Honeycomb Storage Wall: a continuous grid of hexagonal tiles

You hang the first tile on the wall, screw the second one in beside it, and as you bring them together the edges refuse to meet: there's a step between one and the next, or a gap opens up that shows the plaster behind. Then you try a hook and it goes in crooked, or it seats fine but drops off the moment you load it. The system promised a continuous wall of cells where everything clips to everything, and what you've got is a mosaic full of seams. The fault almost never lies in the design of the system; it lies in how the printer has shifted two dimensions that had to come out exactly right: the edge that sits flush with its neighbour, and the catch that holds the accessory.

Honeycomb Storage Wall (HSW) is a wall system: hexagonal tiles that screw to a wall and interlock without seams, forming a continuous pattern that accessories clip into — hooks, trays, brackets, bins. The whole point is repetition: because the entire surface is the same cell over and over, an accessory designed for the pattern clips in anywhere on the wall. This article is about printing tiles and connections that keep that promise in FDM, where the process works against the two dimensions that hold it together.

Why the hexagon tiles the plane and spreads the load

Only three regular polygons tile the plane with no gaps: the triangle, the square and the hexagon. At a vertex of the hexagonal mesh three cells meet at 120°, and those three angles add up to exactly 360°, so nothing is left over and nothing overlaps. That's the geometric reason the tiles nest: the edge of one is, exactly, the edge of the next.

But of the three, the hexagon makes the best use of material, and that's its advantage over an orthogonal grid. For a given area the hexagon has less perimeter than the square, so it uses less wall per cell. And that wall works harder: in a square grid the walls run in two directions, at 0° and 90°, so a diagonal push has only two paths to spread across. In the hexagonal mesh the walls run in three directions, and a point load — the weight hanging off a hook — spreads across three paths instead of two. Gram for gram, the hexagonal wall responds more evenly whichever way the load comes in, without the pronounced weak direction the square grid has along its diagonal. It's the same principle by which a honeycomb carries the weight of the hive on paper-thin walls of wax.

The practical upshot is that the shared wall between cells can be thin and still strong. In FDM the lower limit isn't set by strength but by extrusion: a wall below two perimeters has no core between its two outer lines and splits open at the slightest load. With a 0.4 mm nozzle, two perimeters come to roughly 0.8–0.9 mm, and that's the floor you don't go under even if the model would let you.

The tile pitch is the dimension you can't scale

The whole system hangs on one number: the pitch of the tessellation, the distance that repeats from one cell to the next. That pitch is what an accessory takes for granted when it looks for somewhere to clip. If your tile comes out at the right pitch, any accessory in the ecosystem drops into place; if the pitch drifts half a millimetre every three cells, the error piles up and by the fifth cell the hook no longer goes in.

And here's the FDM trap: the shrinkage on cooling contracts the whole part in proportion to its size. A large tile, several centimetres across, loses an appreciable fraction of its pitch as it cools. That looks like it should doom the system — the pitch comes out short — but it's saved by one thing: the accessory is printed on the same machine with the same material, so it shrinks in the same proportion. The error is common to both parts and cancels; tile and accessory still meet because both have shrunk by the same amount.

That balance only holds if you don't break the common mode, and you break it in two ways: by rescaling one part "to make it fit", or by printing tile and accessory at different scales. So the pitch has to come out at the nominal scale of the official model — the same one for the whole ecosystem — and it isn't fixed with clearance: clearance sorts out a gap between two parts, not a pitch shrunk across the whole surface. If your machine has a genuine scale error and you need to hit absolute dimensions, you compensate with the slicer's shrinkage factor applied equally to every part, never with the flow multiplier — which tunes line width, not scale — and never by rescaling a single tile.

The accessory joint: send the load along the bead, not across the layers

An accessory joins the tile through a projection — a boss or a hook — that enters the pattern. That projection is the first thing to break if you orient it wrong, because in FDM the part is strong within the layer and weak between layers: the print lines bond well side to side but peel apart under relatively little force. That inter-layer bond is the weak axis.

