ERROR
Link
OFFLINE
Source
—
Alarm
—
Recording
—
Network
—
Simulation
OFF
WR offset from master
—
full view
measured
baseline
trip band (±√thr·σ)
◆over threshold
wheel or pinch to zoom · drag to pan · double-tap to fit
Coincidence histogram
—
full range
counts
measured peak
baseline centroid ±1σ
wheel or pinch to zoom · drag to pan · double-tap to fit
Peak tracking — measured vs predicted
—
full view
measured
predicted
trip band (±√thr·σ)
◆anomalous (ticked on axis)
wheel or pinch to zoom · drag to pan · double-tap to fit
—
Coincidence rate tracking — measured vs predicted
—
full view
measured
predicted
trip band (±√thr·σ)
◆over threshold
second detector — moves with source power, not path length
Alarm events—
No alarms recorded.
Detection—
?
WAITING
no frame from the board yet
—× THRESHOLD
PEAK TIMING SCORE vs THRESHOLD—
0
THR
10×
100×
1000×
Peak timing—
—
Coincidence rate—
—
Peak width (jitter)—
—
Measurements—
Count ratesper second
Inject a WR fault→ /api/wr/inject
Channel to disturbchannel
offset
Injected offsetoffset_ns
0.00 ns
The offset is added to the chosen channel's measured value,
so the fault is flagged and classified as that channel — an offset,
a fiber-delay or a round-trip change.
Inject a faultone tap — no dragging
Perturbation detail/api/controls
Perturbed channelperturb_channel
—
Target drift offsettarget_drift_offset_ps
—
Drift time constantdrift_tau_s
—
Jitter σjitter_std_ps
—
WR detector→ /api/wr/config
Confidence levelconfidence_level
—
Baseline windowbaseline_window_s
s
Confidence changes sensitivity at once. A new baseline window
takes effect on the next retrain.
Modelquick — safe mid-demo
Confidence levelconfidence_level
—
Baseline windowbaseline_window_s
s
Sustained drift before alarmsustained_drift_threshold_s
—
Retrain windowretrain_window_m
—
Continuous retrainingcontinuous_retraining
—
Predictorhow it expects the peak to move
Process noise Qpredictor_process_noise
—
Measurement noise Rpredictor_measurement_noise
—
⚠ Slow operation — not during a demo
Window auto-calibration re-searches for the coincidence peak and re-collects the
baseline from scratch. It takes about a minute and the tracking plot goes blank for the whole
of it. Two taps are required on purpose.
Acquisitionone-shot → /api/command
Match windowmatch_window_ps
—
Histogram binsdisplay_bins
—
Data source—
Loading captures…
Link—
Board WiFi—
Checking…
Press SCAN to list networks the board can see.
Join —
Board addressesreach it on any of these
Waiting for the first frame…
White Rabbit link—
Not connected.
Connecting polls the slave servo and
trains a baseline over ~30 s. Then set the source switch to
WHITE RABBIT to watch the link on the main plots.
Live link—
Checking…
This takes the Time Taggers for us alone. While it runs, nobody at Qunnect can use
them — any acquisition of theirs must be stopped first, and will fail to start
until we release.
Local acquisitionBigByte tagger only
One BigByte tagger, no CINT and no White Rabbit: both detectors share
its clock, so no sync offset is needed. Coincidence is NIR × telecom.
CINT acquisition—
Trigger & test—
TEST SIGNAL produces a perfect, entirely artificial coincidence peak. Any measurement
shown while it is on is synthetic.
Remote peak search—
Histogram view—
Recorded data—
Idle.
No recorded files on the board.