Jul 2018: Prototype I

After launching into our mission to build a swim tracker, we decided that such a swim tracker must consist of motion sensors (IMU) and a pressure sensor. Like this, we would be able to track a swimmer’s arm movement and measure how strongly the arm pushed through the water. Our goal was to test such sensors as quickly as possible.

After a couple of weeks, we finished our first prototype: It consisted of a breakout board that contained a Nordic nRF52 chip capable of transmitting data via Bluetooth. For the pressure sensor, we used a force sensitive resistor that we wanted to integrate into the palm of the swimmer’s hand. After we developed one of the most hideously soldered boards, we recorded our very first data! After a couple of tests in the dry however, we realized that the precision of the pressure sensor was too low and that we needed a much more precise solution.

Oct 2018: Prototype II

While the pressure data of the first prototype was too imprecise, the motion data looked good. However, we hadn’t tested it in the water yet. So we split up and while some people were looking for better pressure sensors, the others developed a water proof prototype capable of recording motion data while swimming.

The result of this was the second prototype, which also was very fundamental: At it’s core was an Adafruit Adalogger and an SD card slot for the data. We soldered an IMU to it, wrote the Arduino code, and were then only left with the waterproofing.

Since we didn’t want to lose too much time or money (we didn’t have any), we decided to just glue the electronics into a Hubba Bubba chewing gum case and sealed everything with glue. Finally, we would just attach the case to our hand and swim with it.

This worked great but had one annoying flaw: If we needed to access the electronics, we first needed to scrape off the old glue only to afterwards cover everything in glue again and let it dry. You can imagine how often we cycled to the local swimming pool only to go back home because the device died or there was some bug in the software.

After a lot of gluing however, we finally recorded our very first swimming data. We then discovered that our sampling rate was too low to precisely reconstruct the arm position.

Apr 2019: Prototype III

After the first two prototypes didn’t deliver qualitative data, we wanted to step up the quality for the third one. It not only introduced a custom, multi-layer PCB, but also more sophisticated building blocks such as a small surface-mounted 32MB flash memory, a multi-color LED, a vibration motor, and, most importantly, better sensors.

• For the IMU, we used the Bosh BNO055, which features an IMU that is mounted on the same silicon die as a small Cortex M0 chip and thus enables really high sampling rates with immediate on-chip post-processing. This did the trick and we got high-quality motion data.

• We also found a high precision, water and chlorine resistant pressure sensor, that is accurate enough for our application.

We waterproofed the electronics by molding silicon around them, and attached the band to the hand via a strap. After swimming half a minute in lake Zurich, we had a first look at the data.

At first glance, the recorded data seemed good: When the hand dived underwater, the pressure increased and decreased as the hand resurfaced. However, we didn’t consider the underwater pressure! We couldn’t tell how much pressure was really generated by the swimmer and how much by the water. If we measured a pressure of 65 mBar, it could be that the hand was 60cm in the water and the swimmer generated 5 mBar, or that the hand was 50cm in the water and the swimmer gernerated 15 mBar.

Jan 2020: Prototype IV

To fix the pressure problem of the third prototype, we added a second pressure sensor and developed a fusion algorithm, that extracts the swimmer’s applied power from the measurements of the two sensors. Now a swimmer can see the exact power applied while pushing through the water: The following graph shows four subsequent strokes, each of which was swum harder than the previous one. We can see how both, the applied pressure goes up and the time taken for the stroke goes down.

Parallel to solving the problem with the pressure data, we also made the following progress:

• Our PCB is not completely custom. This allowed us to make it smaller and adapt its shape so it perfectly fits into the band, making it as comfortable as possible for the wearer.

• We introduced a circuit, that lets us shut off the complete power consumption of the board via software. Like this, the board lasts for month when in stand by.

• We have rewritten the code on the band to run in C instead of Arduino. This allows us to sample multiple sensors at the same time as well as to connect to our cloud backend or a user’s mobile phone and upload data.

• We developed an iOS and Android app that can get data from the board and displays all trainings a user has recorded.

• The design is slimmer and more comfortable. It also features a watch-like way of closing the band, that is more elegant than the strap of the third prototype.

The problems remaining with the fourth prototype are mainly in the design of the band:

• By slightly changing the board layout, we can make the band even slimmer and more robust at the same time.

• The lock of the band sometimes pops open. We are in the midst of designing and testing new alternatives to solve that problem.

• The materials used for the prototype are not optimal in feel and durability. We currently intend to use a TPE for the final product.

We also need to improve our data evaluation and visualization.

Plans for the future

We have now reached a state where our technology is mature enough to place the core focus on the design. During our last prototypes, we already gained some experience and tested several designs for comfort and functionality. We went through 16 design iterations, out of which we produced four from silicone and tested them with the electronics in the water.

We now want to take the design process to the next level using more advanced materials and manufacturing methods. We are currently looking for sponsors and funding opportunities that enable us to fine tune and finalize the design for the market.

Parallel to the design, we are working on making our technology more robust, extracting insightful swimming feedback from the data and are refining our business plan and market strategy.