Career

The work does not become less technical. It becomes less bounded.

As a developer, the question is whether the thing I am building is correct. As the person accountable for the whole, correct is necessary and nowhere near sufficient. The question becomes whether it is the right thing at all: whether it can be certified, whether the customer will accept it, whether the engineer who inherits it in four years can maintain it, and whether the decision I am about to take forecloses something we will need later.

Two consequences that took me a while to learn. The first is that most of my expensive decisions happen before there is anything to test, so the feedback loop is measured in years and the only real defense is writing down why. The second is that the work no longer arrives from one direction: my own projects, older products whose problems resurface, and questions from product management and from R&D management that are not attached to any project of mine. Measuring that load and putting a limit on it turned out to be part of the job, and it is the part nobody warns you about.

PPM demodulation of the radio receiver signal, a PID position control loop, electro-hydraulic valve control for the control surfaces, throttle servo control, landing gear deployment, water load and hatches, external digital encoders for position feedback, and remote configuration over RS-232. Plus the PCB design and the complete electrical and data wiring of the aircraft.

Six months, an 8-bit microcontroller, and an aircraft that had to fly. It was also the first PID loop of my career, which draws a straight line from there to the temperature loops I tune now.

DALI and DMX lighting controllers, including a twelve-channel DMX controller. A temperature and humidity probe for matrix displays. A light dimmer. A bootloader and firmware updater. A communication link through light itself. And a test station for the production line.

Three things started there and have followed me ever since: bootloaders, end-of-line test stations, and proprietary communication buses between boards. Fifteen years later I am still doing all three, on considerably larger microcontrollers.

It was a product, not a program, and that is why it still belongs on this page sixteen years later.

The brief was a retro game console and the deliverable was the machine: schematic and PCB, a PIC microcontroller, an LED matrix display, button controls, and the assembler underneath all of it. Two games run on a small base system that handles the display, the controls, and high-score storage, and the base was deliberately designed so that somebody else could add a third game without touching it.

Which is, allowing for sixteen years and a few orders of magnitude, the same argument I make today about a base firmware and a parameter table.

It ended up published twice: the project itself, and a conference paper on introducing retro game console design as a final-year project in the electronics syllabus.

Designing an ABS in 2006 was my first closed loop on a physical system. Measure a wheel speed, compare it against what the vehicle is actually doing, act on a valve, and do all of it fast enough that the physics does not get away from you. A sensor, an actuator, a control law, and a hard real-time constraint.

Twenty years later I tune PID loops on ovens and refrigeration circuits. It is the same problem with a slower plant, and it is why I describe what came next as a change of domain rather than a change of interest.