Cheap, isolated, buildable.
Three small devices give the verification stack its physical senses. This is a costed design — what we will build — not yet a built product. Every measurement is galvanically isolated and non-intrusive.
Poultry has a stun monitor. Fish has nothing.
Poultry processing uses a regulator-accepted inline stun-current monitor; for fish there is no field dose-verification at all — guidance recommends parameters but is silent on how operators confirm delivery, and the required field strength depends on water conductivity. We close that with three cheap devices.
Inline stun-dose + recovery logger
The fish equivalent of the poultry stun-current monitor — a sub-$100 box that clips onto an existing stunner and measures, per batch, the dose delivered and the stun→bleed interval. This is the first device we build.
| Component | Example part | Role | Indicative € |
|---|---|---|---|
| MCU / edge | ESP32-S3 dev board | acquisition, on-device verdict, BLE/MQTT | 8–12 |
| Voltage sense (isolated) | differential probe + ADS1115 16-bit ADC | stun voltage, galvanically isolated | 6–10 |
| Current sense | split-core CT clamp | stun current (rms), non-intrusive | 8–15 |
| Field electrodes | graphite/Ti pair, fixed spacing | in-water field-strength proxy (V/cm) | 5–10 |
| Conductivity (EC) | analog EC probe | conductivity for field interpretation | 12–20 |
| Water temperature | DS18B20 waterproof | context + correction | 2–4 |
| Interval input | button / opto line-trigger | stun→bleed timing | 1–3 |
| Power + isolation | isolated DC-DC, RCD-protected | safety | 8–15 |
| Enclosure | IP65 box + cable glands | wet environment | 6–12 |
Firmware computes rms current, peak field and duration at the stun event, reads conductivity and temperature, runs the recovery-clock, and produces the same signed batch record the software core already generates — buffering offline and syncing when a link appears.
The “dummy-fish” conductivity phantom
A passive strip can’t honestly read field strength in conductivity-varying water — so we don’t use one. Instead, a reusable fish-shaped phantom, tuned to fish-tissue conductivity with embedded electrodes, runs through the stunner periodically to confirm the field actually delivers into a fish-like load.
It is a go/no-go commissioning check: if the internal field is below the species target, the stunner is mis-set before any fish is harmed. No fish-conductivity phantom at stun frequencies exists today — a cheap, reusable, novel opening we can build first.
What it is
Echo-Stun contact evoked-response rig
The bold layer: a cheap proxy for lab-only EEG. A small probe pulse, then we measure the suppression of the evoked muscle/neural response — a 0–100 “insensibility index,” modelled on the human BIS anesthesia monitor that proves such an index is deployable.
| Component | Example part | Role | Indicative € |
|---|---|---|---|
| Biopotential front-end | ADS1299 (8-ch, 24-bit) / AD8232 | µV-scale evoked-response capture | 10–40 |
| Stimulus driver | isolated probe-pulse generator | evoke a response to test suppression | 10–20 |
| Electrodes | Ag/AgCl or dry contact pads | skin-surface pickup | 2–6 |
| Edge compute | Raspberry Pi 5 / Jetson Orin Nano | signal processing, BIS-like index | 70–250 |
| Shielding / isolation | Faraday + medical-grade isolation | SNR + safety | 15–40 |
Safety first. In-water electrical stunning involves lethal voltages and currents. All sensing is galvanically isolated and non-intrusive (clamp/inline taps that never break the stun circuit). Any bench work with energised stunners requires a qualified electrical engineer and proper isolation / RCD protection. The Echo-Stun rig is optional and never gates the cheaper, lower-risk tiers.
Automation levels — not different welfare standards.
The same fail-safe logic runs at every tier. Lower tiers are more conservative (more “check manually”), never less safe. We never relax welfare by budget or country.
Lite
Small / low-resource farms & vessels
- Phone/tablet app
- Manual recovery-clock
- Species checklist & SOP
- Signed batch record
Sensor Box
Medium farms / processors
- Inline dose + recovery logger
- Dummy-fish phantom
- Drift alarms
- Audit-ready reports
Camera / Echo-Stun
Advanced sites
- + camera and/or
- Echo-Stun index
- Edge inference
- Higher assurance
Validation
Labs / certifiers
- EEG/VER collaboration
- Expert labels
- Open benchmark
- Species evidence DB
Recommended first bench build: Item 1 + Item 2 — real welfare value and an auditable record now, with zero live fish.
Deployment cost tiers
Automation levels — not different welfare standards. Log scale (site cost €).
Cost leverage vs the alternative
A sub-$100 verifier rides on top of an existing ~€150k in-water stunner. Log scale (€).
What already exists
In-water stunners; poultry inline current monitors; lab EEG/VER; cheap biopotential front-ends; the ESP32 + sensor ecosystem. We build on mature components.
What we build
The fish inline dose + interval logger; the fish-conductivity phantom; the contact Echo-Stun index — none of which exist for fish today.