Air filter controller

Design decisions

Power supply scheme

The ESP32-S3 uses 3.3V. I want onboard indicator LEDs, and the option to have offboard LEDs as well. Using addressable WS2812-type LEDs is flexible. These need a 5V supply. The fans draw 12V, so this necessitates two voltage regulators.

One option would be to use separate buck converters for both the 5V and 3.3V rails.

With no offboard LEDs, the 5V rail power draw will be low. WS2812B LEDs consume roughly 1mA each when idle. When lit at a fully-saturated color, they will consume about 20mA each. There are four onboard LEDs, so this means 4mA idle, and 80mA when lit.

This is relatively low power consumption. However, offboard LEDs may need significantly more power. If, for example, a 16-LED ring is used, then up to 400mA total may be needed.

During normal operation, the indicator light will be off. So, most of the time, the 5V current draw will be low.

To convert from the 12V supply to 5V, we'll use a buck converter. If we size the inductor for the maximum current draw, then the converter will potentially be noisy during typical operation.

So, we should instead use a linear regulator to convert the 5V to 3.3V, if possible. This will simplify the design, and increase the minimum current draw on the buck converter. Is it feasible?

The ESP32-S3 datasheet says that the power supply should support at least 500mA. The AMS1117-3.3 supports 1A max. The typical power draw of the ESP32-S3 running our ESPHome configuration is 30-120mA. A linear regulator burns the excess voltage as heat. For (worst-case) typical operation, the power consumed by the regulator will be .12A * (5V - 3.3V) = 200mW. This is acceptable.

Using this buck converter calculator, a 100uH inductor means that the maximum ripple current will be about 50mA.

The chosen inductor (SWPA8065S101MT) has a max DC resistance of 0.28Ω, and a max heat-rating current of 1.35A. If the ESP32-S3 draws 500mA, this leaves >500mA for LEDs. This is sufficient.

Filtering

I did a conducted emissions pre-test for v1.0, using a dual 5uH LISN between the board and an adjustable DC power supply. With the board on and idle (30-40mA pulled from the 12V source), I saw a few concerning peaks:

• A sharp peak around 725 kHz, about 10dB above the limit
• Sharp peaks around 3.6kHz, 1.4 MHz, 2.2 MHz, and 3.6 MHz - these are >3dB below the limit
• Broadband noise between 300kHz to 400kHz

Increasing the load on the buck converter slightly by turning on the indicator LED to full white (50-55mA pulled @12V), the peaks get sharper, with the 365 kHz peak exceeding the limit by 6dB, and the 1.4MHz peak exceeding the limit by ~8dB.

There isn't much change with one fan plugged in on full power.