I’ve lost count of how many times I’ve seen this exact sequence:
Engineer finishes the PCBA. Board powers up. All the LEDs blink the right way. Functional tests pass. Everyone high-fives. Then someone says:
“Great, now just put it in the box.”
That “just” is doing a lot of heavy lifting.
What follows is almost always a cascade of mechanical interference, cable routing nightmares, thermal problems that only show up after the lid is closed, and firmware that suddenly behaves differently once it’s inside a real enclosure. I learned most of these lessons the expensive way. This post is me trying to spare you some of that pain.
The Mental Model Shift
PCB assembly is mostly a 2D problem with some height constraints.
Box build is a 3D systems integration problem that mixes electronics, mechanics, thermals, EMI, and human factors.
A populated board is a component.
A finished box build is a product.
That distinction changes almost every decision you make.
What Actually Goes Into a Real Box Build
In the projects I’ve been involved with, the typical bill of materials expands dramatically once you leave the pure PCBA world:
- Multiple PCBAs (main control + power + interface + display driver…)
- Enclosure (sheet metal, ABS/PC, die-cast aluminum, or custom machined)
- Custom cable harnesses (often the most underestimated item)
- Power conversion (AC-DC brick, open-frame supply, or battery + BMS)
- Thermal solution (heat sinks, thermal pads, heat pipes, or forced air)
- User interface elements (TFT, capacitive touch, membrane keypad, status LEDs)
- External I/O (USB-C, Ethernet, industrial connectors, antenna feedthroughs)
- Mechanical hardware (standoffs, brackets, EMI gaskets, vibration dampers)
- Labels, serial number plates, regulatory marks, packaging
The moment you add more than one board or any significant cable length, the complexity jumps.
The Process I Wish Someone Had Forced Me to Follow
Here’s the sequence that actually works in practice, with the technical gotchas I’ve hit along the way.
1. Documentation & DFM Review (Do Not Skip)
Before any parts are ordered, force a proper design review that includes:
- Full BOM with manufacturer part numbers, approved alternates, and criticality flags
- 3D STEP models of the enclosure + boards + major components
- Cable harness drawings with wire gauges, shield terminations, and bend radius callouts
- Clear definition of test points and acceptance criteria
The most common failure mode I see: the mechanical team designs the enclosure around a board outline, then discovers that the tallest component + heat sink + cable connector no longer fits once the real stack-up is considered.
2. Cable Harness Design — The Silent Killer
This is where most projects bleed time and money.
Technical realities that bite:
- High-current power wires and high-speed differential pairs should never share the same bundle without careful separation.
- Every connector needs a defined strain-relief strategy. Zip ties alone are not a strategy.
- Service loops are not optional if the product will ever be repaired.
- Shield termination location matters for EMI. Terminating at both ends vs one end is a real decision, not a religious war.
I’ve seen harnesses that looked perfect on the bench become intermittent after three temperature cycles because the wire was under constant tension inside the enclosure.
3. Thermal Reality Check
Once the lid closes, your carefully simulated junction temperatures often become fiction.
Things that matter more than people expect:
- Airflow path (or lack of it) when the unit is mounted in its final orientation
- Contact pressure and surface flatness of thermal interface materials
- Hot spots created by tightly bundled cables acting as thermal barriers
- Derating of components once they are inside a sealed or semi-sealed metal box
If you have any power density above a few watts, do a thermal prototype early. CFD is useful, but a real thermocouple run inside a closed enclosure is more honest.
4. EMI and Grounding
Putting electronics inside a metal box does not automatically make EMI better. Sometimes it makes it worse.
Common issues:
- Cable shields that are not properly bonded to the chassis
- Floating metal panels that act as antennas
- Power supply common-mode noise coupling into low-level analog sections
- Display flex cables that become unintentional radiators
Good practice I’ve adopted: treat every metal part as a potential radiator or receptor until proven otherwise, and define a single-point or controlled multi-point grounding scheme early.
5. System-Level Testing Hierarchy
Board-level tests are necessary but nowhere near sufficient. A useful testing ladder looks like this:
- Visual & mechanical inspection (torque, connector seating, cable routing)
- Continuity and isolation (especially after harness installation)
- Power-on and basic current draw
- Full functional test (all interfaces, sensors, actuators)
- Firmware configuration and calibration
- Burn-in / stress (temperature cycling, vibration if relevant)
- Final regulatory / compliance checks
The expensive failures almost always happen between steps 3 and 6.
Practical Tips From the Trenches
- Design the cable harness and the enclosure at the same time. Sequential design is how you end up with “creative” routing.
- Leave intentional service access. Future-you (or the field technician) will thank you.
- Define what “done” means in writing before production starts. Vague acceptance criteria create endless arguments.
- Treat firmware loading and unique ID programming as part of the manufacturing process, not an afterthought.
- If the product will ship in volume, invest in a proper test fixture early. Manual probing does not scale.
When Outsourcing Makes Sense
There are projects where keeping everything in-house is fine. There are also projects where the coordination overhead of managing PCB fab, cable shop, sheet metal vendor, and final assembly becomes its own full-time job.
In those cases I’ve found it valuable to work with a single partner that can take responsibility for the full stack — from PCBA through final system test and packaging. One manufacturer that documents their process quite thoroughly (including the stages I outlined above) is OrinewPCB:
They cover documentation requirements, component integration, multi-stage testing, and the typical failure modes in more detail than most service pages.
I’m still learning on every new box build.
What’s the most painful lesson you’ve learned when moving from a working board to a finished product? Cable routing? Thermal? EMI? Something else?
Drop it in the comments — the war stories are usually more educational than the success stories.
Top comments (0)