Budgeting sACN universes before you buy the strip

Every LED job I have seen go wrong on site went wrong in a spreadsheet weeks earlier. The failure looks like flicker, or a section that lags a frame behind its neighbour, and it gets blamed on the controller. It is almost always the universe budget.

The arithmetic

One sACN universe carries 512 channels. For RGB pixels that is 170 pixels per universe, with two channels left stranded. For RGBW it is 128.

pixels_per_universe = floor(512 / channels_per_pixel)
universes           = ceil(total_pixels / pixels_per_universe)
packets_per_second  = universes * target_fps

A 12,000-pixel RGB rig at 44 fps:

Under 3 Mbit/s sounds like nothing. The bandwidth is never the problem — the packet rate is. Consumer gigabit switches will move 3 Mbit/s all day and still drop frames at three thousand small multicast packets per second, because the limit that bites is packets per second and multicast group handling, not line rate.

What I actually check

• Pixels per output, not per controller. A controller rated for 16 universes often means four outputs of four universes, and a run that crosses an output boundary will tear.
• Multicast vs unicast. Multicast is the spec default and the reason cheap switches fall over. If the fixture count is fixed and known, unicast to each node is more predictable.
• IGMP snooping. On if multicast, and verified on — not just enabled in the web UI. Without it every port sees every universe.
• A dedicated VLAN. Lighting data shares a switch with nothing. Not the media server's file copy, definitely not house wifi.
• Headroom. I size for 25% spare universes. Somebody always adds a fixture.

Refresh rate is a choice, not a constant

The reflex is to push 60 fps everywhere. For most architectural and scenic looks, 30 fps halves the packet rate and is visually indistinguishable — except on fast strobes and hard camera pans, where it is very much not. Decide per zone, not per system:

def packet_rate(pixels, channels_per_pixel, fps):
    per_universe = 512 // channels_per_pixel
    universes = -(-pixels // per_universe)   # ceil division
    return universes, universes * fps

for fps in (30, 44, 60):
    print(fps, packet_rate(12000, 3, fps))

Run that before the order goes in. It costs nothing and it has saved me a next-day switch purchase more than once.

Power, briefly

Universe planning and power planning fail together. WS2812B at full white draws roughly 60 mA per pixel; 12,000 pixels is 720 A at 5 V, which is 3.6 kW before losses. Nobody builds for that, and nobody needs to — but the cap has to live in the renderer, not in hope. I clamp total output power in software and inject for the real measured ceiling plus margin.

More on how I measure that ceiling in a later note.