Start here: haitch.diy
Most hardware teams do not fail at PCB design. They fail at everything that happens before PCB design.
Someone has an idea for a connected sensor product. They sketch something on paper, open KiCad, start placing components, realize they have not decided on a microcontroller, switch to a spreadsheet to compare options, lose 4 hours, come back to KiCad, realize the board shape does not fit the enclosure they imagined, and start over.
The PCB is the last 20% of the work. The first 80% is concept exploration, system architecture, component selection, and interface planning. If those happen in disconnected tools, the PCB phase inherits every unresolved decision as rework.
This teardown covers a real 48-hour sprint for a connected environmental sensor product, from rough concept to order-ready outputs, showing exactly what happens in each phase and where most teams lose time.
A 48-hour concept-to-PCB sprint works when you front-load decisions in the right order:
| Phase | Hours | Workspace | Deliverable |
|-------|-------|-----------|-------------|
| Concept exploration | 0-3 | Canvas | Product brief, form factor, user scenario |
| System architecture | 3-8 | System | Block diagram, interface specs, component shortlist |
| Schematic + layout | 8-30 | Electronics | Schematic capture, board layout, DRC-clean design |
| Enclosure | 30-40 | CAD | Housing geometry matched to board outline |
| Integration check | 40-44 | System | Cross-workspace validation, BOM review |
| Export + order prep | 44-48 | All | Gerbers, BOM CSV, STL files, order placed |
The common mistake is spending hours 0-30 on schematic and layout, then discovering at hour 35 that the board does not fit the enclosure, the power budget does not work, or a critical sensor needs an interface you did not plan for.
This teardown assumes basic familiarity with electronic circuits. You do not need to be an EE, but you should understand what a microcontroller, sensor, voltage regulator, and connector are.
For this teardown, the product is a compact indoor environmental sensor that measures temperature, humidity, CO2, and particulate matter. It connects via Wi-Fi, runs on USB-C power, and fits in a small desk-mount enclosure.
This is a representative product that touches every typical hardware discipline: sensing, power, wireless, firmware, and mechanical housing.
Goal: Define what the product does, who it is for, and what the physical form looks like.
Open a Canvas workspace and answer these questions:
Drop reference photos of similar products. Mark what you like and what you would change. This is the product brief that every downstream decision refers back to.
Deliverable: One-page product brief with form factor, sensor list, connectivity, power, and size target.
Goal: Turn the product brief into a system block diagram with specific components and interfaces.
Move into the System workspace and build the architecture:
Block diagram:
Component selection decisions:
| Function | Component | Interface | Why |
|----------|-----------|-----------|-----|
| MCU | ESP32-C3-MINI-1 | — | Wi-Fi built in, low cost, small module |
| Temp/humidity | BME280 | I2C (0x76) | Proven, small, fast response |
| CO2 | SCD41 | I2C (0x62) | True NDIR sensor, no burn-in needed |
| Particulate | PMS5003 | UART (9600 baud) | Industry standard, known data format |
| LED | WS2812B | GPIO (data pin) | Single-wire, RGB, compact |
| Power | AP2112K-3.3 | — | 600 mA LDO, SOT-23-5, low dropout |
| Connector | USB-C (power only) | — | 2-pin power, no data lines needed |
Interface budget:
Power budget:
| Component | Active current | Sleep current |
|-----------|---------------|---------------|
| ESP32-C3 | 80 mA (Wi-Fi TX) | 5 uA |
| BME280 | 3.6 mA (measuring) | 0.1 uA |
| SCD41 | 75 mA (measuring) | 1 uA |
| PMS5003 | 100 mA (fan running) | < 1 mA |
| WS2812B | 20 mA (single color) | 0 |
| AP2112K | Quiescent 55 uA | 55 uA |
| Total active | ~280 mA | — |
280 mA total active draw is well within USB-C 500 mA budget. No battery management needed.
Deliverable: Block diagram, component table, interface allocation, power budget.
Goal: Capture the schematic, place components, route traces, and pass DRC.
This is the longest phase, but it moves faster because every decision about components, interfaces, and power was already made in System.
Schematic capture (hours 8-16):
Work through the schematic in functional blocks:
1. Power input: USB-C connector → AP2112K LDO → 3.3V rail with bulk and decoupling caps
2. MCU: ESP32-C3-MINI-1 with decoupling caps, boot/reset buttons (for development), antenna keep-out
3. Sensors: BME280 on I2C with pull-ups (4.7k), SCD41 on same I2C bus, PMS5003 on UART with enable FET
4. LED: WS2812B with 100 nF decoupling and series resistor on data line
5. Connectors: Programming header (UART), optional debug test points
Board layout (hours 16-28):
DRC and review (hours 28-30):
Deliverable: DRC-clean schematic and board layout, ready for Gerber export.
