The detection of radioactivity, either produced by artificial sources or as part of the natural environment like K-40 or Radon from the inner of the Earth, is a There are commercial offers, but also a number of proposals for do-it-yourself projects. A sound-card of a standard PC is suitable to record the signals, which can then be visualized with any kind of sound-card oscilloscope.
A rather recent proposal of a simple, cheap and easy to operate particle detector based on four silicon PIN photo-diodes(BPW34) is the CERN DIY_particle_detector. A two-stage amplifier with a high-bandwidth op-amp produces large signals of several hundred mV shaped to a width of about 100 µs. Such signals can be easily recorded with a standard sound card. The commercial availability of all necessary parts including the printed circuit board and the large and long output signals make this detector an ideal choice for projects with high-school students. Besides building the device, additional experience is gained in data acquisition software and data analysis of a fascinating phenomenon not directly accessible by the human senses.
For the project presented here, the “electron variant” is used, which detects electrons using four PIN photo-diodes as sensors. Detectable electrons can also be generated by gamma radiation either near or in the PIN diode. When the detector is installed in a tin can, electrons (= beta radiation) from the surroundings are shielded, so that only ambient gamma radiation is registered.
PhyPiDAQ contains several modules supporting data recording, visualization and analysis. An interesting and important aspect to study in this context is the randomness of the occurrence of signals. As the probability to detect a signal is invariant with time, the number of events observed in a time interval is given by Poisson statistics. Typical for such a process is the fact that the time between randomly occurring events follows an exponential distribution.
In addition, the exponentially falling decay rate of a sample of radioactive nuclei is another important aspect that can be studied if a sample of short-lived nuclei can be provided. In Nature, such a source is Radon produced from radioactive decays in the inner of the earth; decay products of Radon, which are themselves radioactive, can be accumulated on an the surface of an electrostatically charged balloon.
Relevant modules of the PhyPiDAQ package are:
-
phypidaq\soundcardOsciwith two classes to record and select and to display waveforms from a PC soundcard -
phypidaq/DisplayPoissonEvent, a class to display a pulse corresponding to a single Poisson event and to show a rate history.
Ready-to use scripts illustrate how to use these classes:
-
examples/oscilloscope/soundcardOsci.pyto record and display wave forms from a sound card. -
examples/scGammaDetector.pyto visualize occurrences of a large signal and also the associated waveform data around the trigger point. A rate history is also shown. The code offers the possibility to store the event times and waveform features in a file for off-line analysis.
Studies of Poisson processes are possible by using the Python scripts
-
examples/poissonFlash.pyto generate, visualize and store data of a simulated Poisson process, and -
examples/poissonLED.pyto produce random flashes of a LED. A photodiode exposed to the light of the LED will produce signals analogous to a detector for gamma rays.
The Python scripts mentioned require a small part of the software contained in the package phypidaq. If the download of the full package is not desired, a simplified option is described below. Ideally, however one should download the entire PhyPiDAQ package and install the Python libraries as follows:
# get git repository with PhyPiDAQ code
cd <workdir>
echo “retrieving https://github.com/PhyPiDAQ/PhyPiDAQ”
git clone https://github.com/PhyPiDAQ/PhyPiDAQ
# install phypidaq modules
cd PhyPiDAQ
python -m pip install .After this step, the Python programs mentioned above can be installed in the directories examples/ and examples/oscilloscope/.
Note: This command sequence also works on the console (“Terminal”) of computers with MS Windows 10 or 11, if Python is installed there.
The Python programs scGammaDector.py and run_scOsci.py can also be executed without downloading the complete PhyPiDAQ package. If a current Python version is available, the libraries can also be downloaded can also be installed directly from the github repository:
python -m pip install git+https://github.com/PhyPiDAQ/PhyPiDAQThe Python programs can then be downloaded via the links run_scOsci.py and scGammaDetector.py into a working directory and executed.
