IoT Environmental Monitor
A wall-mounted indoor air quality monitor measuring CO₂, PM2.5, VOC, temperature and humidity, reporting over Wi-Fi with a two-year calibration interval.
- Client
- Aeris Building Systems
- Sector
- Industrial & IIoT
- Year
- 2025
- Duration
- 18 weeks
- Layers
- 6
- Board size
- 78 × 52 mm
- Placements
- 214
- IPC class
- Class 2
Stack-up: 1.6 mm FR-4 Tg150, 1 oz outer / 0.5 oz inner, ENIG
The challenge
The client's previous unit read 2–3 °C high because the temperature sensor sat in the thermal plume of the Wi-Fi module and the switching regulator. Calibration drift meant field re-calibration every nine months, and the sensor fusion could not distinguish a real CO₂ event from a self-heating artefact.
Our approach
We treated self-heating as a layout problem rather than a firmware correction. The temperature and humidity sensors moved onto a thermally isolated tab connected by a 1.2 mm neck with the copper deliberately starved, and slots were milled either side to break the conduction path. The radio and the buck converter were relocated to the opposite end of the board with a ground pour discontinuity between the zones. Firmware duty-cycles the NDIR lamp and takes the temperature reading 400 ms into the sleep window, after the plume has settled.
The outcome
Self-heating error fell from 2.6 °C to 0.4 °C, which removed the need for the correction table entirely. Drift over the first twelve months of field data stayed inside the sensor's own specification, allowing the calibration interval to be extended to two years. The unit passed EN 55032 Class B radiated emissions on the first chamber visit with 6 dB of margin.
Measured outcome
−2.2°C
Self-heating error
9 → 24months
Calibration interval
6dB
EMC margin
2
Prototype revisions
Why self-heating dominated the design
An NDIR CO₂ sensor compensates against temperature, so an error in the temperature reading propagates directly into the gas reading. A 2.6 °C offset was producing roughly 40 ppm of CO₂ error — enough to trigger ventilation in an empty room.
Zoning the board
The six-layer stack-up gave us a continuous ground reference for the radio while allowing a deliberate discontinuity under the sensor tab. Heat travels through copper far more readily than through FR-4, so the neck carries only the four signals the sensor needs, on 0.15 mm tracks, with no pour.
Firmware that respects the physics
Rather than correcting in software, the sampling schedule avoids the problem: the lamp fires, the radio stays quiet, and the temperature sample is taken once the local gradient has settled. The correction table that the previous product depended on was deleted.
From the bench
The sensor tab, isolated by a starved-copper neck and two milled slots. 
Thermal imaging at worst case confirmed the zone separation.
We had spent nine months correcting a sensor error in firmware. Anode found it was a thermal path in the layout, fixed it in one revision, and handed us the measurements that proved it. The calibration interval more than doubled.
Have a board that needs designing?
Send us the constraints — schematic, mechanical envelope, volume, timeline. You will get a considered response from an engineer within one business day, not a brochure.
We reply within one business day, from an engineer.