a precision of location rather than a matter of whether inner or outer wrist is used – and again this is not related to the watch face orientation. MDPI ElectronicsConfirmed that the reflectance configuration is the enabling technology for wrist-worn wearables precisely because it allows measurement from the skin surface without the tissue-transmission requirement of clinical oximeters. Orientation is physically irrelevant to this sensing approach. The same paper also noted that the wrist is not the ideal site for certain high-precision measurements, but this limitation applies equally at any wrist position, inner or outer, and has nothing to do with, which way the watch face points.
The wrist is not the optimal site for measurement of pulse oximetry as it was at other sites, but this is a precision of location rather than a matter of whether inner or outer wrist is used – and again this is not related to the watch face orientation.MDPI ElectronicsConfirmed that the reflectance configuration is the enabling technology for wrist-worn wearables precisely because it allows measurement from the skin surface without the tissue-transmission requirement of clinical oximeters.
Orientation is physically irrelevant to this sensing approach. The same paper also noted that the wrist is not the ideal site for certain high-precision measurements, but this limitation applies equally at any wrist position, inner or outer, and has nothing to do with, which way the watch face points.
| PPG mode | LED and photodiode placement | Orientation dependent? |
|---|---|---|
| Transmission PPG (hospital clip, ear probe) | Opposite sides of tissue; light passes through | Yes - specific placement required |
| Reflectance PPG (all wrist smartwatches) | Same side of skin; reflected light measured | No - direction-agnostic by design |
Accelerometers and gyroscopes: three-dimensional by design
Accelerometer and gyroscope are quite different from the PPG optical sensor. Accelerometer and gyroscope are used for the measurement of the motion and orientation of the device. In the 3 axis accelerometer, the force of gravity is measured along each of the 3 perpendicular axes. This enables the device to understand, which axis is oriented towards the ground, in order to turn the screen when the watch is tilted, and also to identify the characteristic acceleration patterns that correspond to a fall or a car accident.
Not to be confused with always having a defined orientation, an orientation aware device can be in any orientation at any given time. At the core of this capability is the 3D structure of the accelerometer. This sensor measures all three axes in any desired orientation and uses the relative measurement of these axes to calculate the exact orientation of the device at any given time.
Apple’s watchOS for example, constantly asks the user to orientate their Apple Watch in correct positions such as wrist left/right and crown left/right. This is based on the assumption that the Apple Watch can be placed in any of these defined orientations at any given time, which is true to its design.
According toApple's official support documentationIt
According to Apple's official support documentationIt turns out that Apple has a support page on “Accurate Measurements with Watch” that talks about how the sensors work on the watch when it is worn on the wrist. We also know that the configurable orientation setting on the Samsung Galaxy Watch can be controlled via the Galaxy Wearable app. So it turns out that all major smartwatch platforms have 3D orientation independence as a first class engineering requirement. Galaxy Wearable app. These design decisions confirm that three-dimensional orientation independence is a core engineering requirement for all major smartwatch platforms.
So it turns out that all major smartwatch platforms have 3D orientation independence as a first class engineering requirement. Galaxy Wearable app. These design decisions confirm that three-dimensional orientation independence is a core engineering requirement for all major smartwatch platforms.
For Smartlet users: step counting, activity classification, sleep tracking, fall detection and crash detection will still work whether the watch face is pointed up or down. Our firmware doesn’t depend on any assumptions of the hardware orientation. It depends on the declared configuration.
What peer-reviewed research says about smartwatch accuracy
There are 82 studies and 430,052 participants in the living review. The mean bias against clinical ECG is 0.27 bpm. The orientation debate was settled by the physics before it started.
Smartwatch heart rate and blood pressure readings have been a topic of great interest and controversy in recent scientific literature. We summarize two of the most relevant studies.
