Here’s a look at the electronics and sensors deployed inside a widely used fitness watch.
Leland Teschler
Executive Editor
The fitness tracking firm Fitbit says it sold something over 21 million fitness watches last year. To get an idea of how fitness watches work their magic, we tore down one of the more popular fitness watches, the Fitbit Charge.
The Charge tracks steps, distance, calories burned, floors climbed and active minutes, monitors sleep, and has caller ID that can read out incoming calls when your phone is nearby. The watch also syncs with Bluetooth and comes with a charging cable. It’s said to have a battery life of seven to ten days – considered good for smart watches like this — and can monitor sleep patterns. The watch works in conjunction with a smart phone app that reads out all these factors and adds niceties such as badges for making progress toward a fitness goal.

An examination of the Charge internals begins with prying off the watch’s plastic housing to reveal the display, circuit board, and vibration motor (from Jinlong Machinery Electronics in China) used for signaling. These watch components sit on top of a substrate that goes against the wear’s skin. The substrate includes a C-shaped metal strap that braces the charging cable plug while also serving as a cradle for centering the PCB and display in the plastic watch housing.
The substrate of the watch is up against the wear’s skin in normal use. The surface that is in intimate contact with the wear’s wrist holds the heartbeat sensor. An examination of the surface touching the wrist reveals both the sensor chip itself and two dots which are infrared LEDs. To one end of the heartbeat sensor sit two prongs which comprise the connection plug for the charging cable.



The heartbeat sensor actually consists of both the optical sensor chip and the two infrared LEDs. In normal use the optical sensor detects light from the LEDs that reflects from the wear’s skin. Heartbeat detection is via a technique called pulse oximetry. This measurement method takes advantage of the fact that oxygenated and de-oxygenated hemoglobin have different optical properties. With every heartbeat, there is a spike in arterial oxygenated blood which the optical sensor detects as a change in the skin’s absorbance or reflectance of LED light.
Put another way, the optical sensor measures the amount of LED light reflecting from the blood under the wearer’s skin. The reflectance is a bit different during a heartbeat, and the sensor will detect this periodic signal from which the heart rate is extracted.
The substrate holding the optical sensor and LEDs also holds an integrated circuit. This chip carries CMOS op amps optimized for amplifying super-small sensor signals and is made by Texas Instruments (OPA2363). One supposes that Charge designers located the op-amp chip on the same substrate as the optical sensor, rather than on the main Charge circuit board, to reduce the possibility of noise garbling the signals from the optical sensor.
Another interesting construction detail is that the connector from the heart rate monitor substrate to the main circuit board lies directly overtop the optical sensor. We might surmise the Charge designers wished to further minimize the chance of noise impacting the heart rate signals; sitting where it does, the connector might provide some shielding against other signals emanating from the main circuit board.
Flex cables
There are two flat flex-type cables in the Charge. One connects the main PCB to the optical sensor substrate. The other connects the display to the main circuit board. The display is an OLED, logical because OLEDs may consume relatively little power compared to alternatives.
Both flex cables attach to the bottom of the Charge circuit board. Also nestled on the bottom of the PCB is the lithium-polymer battery, a 3.7-V cell capable of providing 140 mA. Three other components on that side of the board are noteworthy.



