Chl-a Rhodamine Ladder — high-drive response test

Operating point & window

All units run v0.1.21-configurable with Notehub env config: led_currents_mask=111000100000 (LEDs 32 / 512 / 1024 / 2048), SiPM bias 34.0–48.5 V × 27 steps (data counts → volts: V ≈ −1.385 + 0.01262·counts), 1-min samples / 20-min reporting. Config verified live on 500185 before the run.

Protocol — rungs & doses

Working stock: 100 µL of 20% concentrate (~200 g/L) into 100 mL DI → 200 mg/L (same stock concentration as the 1 September run). Cumulative additions into 35 L; mix ~10 min after each dose, dwell ≥45 min per rung (one full 20-sample sync). Ascending only. Record water temperature per rung (rhodamine ≈ −2%/°C).

Rungs 1–3 replicate the 1 Sep ladder where the In-Situ AquaTroll Chl-a channel detected at σ≈47 (Jamison 0.0059→0.0545 RFU by 103.5 ppb). Rung 4 replicates the 2 mg/L screen dose (500151 lit; every Chl-a unit dark) at 4× the LED drive. Optional coda if all dark: +3.15 mL neat → 20 mg/L. Dose timestamps appear as markers on the charts once logged.

Reference instruments: three In-Situ AquaTroll sondes are co-deployed in the tank (in at 3:23 PM, teal marker on the charts), giving a direct Lume-vs-sonde comparison in the same water at every rung. Their data reaches us on a lagging upload path (dropouts of hours are normal while the instruments stay healthy), so the sonde comparison is joined offline after the run rather than plotted live here. Note for reading the Lume traces: with several optical instruments flashing in 35 L, an isolated single-sample spike coincident with nothing else is plausibly a flash collision, not a response — a real response holds across consecutive samples.

Fleet status — structure check

Each unit must sweep 4 LEDs × 27 bias steps before dosing starts; the pre-dose batch is its blank at the new operating points. “Net” below = mon2 at the top bias step minus mon2 at the bottom (pedestal) step, per LED, per sample.

Response — net signal at LED 2048, all units

Findings: ladder complete

No response on any of the eight units, at any LED level, through 2 mg/L. Per-rung medians of the net bias-arc signal sit at 7–11 counts on every unit with 0–1-count shifts from blank, and — the deeper result — absolute mon2 is pinned at the pedestal (~180 counts) across the entire 34–48.3 V bias arc at every LED level including 2048. The signal chain registered no light at all: no fluorescence, and no LED-current scaling anywhere (512 ≈ 1024 ≈ 2048).

Two things sharpen this beyond “still dark”: (1) 500151 lit at 1.99 mg/L in the 2 September screen at one-quarter of today’s drive, and today it is pedestal-flat at the same concentration — and (2) the September in-water medians for this batch were 68–111 counts net, an order of magnitude above today’s 7–11. Raising LED current and bias did not rescue these units, and the batch reads darker under v0.1.21 than it did under the September build. The firmware-timing question this raised has since been closed: on 2026-09-11 evening, TLF unit 50091 (identical electronics, FDOM solid-standard puck) ran the same v0.1.21 build and read its puck at ~3,030 counts vs the ~180 pedestal — v0.1.21 measures light normally. The Cl-A null therefore stands as a sensor-side excitation failure of this batch, now demonstrated up to LED 2048. (The September–today difference in in-water medians reflects metric and build differences, not a firmware light-capture defect.)

The two AquaTroll sondes that carry fluorescence channels (Marina and LakeviewEast) confirm the 2 mg/L rung. At the 21:40Z dose their Chl-a channel rose from 0.004 and 0.021 RFU to about 11 and 8.5 RFU within one 5-minute sample, and FDOM from 0.01–0.02 to 0.5–0.65 RFU, while all eight Lume units stayed at the pedestal in the same water. Both sondes read near zero from 22:05Z with temperature falling toward room temperature, so they cover this rung from 21:40 to 22:00Z. They went in after the rung 3 dose and say nothing about rungs 1 and 2. At 100 ppb they read 0.002–0.009 RFU (Marina) and 0.022 RFU (LakeviewEast), below the 0.036–0.055 RFU the 1 September run recorded near 100 ppb.

Reference sondes — In-Situ AquaTroll, same tank

Loading sonde dataset…

Per-unit dose response — all four LED levels

Net mon2 per LED level. Grey 32 · blue 512 · purple 1024 · red 2048. Dashed verticals = dose events.

Tank temperature