OFFICINA TURINI / otSPADPrototype overview ↗
EXPERIMENT 01 / LOW-LIGHT DETECTION

Light beyond
the visible.

A phosphorescent specimen associated with americium-241 was left undisturbed for more than seven hours. A cooled silicon photomultiplier could still detect its presence — and the count rate returned to its dark baseline when the specimen was removed.

Specimen present
13,708 cps
Stabilized 5-minute mean
Specimen absent
11,615 cps
Stabilized 5-minute mean
Measured excess
+2,093 cps
18.0% above the dark baseline
Elapsed since earlier test
7+ hours
Persistent contrast at the next session
Experiment date · 10 November 2026Detector · onsemi MICROFC-30050-SMT-TRCold plate · approximately −20 °CActual experimental data · not simulated
The experiment

Can a cooled SiPM reveal a signal long after the light is gone?

otSPAD was developed to measure extremely weak optical signals using a programmable silicon photomultiplier, active temperature control, analog signal conditioning, a fast comparator, an FPGA counter and a DSP-based acquisition system. This experiment tests the complete measurement chain under deliberately simple laboratory conditions.

A small, yellow-green phosphorescent coating associated with an intact Am-241-bearing specimen was observed through a 55 mm focal-length objective in a light-tight enclosure. An earlier one-hour run recorded a strong, slowly decreasing count rate. More than seven hours after that run ended, a new measurement was made without re-exciting or repositioning the specimen prior to its first phase.

01 / OBSERVESpecimen in view

Record a stable count rate for about 1,001 seconds with the sample in the optical setup.

02 / ISOLATEPause and remove

Pause counting, turn off SiPM bias and remove the specimen from the enclosure.

03 / RESTORECalibrate and settle

Recalibrate and allow the previously characterized startup transient to settle before resuming.

04 / VERIFYMeasure the dark baseline

Record a further 867 seconds with the specimen absent and compare stabilized sections.

Direct observation

The signal drops back to the instrument baseline.

Across the first phase, the event rate remained close to 13.7 kcps. During the deliberate pause, the specimen was removed and the detector was recalibrated. The resumed measurement was around 11.6 kcps — the previously characterized dark count range of this cooled setup.

Evening run · source-present / source-absent comparison
2026-11-10T22:57:48.891795 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/
Each line is averaged into 10-second bins from the raw 100 ms gates. The gray interruption is intentional: no count data were recorded while the bias was disabled, the specimen removed and the instrument calibrated. The dotted line is the measured, stabilized post-removal baseline.
MeasureSpecimen presentAfter specimen removal
Recorded gates10,0108,673
Recorded measurement duration1,001.0 s867.3 s
Mean count rate · final 300 s of each phase13,708.0 cps11,614.6 cps
Fano factor · final 300 s (raw 100 ms gates)0.9550.998
Cold-plate temperature · stabilized mean~−19.95 °C~−19.99 °C
Commanded bias28.822 V28.819 V
Comparator threshold2.300 V2.300 V

Stabilized means use the last 300 seconds of each phase, to avoid overweighting an early post-calibration settling component. Temperature refers to the cold plate, not a direct junction-temperature measurement.

“The important result is not that the counter detected a bright source. It is that a small, repeatable count-rate excess remained measurable after hours — and disappeared when the specimen was taken away.”

A longer view

From afterglow to a controlled removal test.

The earlier acquisition began near 68 kcps and decayed over an hour, reaching approximately 21.9 kcps in its final five minutes. The following session started more than seven hours later at approximately 13.7 kcps with the specimen still present. This is consistent with a very slowly fading optical contribution, but it is not by itself a measurement of the radiative excitation mechanism.

Two separate sessions, shown side by side
2026-11-10T22:57:48.998704 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/
Left: the preceding one-hour observation. Right: the subsequent evening experiment with a deliberate intervention. The break between sessions is more than seven hours and is not plotted as continuous data. The two sessions are not combined into a single decay fit.
Signal quality

The difference is large compared with counting noise.

