A PCB design mistake caught on the breadboard costs nothing. The same mistake caught after fabrication costs a respin — typically one to three weeks and several thousand rupees, sometimes more if the board has already been assembled. Caught after the product has shipped, it can mean a recall.
Most PCB failures in IoT products trace back to a small, repeating set of mistakes. None of them are exotic — they're well-documented, well-understood, and almost always avoidable with a careful design review before sending files to fabrication. If you're earlier in the process, our guide to IoT product prototyping in India covers the full journey from concept to working unit.
Why PCB mistakes are so expensive
Unlike software, a hardware bug isn't fixed by pushing an update. A PCB design error usually means: redesign the board, re-fabricate it (one to three weeks depending on the manufacturer), re-assemble it, and re-test the whole system. If the error wasn't caught until after units were already manufactured in volume, the cost multiplies by however many units are sitting in inventory.
This is why a disciplined design review — checking for these specific mistakes before submitting Gerber files — is one of the highest-leverage steps in the entire hardware development process.
Mistake 1 — Skipping decoupling capacitors
This is the single most common mistake in first-time PCB designs, and the one with the most confusing symptoms when it goes wrong.
Every integrated circuit on a board draws current in sudden, sharp bursts as it switches states internally. If the power supply can't respond to these bursts instantly — and physically, due to trace inductance, it can't — the local voltage at the chip's power pin dips momentarily. This is called a power supply transient, and it can cause anything from random resets, to corrupted data on a communication bus, to a microcontroller that simply refuses to boot reliably.
The fix: place a decoupling capacitor — typically 100nF ceramic — as close as physically possible to every power pin of every IC, with a short, low-inductance trace directly to the pin. For ICs with higher current demands, add a larger bulk capacitor (often 10µF or more) nearby as well, to handle slower variations in current draw that the small capacitor can't fully absorb.
The symptom that points directly to this mistake: a circuit that works fine on the breadboard with long, messy jumper wires, but behaves erratically once it's on a tightly laid-out PCB. Counter-intuitively, the breadboard's parasitic inductance from long wires sometimes accidentally provides enough filtering that the PCB — built correctly in every other respect but missing decoupling — does not.
Mistake 2 — Poor ground plane design
A solid, continuous ground plane is one of the simplest things to get right and one of the most common things to get wrong — usually by accident, through routing decisions made without realising their effect on the ground plane underneath.
Common ways a ground plane gets accidentally broken: routing a trace across a layer in a way that splits the ground plane into two isolated islands, placing too many vias in a tight cluster that perforates the plane, or routing a high-speed signal trace across a gap in the ground plane beneath it — which forces the return current to take a long detour, turning the trace into an unintentional antenna.
The fix: use a dedicated ground layer wherever the board has more than two layers, and on two-layer boards, pour ground generously on both layers and stitch them together with vias at regular intervals. Before finalising the layout, visually inspect the ground plane for unintended splits — most PCB design tools can highlight plane continuity, and it takes a few minutes to check.
The downstream symptoms of a broken ground plane are often subtle: increased electromagnetic interference, noisy analog readings, or communication errors on high-speed buses that only appear intermittently — exactly the kind of bug that's expensive to debug after the fact because nothing is obviously wrong with the schematic.
Mistake 3 — Ignoring trace width and current capacity
Every PCB trace has a maximum current it can carry before it heats up excessively — and at high enough current, before it can be damaged permanently. This is a function of trace width, copper thickness, and how much the trace is allowed to heat up.
This mistake shows up most often on power traces feeding motors, LEDs in quantity, charging circuits, or any component that draws meaningfully more current than typical microcontroller logic. A trace sized fine for 100mA can become a real problem at 2A — heating up, causing voltage drop along its length, and in extreme cases, lifting off the board entirely.
The fix: for any trace carrying significant current, calculate the required width using a trace width calculator (widely available as free online tools, based on IPC-2221 standards) rather than guessing. As a rough starting reference for a 1oz copper outer layer at room temperature, a 0.5mm trace handles roughly 0.5–1A depending on allowed temperature rise — but always verify with a proper calculation rather than relying on a rule of thumb, since the relationship isn't linear and depends on several variables.
