Why 2.5 m and not 3 m

A 3 m globe is a better object. It is also 28.3 m² of screen, 5.5 million pixels at P2.5, and around 500 kg of shell before the base — which in this hall meant a structural report, a floor survey and a four-week delay. At 2.5 m the shell is 19.6 m² and roughly 255 kg, and the existing floor slab took it without reinforcement.

The client asked whether 2.5 m would look small. It does not, because the hall is 7 m wide and the globe sits at eye height: from the entrance you see the earth fill about 60% of your field of view at three metres. Size on a screen is a function of how close you can get, not how big the diagonal is.

That argument saved the project four weeks and a structural report. It is also the argument that costs us money, because a 3 m sphere is a bigger invoice. We make it anyway.

One seam, and where it goes

A sphere has a seam whether you like it or not — the place where the last band closes. On a globe showing continents, you want that seam somewhere nobody looks: the middle of the Pacific, or straight down the back if the sphere is on a column against a wall.

Here the seam runs down the service side, facing the wall, 40 degrees off the vertical. The pixel map is generated from the same origin, so content that crosses the seam still lines up. What does not line up is a stock 360° video from a stock library, which almost always starts at Greenwich. We re-mapped their loop and lost half a day.

The last band closing the pole is cut by hand on site. Every sphere is slightly different by the time it is up — welding draws, the shell settles, the anchor plate is never exactly level — so the final modules are trimmed to fit what is actually there rather than what the drawing said. Two technicians, four hours, and a box of spare modules for the ones that get cut wrong the first time.

Looking up at the top pole. The last band closes by hand on site — no two spheres close the same way, so the final modules are trimmed to what is actually there.
Looking up at the top pole. The last band closes by hand on site — no two spheres close the same way, so the final modules are trimmed to what is actually there.

The 40 mm gap

The sphere sits on a 1,100 mm steel column. If the sphere meets the column directly, you have a hard line where two different radii meet, and any movement in the column telegraphs straight into the shell. If you caulk it, the caulk cracks in eighteen months and looks worse than the line.

So we designed a shadow gap: a 40 mm recessed ring, black anodised, that the light from the bottom band of modules does not reach. From a metre away it reads as a floating sphere. It also gives the shell somewhere to move, and it hides the cable pass-through, which on this build was 12 cables plus two power feeds.

That ring cost about the same as three modules and took two days to design. It is the single detail visitors comment on, and nobody knows it is there.

Power, heat and the slow rotation

At 800 W/m² the globe draws 15.7 kW at full white. In practice the content is a map, not a white field, so it sits nearer 6 kW — but the power feed is sized for the peak, because the client will eventually play something bright and we are not going to be the reason it browns out.

The hall has no mechanical cooling at the globe's height. A sphere that size in a closed room with no air movement will run its modules about 12 °C above ambient, which is fine indoors but leaves nothing in reserve. We specified the rear of the shell as a vented plenum and put a slow extraction fan at the top pole, running at 30% speed so it cannot be heard from the floor.

The rotation is 0.5 rpm, which is one turn every two minutes. Faster and it reads as a machine; slower and visitors think it is broken. The bearing is a slewing ring rated for 1.2 t, ten times what it carries, because a globe that stops is a maintenance call in a public hall.

What we would change

The column should have been ours. It came from the fit-out contractor, 6 mm out of plumb over 1,100 mm, which is inside their tolerance and outside ours. We shimmed it and nobody will ever know, but the shim is now part of the load path and I would rather it were not.

The other thing: the client wanted the globe to start on a button. It starts on a button. Nobody presses the button, because the hall is staffed and the staff turn it on at opening. We built a control that was never used, and that is on us for not asking who would be standing there at 9 a.m.