Goal: Build an enclosure that fits the board outline, exposes sensors to air, and mounts cleanly.
Move into the CAD workspace. The board outline and mounting hole positions carry over from the electronics phase, so you do not need to re-measure anything.
Enclosure requirements from the product brief:
Generate the base shell, add features, and export STL for 3D printing. Follow the same Canvas → CAD flow covered in Build a Sensor Enclosure from Concept to STL in One Session.
Deliverable: Top and bottom shell STLs, ready for print.
Goal: Verify that all outputs are consistent and nothing was lost in handoffs.
Go back to the System workspace and run through:
Fix any mismatches now. A 30-minute correction here saves a week of waiting for a wrong board to arrive.
Goal: Generate all manufacturing outputs and place orders.
| Output | Format | Destination |
|--------|--------|-------------|
| PCB Gerbers | .zip (Gerber 274X) | JLCPCB, PCBWay, OSH Park |
| BOM | .csv | LCSC, Digi-Key, Mouser |
| Pick and place | .csv | Assembly house (if using SMT service) |
| Enclosure top | .stl | 3D printer or print service |
| Enclosure bottom | .stl | 3D printer or print service |
| Enclosure STEP | .step | CNC or injection mold (future) |
Place the PCB order. Print the enclosure. Order components. While you wait for boards to arrive, you can start firmware in the Firmware workspace using the same system context.
Deliverable: All orders placed, firmware development started.
Before starting the 48-hour sprint, have these ready:
Time commitment: 48 hours of focused work, not elapsed time. This can be two intense days, a long weekend, or a week of evening sessions. The workflow is the same regardless of pacing.
Jumping straight to schematic without a system architecture. You will pick a microcontroller, then discover it does not have enough I2C buses, UART peripherals, or GPIO pins for your sensor set. The System phase takes 5 hours. Re-spinning a PCB takes 2 weeks and $50.
Designing the PCB before deciding on the enclosure. The board outline, mounting holes, and connector positions all depend on the enclosure form factor. If you finalize the PCB first, the enclosure either does not fit or requires an awkward compromise. Decide the enclosure form in Canvas, size the board to match, then do detailed layout.
Skipping the integration check. The most common failure is a board that does not fit its enclosure because one team member moved a connector during routing and nobody re-checked the mechanical alignment. The 4-hour integration check at hour 40 catches these before you spend money.
Over-scoping the first board. Your first custom PCB should validate the core sensing and connectivity. Do not add displays, batteries, Bluetooth, and cellular to the first version. Each feature adds layout complexity, power budget risk, and debug surface area. Ship the simple board, confirm the fundamentals work, then add features in revision 2.
The Canvas → System → Electronics → CAD flow in Haitch keeps every phase in one connected workspace. The product brief from Canvas informs the system architecture. The component decisions from System carry into the schematic. The board outline from Electronics feeds directly into the CAD enclosure. No copy-pasting between tools, no outdated spreadsheets, no "wait, which version of the board outline are we using?" conversations.
For teams comparing this approach to a traditional fragmented tool stack, see Best Hardware Product Development Software for Startups. For a deeper look at how AI CAD fits into the enclosure phase, read AI CAD Software for Physical Product Teams.
Do I really need all four workspaces?
For a product that combines electronics and mechanical housing, yes. Skipping Canvas means guessing at the form factor. Skipping System means guessing at the architecture. You can skip the CAD phase if you do not need an enclosure yet, but most teams want at least a rough housing for testing.
What if I am not using an ESP32?
The workflow is the same for any microcontroller. The System phase is where you select the MCU based on your interface, connectivity, and power requirements. STM32, nRF52, RP2040, ATmega, or any other platform follows the same concept → architecture → schematic → layout → enclosure sequence.
Can I do this solo?
Yes. The 48-hour timeline assumes one person doing all four phases. A two-person team (one electronics, one mechanical) can run the electronics and CAD phases in parallel after the System phase, potentially finishing in 30 hours.
What if I need to change a component after I start layout?
Go back to the System workspace, update the component and interface allocation, then propagate the change forward. Because the system architecture is the source of truth, you only update it in one place instead of hunting through multiple files.
How does this compare to using KiCad + FreeCAD + spreadsheets?
The individual tools are capable. The cost is in the handoffs. Every time you move between KiCad, FreeCAD, a spreadsheet BOM, and a separate project document, you lose context and risk version drift. The 48-hour timeline is achievable with separate tools, but the integration check at hour 40 becomes much harder because there is no shared source of truth. See Haitch vs Traditional Hardware Tool Stack for a detailed comparison.
Front-load decisions. Spend 8 hours on concept and architecture before touching a schematic. The PCB phase moves twice as fast when every component, interface, and mechanical constraint is already decided. Export Gerbers by hour 48, place the order, and start firmware while you wait for boards.
Start building: haitch.diy