Bioaerosol Sampler Control System

The problem

The instrument is a virtual impactor — a high-flow-rate bioaerosol sampler that separates particles by size using airflow rather than a physical filter barrier, concentrating the fraction you care about into a smaller sample stream.

That separation depends on the flow rate being what you think it is. A virtual impactor pulling 225 L/min is only cutting at the particle size it’s designed for while it’s actually moving 225 L/min — and over a long outdoor run, filter loading, temperature swings, and supply voltage sag all drag the real flow away from the setpoint.

Set the flow once at the start of a run and you don’t get a slightly worse sample. You get a sample whose size cut drifted, taken at a flow rate you can no longer state.

Where I started

I didn’t start from nothing. There was already a basic controller for the sampler — a DAC driving the motor, with buttons and a screen for setting a speed by hand. What it couldn’t do was know or hold a flow rate.

The original controller rig: Arduino, screen, buttons, and breadboard on a plywood panel, wired to the sampler motor

Everything from there was mine to add. I implemented the flow sensor and a relay for switching the sampler on that breadboard rig first, so the sensing and switching were proven before anything became permanent.

Electronics

With the circuit working, I committed it to a schematic and board layout.

The control circuit schematic

The board was laid out in KiCad — pressure and flow sensor inputs, tach and signal lines for the sampler, a relay output, front-panel buttons, a display header, and separate 9 V and 24 V rails.

The KiCad board layout

Then fabricated and populated: screw terminals for the sensor and tach lines, a relay for switching the sampler, and headers for the controller and I²C bus.

The fabricated and populated control board

Then I wrote the firmware and integrated the pressure sensor, which is what turns a raw reading into a flow rate the loop can act on.

The enclosure came last, once the board and its connectors were fixed — which is the right order. Designing a watertight box around electronics that are still changing means redrawing it every time a connector moves.

Exploded CAD view of the watertight enclosure, control board, and front panel

The control electronics in their enclosure

Characterising the instrument

Before the loop could hold a flow rate, we needed to know how the impactor behaves — how pressure drop across the minor and major flows varies with both the motor flow and the vacuum flow driving the sample stream.

Contour maps of minor and major flow pressure drop against motor flow and vacuum flow

Those maps are what turn a pressure reading into a flow rate, and they define the operating region the controller has to work within.

The control loop

On top of that hardware I implemented and tuned a PID feedback controller that reads live pressure and flow data and continuously corrects the flow rate, rather than trusting an open-loop setting made once at the start of a run.

Tuning is where the work actually is. Too aggressive and the loop chases sensor noise and hunts around the setpoint; too soft and it never catches a slow drift from a loading filter — exactly the disturbance the loop exists to reject.

Result

Step-testing the controller between roughly 140 and 200 L/min, the measured flow tracks the commanded setpoint, settles without sustained hunting, and holds there.

Measured flow tracking commanded setpoint through repeated step changes between 140 and 200 L/min

That plot is the whole point of the project. It means the sampler holds its target flow in real time through changing conditions — which keeps the size cut consistent and the collected samples comparable across a run and between deployments.