Using the final five minutes from each phase, the excess rate is 2,093.4 cps, corresponding to 18.0% of the measured dark baseline. The Poisson-scale fluctuations of 100 ms gates are only a few tens of counts, much smaller than the mean separation between the two populations.

Raw-count distributions · stabilized windows
2026-11-10T22:57:49.083188 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/
Empirical histograms for the final 300 seconds of each phase, overlaid with ideal Poisson distributions having the corresponding sample means. This is a comparison of detector event statistics, not a spectroscopic identification of the light source.
Minute-averaged stability within each phase
2026-11-10T22:57:49.144069 image/svg+xml Matplotlib v3.10.8, https://matplotlib.org/
The minute averages remain well separated. A small slow drift in the post-removal baseline is visible, supporting the decision to compare stabilized final windows rather than all gates indiscriminately.

Using non-overlapping 10-second blocks, the within-session statistical 95% confidence interval for the mean excess is approximately 2,079–2,108 cps. This narrow interval quantifies counting statistics and short-term variation; it does not account for systematic uncertainty introduced by calibration, optical alignment or changes to the specimen.

Instrumentation

A real instrument — not a simulated detector.

The measurement used the Officina Turini otSPAD prototype, based on an onsemi C-Series SiPM and a programmable analog front end. The comparator had already been stabilized with positive feedback following earlier high-rate testing.

The counter is implemented in an FPGA and read through a Blackfin DSP acquisition system. This experiment recorded 18,683 individual gate values with an approximately 100 ms integration time, preserving time ordering and monitoring hardware state throughout the session.

Operating conditions

SiPM: MICROFC-30050-SMT-TR (3 × 3 mm²) · TLV3501 comparator with 10 kΩ feedback · threshold 2.300 V · programmed overvoltage 2.5 V · cold plate ≈ −20 °C · gate 100 ms · 55 mm objective · light-tight improvised enclosure.

Data-integrity detail. The log records a deliberate pause at ~1,001 s, a bias-off command at ~1,006 s, calibration at ~1,056 s and resume at ~1,136 s. The recorded sequence and FIFO-overflow counters jump during this pause because queued and partial gates were discarded. This is not represented as missing active measurement data, and the gap is not interpolated.
What the test establishes

Strong evidence of a specimen-associated signal. No exaggerated claims.

What was demonstrated

The prototype reliably discriminated two different count-rate levels in one instrument session: one with the specimen present and one after it was removed. The difference survived for more than seven hours after the earlier run, and the detector returned close to its dark baseline on removal.

What remains to be isolated

The test does not uniquely identify the cause of the excess. A stored afterglow, continuous radiation-driven emission, direct radiation interactions in the SiPM, residual ambient light, and any calibration-related systematic effects cannot be fully separated without further controls.

In particular, the same intervention changed both specimen presence and the detector calibration state. A matched nonradioactive phosphor control or an optically opaque control with the specimen otherwise unchanged would be needed to specifically establish radiation-induced luminescence. The experiment also provides no calibrated photon flux, scintillation yield or source activity measurement.

Why it matters for otSPAD

A detector designed for very low photon flux should deliver more than impressive specifications. It should reveal a real, faint and persistent signal — and then return to a reproducible reference. This experiment is a practical demonstration of the prototype's sensitivity, electronic stability and ability to make a meaningful differential measurement.

Explore the otSPAD prototype and its technical documentation
Method & references

Reproducible numbers, explicit limits.

Source files: longRun.zip (earlier one-hour observation) and finalTest.zip (subsequent controlled removal measurement). For the principal before/after comparison, rates were computed from the raw gate counts and recorded live-gate ticks; descriptive statistics use the final 300 s of each phase. Recorded UTC timestamps and events locate the pause and calibration. No values are fabricated or extrapolated across the pause.

  1. Officina Turini — Programmable system for single-photon detection, prototype overview.
  2. onsemi — C-Series SiPM datasheet, MICROFC-30050 family specifications.
  3. ATSDR / NCBI — Americium-241 decay scheme, radionuclide background; not a measurement of this specimen.

Experimental note, 10 November 2026. All plotted data are from the prototype. Statistical reference curves are calculated from measured means and explicitly labeled as references.