Great question. We looked at a fair amount of options when we chose the chipset. The major ones we looked at were CSR, TI, and Nordic.
The problem with CSR is that they are pretty closed. Generally, you have to buy a very large amount of devices before they would even talk to you. And then, most of their libraries and stack are closed. I looked into them about a year ago and basically hit a brick wall, they simply don’t want to talk to you unless you are going to buy 1M devices.
TI was also an option, but not a very attractive one. The main chipset uses an 8051 MCU, not ARM, and it had limited RAM. We needed at least 16K RAM for the Particle cloud encryption, so this wouldn’t work. Plus, they usually require expensive compilers and tools like Keil.
Nordic, on the other hand, works with gcc. Plus, they are now offering the new 32K RAM option which finally allows us to make this work. Before the 32K option was available, we were actually using a second MCU to do encryption for us, which added to the cost and complexity. The new nrf51822 is the first BLE SoC we could get our hands on that would allow us to support it.
The nrf51822 has also been around for a long time. It is a stable, reliable solution with a good ecosystem of modules to choose from. They also are moving more towards open source, so we shouldn’t have any issues opening up the whole codebase (except for the compiled BLE stack, but that comes pre-loaded anyway).
So really, Nordic hit all the right points, and it was kind of a no-brainer. Without the new nrf51822, this would actually be a much more complicated solution.