ECG and the digital crown
A 2026 living systematic review and meta-analysis published in npj Digital Medicineusing data from 82 studies with 430,052 participants reported Apple Watch mean bias from clinical ECG reference was 0.27 bpm with limits of agreement of approximately 7 bpm either side. “This represents the most comprehensive assessment of Apple Watch measurement accuracy published to date” state the authors. npj Digital Medicine, covering 82 studies and 430,052 participants, found that Apple Watch heart rate measurements show a mean bias of 0.27 bpmVersus clinical ECG reference, with limits of agreement of approximately 7 bpm in each direction. The authors describe this as the most comprehensive assessment of Apple Watch measurement accuracy in the scientific literature to date.
“This represents the most comprehensive assessment of Apple Watch measurement accuracy published to date” state the authors.npj Digital Medicine, covering 82 studies and 430,052 participants, found that Apple Watch heart rate measurements show a mean bias of0.27 bpmVersus clinical ECG reference, with limits of agreement of approximately 7 bpm in each direction. The authors describe this as the most comprehensive assessment of Apple Watch measurement accuracy in the scientific literature to date.
Have all seen fitness trackers
We have all seen fitness trackers and wearables pop up everywhere from runner’s wrists to social media. According to a 2024 integrative review of 55 research papers published in the Journal of Sports Sciences (as reported by Runners Connect), heart rate monitoring is an incredibly accurate feature on devices such as the Apple Watch, with an error of about 2–3 beats per minute (bpm) when sitting still and about 5 bpm or less about 87% of the time during dynamic activities. While this may not be accurate enough for scientific use or very exacting fitness tracking, it is certainly accurate enough for general health and fitness training purposes. Runners Connect, found Apple Watch accurate within 2 to 3 bpm at rest and within 5 bpm approximately 87 percent of the time during exercise. For most health monitoring and training applications, this accuracy is clinically sufficient.
While this may not be accurate enough for scientific use or very exacting fitness tracking, it is certainly accurate enough for general health and fitness training purposes. Runners Connect, found Apple Watch accurate within 2 to 3 bpm at rest and within 5 bpm approximately 87 percent of the time during exercise. For most health monitoring and training applications, this accuracy is clinically sufficient.
| Metric | Performance | Source |
|---|---|---|
| Heart rate mean bias vs ECG | 0.27 bpm | npj Digital Medicine 2026 (82 studies, 430,052 participants) |
| Heart rate accuracy during exercise | Within 5 bpm, 87% of the time | 2024 integrative review, 55 studies |
| AFib detection sensitivity | 95% | JACC Advances 2025 (26 studies, 17,379 patients) |
| AFib detection specificity | 97% | JACC Advances 2025 |
These statements only apply to wellness-grade data on a smartwatch and are not diagnostic. The ACC’s 2024 Expert Analysis: Arrhythmias and Implications for Electronic Health Records and Wearables determined that consumer-grade smartwatches are able to reliably and accurately detect A-Fib and that all detections should be reviewed by a doctor prior to taking any action in any orientation and is a characteristic of the technology category. 2024 expert analysisConfirmed that consumer smartwatches achieve high accuracy for atrial fibrillation screening, but that findings require physician confirmation before clinical action. This is true at any wearing orientation and is a property of the technology category.
ACC’s 2024 Expert Analysis Arrhythmias
The ACC’s 2024 Expert Analysis: Arrhythmias and Implications for Electronic Health Records and Wearables determined that consumer-grade smartwatches are able to reliably and accurately detect A-Fib and that all detections should be reviewed by a doctor prior to taking any action in any orientation and is a characteristic of the technology category.2024 expert analysisConfirmed that consumer smartwatches achieve high accuracy for atrial fibrillation screening, but that findings require physician confirmation before clinical action.
This is true at any wearing orientation and is a property of the technology category.
The one variable that actually matters
The orientation of the sensor doesn’t appear to make any difference in the accuracy of the PPG readings (nor does it make any difference to the state of the accelerometers, which are always on and sampling no matter, which way the sensor is worn). So, if we look at the factors that really do affect wearable activity tracking what have we left? In fact, the answer is still the same as it ever was: skin contact. And the large scale studies agree on this point, just as they have on every other occasion. skin contact.
Reflectance PPG sensors measure from the microvasculature distributed throughout the wrist tissue. Orientation is irrelevant. Skin contact is everything.
