A particle detector built for cost-effectiveness when ordered at O(100) pieces.
The detector concept is adapted from the DIY Particle Detector Project by Oliver Keller.
Revision 1 | Revision 2 |
---|---|
![]() |
![]() |
Schematic Layout | Schematic Layout Component Positions |
Initial protoype: Working with adjustemtns to a few component values. Requires a lot of electrical and conductive tape to assemble. Designed for 16 diodes with the option to measure coincidences. | Streamlined redesign. Mainly adding a PCB casing to block light and radio imission. Attempting to add a simple alpha particle detector channel by stripping the isolation off of a diode (not successfull yet). Single channel with 6 diodes in parallel. Using additional ADC channels with less gain for debugging and potential experiments with scintillators. |
- Time distribution of pulses
- Rate vs distance (1/d^2 + offset)
- Absorption coefficient
- Increase material thickness between detector and source
- Spectroscopy?? (alphas, sand down diode?
- Balloon in cellar
T
T <block_index> <timestamp_us> <overflow> # <sample_0> <sample_1> .. <sample_N>
block_index
: uint16_t, mainly relevant for debuggingtimestamp_us
: uint64_t, Trigger point timestamp in micro seconds since start of the measurementoverflow
: bool, set to 1 if an overflow happend prior to this samplesample_n
: int16_t, samples from the triggered waveform
The goal was to keep it small to keep the PCB cost low and to make it fit easily. Components were chosen to keep the price low and based on availablility at the chosen manufacturer.
- Strip detector with 4 pixels?
- File down SMD LEDs for less absorption?
- 3D-Printed Case?
- Interconnect multiple detectors?
- Put VCC on interconnect instead of 3V3
- Review diode cutout size
- Increase power supply castellated hole distance
- Increase LDO heat sink plane size
- Think about how usefull the coincidence concept is: 2x2 = 2+2 = 4!
- Silk to highlight Diode grouping
- Seperate bulk capacitor on 5V rail for analog
- Spice simulation: peak of ca. 40
$\mu\text{V}/\sqrt{\text{Hz}}$ - Resistor thermal noise:
$\sqrt{4k_BTR}\approx 400 ~ \text{nV/}\sqrt{\text{Hz}}$ - LM358 input noise is ca. 40
$\text{nV}/\sqrt{\text{Hz}}$ - Gain of second stage is ca. 100 for the peak frequency
- => Resistor thermal noise is dominant part in simulation, LM385 noise should be negligible
$100 \cdot 400~\text{nV}/\sqrt{\text{Hz}} \approx 40~\mu\text{V}/\sqrt{\text{Hz}}$
- Resistor thermal noise:
- 10g KCl has ca. 164 Bq.
- Estimate of geometric acceptency at 2 cm distance:
$A = \frac{4 \cdot 7.5~\text{mm}^2} {4\pi\cdot (20~\text{mm})^2} \approx 0.006$ - With perfect efficiency: ca. 1 Hz