I built a GPS bike tracker and I don’t know what to do with it

Over the past years several bicycles have been stolen from me and my family. We bought locks that were more expensive than the bikes. But it just doesn’t help. Angle grinders are somehow always stronger than bike locks. We had to step up the game and enter level two.

I not a material engineer and I’m not sure if its even possible to build a lock that is stronger than an angle grinder but I studied electrical engineering, I’m a full stack software developer and I love my bike.

Instead of preventing the bike from getting stolen I decided to build a device that gets stolen with the bike alongside and can be tracked back and will never run out of power as long the bike is moving on its own wheels for a few minutes.

Building a GPS bike tracker

It took serveral iterations to get a design that worked as planned. So I’ll only show the last design that is the current (working-) state and can be mass manufactured. The first prototypes used the GSM network and a SIM card for connectivity but as the contracts are getting expensive for several bikes, I decided to use the free LORAWAN and The Things Network as the data exchange service.

Some more design choices:

• Arduino based (Atmega 368 @ 8Mhz)
• super low power in standby
• GPS for localization
• RFM95 / SIM800l for communication
• 3.3v logic voltage
• 3.7v buffer battery
• charging from 6V bike dynamo
• embedded into rear light

Circuit design of my bike tracker

PCB design

The rear light was the best location for the tracker. We have power, a waterproof enclosure and every bike has to have a rear light in Germany. And the potential thief doesn’t recognize the tracker. The light I used is available on Amazon for < 10€ and is easy to open.

The rear light I used for the enclosure

The design was sent to china for manufacturing. Less than three weeks later I had the first prototypes on my desk. Components were added and tested. Due to covid I had to wait several months for the delivery of the GPS chip and some smaller parts. But the proof of concept is built and it passed the functionality test.

First prototype implemenation of the final product. (Some changes have to be made. Components have the wrong size and the layout has it’s weak spots.)

On the top right we have the GPS receiver, the GPS antenna and a reset button for development purposes. In the center there is the Atmega with Arduino bootloader and the power saving 8Mhz crystal. On the left we have all the power and charging components and the RFM95 LORA radio module. On the backside we have RX, TX, GND, Power and some breakout connections for debugging. The charge and state LEDs of the LIPO management chip and the GPS TX LED are also still on the board. The three red LED rear light LEDs are powered by a small mosfet (even though I somehow flipped the layout 🤦… but now I know that you can solder SOT-23 casings upside down (not recommended for production)).

The backside is boring.

Yes it fits into the casing! (No, the yellow wire and the cut traces are not part of the original design)

Super odd problems I haven’t solved

1. Switch mode regulators are complicated. I used the TPS62291 3.3V buck switchmode converter. Followed the datasheets advices but it gets hot and shuts off after a few seconds. (Temporary solution: cut all 3.3V traces and connect to LIPO+ rail (LIPO protection shuts off GND rail))
2. High frequency is black magic too. The LoRaWAN had bad range in my experiments. Maybe the ground plane is too large and shields the coil antenna (or some funny resonance).
3. Washing the board is bad when you have the GPS antenna installed. (I had to replace it after washing)
4. The LiPo management IC TP4056 broke at some point. Also had to replace it.
5. I ignored the accelerometer as the parts never arrived and it’s not important for the proof of concept

Things I leared

1. Building PCBs with SMD parts is not impossible.
2. SIM28L chips are actually easy to use
3. Having as many debug pins as possible makes life easier
4. High frequency stuff is black magic (make use of datasheets!)
5. Antennas need space (let them point up and keep stay away from GND and HF)
6. You can get free sample SMD parts from some manufacturers if you ask kindly

OK cool. And now?

This project was fun. I learned alot but I lost interest in commercializing the product. If anyone is interested in building on top of this, feel free. But please don’t copy and say its yours.

Take a look at the GitHub repo. https://github.com/Neumi/trackix

This tracker can be mass produced for around 30€ each and it would be possible to flood the market with GPS trackable bikes. Let me know what you think. Is it a cool idea to build a “map of thiefs” by collecting all the location data from many stolen bikes and smoking out the places where people hide stolen bikes?