A study published in PMCPaper published on BioRxiv We recently published our paper to PMC, here is the summary: In this work, we evaluated the impact of signals from a commercial smartwatch embedded photodetector (PPG) over a full 24 hours of free living activity. Of particular interest was understanding why the signal from the PPG would occasionally “lose lock” and what effects this had on the quality of the data. Through analysis of both motion data and skin motion relative to the sensor, we determined that the primary cause of degradation of the signal measured by a smartwatch worn on the wrist is relative motion of the sensor to the skin surface, such as during large movements that cause the sensor to move on the skin. This results in a “rubbing” of the skin that increases the reflection from the epidermis while decreasing the reflection from underlying tissues thereby reducing the peak to peak amplitude of the signal, and hence decreasing the quality of the ECG measurement. We determine that a wristband that is firmly fastened provides a better quality measurement compared to a wristband that is too loose. We also found that the exact location on the wrist had a less significant effect due to the small differences in motion, as long as the wristband is tightened sufficiently for optimal sensor-skin contact.
Through analysis of both motion data and skin motion relative to the sensor, we determined that the primary cause of degradation of the signal measured by a smartwatch worn on the wrist is relative motion of the sensor to the skin surface, such as during large movements that cause the sensor to move on the skin.
Results “rubbing” skin
This results in a “rubbing” of the skin that increases the reflection from the epidermis while decreasing the reflection from underlying tissues thereby reducing the peak to peak amplitude of the signal, and hence decreasing the quality of the ECG measurement.
We determine that a wristband that is firmly fastened provides a better quality measurement compared to a wristband that is too loose. We also found that the exact location on the wrist had a less significant effect due to the small differences in motion, as long as the wristband is tightened sufficiently for optimal sensor-skin contact.
Same rules as choosing a good strap. Scientific Data(Nature) specifically documented that wrist posture changes - not orientation, but changes in contact pressure as the wrist flexes - produce measurable variations in PPG waveform quality. The recommendation from this research: wear the device snugly enough to maintain contact without restricting circulation. This is the same fitting guidance for any quality strap.
Apple's support documentationOn accurate measurements Apple Watch must be in contact with your skin and fit snugly to the body during workouts. The Smartlet is designed to accomplish this. The pod cradles the Apple Watch keeping it in constant contact with the skin on the inside of the wrist. Proper fit ensures the sensor array maintains contact on the skin of the wrist for all ranges of motion during the workout.

The Smartlet pod keeps the smartwatch in continuous skin contact in the inner wrist area, the only variable cited in peer-reviewed literature as affecting the accuracy of PPG readings.
Smartlet and sensor performance in practice
The practical summary covers three areas.
I hear this question a lot. So, people have the idea that the display has to be facing outwards for an accurate reading. So they try it outwards and then inwards, and they notice that there is much less variation when the display is facing inwards and also that the outer watch face stops you from raising your wrist up slightly when you bend your arm, which is what the PPG measures.
Yeah, it’s a common misconception and one that’s very understandable, but unfortunately it’s not correct. The PPG science behind this works out perfectly and it’s totally fine to wear your Smartlet inwards. The contact between the skin and the Smartlet is very good so any benefit or disadvantage from wearing it inwards or outwards is negligible.
Heart rate and all PPG-based metrics:Heart rate and all other PPG-based metrics The light on the inside of the wrist follows the same reflectance principles as the light on the outside of the wrist. The blood volume pulsation in the microvasculature is present at both surfaces. It’s just a matter of skin contact, not the orientation.
Motion-based metrics:Accelerometer Gyroscope Motion-Based Metrics
Motion-based metrics:Accelerometer and Gyroscope Motion-Based Metrics The accelerometer and gyroscope are 3-dimensional sensors and are orientation independent meaning they work in any orientation of the device. Thus features such as step counting, activity recognition, sleep detection, fall detection and crash detection work whether watch face is up or down. WatchOS and Wear OS, as well as Garmin Connect and other major smartwatch platforms are also all orientation independent.
