Precision systems · forensic analysis console

Clarity inthe dark.

One corridor. One cold light. One buried signal. Scroll to run the analysis: the noise floor falls, the signal isolates, and the tolerance tightens until only what is true remains.

Scroll to begin the run

SampleCorridor B · frame 0931 Capture1 × 4000 K fluorescent MethodProgressive noise rejection InstrumentTrace + spectral + reticle Target±2 µm · 99.7 % confidence

01 — Acquire

Everything is noiseuntil you measure it.

The raw capture is honest and useless. Sensor grain, mains hum, thermal drift: the trace behind this page carries all of it, at a noise floor of −12 dB.

We do not clean an image. We reject what does not repeat. Each stage below removes one class of noise and records the tolerance we can now defend.

A dark laboratory corridor, one fluorescent tube lit near the ceiling, the far door barely visible. — **Frame A** full fieldCrop 0.00,0.00 → 1.00,1.00

02 — Isolate

One light.Lock on it.

The corridor has exactly one source. The reticle finds it, and every other reading in the frame is now referenced to that 4000 K tube.

Mains hum at 50 Hz and its harmonics leave first. The noise floor drops past −48 dB and the trace behind you stops shivering. This is the moment the page is built around.

Close crop of the corridor's single fluorescent tube, the only light source in the frame. — **Frame B** sourceCrop 0.78,0.02 → 1.00,0.24

03 — Calibrate

Measured,not guessed.

Four references, one frame. Each is fitted against the locked source and the residual is written down, not rounded away.

A tolerance is a promise about the worst case. Ours narrows from half a millimetre to two microns over the length of this run, and the bars record how much of that promise is already kept.

ReferenceNominalResidualFit

Tube pitch1200.00 mm—

Ceiling grid600.00 mm—

Door reveal2100.00 mm—

Floor joint900.00 mm—

04 — Verify

The far doorwas always there.

At −90 dB the spectrum is a single clean peak and the darkest crop in the frame resolves into an edge, a handle, a threshold.

Nothing was added. The door is not enhanced; the noise that hid it is gone. That is the difference between a picture and a measurement.

Deep crop of the corridor's far end where the door edge emerges from darkness. — **Frame C** far doorCrop 0.48,0.40 → 0.76,0.76

0 HzPeak · 412 Hz · signal2 kHz

05 — Report · run 0931 complete

See itclearly.

Vanta Labs builds instruments for people who cannot afford to be approximately right. Send us the frame, the part, or the process, and we will tell you exactly what it is.

Request a calibration

Noise floor−96.0 dB

Signal to noise68.0 dB

Tolerance±2 µm

Confidence99.7 %

ResultClear