Multicast

Multicast has always been our fundamental requirement when designing low power wireless networks. It brings enormous efficiency into a system. A single message, such as a sensor reading or a dimmer setting action is sent out in the air and picked up by multiple receivers, which act upon such message.

The key problem here is reliability. Usually (and this is a human nature) we want to hear a confirmation. But that can be very costly and slow. Up to a point of not serving the initial purpose at all. Like in a lighting system.

A typical ceiling in a commercial building today has dozens of lights. Often more than a hundred. But even with small rooms equipped with as few as 20 lights, multicast is orders of magnitude more reliable, compared to acknowledged unicast traffic.

After all it is all very simple. Let's assume a probability of delivering a single message to a destination is 90%. Meaning there is 10% chance the message will not be received (for many reasons: interference, collisions, ...). If two messages are sent, the probability that none of them is received is 10%^2 = 1%. With three messages in a row the failure rate drops to 0.1%, so we have 99.9 reliability in such system. Essentially the number of messages translates into the number of nines in the reliability figure. Want five nines? Send a message five times.

In an acknowledged scheme, when every recipient responds with a confirmation, the math is much worse. With 20 lights, we have 21 messages if all goes well and a few less when some responses are not sent or are lost, which triggers a retry. Such retry would make a sense in an isolated, wired or optical transmission system. In wireless even if you retry to a one node, you still occupy the whole radio space. The cost of sending a message to a single recipient is the same as the cost of sending it to multiple recipients.

This is why wireless multicast systems shine. And especially in lighting control multicast is the most natural way of setting up a system: one switch - many lights, a couple of sensors - many lights.

Somebody, some time ago, for a reason unknown to me, came with an idea of applying tightly routing and unicast wireless transports, such as ZigBee or Thread to lighting domain. It was a bad idea. Very likely the one that limited widespread proliferation of wireless lighting control systems. It is about to change. With Bluetooth Mesh around the corner, wireless lighting is now getting a second chance. It will be very exciting to see how it plays out.