Extruding into the third dimension

You are trying to see a stack of films as a little building, not as overlapping colored rectangles. The screen is a map. The object has height. Extrusion will not tell you whether the gap under a beam is air after a release or insulator that stays. You decide that. Extrusion then gives each layer a thickness and an order, and a cut can show whether a via lands, whether that gap is open, or whether a board hole reaches its plane. The picture at the top is the standard for that question. Thin films lie on a thick carrier. Equal bricks would be a diagram of the layer list, not the chip.

The bricks are only as good as the table

You need three facts per layer: what the color means, how thick the film is or how deep the etch is, and the order of deposit and etch. A technology file or a stackup is where those facts should live. Typing “1 µm for everything” makes a neat toy and a wrong building. Board copper at 35 µm next to a 1.6 mm core should not look like two equal bricks. A gate oxide and a thick MEMS structural film should not share a height because they shared a screen.

The second picture is a cleaved edge, the same story without a render: bright lines in the break, each line a film whose thickness was set in the process, not in the zoom level. A via that looks enclosed in top view and is an island in this cliff wastes a mask. A comb that looks free in top view and is a ridge on the substrate wastes a wafer.

What extrusion will not decide

Boolean leftovers in the top view become threads in 3D. A sliver you ignored becomes a wall. The viewer will not show undercut, sloped resist, or a void in a via unless a real process simulator is in the loop.

The failure mode of skipping the side view is a meeting where two people point at the same rectangle and mean different floors. The failure mode of trusting a pretty extrusion is believing the cartoon. Use the photographs as the standard: thin films in a real cliff, thicknesses from the recipe. Then let the viewer draw those thicknesses, and throw the picture away if a note layer — a pin, a blockage — has been extruded into fake matter.

Cleave a scrap from the same wafer, or take the published cross-section of this exact process, and put it next to the editor. Pick one beam that is supposed to move. On the cleaved edge, point at the bright line directly under that beam. Name the layout layer that line is supposed to be, and read the run sheet for that layer. If the sheet says the film is sacrificed, the gap is air, and the thickness you type is the height of that air. If the viewer still draws the beam sitting on the silicon, the sacrificial thickness is missing or it is stacked in the wrong place. Fix the table and look again.

Then point at any slab in the viewer that you cannot find as a line in the cleave. A pin, a blockage, or a keep-out given a thickness by accident shows up as a floating brick no wafer grew. Delete that thickness. The second photograph, the broken edge, will not show it, and that absence is the correct result. Write that thickness in the units the run sheet uses, and put the layer name in the same row. Then ask which process step the scrap was pulled from. If the line under the beam is still there and the sheet says that film is sacrificed, the cleave was made before release. The finished part still gets a gap of that thickness, and the viewer has to show the gap even though this particular scrap still shows the line. A cleave from after release should show the gap itself. Match the viewer to the step the scrap actually came from, not to a wish that every photograph is the finished part. Count the bright lines from the silicon upward until you reach the line under that beam. The count is the order of the films. If the table has more rows than lines, something you typed is not in this cleave, and it is usually a note that was given a thickness. If the table has fewer rows, a film in the cliff has no layer in the file. Either mismatch is the row you fix before you trust what the viewer draws.

One beam, one thickness from the sheet

Pick the beam that is supposed to move, and read one row of the run sheet: the layer name, the thickness, and whether that film is sacrificed. If the sheet says a 2 micrometer structural film over a sacrificial film, those are two rows, not one brick labeled “MEMS.” Type the sacrificial thickness as the height of the air, and the structural thickness as the beam. If the viewer draws the beam sitting on the silicon, the sacrificial row is missing or it is stacked under the wrong name. The cleaved edge in the second photograph is the check: a bright line under the beam means the film is still there. A gap means it is gone. Match the viewer to the step the scrap came from.

Board copper makes the same mistake in different units. Copper at 35 µm on a 1.6 mm core is a thin film on a thick carrier, which is what the picture at the top is for. Typing 1 µm for every layer turns that core and that foil into equal bricks, and a via through the core becomes as tall as the foil. The cliff will not look like that. Delete the thickness on any pin, blockage, or keep-out. Those rows are notes. A floating brick you cannot find as a line in the break is the row you remove before you show anyone the extrusion.