Workshop notes

I don't model for renderings

I don’t model for renderings

A few years ago I took a job from a designer who had already modeled their product. It looked incredible in the screenshots. Chamfered edges, perfect material textures, ambient shadows. The problem was the file couldn’t be made.

The walls were 1.2mm thick in places. The internal ribs had zero draft. There was an undercut that only a five-axis machine could reach, and the client had a small batch budget. Not five-axis budget. I had to explain that the model was a picture, not a plan. That conversation is why I now lead every project with the same question: “What machine is this going to be cut or printed on?”

Pretty CAD is a trap

There is nothing wrong with a clean render. Clients need to see what they’re buying. But when the model is built to look good first, fabrication becomes an afterthought. Corners get rounded because they look nice, not because a bit can reach them. Walls get thinner because the render reads as elegant, not because the material can hold them.

I work the other way around. I start with the process, then design inside its limits. For plywood furniture, that means asking about sheet size, bit diameter, and whether the shop has a vacuum table or needs tabs. For 3D prints, it means wall thickness, overhang angles, and orientation. For CNC machined parts, it means tool access, clamping surfaces, and whether a right-angle head is available.

This is what design for manufacturing CAD service small batch production actually means to me. It is not a label. It is a workflow that respects the person standing at the machine.

The tolerance budget

Every project has a tolerance budget, even if nobody writes it down. If a cabinet carcass is assembled with biscuits and screws, the joints can tolerate ±0.5mm. If a sliding lid needs to glide, maybe ±0.2mm. If two aluminum parts bolt together with dowel pins, now we are talking ±0.05mm or tighter.

I decide the budget before I draw the first feature. Not after. Because once the model is full of detail, it’s psychologically harder to loosen a fit. It feels like giving up. It’s not. It’s giving the workshop a file they can actually use.

On a recent batch of ten custom brackets, the client wanted a press-fit for a 10mm steel pin into an aluminum hub. I asked what press meant to them. They said “tight.” I made the first holes 9.97mm — that gives a 0.03mm interference, right in the middle of what works for aluminum without galling. I added a note: “9.95mm available if you need a heavier drive fit; ream if needed depending on pin tolerance.” That range saved us from ten parts that might have been too tight or too loose. The shop could adjust on their floor, not guess on mine.

Manufacturing-ready means readable

A manufacturing-ready CAD file for CNC cutting is not just a clean outline. It is a file that answers questions before they are asked. I include:

  • A setup sheet with material, thickness, and bit details.
  • Layers or colors that distinguish cut lines, engraving, and drill points.
  • Tabs where the sheet needs them, marked clearly.
  • Dogbones or T-bones on inside corners, sized for the actual bit.
  • A PDF preview so the operator can see the assembly intent without opening the CAD file.

I also avoid sending STEP files when the job is 2D cutting. STEP is great for machined parts. It is unnecessary overhead for a plywood sheet layout. Match the file format to the process. Sounds obvious, but I still receive IGES files for laser jobs sometimes.

Iteration is not failure

People think design for manufacturing means getting it right the first time. It doesn’t. It means building the model so that when something is wrong, it is easy to change without breaking everything else.

Last winter I designed a small desk with cable routing channels. The first prototype came back from the CNC shop with a channel that was too shallow — the cable connector wouldn’t fit through. The client hadn’t specified the connector diameter. I had assumed a common size. Wrong.

Because the channel depth was driven by a single parameter, I changed it, updated the DXF, and the shop cut the second iteration the next morning. Total delay: one day. If I had modeled the channels as fixed extrusions with hand-drawn fillets everywhere, it would have been a half-day of cleanup for a one-line change.

What I tell clients now

I tell them the render will come, but it comes after the file is manufacturable. Sometimes the render looks slightly less dramatic as a result. A real edge has a radius because the tool has a radius. A real hole has clearance because parts need to assemble. That honesty is the whole point.

If you’re hiring someone to prepare CAD for a workshop, ask them what bit diameter they assumed. Ask them what tolerance they used for press-fits. If they can’t answer quickly, they modeled for the screen. I’ve been that person. I don’t want to be again.