Wireless Mesh Physics

Last week I posted on the airline industry racing for newly designed aircraft built of new composite materials to shave hundreds of tonnes of weight that do not have to be kept airborne. This is pure physics that dictates technology choices when designing a well performing airliner. Steel and aluminum are the two most popular materials used in ground and sea transportation today: cars, trucks, railroads, trains, ships. But you simply cannot use steel and even using aluminum becomes less of an option today when building an airliner.

The same is true when trying to take technologies used in wired communication networks and apply them to wireless networks. This just does not work, because the physics of electromagnetic radiation in radio frequencies are fundamentally different from the physics of a copper wire or of a strand of glass.

Fundamentally it starts with wireless being much more lossy, due to interference. Smartscrapers surround us. The density of wireless IoT radios today exceeds the design assumptions engineers had just a few years back. One example is the Thread protocol is designed to handle up to 200 nodes in a network while the prediction is a typical home will have 500. Applications such as WiFi, GPS, LTE, Bluetooth are growing and are fighting for spectrum and interference / collisions are a fact today, contributing typically to about 10%-20% of wireless packets not being delivered on a 1st attempt.

Secondly, the common misconception is about isolation of wireless transmissions. In a wired world, when a signal is sent down one switch port, it never appears on an adjacent port. The ports are isolated. They can carry different signals at the same time and not interfere. This is not the case in wireless. Wireless is a broadcast, by nature. Everybody within the radio range receives that broadcast. And if anyone within the range transmits at the same time, we have interference and collisions that result in packet loss.

Interference and lack of isolation are the physics of wireless. These physics are fundamentally different from the physics of wired connections. The physics dictate selection of appropriate technologies. In low power wireless it is the radio, where the choice today is clear: Bluetooth Low Energy is by far the best available and getting even better with every new release. It is like the composites both Boeing and Airbus have selected for their winning new aircraft designs.

Then it comes down to carefully evaluating the design architecture for a network. Single - path routing is not an option due to the inherent packet loss - multipath is a must. And before starting to evaluate any routing algorithms, one has to accept there is no isolation, no switching in wireless (unless we move to multi-channel scenarios or other sharing schemes like TDMA or CDMA).

Finally we have to remember reliability is of an utmost importance. Having a light switch that turns on lights only sometimes or only some of them at a time is the worst thing that can happen. And would prevent the market from widespread adoption of wireless technologies. This is what has been plaguing wireless IoT for years, especially the 802.15.4 - based standards. As recently as at the 2017 CES at the prominent booth of a heavily promoted low power networking standard I heard "sorry, this does not work because there is too much WiFi in the hall...". What message is this to the market? What message is this to a consumer? To the industry? It gets down to the physics again. We have to accept wireless is different. To build a wireless network you need the lightest radio available and an architecture on top of it that can fly, not just taxi around an airfield.