A typical waveform recorded after issuing the command scGammaDetector.py -o on the command line is shown in the figure below. The signal is clearly visible above the noise level of approx. 3500 ADC counts. It is sufficiently large to be directly connected to a earphone so that the signal clicks can also be acoustically perceived.
Note that some soundcards invert the signal. The original output pulse is negative with a clearly visible positive overshoot.
{width=80%}
The script provides a number of command line options to control the visual output, enable storage of results to a file, and to set up the parameters of the soundcard and the trigger options.
The output of the command ./scGammaDetector -h is shown here:
usage: scGammaDetector.py [-h] [-q] [-o] [-n] [-f FILE] [-w] [-t TIME] [-z SAMPLESIZE]
[-s {48000,96000,192000,44100}] [-c {1,2}] [-l TRGLEVEL] [--trgfalling] [-d]
[--overshoot OVERSHOOT] [-r RANGE] [-i INTERVAL]
Read waveforms from soundcard and display and optionally store data
options:
-h, --help show this help message and exit
-q, --quiet no status output to terminal
-o, --oscilloscope oscilloscope display
-n, --noeventdisplay deactivate event display
-f FILE, --file FILE base filename to store results
-w, --write_raw write raw wave forms
-t TIME, --time TIME run time in seconds
-z SAMPLESIZE, --samplesize SAMPLESIZE
number of samples per read
-s {48000,96000,192000,44100}, --samplingrate {48000,96000,192000,44100}
sampling rate
-c {1,2}, --channels {1,2}
number of channels
-l TRGLEVEL, --trglevel TRGLEVEL
level of trigger
--trgfalling trigger falling edge
-d, --trgdeactivate deactivate triggering
--overshoot OVERSHOOT
minimum overshoot fraction
-r RANGE, --range RANGE
display range
-i INTERVAL, --interval INTERVAL
time bin for rate display
Because the signal rate is very low in normal environments without a radioactive source,
the trigger level should be set to be just above noise level so that some noise pulses
become visible. It is also advisable to use the option -o to switch on the oscilloscope
view.
Note that the signal level depends on the settings of the soundcard. Use the appropriate
tool of your PC operating system to select the standard input device used for sound recording
and adjust the volume control.
To register the signal shown above, the trigger options --trgfalling -l -5000 were used.
To become acquainted with the software, in particular the selection of data by setting appropriate trigger conditions, it is useful to use a microphone signal as input source. Initially, only the oscilloscope with a very low trigger level is started:
python3 scGammaDetector.py -n -o -l 100
This will show raw data from the soundcard in the oscilloscope display. Now make some noise,
e .g. by clapping your hands or snipping your fingers, and you will see some short signals
well above the average noise level. Remember the typical signal level of the background noise.
Stop the program by typing 'E' on the command line or click the 'End' button in the graphical
control interface. Then, restart at a higher trigger level, this time also enabling the
event display:
python3 scGammaDetector.py -o -l 1500
You should see no signals at all - unless you create a loud sound signal, which then is displayed in the oscilloscope and also in the event display window.
Detecting the very small signals from the CERN DIY particle detector works exactly
the same way. Connecting the detector output to a microphone input of a soundcard and
repeat the the same procedure just described to find the right trigger level to
separate noise from true signals of detected particles. Note that the signal level depends
on the settings of your soundcard, most importantly the volume.
If possible, increase the sampling rate to the highest possible value supported by
your soundcard - possible values are 44100, 48000, 96000 or 192000 samples/s.
Also consider adjusting the sample size of a single recording using the option
-z<n> - 256 or 512 are optimal settings for the short pulses of the particle detector,
but some sound drivers only support a minimum setting of 1024. If the sample size is too
large, more than one signal may be contained in the sample, but only the first one would
be triggered and counted.
It is obvious that the trigger level has a strong influence on the recorded signal rate.
If it is too low, most true signal will be detected, but some noise pulses (called "background")
will also be present. It the trigger level is too high, noise pulses will mostly be suppressed,
but also some signal pulses will be lost. There is no way out here - the detection efficiency
and the background suppression cannot both be 100 %.! If an absolute rate is to be determined,
corrections for signal efficiency and background contamination of a selected signal sample
must be applied.
