
The Story
Pick up any smartwatch spec sheet and you’ll see a wall of sensors. Optical heart rate. ECG. Blood oxygen. Skin temperature. Some now claim blood pressure. The marketing implies that more sensors equals a smarter, healthier you.
관련해서 what each sensor measures도 함께 참고하시면 좋습니다.
관련해서 a wrist sensor tackling pulse-oximeter skin-tone bias도 함께 참고하시면 좋습니다.
관련해서 reading oxygen far deeper than the wrist도 함께 참고하시면 좋습니다.
Wondering if those sleep stages are real? See how to read smartwatch sleep tracking.
Not sure a full watch is even right for you? See smartwatch vs fitness tracker.
To actually train with those numbers, see heart rate zone training on a smartwatch.
Deciding between ecosystems? See the cross-brand Apple vs Galaxy comparison.
Here’s the thing: a sensor is just hardware that collects a signal. Whether that signal is accurate, whether it’s medically meaningful, and whether it’s even cleared by a regulator are three completely separate questions. A watch can have six sensors and still tell you almost nothing a doctor would act on. So let’s walk through what each one actually does, where it breaks, and what “medical grade” really means.
Optical heart rate (PPG). This is the green light on the back of the watch. PPG stands for photoplethysmography — don’t let the name scare you. The watch shines light into your skin and measures how much bounces back. Blood absorbs light, and the amount pulsing through your wrist changes with each heartbeat, so the watch counts those pulses. At rest, on a good fit, wrist PPG is genuinely reliable — often within a beat or two of a chest strap.
The trouble starts when you move. Motion is the number one source of error in wrist PPG. Every runner has watched their watch report a wild heart rate during intervals, because the sensor is trying to read a signal through a wrist that’s bouncing. Two other factors matter more than most people realize. Tattoos can wreck the reading — one study on a wrist-worn optical sensor found heart-rate error as high as 22.9% over tattooed skin, and counter-intuitively the error was worst at rest and actually shrank as exercise intensity rose. And skin tone plays a smaller role than the headlines suggest: at rest the difference is minimal, but in studies error tends to run a bit larger on darker skin at high exercise intensity, because melanin absorbs some of the green light the sensor depends on. The fix the whole industry uses is a chest strap for serious training. PPG is a lifestyle sensor, not a lab instrument.
ECG (or EKG). This one’s different, and it’s the closest thing on your wrist to a real medical tool. When you touch the crown or bezel with a finger, you close an electrical circuit across your heart and the watch records the actual electrical rhythm — a single-lead ECG. A hospital uses twelve leads, so this is a simplified version, but it’s real electrophysiology, not an optical guess.
This is also where regulation actually shows up. The Apple Watch ECG app received FDA De Novo authorization — that’s the pathway for a genuinely new class of low-risk device — and so did its irregular rhythm notification feature. What that clears it to do is narrow: spot signs of atrial fibrillation, the most common serious arrhythmia. It is not cleared to detect heart attacks, and Apple limits both features to users 22 and older. Samsung’s Galaxy Watch also offers ECG, and it’s cleared in the US through the FDA’s 510(k) pathway; outside the US, availability varies by country because each region’s health regulator has to sign off separately. That’s the recurring theme with wearable health: a feature can exist in the hardware everywhere and still be switched off where it isn’t approved.
Blood oxygen (SpO2). Same optical trick as heart rate, but using red and infrared light, since oxygen-rich and oxygen-poor blood absorb those colors differently. It estimates the percentage of oxygen your blood is carrying. Useful context for altitude, sleep, or general wellness — but wrist SpO2 is an estimate, not the fingertip pulse oximeter a nurse clips on you.
SpO2 is also the poster child for how messy this space gets. Because of a patent fight with Masimo, Apple had to disable blood oxygen on new Apple Watches sold in the United States, then brought it back with a workaround that moves the calculation off the watch and onto your paired iPhone. Nothing changed about your health — the sensor was always physically there — but the feature was legally unavailable for well over a year in one country. It’s a clean reminder that what your watch can show you is shaped as much by courtrooms as by science.
