13 — Molding and Casting

1. How I Will Use Molding and Casting

Both halves of my final project have a rotary encoder on them: the transmitter routed into the guitar body, and the receiver that forwards the MIDI to a computer. Both are still wearing the generic caps that came with the encoders.

I want a matching pair instead, which is what makes casting the right process. One master, one silicone negative, then as many identical knobs as I need.

I would rather have them in something rubberized so they grip under a finger. I am not sure the Fablab stocks a casting material like that, so plastic resin is the fallback and would be fine.

2. Designing the Knob

A short cylinder with a blind bore for the encoder shaft, built in one Part Design body: a single Pad, then two Chamfers and three Fillets so no sharp edge is left on it. Sharp edges are where silicone tears and where resin traps air.

The model is driven from a Spreadsheet, same as the table in Assignment 11, so the knob can be resized without reworking the features.

3. Modelling the Negative Mold

The idea is simple: subtract the knob from a block. What made it slow is that Part Design does not let one body reference another body's geometry.

First attempt was a second body in the same document, sketched against the knob.

The dialog suggests a ShapeBinder, which copies external geometry into the active body. But it brings in a face or an edge, not the solid, so the profiles I derived from it were open. Padding an open profile gives Wire is not closed.

Next was Up to face: pad the block and let it stop against the knob's surface. That is the operation I wanted.

Picking a face from another body fills the field in and then fails on recompute. A feature can only reference geometry inside its own body.

What worked was dropping cross-body references entirely: build the block as its own body (MoldBase), make a Clone of the knob, and subtract it with a Part Design Boolean set to Cut. The Clone is a real solid inside the target body, so nothing reaches outside.

The Clone stays linked to the original, so the cavity is still parametric. Changing the knob updates the cavity in one recompute.

The cavity walls have no draft. That is fine for silicone, which flexes off the master, but a rigid mold milled straight from this block would need a degree or two of taper to release.

4. Fitting the Mold to the Stock

The mold block on its own is not enough to machine. The mill needs the wax stock around it, and the block needs to sit in the middle of that stock. So the last body is Stock Base: a ShapeBinder of the mold's face, a rectangle sketched around it, and a Pad out to the stock thickness.

The rectangle went in at 110 by 88 mm, the footprint of the wax block. Constraining it took a few passes.

To centre the block I ran lines from each side of the rectangle to the corresponding side of the mold face and set the opposite pairs equal. Two equalities, one per axis, and the block is centred without a single hard-coded offset. The first attempt piled on far too many constraints.

Those four lines are geometry the Pad would try to use, so they get toggled to construction. They still hold the sketch together and the Pad ignores them.

Then the Pad, which came out on the wrong side of the sketch plane.

The length itself is not typed in. It is bound to the spreadsheet, same as everything else on this part, so a different block of wax is a cell edit rather than a remodel.

Last step, the same Boolean as before. The operation dropdown defaults to Fuse, which would weld the mold into the stock instead of taking it out.

That is the file the mill gets: a part positioned in a stock of a known size, both driven from the spreadsheet.

5. Milling the Wax Master

The mill is the Fablab's Roland MDX-40, driven from VPanel in NC code mode, with the CAM done in VCarve Pro. I leaned on the Aalto Fablab tutorials on YouTube for the workflow and stopped every few steps to redo the numbers for my own job.

The wax blocks in the lab get reused, so the first session was reclaiming stock: skimming the leftover geometry off the top with a 22 mm end mill until the surface was flat again.

Saving that toolpath taught me the post processor lesson of the week. The Roland MDX-40 post writes a .rol file, and the VPanel on the lab machine runs in NC code mode, which does not take it. The combination that works is the USBCNC (mm) post, which writes the .nc the panel expects. Both versions are in the files below because I only found out by trying.

Then the mold itself. The Boolean result from section 4 went out of FreeCAD as an STL and into VCarve through Import 3D Model, sunk 27.52 mm below the top of the material so the cavity sits at the right depth in the block.

The cutter is a 2 mm ball nose. I set it up in the tool database with the numbers worked out from the tutorials: 14000 rpm, 1120 mm/min feed, 280 mm/min plunge, 0.8 mm pass depth, 0.2 mm stepover. Writing them here so I do not have to derive them again.