The output from the particle detector seen under measurement conditions at low rates is
shown in the figure below. A flashing circle indicates the occurrence of triggered event,
and the corresponding (normalized) wave form with 100 sampling points around the trigger
time is displayed. A rate history is also shown; the bin width in seconds can be set using
the option --interval <n>.
{width=66%}
The determination of the background level and the signal efficiency is not always easy to
do with small uncertainties. the background level can be determined rather precisely by
performing measurements without the signal source. The determination of the signal efficiency
requires precise knowledge of the detector and the signal characteristics, typically gained
by detailed modeling of the physics processes in the detector and the response of the
front-end electronics. In case such studies signal efficiency vs. purity are interesting,
data recorded with scGammaDetector.py can be analyzed to gain more insight. Recording of
detected pulses is switched on with the option -f <filename>; for each detected signal,
the "event number", the time of occurrence in seconds since program start and the pulse
height in ADC counts are stored in the file. In an off-line analysis, a pulse-height
spectrum, i. e. the frequency at which pulse heights in a given interval occur, can be
derived from data taken at a low trigger threshold. This will show a large number of very
small pulses, but also a clear accumulation arising from true signals at higher values.
From data taken without a particle source a spectrum of expected background signals
during the measurement can be determined and subtracted from the data taken with a
signal source.
The sensor surface area of the DIY particle detector is very small at just 28 mm², and and the sensitive layer is very thin. Of the typically several Hz gamma rate at normal ambient radioactivity of typically 0.1 µS/h, and therefore only a small fraction is registered. By comparison with a dosimeter, in this case a Radiacode 102, see Instructions, an approximate calibration can be made:
A dose rate of 0.1 µS/h corresponds to 1.3 registered events per minute.
An example of a measurement of environmental radioactivity has been shown above. If sufficiently long measurement times are foreseen, the do-it-yourself detector can be used to study radioactivity in different environments. The difference in dose rate outdoors, in living rooms or in rooms with tiles or even granite differ by factors of two to three. With sufficiently large measurement times of a few tens of minutes such differences can be shown with significant statistical precision. The effects of weakly radioactive rock samples or the decay products of Radon on the surface of a balloon, electrically charged by rubbing, are also detectable.
The data stored in a file during data collection is used for the statistical
evaluation of pulse heights. Data recording is switched on with the option
-f <filename>. For each recorded signal, a consecutively incremented event
event number, the time of occurrence in seconds since the start of the program,
the peak-to-peak pulse height in ADC counts of the bipolar signal, the ratio of the
heights of the second and first peak, the temporal distance of the peaks
and the full widths at half maximum of the peaks are stored in the file.
The file in csv format contains the columns
event_number,event_time,pp_height,p_ratio,p_dist,fwhm1,fwhm2
To display the spectrum of pulse heights, i.e. the frequency with which pulse heights occur in a certain time interval, only the third column is used.
The result of a measurement with ambient radioactivity is shown in the figure below. The volume setting of the sound card was selected so that the trigger threshold was still within the range of the noise signals. Note that the pulse heights are defined as the difference between the maximum and minimum values of the ADC counts in the signal range (“peak-to-peak” pulse heights).
{width=66%}
The logarithmic representation shows a very large number of pulse heights just above the trigger threshold, which drops rapidly towards larger values. From around 10000 ADC counts, you can see the flat distribution of the signals caused by the ambient gamma radiation.
If the trigger threshold is set above the transition point from the noise to the signal distribution, noise signals are suppressed and only real gamma rays are registered. The trigger rate then corresponds directly to the rate of detected gammas.
The Python* program data/GammaAnalysis.py described below was used to display the
pulse height distribution.