Skin temperature. Two sensors — one near your skin, one under the display — let the watch isolate body warmth from room air, with claimed precision around 0.1°C. But here’s the part people get wrong: this is not a fever thermometer. It’s tuned to detect the small overnight shift in wrist temperature that tracks your menstrual cycle and helps estimate ovulation retrospectively. It reads relative change over nights, not an absolute “you have a 38.5°C fever” number. Judge it as a cycle-tracking aid, not an illness detector.
Blood pressure. The headline feature everyone wants, and the one to be most skeptical of. There are two ways to do it. A real medical monitor uses a cuff that physically squeezes your arm. A watch can’t do that, so it uses the optical sensor to estimate pressure from the shape of your pulse wave — which only works if you first calibrate it against a real cuff. Samsung’s implementation, for example, needs a traditional cuff to calibrate and then re-calibration roughly every 28 days, and Samsung positions it as a wellness feature, not a diagnostic one — in the US it rolls out under the FDA’s general-wellness policy rather than as a cleared medical device. Independent research has repeatedly found cuffless, wrist-based blood pressure to fall short of clinical accuracy. Treat the number as a rough trend line, never as a reading you’d change medication over.
Accelerometer and gyroscope. The unglamorous motion sensors, and quietly the most dependable of the lot. They power step counting, sleep-stage estimates, workout detection, and crash or fall detection. Fall detection in particular has a real track record of getting help to people who couldn’t call for it themselves. No regulator needed — measuring motion is just physics, and physics doesn’t argue.
Here’s the whole picture in one table:
| Sensor | What it measures | Accuracy reality | Medical clearance |
|---|---|---|---|
| Optical heart rate (PPG) | Pulse via reflected green light | Solid at rest; degrades with motion, tattoos, high-intensity exercise | Not a regulated diagnostic; wellness use |
| ECG (single-lead) | Heart’s electrical rhythm via finger contact | Genuine electrophysiology; single-lead, simplified | FDA De Novo (AFib); varies by country |
| Blood oxygen (SpO2) | Blood oxygen % via red/infrared light | Estimate; below fingertip oximeter precision | Wellness feature; availability affected by patent disputes |
| Skin temperature | Overnight wrist temperature shift | ~0.1°C relative; not an absolute fever reading | Cycle-tracking aid, not a thermometer |
| Blood pressure | Pulse-wave estimate (optical) | Needs cuff calibration; below clinical accuracy | Wellness only in most regions; not a diagnostic cuff |
| Accelerometer / gyroscope | Motion, steps, sleep, falls | Reliable | No clearance needed; fall detection well proven |
The Takeaway
If there’s one idea to carry out of this, it’s that the sensor count on the box is close to meaningless. What matters is the pairing of hardware quality and the regulatory clearance behind a specific feature. An ECG that’s been through FDA De Novo review is a different animal from a blood-pressure estimate marketed as “wellness” — even though both live under the same watch glass.
I’ve made this point in a couple of other pieces — one that breaks the heart, ECG, and blood-pressure numbers down by the actual figures, and one that sorts every brand’s health features into three tiers of what actually works. The pattern is consistent: the features people buy for (blood pressure, “medical” claims) are usually the least trustworthy, and the boring ones (heart rate at rest, motion, ECG for AFib) are the ones that hold up.
So how should this change what you buy? First, decide which one health job you actually care about. If it’s arrhythmia screening, you want a cleared ECG in your country — full stop. If it’s fitness, resting heart rate and a good accelerometer cover almost everything, and a chest strap covers the rest. If it’s cycle tracking, temperature helps. If it’s blood pressure for a diagnosed condition, buy an actual cuff and treat the watch as a curiosity.
And don’t let the sensor list pull your eye away from the boring specs that decide whether you’ll wear the thing at all. A watch loaded with sensors that dies by dinner, or that you’re scared to swim in, gets left in a drawer — and a sensor in a drawer measures nothing. Before you weigh sensors, it’s worth reading up on what really drains a smartwatch battery and what those water-resistance ratings actually promise. Sensors are the flashy part of the pitch. Battery and waterproofing are what determine whether the watch is on your wrist when the sensors would matter.
This article is for informational purposes only and is not medical advice. Wearable health data is for reference; consult a qualified clinician for diagnosis or treatment.
Photo: Luke Chesser / Unsplash
댓글 남기기