Start here: haitch.diy
Most sensor enclosures start the same way. You know roughly what goes inside. You have a board dimension, a sensor window, and some mounting idea. But instead of moving straight into geometry, you open a CAD tool, stare at the empty canvas, and spend the first hour fighting the tool instead of the design.
The problem is not the CAD step. The problem is skipping the concept step.
This walkthrough covers a single working session, roughly 90 minutes to 2 hours, where you go from a rough sensor enclosure idea to an export-ready STL. No multi-day pipeline. No switching between three different apps. One session, one connected context, one file at the end.
You can build a sensor enclosure from scratch and export a printable STL in one session if you split the work into two phases:
1. Canvas first (20-30 minutes): nail down the form factor, sensor placement, mounting strategy, and critical dimensions before touching geometry.
2. CAD second (60-90 minutes): generate the shell, add features, refine tolerances, and export.
Skipping Canvas and jumping straight to CAD is how you end up rebuilding the enclosure three times because you forgot the cable routing or picked the wrong split line.
If you are designing injection-molded production enclosures with complex draft angles and snap fits, this walkthrough covers the concept-to-prototype phase. Production tooling is a separate step.
The goal of the Canvas phase is not to draw a pretty enclosure. It is to make the decisions that prevent rework in CAD.
Step 1: Drop your constraints (5 minutes)
Open a Canvas workspace and lay out the hard constraints:
Use text blocks, not drawings. You are making a decision log, not a sketch.
Step 2: Pick a form factor (10 minutes)
Decide the enclosure shape and split strategy:
| Form factor | Best for | Split strategy |
|-------------|----------|----------------|
| Rectangular box | Most sensor boards | Top/bottom shell with screw bosses |
| Cylindrical tube | Pipe-mount sensors | Two halves with alignment pins |
| Flat puck | Desk or wall sensors | Base plate + snap-on lid |
| Clip-on saddle | Retrofit sensors | One-piece with flex living hinge |
Pick one. Write down why. Add the target wall thickness (1.6 mm for FDM, 1.2 mm for SLA).
Step 3: Mark the critical features (10 minutes)
On the Canvas, call out:
By the end of Canvas, you should have a one-page decision sheet that answers every question CAD will ask you.
Now you move into CAD with a clear brief. No guessing.
Step 4: Generate the base shell (15 minutes)
Describe the enclosure to the CAD workspace using the constraints from Canvas:
> "Rectangular enclosure, 80 x 50 x 25 mm internal cavity. 1.6 mm wall thickness. Top/bottom split at the midline. Four M3 screw bosses at the corners, 5 mm diameter, 8 mm tall. 2 mm fillet on all external edges."
The AI generates the initial geometry. Review the output:
Fix any issues with follow-up prompts before moving on.
Step 5: Add the sensor window (10 minutes)
Cut the sensor aperture into the correct face:
> "Add a 15 x 10 mm rectangular cutout on the top face, centered, offset 12 mm from the front edge. Add a 0.5 mm lip around the cutout for a gasket or press-fit lens."
Check that the cutout clears the sensor module and any adjacent components on the board.
Step 6: Add cable routing and ventilation (10 minutes)
Step 7: Add mounting features (10 minutes)
Based on Canvas decisions:
| Mount type | CAD feature |
|------------|-------------|
| Screw mount | Through-holes or keyhole slots on a flange |
| Magnet mount | Recessed pockets sized for disc magnets |
| DIN rail clip | Rail profile extruded on the back face |
| Adhesive pad | Flat recessed area on the bottom |
Step 8: Refine and check (15 minutes)
Run through the checklist:
Step 9: Export STL (5 minutes)
Export the top and bottom shells as separate STL files. Name them clearly:
If you need STEP for a colleague or manufacturer, export that too. Both formats export from the same geometry in Haitch.
Before starting, have these ready:
| Printer type | Min wall | Min hole | Max overhang |
|-------------|----------|----------|--------------|
| FDM (PLA/PETG) | 1.2 mm | 2.0 mm | 45 degrees |
| SLA (resin) | 0.8 mm | 1.0 mm | Supports needed |
| SLS (nylon) | 0.7 mm | 1.5 mm | No limit |
Starting in CAD without a Canvas brief. You end up guessing dimensions, picking arbitrary wall thickness, and rebuilding after the board does not fit. Canvas takes 20 minutes. Rebuilding takes 60.
Forgetting the cable exit. The enclosure looks perfect until you realize there is no way to route the USB cable. Always mark cable exits in Canvas before generating geometry.
Ignoring component height. Board dimensions are not the same as board-plus-component dimensions. A BME280 sensor breakout is 2.5 mm tall, but the pin header underneath adds 8.5 mm. Measure the total stack height.
Over-engineering the first version. The goal is a working enclosure for testing, not a production part. Skip snap fits, gaskets, and complex draft angles on the first iteration. Get the fit right, then refine.
The Canvas-to-CAD flow in Haitch keeps concept decisions and geometry in one connected workspace. The constraints you define in Canvas carry context into CAD, so the AI generates geometry that already accounts for your board dimensions, sensor window, and mounting strategy.
You do not need to re-explain the project every time you refine the geometry. The context survives across the session.
For teams evaluating how AI CAD fits into a physical product workflow, see the AI CAD Software for Physical Product Teams comparison. For a broader view of how connected workflows compare to fragmented tool stacks, read Haitch vs Traditional Hardware Tool Stack.
How long does this actually take?
90 minutes to 2 hours for a straightforward rectangular enclosure. Cylindrical or complex organic shapes may take longer because the geometry requires more iteration. The Canvas phase is always 20-30 minutes regardless of complexity.
What if I do not have exact board dimensions?
Measure the board with calipers. If you only have a rough idea, add 2-3 mm clearance on each side and plan to refine after a test print. Getting close is better than waiting for a perfect datasheet.
Can I use this workflow for multi-board enclosures?
Yes, but spend more time in Canvas. Multi-board enclosures need a stacking plan, inter-board clearance, and connector alignment before you touch CAD. The Canvas phase might take 45 minutes instead of 20.
What if the first print does not fit?
That is normal for a first prototype. Measure where it fails, adjust the CAD parameters, and re-export. Because the geometry is parametric, adjusting one dimension does not require rebuilding the entire enclosure.
Spend the first 20 minutes in Canvas making decisions. Spend the next 90 minutes in CAD executing those decisions. Export the STL by the end of the session and send it to the printer. The enclosure will not be perfect, but it will be real, and you can iterate from there.
Start building: haitch.diy