Thermal gradients within a test environment introduces artificial humidity measurement errors. For example, as an environmental chamber reaches its target temperature, hot or cold spots may form due to uneven airflow or poor thermal distribution. Even minor differences (for example, a 0.2°C variation between the warmest and coolest points) can result in an apparent RH error of up to 0.8% due to RH's inverse relationship with temperature. Such discrepancies may influence whether a device passes or fails evaluation, even if the sensor itself is operating correctly. This is because the temperature gradient causes the sensor and reference to actually experience different RH conditions, causing mismatched measurement results.

While complete temperature uniformity is ideal in a RH chamber, that is often not realistic. The temperature gradient should not exceed the typical temperature accuracy of the RH sensor under test. For example, if testing the HDC3020, the typical temperature accuracy is ±0.1°C. So the maximum acceptable thermal gradient for a RH chamber would be ±0.1°C between the coolest and warmest points. This will maintain a %RH difference of no more than ±0.4%RH, which is within the ±0.5% typical RH accuracy.

Figure 4-6 shows a real-world example where thermal gradients resulted to RH errors. The error was more pronounced without airflow, and reduced (but not eliminated) when low-speed fans were added to the chamber in order to circulate air and reduce thermal gradients. Red indicates temperature, blue indicates RH.

To minimize thermal gradients:

• Place sensors as close as possible to the RH reference.
• Place sensors as close together as practical within the chamber.
• Use internal fans to improve air circulation and temperature uniformity.
• Limit airflow at sensor level to ≤1 m/s. If higher fan speeds are necessary to achieve mixing, place fans away from the sensors to avoid localized airflow effects.

Figure 4-7 shows a fan setup used to enhance circulation and reduce thermal variation in the chamber.