Two toolpaths over the model: a Z-level roughing pass leaving a 0.2 mm skin, then an offset finishing pass to take the skin off. The knob is small and there is not much wax around the cavity, so the walls of the future silicone mold are thinner than I would like. I saw that in the preview and went with it anyway. Redoing the model at that point was a worse deal than living with thin walls on a first iteration.

The first attempt went wrong in a way the preview cannot show. The spindle was turning too fast for wax, so instead of clearing as chips the material heated up, smeared and welded itself back onto the walls. The rim of the pocket tore instead of cutting.

For the second attempt I dropped the spindle override in VPanel to 20 to 30 percent for the wax. Even 30 percent felt harsh once the cutter was working at the bottom of the pocket. The other change was babysitting the job: instead of letting it run and melt, I kept quitting the cut, cleaning the packed wax out of the cavity, and restarting. The same shape took far longer the second time, but the edges stayed edges.

The first cavity is scrap. The second one is a bit rough around the edges but the knob, its bore and the fillets all read clearly in the wax. For a first iteration of modelling, milling and casting on my own, I am counting it as a win.

6. Pouring the Silicone Mold

The wax block is the positive. Silicone poured over it becomes the flexible negative that the actual knobs get cast in later. In the casting room I looked for a kit that was already open before breaking the seal on a new one, and found SORTA-Clear 37, a platinum silicone that happens to be food safe. Less than half of it was left and the material had gone noticeably stiff in the bottle. I am not sure how much that matters, but it did still mix.

The ratio is 1:1 by volume. I measured the parts into separate cups, combined them and stirred. I misjudged how much the cavity needed and mixed roughly double, which is the note to self for next time.

The cup went into the vacuum chamber to degas before pouring.

After twenty minutes there were still bubbles coming up, probably helped by how stiff the material was to begin with. I had to leave, so it got poured as it was.

Demolding happens tomorrow. If the bubbles ruined it, the redo is Mold Star 30: the mold itself has no reason to be food safe, and a second pass through the same process with fresh material and a proper degas should come out better. The knobs themselves come after that, and picking their resin is still open.

7. Demolding and Trimming

Next day. The silicone had set overnight, firm under a finger, so the bubbles did not ruin it and the Mold Star redo stays on the shelf.

Both cavities released without a fight. Silicone flexes and wax does not grab, so each mold came out whole, wearing a thin skirt of flash where the overpour had spread across the top of the block.

The detail transfer is the part that sold me on the process. The tool marks from the milling read on the silicone walls, and a few crumbs of purple wax came along embedded in the surface. The air I never got out in the degas shows as pinprick bubbles in the flash and speckled through the body of each block, but the cavity surfaces came out clean enough to cast in.

Then the trim: the flash cut away until both blocks stood flat on their own.

8. Casting with Smooth-Cast 305

For the knobs themselves the casting room had Smooth-Cast 305, a liquid plastic that cures white: 1A:1B by volume, seven minutes of pot life, thirty minutes to demold. That is the plastic fallback from section 1, so these knobs come out hard rather than rubberized.

The kit fought back before it opened. The part A container was glued shut where the material had hardened around the lid on its own, and it took a pair of scissors to cut through the crust and crack it open.

Equal volumes into two cups, then both into one and stirred.

The pour went into the mold that came from the better of the two cavities.

There was resin left over, and with seven minutes of pot life it was use it or bin it. Other molds from other people's projects were lying around the room, and some of the shapes looked interesting, so the leftover got shared out across those instead of the bin.

Thirty minutes later the pour had gone from liquid to solid white, in the knob mold and in every found mold that got a share of the leftover.

The knob pulled straight out of the silicone. The milling ridges from the wax made it through the silicone and into the plastic walls. The open face cured rough where the remaining bubbles collected, since that side of the pour sat against air instead of silicone. Whether the bore underneath seats on the encoder shaft is the next thing to test.

9. Files

Download Knob.crv, the VCarve job for the mold with the roughing and finishing toolpaths.

Download FacingOperation.crv, Facing.rol and Facing.nc, the first facing job. The .rol is the export that VPanel's NC mode refused, kept as a reminder.

Download FacingWaxOperation.crv and FacingWax.nc, the facing job for the block the mold went into.

Download 3D Roughing 1.nc and 3D Finish 1.nc, the exported roughing and finishing passes for the mold cavity.