Research / Controlled sniffing
Research pillar 02

Moving air through the system in a repeatable way.

A machine cannot smell well if air reaches the sensor randomly. It needs a controlled breath — a repeatable path from sample to chamber and back to a clean baseline.

What we research

Airflow, chamber shape, pump timing, clean-air purging, and recovery time often decide whether a smell reading is useful at all. Aeralyte treats sampling, sensing, and AI as one system — and puts more early effort into active sampling than into exotic sensors.

The path

Clean-air baselineEstablish a reference before each exposure.
Valve switchSelect clean path or sample environment.
Pump + flow controlDraw the sample at a controlled rate.
Sensor chamberExpose the array; record the response.
Purge & recoveryReturn to baseline; check repeatability.

Where this stands

The loop has run for realBench · recorded runs

Every run records its sampling context — room volume, ventilation, pump flow, sample period — alongside the sensor stream, and the analysis is built so context can explain a reading but never substitute for the sniff.

The sample–expose–purge loop is no longer only a diagram — the bench records controlled exposure runs, with blanks and confounder controls alongside every exposure. These are recorded runs — analysis stays behind the truth gates. The rig and airflow casing this breath will live in are in engineering now.

Run metadata schema — fields recorded with every run

room_volume_m3Enclosure or room volume
ventilation_achAir changes per hour
pump_flow_lpmSample draw rate
sample_period_sSeconds between samples

Further reading

  • →Machine smell — the array the sample is drawn across.
  • →Drift & reality — why we track context against real-world change.
  • →Truth gates — where the sampling sequences become gated protocols.