I have been reviewing the documentation for pretty much every Zigbee and Zigbee-like mesh networking hardware solution I can find. So far I have been unable to find a solution which meets my target requirements: simple to implement, self-healing mesh networking, low enough in cost for modestly profitable low volume designs, a migration path to a high volume lower cost design, and the ability to programmatically control the node type at run time.The last requirement is driven by my goal of making a single device platform with with a single hardware connector. A mini USB connector would provide power and data for the coordinator device (the one at the PC) and just power for all other devices, which would be routers.

Jennic looked pretty good right up to the last requirement. Unfortunately there is no way to build an application which can programmatically perform router OR coordinator functions depending on the USB status. Which is a real shame, as the whole system seems otherwise nicely thought through. An integrated micro/radio has a ton of RAM and ROM. The user application is loaded into RAM on startup. A high volume product can store the application in ROM. A low volume or development system can store the program in external serial flash. The micro/radio can be purchased inexpensively. A complete module can be constructed for about ten or twelve bucks in relatively small quantities. A readymade module can be purchased for about twenty-five bucks. All the tools, including the stack, C compiler and IDE, are free. The dev systems are reasonably affordable, and a recent press release suggests that they will soon be cheaper still. A very nice touch is the API programming model, where the hardware is completely insulated from the user app, accessed only by API calls. Very clean. and much more likely to work! This approach is also used by the Ember 260, but it is ridiculously expensive.

Sadly, I am not willing to sacrifice the ability to construct a single product which acts as coordinator or router depending on environmental context. So I will have to forgo Jennic for now, and continue to look at other systems.

The MaxStream modules look like they contain both router and coordinator functionality. Unfortunately they are expensive and offer no path to chip-level integration. I may end up using them for simplicity if nothing else.

The Meshnetics system has some compelling features. It is based on a standard platform, the Atmel radio and an Atmel micro. The prefab modules seem well thought out, and well designed, both hardware and software. The system implies that a high volume product could emulate the prefab modules, and run the same application, but it is not clear how that works financially. I can find no offering of their stack independently, and their applications info suggests that their development environment supports only their modules. And of course, the documentation, tools and support from Atmel is awful, as always. Finally, their dev kits are way too expensive.