If, in addition to the peak-to-peak pulse height, the other typical characteristics of the pulse shapes are considered, it is possible to distinguish noise and signal pulses on the basis of their typical shapes. In this way, the course of the pulse height spectrum can also be displayed in the noise region. This is shown below in Figure 4. The classification of the noise pulses shown in orange was achieved using an artificial neural network network that was trained on signal shapes for large and very small pulses.
{width=66%}
Translated with DeepL.com (free version)
The analysis of registered gamma quanta from a small sample of pitchblende recorded with scGammaDetector.py is shown in
the figure below. Signal pulses were registered at a rate of approximately 1.12 HZ. The file
GammaStrahlung_Pechblende.csv contains approx. 10000 recorded events with the columns
event_number, event_time[s], pulse_height[adc]
The graphs below were generated with the the Python code data/GammaAnalysis.py.
They show the number of events in intervals of 10 s duration, the frequency distribution
of the observed numbers of events and the time between two events.
The expected distributions resulting from the mean rate are also plotted, i.e. a uniform distribution for an average number of events of 1.11 in every 10 s interval, the corresponding Poisson distribution and an exponential distribution for a mean time interval of 1/1.11 s = 0.90 s between the events are also shown. The graphics show very nicely the properties expected for a Poisson process.
{width=66%}
The radioactive noble gas radon-222 is formed in nuclear decay in the earth's interior and reaches the surface through fissures in the ground. In particular, it can accumulate in the air in buildings if ventilation is poor. A Radon sample can be taken from the air by collecting the Radon decay products on a balloon that is electrostatically charged by rubbing it. The isotopic mixture is dominated by the short-lived daughter nuclei Po-218, Pb-14, Bi-214 and Po-214, as shown in the diagram below.
{width=50%}
A radon sample from air thus contains a series of short-lived daughter nuclei and is
therefore well suited to illustrate the time dependence of the activity. The rate as a
function of time is shown below; the data in the file Radon.csv have been
again analyzed with the Python script RadonAnalysis.py described below and
displayed graphically.
{width=80%}
Since several daughter nuclei are involved, the reduction of the original activity does not follow the exponential decay law. A clear increase in activity can be seen above the background level after about 12 min, when the balloon was placed on the detector after about 15 min of accumulation time. The initial mixture of the isotopes Pb-214 and Bi-214 is unknown. The decay of Pb-214 and the production of Bi-214 and its subsequent decay with lifetimes of 27 min and 20 min result in the the gamma rate over time shown above, which drops back to the level of ambient radiation after approx. 2.5 hours.
The code for generating the results graphics shown above can be found in the
Python script data/Analysis.py, which can be used as an example for your own
evaluations or as a ready-to-use program. The two data files data/ambient_lowTrigger.csv
and data/Pechblende.csv contain data recorded with the CERN DIY detector and the program scGammaDetector.py and can be analyzed with Analysis.py.
The first data set was recorded with a low trigger threshold with ambient radioactivity
only, while the second measurement was made with a sample of radioactive pitchblende
with trigger threshold well above the noise level.
Input of
python3 GammaAnalysis -h
shows the available options and parameters:
*==* script ./Analysis.py executing, parameters: ['-h']
usage: GammaAnalysis.py [-h] [-b BINS] [-i INTERVAL] [-c CUT] inFileName
Analysis of DIY Detector
positional arguments:
inFileName input file name (CSV format)
options:
-h, --help show this help message and exit
-b BINS, --bins BINS bins for Pulse Height Histogram
-i INTERVAL, --interval INTERVAL
time interval for Rate Histogram
-c CUT, --cut CUT cut on minimal pulse heightThe pulse height distribution is displayed in BINS intervals as well as the rates in time intervals of length INTERVAL; the parameters BINS and INTERVAL are specified as options. If noise pulses are contained in the data, the “CUT” parameter can be used to suppress smaller pulses when analyzing the rates.
The graphics shown above are generated with the commands
python3 GammaAnalysis.py -c 11000 -i 60 environment_lowTrigger
or
python3 GammaAnalysis.py -i 10 pitchblende.csv
resp.
python RadonAnalysis.py -i 180 Radon.csv.