The part that arrived in a ziplock bag
The part that arrived in a ziplock bag
A fabricator sent me a small metal bracket in a ziplock bag. No drawing, no CAD, just a worn part and a note: “We need fifty of these for next month. Can you make STEP files?” That is a common reason people look for someone to create STEP files from my sketches or physical parts.
The bracket was aluminum, roughly 85 mm by 40 mm, with three holes and a curved pocket. It had been drilled, tapped, and milled. Nothing exotic, but the worn surfaces made every measurement a decision. Do I model the part as it is now, with scratches and rounded edges? Or do I model what it was supposed to be when it was new?
Capture first, interpret second
I started with calipers and a micrometer. Calipers are fast for overall dimensions, good to about ±0.05 mm. The micrometer confirmed thickness: 8.02 mm. The calipers said 8.0 mm. That 0.02 mm difference is the kind of thing you would miss if you trusted only one tool.
For the curved pocket, calipers are useless. I used a structured-light 3D scanner, which is typical for this kind of work. Scanner accuracy is usually around ±0.02 to ±0.1 mm, but the final CAD accuracy depends on how well you clean the scan and interpret the surfaces. I treat scan data as a guide, not as the final model.
Design intent vs as-built
The three holes were the most important features. Two were clearance holes for M6 bolts. One was a threaded M8 hole. The thread pitch was 1.25 mm, standard coarse. I measured the center distances with calipers, then modeled the hole pattern as symmetric because the part was clearly meant to be symmetric. The actual worn part was not perfectly symmetric — drilling had shifted one hole by about 0.3 mm — but a fabricator wants the intended pattern, not the accumulated wear.
That is the central choice in reverse engineering. You are not copying the old part exactly. You are reconstructing the design intent. Rounds go back to nominal radii. Hole centers go back to symmetry. Tapped holes stay threaded. The client does not want a model of a worn bracket. They want a bracket that can be made again.
Building the STEP file
I rebuilt the part as a solid model in CAD. Extruded base plate, cut the pocket, drilled the clearance holes, added the threaded hole, applied edge breaks. Then I exported as STEP AP214, which is the safest flavor for general machining. STEP carries mathematically defined solid geometry, so the fabricator gets real faces and edges, not a faceted approximation like STL.
I also added a 2D drawing with the critical dimensions and tolerances. The drawing said the M6 clearance holes were 6.6 mm, which is the standard normal clearance fit. The overall length was toleranced at ±0.2 mm because the bracket mates with stamped sheet metal, not precision ground parts. The M8 tapped hole got a note: tap to depth, 12 mm minimum.
What almost went wrong
The pocket had a radius blend in one corner that looked decorative. I almost ignored it. Then I noticed the matching part in the assembly had a protrusion that nested into that blend. If I had left it as a sharp corner, the new brackets would have interfered. The scan saved me there. The calipers never would have caught it.
I also had to ask the client about the material. The old part was aluminum, but they did not know the alloy. 6061-T6 is the default guess for machined aluminum brackets, and it machines well. I noted that assumption on the drawing. If they had wanted steel, the holes and taps would have stayed the same size, but the manufacturing note would change.
Delivery
The final delivery was a STEP file, a PDF drawing, and a short note explaining the design-intent assumptions. The fabricator loaded the STEP into their CAM system, programmed the toolpaths, and cut the first sample in a day. The bracket fit. That is the whole point of custom CAD design service for CNC cutting — turning a bagged sample into something a machine can repeat.
Reverse engineering is not magic. It is just careful measurement plus the decision to model what the part was supposed to be, not what time made it.
[IMAGE: A dark technical flat-lay showing a worn aluminum bracket on a graph-paper surface next to a caliper, a 3D scanner, and a ghosted CAD wireframe overlay of the reconstructed part. Amber highlight on critical hole centers and edge breaks. Small STEP file icon in the corner. Clean, premium workshop aesthetic. No text. 3:2 aspect ratio.]