Selecting a mesh network for an outdoor lighting installation

[AlexStewart]

Hello forum!

I am in the planning phase of a new project where I want to build a large outdoor interactive lighting installation. The concept involes building a large metal dome spaceframe (a subdivided icosahedron) about 4 meters tall and 6 meters wide. Each edge of the framework will house an addressable LED pixel strip, and each vertex of the framework will house a microcontroller node responsible for controlling them. See image for a simplified example:

To control the nodes, I think I need to build a mesh network. Could anyone recommend a network technology I could use here?

A bit more context below:

• The intallation will be used at music gigs, and will display precomputed LED animations.
• The system will be controlled from a computer running TouchDesigner.
• The computer could connect to the network direct or through a head node gateway. This can be worked out later; the interface to the computer should not drive the network technology decision.
• The nodes will need to be addressable so that computer can send the correct animation sequence to node.
• The whole setup will be "portable". It'll be put up and down repeatedly, so hard coding node addresses is not preffered as it would complicate construction.
• Some sort of address allocation during an initialisation would be preffered,
• There will be many verticies in the frame, around 100, so cost per node is a primary consideration.
• I intend the node to be as simple as possible: ie a microcontroller, power regulator, network interface, and the LED output channels
• I have experience designing PCBs so I am anticipating making a custom board for the nodes
• I used to target AVR chips but nowadays I target the RP2040, so I'd aim to use that as the MCU
• I've used Arduino for years so sticking to Arduino compatible hardware is always my preference
• The network could be wired, as I'll already be routing power cabling to the nodes anyway
• OR it would be wireless but I'm anxious that'll be too expensive and may not work within the context of the metal framework

My initial thoughts have been:

• I have experience with CAN-Bus (I work in automotive), but i've already dicounted this as it doesn't lend itself well to a mesh topology
• While networks like I2C and SPI might work off-board, I've discounted it as unsuitable for this scale of installation
• Wireless like WiFi or XBee might turn out to be ideal, but it'd have to be very cheap, prefferably use trace antennas, and have some way of allocating addresses in a predictable way.

If anyone has any experience, ideas or thoughts I'd love to hear them!

[oldcurmudgeon]

Each edge connects to two vertices, so first you need to define which vertex controls which edge. This may simplify things if some vertices can be passive and not have any electronics.

[mikb55]

@AlexStewart

If anyone has any experience, ideas or thoughts I'd love to hear them!

How many bytes per second will you need to distribute via the network and what latency (in milliseconds) is acceptable, bearing in mind that mesh networks have variable latency.

Wireless will probably work OK in the lab, but what about at a venue where there are literally thousands of mobile phones transmitting on cellular and WiFi frequencies? Do you have a budget to operate in the licensed radio spectrum?

The approach used by "Mikes's electric stuff" AKA whitewing.co.uk is to have relatively dumb PIC microcontroller nodes that interpret data sent on multiple high speed RS485 links. Effectively each node is a de-multiplexer. You can find his projects on youtube.

Smart nodes sound attractive until you have to deal with the headache of rolling out firmware updates to multiple devices every time you need to make a minor change.

[jremington]

+1 for "Mike's Electric Stuff". His lighting projects are amazing. I like this recent example for Deep Mind, where he did the node design/firmware and power: Art — Jason Bruges Studio

Very ambitious project. Have fun and good luck!