For higher current paths, consider using thicker copper (2oz instead of the default 1oz), wider traces, or copper pours instead of traces entirely.
Mistake 4 — No test points or debug access
This isn't a mistake that causes the board to fail — it's a mistake that makes every other failure dramatically harder to diagnose.
A board with no exposed test points on key signals — power rails, communication buses, reset lines, critical sensor outputs — forces debugging to happen by probing directly on tiny IC pins or via pads, which is slow, error-prone, and can damage components if a probe slips.
The fix: add small, clearly labelled test points on every signal that's likely to need probing during bring-up — every power rail, every communication bus (I2C, SPI, UART), the reset line, and any signal that's historically been a source of problems in similar designs. These cost almost nothing in board space and save hours during the bring-up and debugging phase, which is exactly when time pressure is highest.
Equally important: include a proper programming and debug header — JTAG, SWD, or whatever your microcontroller requires — accessible without disassembling the enclosure. A board that requires removing screws and connectors just to reflash firmware during development slows down every single iteration cycle.
Mistake 5 — Designing for the lab, not manufacture
A board that works perfectly as a one-off prototype can still fail to be manufacturable at scale, for reasons that have nothing to do with the circuit itself.
Common manufacturability mistakes: components placed too close together for automated pick-and-place machines to assemble reliably, inconsistent component orientation that increases assembly errors, missing or incorrect silkscreen markings that make manual inspection difficult, and components selected that are in low stock or single-sourced, creating a supply chain risk the moment volume production begins.
The fix: run a Design for Manufacture (DFM) check with your chosen fabricator before finalising the design — most PCB manufacturers, including Indian fabricators like PCBPower, offer this as a service or at minimum a basic automated check. Keep consistent component orientation where possible. Verify component availability and lead times for every part in the Bill of Materials, not just the exotic ones — generic passives go out of stock too, especially during supply chain disruptions.
Two more worth knowing — thermal and connector mistakes
Thermal management. Components that dissipate meaningful heat — voltage regulators, power MOSFETs, high-current drivers — need a thermal path off the board, whether through copper pour acting as a heatsink, thermal vias connecting to an internal ground plane, or in more demanding cases, an actual heatsink. A regulator that works fine on an open bench can overheat and shut down once it's inside a sealed enclosure with no airflow — a mismatch between lab conditions and real deployment conditions that's easy to miss.
Connector and mounting mistakes. Connectors placed too close to board edges or enclosure walls, mounting holes that don't match the actual enclosure (a surprisingly common and entirely preventable mistake from designing the PCB and the enclosure on separate timelines), and connectors oriented in a way that makes cable routing awkward inside the final product. These aren't electrical mistakes, but they cause real delays — usually discovered at the worst possible moment, when the first enclosure prototype arrives and the board doesn't fit.
A pre-fabrication checklist
| Check | What to verify |
|---|---|
| Decoupling capacitors | 100nF ceramic on every power pin, placed as close as physically possible |
| Ground plane continuity | No unintended splits; high-speed traces don't cross plane gaps |
| Trace widths | Calculated against actual current requirements, not guessed |
| Test points | Power rails, comms buses, reset line all accessible for probing |
| Debug header | Accessible without disassembling the enclosure |
| DFM check | Run with the actual fabricator before finalising the design |
| BOM availability | Every component verified in stock, including generic passives |
| Thermal path | Heat-generating components have a path to dissipate heat inside an enclosure |
| Enclosure fit | Mounting holes and connector positions verified against actual enclosure CAD |
None of these mistakes require advanced expertise to avoid — they require a deliberate review pass before submitting files to fabrication, checking specifically for each of these categories rather than just confirming the circuit "looks right." The cost of that review is an afternoon. The cost of skipping it is usually a respin, sometimes a redesign, and occasionally a much harder conversation with whoever is waiting for the product to ship.
If you're working through a PCB design for an IoT product and want a second pair of eyes on the layout, get in touch with us. PCB design and embedded systems are core parts of what we build at Manthrix.