Designing the Equisense tracker

I am a graduate student at the University of Technology of Compiègne, France. At the same time, I work as Firmware engineer for Equisense and I am helping to build the first connected equine tracker.

As I don’t have knowledge in designing PCB, we worked with a 3rd party for our first PCB prototype to do the schematics/layout/gerber and to assemble our first batch of 5 PCBs. Together, we defined specifications of the product at the hardware and firmware levels. I took charge of choosing the components and wired them during the elaboration of the schematics (microcontroller, BLE module, sensor and external Flash). I learnt a lot about how to design schematics thanks to Upverter and I drew my first schematics with their online tool. I didn’t know enough about rechargeable battery and voltage regulation to handle this part but I learnt a lot during the process. Also, I didn’t spend time dimensionning resistors and capacitors as I am not an electronics engineer…

First Equisense PCB

Then I made my first wrong assumption that we had to deal with. Due to a very satisfaying Kickstarter campaign we had to accelerate the development of the product so we contacted manufacturers to work on Design For Manufacturing (DFM) — mostly on the mechanical design as we had a PCB prototype and only a skin mechanical design. But it wasn’t that simple. We contacted a company capable of working on the whole product: electronics, firmware and mechanical parts. During the first call, the CEO give us a hard and bad review of the PCB we designed. Besides loosing a few months of hard work, I liked his arguments.

First wrong assumption — BLE Module

Our first prediction was to sell about 3,000 units the first year. I’ve read a lot about the pros and cons of using a pre-certified BLE module. Most websites (Texas Instruments, Predictable Designs (previously Teel Engineering) and others) claim that designing an RF product from a BLE chipset is expensive and should be considered only if the predictions are over 10,000 units, for these main reasons:

  • Certified for FCC, CE and more (about $7,000)
  • Sometimes certified Bluetooth SIG ($4000) (Cypress or LSR modules)
  • Shorter software development, decreased time-to-market

I agree, but:

  • BLE modules cost about $6 — I choosed Silabs’ BGM111 new module which is great (custom BLE service, useful library, low power) but comes with a price.
  • We need a host processor, because we are embedding some detection algorithms (~$2)

Added together, those costs are about $4 more expensive than using a BLE chipset like the nRF52 (powerful 64Mhz Cortex M4F BLE chipset).

  • $4 x 3,000 units = $12,000
  • FCC, CE and Bluetooth SIG certifications are about 7,000+4,000= $11,000. For small start-ups, Bluetooth SIG charges $2,500 to become a member, so count about $9,500 for certifications.
The Innovation Incentive Program (IIP) is designed to enable small start-up companies to commercialize their first Bluetooth product. The company is required to have an annual revenue of less than USD $1 million and no prior QDLs, EPLs or Declarations. A member who qualifies for the IIP program can receive up to two declarations at the reduced price. — Bluetooth SIG

We are saving up to $2,500 with our own design for only 3,000 units. Still, we have to pay more to get our RF design from a 3rd party company (engineering costs) but I think it’s worth it for several reasons. Oh and, by the way, the design I worked on very hard for a few months is just garbage now.

Nordic Semiconductor’s nRF52 chipset

Without this chip, I’m not sure I would be that confident with designing our own chipset/antenna solution.

I’ve worked on the nRF51 for 6 months at Spire and know this chip pretty well. Nordic Semiconductor released the nRF52 chip, a 64Mhz Cortex M4F upgrade of the nRF51, also including the Bluetooth stack. The SDK provided is very very complete for many IoT devices, so I was able to implement my own BLE service with one characteristic in a day thanks to very useful examples (I didn’t do that much with the BGM111 within a day). The SDK and SoftDevice (BLE Stack) is quite the same between the nRF51 and nRF52 so I should be able to implement main features very quickly. Thus, we can cut some development costs.

We now have one MCU which is less complex than having a host MCU + BLE module (embedding another MCU). Silabs has a great library to make communication between the host and BLE module easy to implement and less prone to errors. It’s based on UART and there are request and response packets transfered for each operation. I’m not sure this solution is really efficient (power consumption, overhead in communication).

Another advantage of the nRF52: Firmware updates. Nordic’s SDK includes a bootloader capable of upgrading the application firmware, BLE stack and the bootloader itself. Nordic also provides a smartphone application (iOS and Android) that can be used with the bootloader. I already told you about development costs earlier but choosing the nRF52 decreases our development costs for many features.

Going forward

We sold 1,200 units during our Kickstarter campaign so our predictions can be adjusted and we can expect to sell about 5,000 units during the first year.

We will receive our first PCB prototypes at the end of January. Plastic molds are currently being designed, and we hope to be able to order them at the end of January.

Now I have to work for my final exams. In two weeks, I will work full-time on the project for 4 weeks so I may give you some interesting updates.