Experiment Capture

experiment mode

a guided bench instrument · plan → capture → notes · raw readings preserved · isolated from the living log

app v0.7.1

{{ port }} ESP32 · 4 probes 115200 baud {{ ownerLabel }}

provenance {{ expId }} {{ startLocal }} {{ durationLabel }} {{ rateLabel }} auto-stop {{ durationLabelPlanned }} {{ sourceLabel }} fw {{ fwVersion }} app v0.7.1 raw {{ rawRangeLabel }}

Fields marked — unavailable are designed-in but not yet emitted by the capture stream; they fill in when their slice lands (no redesign). Times: local Chicago for the human, UTC for machines.

A plan is waiting for you drafted by {{ draftBy }} · review before arming

suggestion

subject

rate

duration

{{ c.id }} · {{ c.label }}

hypothesis

✓ loaded {{ draftBy }}'s draft Everything below is editable — it's yours now. Nothing records until you arm and confirm.

Draft dismissed — building the plan by hand.

What you're capturing

Subject — a human note, in your words

A subject is required — it names the run and its folder.

saved as{{ expId }}· your words stay intact; the id is the machine-safe slug

Intent — what you expect to see

Probe labels

what each probe is, and where it sits

primary

Labels travel with the capture — they tag the live channel cards, the chart legend, and the saved readings. They're notes, not calibration.

Sample rate

faster → finer detail, bigger file

Why {{ rateLabel }}? {{ rateWhy }}

Duration

a bounded session — it auto-stops here

Auto-stops at {{ durationLabelPlanned }} (≈{{ plannedSamples }} samples at {{ rateLabel }}). The capture process owns the stop — it ends even if the browser closes.

Source

⚠Synthetic data is excluded from evidence and never used for real bench work. The capture is tagged source=synthetic so it can't be mistaken for a reading.

Destination

🛡{{ destPath }}

Isolated — your living monitor log is never touched. {{ overwriteNote }}

The Monitor logger holds {{ port }} for the living baseline. Capture needs exclusive ownership so the two streams can't silently contaminate each other.

ARMEDreview, then begin — nothing is recorded until you confirm.

subject

rate

duration

source

writes to

{{ p.id }} · {{ p.labelOrDash }}

{{ curRaw }}raw · {{ primaryTag }}

{{ elapsedLabel }} · {{ sampleCount }} samples

auto-stops in {{ autoStopIn }}

Live channels

your four labelled probes · raw counts

primary

{{ c.bandName }}±{{ c.spread }}

raw ADC · 12-bit (0–4095) · HIGH = drier · no calibrated %, ever

primary {{ primaryLabel }} · {{ primaryTag }}

raw preserved continuously → {{ destPath }}readings.csv

{{ stopReasonLabel }} · {{ sampleCount }} samples kept

The readings 🔒 raw · immutable {{ sampleCount }} samples · {{ durationLabel }} · {{ stopReasonLabel }}

{{ p.id }} · {{ p.labelOrDash }}

time in each band · primary {{ primaryLabel }} · seconds, not %

Your reading of it interpretation — kept separate from the raw {{ autosaveLabel }}

Hypothesis Method Findings Conclusion

{{ savedLabel }} drafts autosave locally — they survive a reload or a lost connection

raw + band are the reading · interpretation stays in the notes layer · times: local Chicago + UTC tend well.