v2 board ・ Arduino Nano ESP32
ACCEL KITCHEN — COSMIC RAY DETECTOR v2
A SiPM picks up the scintillator's light and a peak-hold circuit captures the pulse height. An Arduino Nano ESP32 logs pulse height, time, temperature, humidity and pressure as one line per event, and outputs it to USB, Bluetooth and microSD at once. Connect up to three external detectors and measure in coincidence.
01 — INSIDE
From receiving the light to recording it. The camera moves to each part in turn.
When a muon passes through the scintillator, part of the energy it deposits becomes light. A 6 mm-square SiPM (onsemi MICROFC-60035) turns that light into an electrical pulse. The SiPM sits on a 40 mm-square daughter board, and the whole daughter board can be swapped for another type.
An LT3461 on the back steps the USB 5 V up to 29.0 V, which is 4.3–4.8 V above the breakdown voltage. The value is fixed by a resistor divider. The daughter board's 49.9 Ω and 10 µF filter out noise before the voltage reaches the SiPM.
The SiPM current becomes a voltage across R12 (49.9 Ω) on the daughter board and is amplified by the non-inverting amplifier U2A. The gain is ×5.0–×45, set by trimmer R10 on the back. The input is protected by 1 kΩ and a BAT54S. The amplifier output can be measured at TP2.
The peak hold around U2B doubles the peak of a pulse a few hundred ns long and stores it on capacitor C6. The stored voltage appears at A0. Once the microcontroller has read it, D8 sends a 20 µs pulse that discharges C6 through Q1, ready for the next pulse.
It keeps reading A0 (the peak-hold output) and treats anything above the threshold as an event. After waiting for the external detectors' coincidence flags, it reads A0 with the 12 bit ADC, and that value is the pulse height. At every start-up it watches A0 for one second and sets the threshold to the noise maximum + 50.
Logs are written to a FAT32 card starting from /LOG0001.TXT. Each line is committed as it is written, so even if you pull the USB cable you lose at most the last line. The lines are exactly the ones sent over USB and Bluetooth.
Temperature, humidity and pressure are read every 2 seconds and written into every event line. You can study the “barometric effect”, where the count rate drops as pressure rises. The SiPM breakdown voltage shifts by about 21.5 mV per °C, so temperature is also what you use to re-examine pulse heights later.
A 0.96-inch OLED standing at the edge of the board shows the count, rate, threshold and environment values every 2 seconds. An LED flashes on every event and the buzzer beeps at a pitch that follows the pulse height (500–4000 Hz, logarithmic). Add a vibration motor on A7 and bigger pulses shake for longer.
One OUT and three INs. When it triggers, it outputs COIN_OUT (3.3 V, 100 µs), waits 50 µs and reads the others' COIN_OUT. If a partner reacted at almost the same moment, that jack's column becomes 1. The analog pulse height can be read too.
A two-piece design: the board drops into a box with a floor, and a lid is fixed with four M2 screws. The front has a window for the OLED and holes for RESET and the LED; the side has USB and a square opening holding the four jacks and the microSD slot. A clear light pipe carries the LED's light to the hole on the front. The STL files are in cad/.
02 — EXPLORE
Tap “Interact” to rotate it with your finger.
PART
More than 80 parts, from boards and case to screws, come with descriptions.
SCINTILLATOR
The board and case are the same; only the part that emits light differs. Choose by whether you want to collect widely or carry it small.
A wide, plate-like shape. Its 25 cm² of area catches plenty of cosmic rays, so counts add up even in a short time. It suits following changes in the count rate and coincidence with two units stacked.
A small, die-like shape. It is compact, suited to carrying around and to installing in tight spots. With less area it counts fewer cosmic rays.
Both work the same way: a passing cosmic ray makes light, and the SiPM turns that light into an electrical pulse. The horizontal axis is light intensity (= deposited energy) and the vertical axis is the number of counts. The count rate is roughly proportional to the scintillator's area (as a guide, about 1 per cm² per minute at sea level).