Data grade:Despite wearing the smartwatch at an unusual angle the wellness grade of the collected data will not be adversely affected. The data will still include trends, risk notifications and wellness measurements. The Smartlet does not transform the smartwatch into a medical diagnostic device, and this is still true when wearing the smartwatch at an unusual angle to collect more data.
Unmatched in terms of breadth of health signals tracked on any Smartlet-enabled wearable device: - Apple Watch Series 10 & Apple Watch Ultra 2 (Heart Rate & RR variability through PPG (photoplethysmography) signal, ECG (electrocardiogram) signal, SpO2 (saturation of oxygen in arterial blood) through pulse oximetry, body temperature, fall detection, & crash detection) - Garmin Fenix 8 Pro (Heart Rate & RR variability through PPG (photoplethysmography) signal, SpO2 through pulse oximetry, heart rate & RR variability tracking) - Withings ScanWatch Nova (innovative analog watch style hybrid ECG continuous signal, SpO2) Apple Watch Series 10 and Ultra 2 (PPG, ECG, SpO2, temperature, fall detection, crash detection), Garmin Fenix 8 Pro (PPG, SpO2, pulse oximetry, HRV, satellite communication), and Withings ScanWatch Nova (continuous ECG, SpO2, in a hybrid analog watch form factor).
Explore compatible configurations
The complete smartwatch indexThe complete smartwatch index can be found here. For mechanical watch pairing we are referencing the Seiko brand page, the Hamilton brand page and the Tudor Black Bay model page. This gives us a good idea of the most popular collector segments in Europe, Asia and North America. Seiko brand page, the Hamilton brand page, and the Tudor Black Bay model pageAre useful references for the most popular collector segments in Europe, Asia, and North America.
Frequently asked questions
Yes. Reflectance PPG is using the LED and photodiode on the same side of the skin. The blood volume changes from the capillary microvasculature in the wrist tissue are due to the vasculature dispersed in the wrist tissue. It works on the inner wrist as well, it is the skin contact that matters, not the angle. Source:Charlton and Marozas, NIH/PMC, 2021. Charlton and Marozas, NIH/PMC, 2021.
Yes. The accelerometer can record the gravitational acceleration on the 3 axes at the same time. Apple also has wrist and crown orientation settings, which should cover all conditions. Apple confirmed this in their official support documents. 3D orientation independence was a design goal of the iPhone 4. Apple's support documentation.
Does fall detection still work with Smartlet?
Yes. A fall is detected based on the unique multi-axis acceleration pattern of an uncontrolled fall, which is registered by the 3-axis accelerometer and gyroscope in the smartwatch. The sensors are orientation independent, so it doesn’t matter, which side of the watch face you hold.
Is smartwatch health data reliable enough to act on?
Smartwatch data is wellness-grade. An npj Digital Medicine living review of smartwatch ECG and heart rate variability considered 26 studies to determine the clinical accuracy of wellness-grade smartwatch measurements, and found Apple Watch heart rate monitoring to be clinically accurate at the wellness-grade level.
The JACC Advances meta-analysis of 5 studies of other Apple Watch studies found that the Apple Watch algorithm for detection of atrial fibrillation had a sensitivity of 95% and specificity of 97%. Smartwatch data is a flag or alert that can trigger a doctor visit, but is not intended to be used for a medical diagnosis. Our smartwatch will work at any wearing orientation.
Which Smartlet-compatible smartwatch has the most complete sensor suite?
For cardiac and respiratory, motion and emergency detection: Apple Watch Ultra 2 and Apple Watch Series 10 (compatibility page) For biometrics for training: Garmin Fenix 8 Pro (compatibility page) For continuous ECG in a slim form factor: Withings ScanWatch Nova (compatibility page) compatibility page). For training-focused biometrics: Garmin Fenix 8 Pro (compatibility page). For continuous ECG in a slim form factor: Withings ScanWatch Nova (compatibility page).
Reflectance PPG does not care which way is up. Skin contact is the only variable that matters. Smartlet keeps the sensor pressed against the inner wrist - precisely where it performs best.
Ausgewählte Artikel
Keine Personen gefunden
Wir konnten nichts mit diesem Begriff finden. Bitte versuchen Sie es erneut.