One is a battery charger IC from TI (BQ24232) that also lets the watch work from power coming via the USB connection if necessary. Another point of interest is an altitude sensor from Measurement Specialties. This altitude sensor is basically a MEMS pressure sensor calibrated for altitude. Its spec sheet says it has a resolution of 20 cm. The sensor module includes an ultra-low-power 24-bit analog/digital converter.
A chip sitting next to the battery charger IC is a bit of a mystery. Its markings aren’t definitive, but there are clues to its identity. One of those hints is that the chip sits super-close to the altitude sensor. Thus it could easily have something to do with altitude sensor readings. Also, a close look at the circuit board reveals there is a connection from the mystery chip to the side button the user pushes for getting readings. There’s another connection from the chip to the TI battery charger IC. So one might surmise that the mystery chip might have a role in managing readouts of the altitude sensor.
The other side of the PCB is more densely populated. This is the side of the PCB one sees when first popping the plastic cover off the watch. The OLED display sits on top, pressed against a piece of plastic that both supports the display and doubles as an antenna for the Bluetooth connection. The metal bracket which serves as a supporting frame for the recharger cable, wraps around and seems to help keep the OLED readout stable as well.
Adhesive attaches the OLED to the Bluetooth antenna assembly. And the antenna assembly is held to the circuit board by two tiny torx screws. Unscrewing these releases part of the antenna assembly supporting the OLED. The rest of the antenna assembly then can be pulled off the circuit board to reveal the components on the board.
There were several smaller ICs on this side of the board whose markings and PCB traces just weren’t definitive. One of them, based on the pin-out connections, might be a low-drop out linear regulator from TI.
However, several other chips were readily identifiable. One is a Texas Instruments boost dc-dc converter (TPS61093) that sits at the end of the PCB near the solder connections for the battery and vibration motor. This chip serves as a power supply for the OLED display.
The board also contains two different processors. One is an eight-bit unit (STM8L151 from ST Microsystems). We can speculate the eight-processor is there to handle mundane tasks such as display management, perhaps dealing with the Bluetooth connection (via a Nordic nRE8001 chip), and the user input.
The other is a 32-bit device (STM32L1 also from ST Microsystems). One might wonder why the watch needs two processors. The answer seems to be that there’s a lot going on in a Charge. The 32-bit chip, for example, could be required because there is a significant amount of signal processing associated with pulling a valid heart rate out of the data coming in from the optical sensor. Ditto for calculating the number of steps the wearer takes. On the Charge, this task apparently requires data from two accelerometers.
Indeed, it was a surprise to find two identical STM LIS2DH (also from ST Microsystems) three-axis accelerometers on the Charge circuit board. According to the spec sheet, these are low-power devices capable of measuring accelerations with output data rates from 1 Hz to 5.3 kHz.
The accelerometers can measure motion in three axes, so it is a bit of a puzzle as to why the watch would need two of them. But a look in the academic literature on biomechanics reveals a possible reason. There is a lot of work being done on using arrays of redundant 3D accelerometers to more accurately estimate the motion of joints and angular velocity errors.
With that in mind, recall that the Charge estimates the number of steps taken based on the motion of the wearer’s arm. Fitbit’s literature says its algorithm for counting steps is designed to look for intensity and motion patterns most indicative of people walking and running. A tracker on a wrist can pick up extra steps if it interprets arm movements when working at a desk, cooking, or doing other tasks as walking.
From these provisos, one can surmise that it’s tough to judge which accelerations result from taking a step and which come from raising your arm for any number of other reasons. Apparently, then, those kinds of distinctions require two accelerometers. We’d further bet that the designers of the Charge wish they had room to include even more accelerometers on the circuit board to make that job easier.
All in all, there is a lot of technology crammed inside a fitness watch. It is probably no surprise that it takes two processors and two accelerometers to make some of the magic happen.
References
Fitbit Charge
Jinlong Machinery & Electronics Co., LTD
Measurement Specialties altimeter
STM LIS2DH accelerometer
STM Microsystems STM8L151 microprocessor
STM Microsystems STM32L1 microprocessor
Texas Instruments OPA2363 precision op amp
Texas Instruments BQ24232 battery charger
Texas Instruments TPS61093 boost converter
HI
Very nice review.
How do you connect the flex cable to the heart beat sensor. I struggle to get the cable in the socket and to keep it in place.
Thanks
In a teardown all we are concerned about is getting things apart for a good look. So reconnecting the flex cable to the heart beat sensor wasn’t on our to-do list. We’ve never tried to reconnect it. Good luck!
Just tried a battery change and reassemble, thanks for the Teardown helped lots, and the repair works perfectly.
As for re-attaching the flex cable to the HR sensor. The small black lever needs to be in the up position (yes I know you knew that), The flex cable slides in on the opposite side to the lever.
This is tricky as it is a tight fit and just a little more pressure than expected is required to seat it securely.
Make sure you clip the lever back down, Before you fit the assembly back into the wrist band check your connection by turning the Fitbit on and give it a gentle shake. If the green lights are pulsing very quickly you have a sweet connection. If you have a slow on / off flash or no lights at all you probably need to try again, the best of luck.