ACTUAL UNIT
Photos of the prototype. The square block wrapped in black tape is the scintillator and SiPM. The count appears on the blue OLED in front, and a green LED lights at the board's edge.
03 — PHYSICS
A primary cosmic ray (mostly a proton) collides with an atomic nucleus high in the atmosphere and makes a shower of particles. Almost everything that reaches the ground is a muon: about 1 per cm² per minute at sea level.
As a muon crosses the scintillator it loses energy, and part of that becomes light. About 2 MeV per cm in plastic (blue light), and about 5–6 MeV per cm in CsI.
A SiPM with 18,980 microcells of 35 µm. A cell hit by a photon undergoes avalanche multiplication.
R12 (49.9 Ω) on the daughter board turns it into a voltage: a fast pulse of a few hundred ns.
It is amplified ×5–×45, and the peak hold stores double the peak height. When the voltage at A0 exceeds the threshold, it is read as an event.
About 60 µs after the trigger, A0 is read with the 12 bit ADC and output as one line together with the time and environment values. Once read, D8 empties the hold.
04 — SIGNAL
Change the pulse size, the amplifier gain and the threshold, and the amplifier output, the peak hold and the recorded pulse height (the ADC column of the log) change. The hold capacitor can only charge at 1.36 V per µs, so faster pulses lose more of their peak.
A rough guide using the same first-order model as arduino/shaping.html (SiPM 60035 + GAGG:Ce, R12 = 49.9 Ω, TP2 half-width 432 ns), plus a peak hold limited by charge (U2B output current 30 mA ÷ C6 22 nF). “Muon” is a size that would give 7.9 V at TP2 if the amplifier had no rails; in practice it saturates at 4.9 V. The hold discharge has a 300 µs time constant, and the ADC is approximated with about 3.1 V as full scale.
05 — COINCIDENCE
Muons travel near the speed of light, so they pass two stacked units almost simultaneously. SiPM noise and environmental radiation mostly hit just one. Keep only events where “both reacted at almost the same time” and you pick out muons more reliably. The wider you space them, the more only near-vertical rays remain.
Two unrelated events happen to fall inside the window (±W) at a rate of 2 × W × R1 × R2. The v2 window is about ±50 µs.
About 0.36 per hour
COIN_OUT (D7, 3.3 V, 100 µs) and its own peak-hold output ANA_OUT
Reads the partner's COIN_OUT on D6 and the analog on A1
Likewise D5 / A2
Likewise D4 / A6
06 — USE CASES
The number of cosmic rays changes with place, height and solar activity. Here are examples of real measurements.
Counts were recorded every minute while climbing. The higher you go, the thinner the air and the more cosmic rays arrive. From about 7 per minute at the foot, it rose to about 20 near the summit.
Dots are the cosmic-ray counts; light blue is the height climbed. Measured on a two-day climb with one night on the mountain.
It runs on a mobile battery, so you can take it on board. About 45 per minute on the ground becomes about 400 at around 10 km altitude. Up in the sky, roughly 9 times as many cosmic rays fall as on the ground.
A record of about 90 minutes, from takeoff to landing.
When the Sun has a large eruption, the cosmic rays reaching Earth can drop for a while. A detector at a school caught a drop from about 29.5 to about 28.3 per second right after particles from the Sun arrived.
Top: particles arriving from the Sun (GOES satellite). Bottom: cosmic rays counted at the school.
Cosmic rays weaken as they pass through concrete bit by bit. Measuring floor by floor in a school building, the count rose on higher floors: about 3.5 per minute on the 1st floor against about 5.4 per minute on the roof, because fewer ceilings and floors lie above your head. From this difference you can work out the thickness of walls and ceilings.
Measured floor by floor, from the 1st floor to the roof, in three wings of a school building.
The graphs are real measurement examples. The numbers you count depend on the type and size of the detector. The moving pictures are illustrations based on measurement trends (the Forbush decrease is exaggerated to be easier to see).
