2023

[028] FAN CONTROL VIA DIFFERENTIAL PRESSURE

by

On paper: read a sensor, drive a fan, close the loop. In practice, it’s a lot more interesting than that.

The differential pressure sensor outputs an analog voltage proportional to the pressure difference across its two ports. Picking the right sensor range is already a meaningful decision — too wide and you lose resolution, too narrow and you saturate on the first transient.

The analog front end needed real attention: RC low-pass filter at the ADC input, proper decoupling on the sensor supply. The STM32F030F4P6 — a small Cortex-M0 in a TSSOP-20 package — handles acquisition and the discrete PID. A few details make all the difference: integral anti-windup, a sampling period matched to the system dynamics, filtered derivative. Tuning was done experimentally — P first, then I, then D only when needed. A classic project, but a genuinely instructive one.

[027] REVERSE-ENGINEERED INTERFACE KEY

by

No new features here — just making an existing system more reliable and easier to interface with.

The goal was to design a compact key-like device that fits into existing machines with a very specific mechanical footprint, while providing a clean and repeatable electrical connection.

The housing geometry was entirely derived through reverse engineering: every dimension, alignment point, and insertion behavior had to match the host machine closely enough to guarantee consistent positioning and reliable operation. For the contacts, pogo pins were the natural choice — robust, tolerant of small misalignments, rated for many insertion cycles. An integrated LED gives immediate visual feedback on status.

What’s interesting about this kind of project is that the challenge isn’t inventing something new. It’s designing something new that fits perfectly within the constraints of something old.

[026] CUSTOM SOLID STATE RELAYS

by

Replacing mechanical relays with MOSFET-based SSRs is always appealing. No moving parts, no contact wear, silent switching. But a custom SSR power distribution board is a lot more than a simple switching circuit.

Each channel has a MOSFET running in full conduction. The dominant loss is conduction loss: P = I² × Rds(on). That makes Rds(on) a critical parameter, directly tied to thermal performance. The target was to keep junction temperature rise under +50°C above ambient with passive cooling only — which meant careful heatsink selection, thermal interface material, and mounting quality.

Current sensing is done with Hall-effect sensors: no resistive losses in the power path, natural galvanic isolation. The ATSAMD20 uses those readings for overcurrent protection and RS485 telemetry. One of those projects where every design choice genuinely affects every other one.

[025] THREE-PHASE VOLTAGE SENSOR

by

Measuring multiple voltages on a three-phase system stacks up constraints quickly: high signal levels, mandatory isolation, self-powering from the measured bus, and everything has to fit inside a DIN rail enclosure.

Voltage transformers were chosen over resistive dividers — for safety, galvanic isolation, and signal quality. Outputs are conditioned before hitting the STM32G0B0CET6’s ADC. Analog multiplexing lets several channels share a single measurement path, keeping the hardware compact.

The board powers itself via a buck converter fed from the bus it measures. That sounds straightforward, but switching noise on an analog measurement board is a genuine concern — placement, supply separation, grounding strategy, and filtering were all handled carefully. RS485 telemetry for reporting. The DIN rail enclosure imposed mechanical constraints from day one.