Orient and print the connection so the load runs along the bead, in the plane of the layer, not across the joint between layers. A hook printed lying down, with the weight pulling perpendicular to the layer lines, delaminates: the load parts two layers along a single line and the hook opens like a book, almost always on the first loading. Printed standing up, with the same weight working in the plane of the layers, the load is carried by the solid material of the bead, not by the seam between beads. It's the same weight, but in one case the part's strong axis carries it and in the other its weak axis does.

The second point is the root of the projection. A boss rising from a base with a sharp internal corner concentrates all the stress on that edge: that's where the crack starts. A fillet at the root spreads that stress over an arc instead of a point, and in FDM it has an added benefit — it avoids the gap between perimeters that a sharp internal corner leaves, that little void where the part begins to crack. You don't need a big radius; it's enough that the right angle stops being one.

Printing tiles that come out flush

Whether two neighbouring tiles sit flush comes down to two geometric factors, and both are decided by the first layer.

The first is flatness. You print the tile resting on its mounting face, and that face is what seats and acts as the reference against the wall and against its neighbour. If the first layer squishes unevenly — bed out of calibration, first-layer height set wrong — the tile comes out with a slight bow, and a warped tile won't seat flat: it rocks on three points and leaves its edges lifted. The edge that was meant to sit flush with the neighbour ends up above or below it. A levelled bed and a uniform first layer aren't a luxury here; they're the condition for a mosaic with no steps.

The second is elephant's foot. The squish of the first few layers spreads the part out at its base, so the bottom edge of the tile comes out thicker than the rest of the flank. You push two tiles together and they touch first at the bottom, on that swollen lip, leaving a gap at the top or riding up over each other. That's why the edge isn't drawn to its exact nominal size against the neighbour: you leave a small edge tolerance, on the order of a tenth or two per side, or a chamfer on the flank, so the tiles meet flush without fighting over that bulging foot. If your slicer has elephant's-foot compensation, turn it on here — but in one place only; don't double it up with the model's tolerance, or you'll end up with clearance to spare and the tiles will rattle.

The same FDM bias that narrows holes governs the catch and the fixing: pockets shrink and projections swell. The accessory's projection printed at nominal size comes out a touch bigger than drawn, and the tile's pocket comes out a touch smaller, so the nominal fit — the one with zero clearance in CAD — binds. The screw holes that fix the tile to the wall shrink the same way, so open those out on purpose too, and orient their countersink so as not to leave a sharp corner where a mounting crack can start. How much you open each hole depends on whether the catch slides, turns or clicks, and the physics behind that bias is in Real printed clearances.

The dimensions that matter and what you can honestly claim

HSW is a community ecosystem, not an industrial standard with a published spec that pins down every dimension. The quantities that govern the fit — the exact tessellation pitch and the dimensions of the catch — are defined by the system's official model, and the honest way to work is to measure them off the published model or consult the system spec, not assume a number. Any pitch or catch figure I gave you here to three decimal places would be misleading precision. What I can give you with a solid basis are the FDM clearances that derive from those dimensions, because they come out of how the process behaves, not out of the system:

Those numbers are a starting point for well-calibrated PLA. In PETG, add 0.05–0.10 mm per side to the moving catches and be wary of press-fit clips, because the material creeps under load and a catch that goes in perfectly can work itself loose within a few days. The only reliable value comes from your own printer with a coupon oriented the way the final part will run.

An HSW tile is, in essence, hexagonal storage taken to the wall: the same cell that in Modular hexagonal storage organises a drawer or a worktop, here set on edge and repeated across a wall. If those tenths of a millimetre of clearance sound like numbers pulled out of thin air, they aren't — where they come from and how to measure your own is in Real printed clearances, and that's the next step before you print a whole wall and discover the step at the fifth cell.