07 — WEB APPS
Nothing to install. Open the console in Chrome or Edge and connect over USB (Web Serial) or Bluetooth (Web Bluetooth). Below is a demo of that screen; press the buttons to try it.
Data recording app for Cosmic Ray Detector v2
Let's count cosmic rays
Connect the detector to your computer and the cosmic rays that arrive turn into graphs on the spot.
Energy distributionADC histogram, bin width 64
Count rate over timeEvery 20 seconds
A0 waveform read every 12 µs (32 points before the trigger, 128 after). Counting stops while it is shown.
Waiting for data…
Web Serial and Web Bluetooth (Nordic UART). They work with the same commands and responses, and Bluetooth reconnects automatically if it drops.
Pick a folder and incoming data is written continuously as .dat and .csv. The file splits when the date changes.
Pulse-height distribution, count-rate history, and statistics for temperature, humidity and pressure. The count rate is computed with dead time subtracted.
Reads A0 by DMA at 83.3 kHz (every 12 µs) and shows each event's waveform in 161 points. Counting stops while it is open.
Signal threshold (with auto-adjust at start-up), and switches for microSD, external detectors, OLED, buzzer and vibration, all by button.
Upload to the measurement server in real time, for comparing multiple sites. The display can switch between Japanese and English.
Load saved .dat, .csv and microSD logs for an ADC histogram, zenith-angle distribution and energy calibration.
console/plotter.html ManualAn explanation that follows the flow of the console screens, with screenshots and figures; technical material is collected in the appendix.
console/manual.html Waveform ProbeA waveform viewer, used with the waveform firmware, for looking at the peak-hold waveform and baseline noise.
wave_plotter.html Legacy dashboardThe dashboard before the console. Same functions, with everything laid out on a single screen.
v2/index.html Pulse analysisA report examining, by calculation and measurement, the analog front end's time constants, saturation and peak-hold losses.
shaping.html08 — DATA
Each detection makes one line of record. It is plain text separated by spaces, so spreadsheets and Python can open it as is.
How much energy the passing particle deposited. As a histogram, you see the peaks for cosmic rays.
Time since start-up, in microseconds. You can follow changes in the count rate and the gap between individual events.
The local environment is recorded with every event. You can check the “barometric effect”, where the count rate falls as pressure rises, with your own data.
For each event, it keeps whether the connected detectors (up to 3) reacted at the same time. Stack them and you can select only cosmic rays coming from straight above.
Also the size of the particle seen by the partner that reacted together, useful for comparing the same particle measured by two units.
Time when it could not count, such as while writing data, is recorded too, so you can compute the correct count rate.
measuring at once, at most
places to record (USB, Bluetooth, microSD)
of environment values (temperature, humidity, pressure)
cosmic rays counted and recorded
09 — START
With the firmware written, it starts measuring as soon as you connect USB-C.
With USB unplugged, plug the daughter board's J1 into the readout board's J2 and fix it with M2 spacers and screws.
Wrap it in reflector, seal out the light with black tape and set it on the SiPM. Any gap lets fluorescent light spike the count.
Put a FAT32 card into J8 through the opening on the side. Without it, output still goes to USB and Bluetooth.
To a PC or a mobile battery. At each start-up it watches A0 for one second to decide the threshold, so leave it still while it does.
If Cnt rises, the LED flashes and the buzzer beeps, it is working. Thr: is the threshold.
Open arduino/v2/console/index.html in Chrome / Edge and press “Connect by USB”. Over time the histogram grows and the count-rate error shrinks.
10 — SPEC
Values come from the schematic (KiCad 9), the firmware (COSMICRAY_DETECTOR_ESP32-S3-Nano-v2), arduino/description.md and shaping.html. The successor v2.1, which reflects a circuit review, replaces the shaping stage and uses a DAC-set threshold. This device is for counting natural cosmic rays and is not calibrated as a dosimeter.
11 — NEXT