Quantum random number generator

[rdagger]

I built your particle detector (BPX61 version with glass removed). Thanks for taking the time to publish this great project! My goal is to build a quantum random number generator based on radioactive decay as a source of entropy. I’m trying to test the detector using a 25 mm diameter glass marble doped with 3% Uranium 238 that I purchased from United Nuclear. I’m not getting any pulses on my hardware oscilloscope and I don’t hear anything but noise when piped to my computer’s sound card. I followed your troubleshooting guide and I believe the hardware is assembled correctly.

I’m assuming that there is a problem with my testing methodology or that the uranium concentration in the marble is too low to register on the detector. I’d really appreciate any suggestions.

Here are some pics:

[ozel]

That is a great build! it looks totally awesome. Easily wins the MacGyver award, hands down! :D
May I resued the pictures? If so, please let me know how to cite them.

From uranium glass, there are not many alphas if at all, usually. But you should still be able to detect quite some electrons and gamma photons with the BPX61 version. The electron detector variant with more diodes would give you 4 times more events.

Which troubleshooting steps do not work for you? Does the detector react to light when the case is opened?
Also check if the BPX61metal case is maybe touching the circuit board. There is a chance for a short circuit if it rests evenly on the board surface. If so, melt the solder again and lift it up just for a millimeter.

BTW, later for analysing the randomness of you detector data, you can checkout this nice website: https://www.fourmilab.ch/hotbits/

[rdagger]

Please feel free to use the pictures. You can cite them to rdagger. Once I get it working I will post a video on my YouTube channel: https://www.youtube.com/rdagger68 and on my website www.rototron.info.

I didn't have any issues troubleshooting that's why I think it is the marble. Although I am seeing a lot of noise. I'll post a few oscilloscope pictures in a few days. There is 10K ohms across the battery leads. I see a difference on the scope when light is present and there is a ping on the scope when I tap the case. I used a 3D printed spacer when I soldered the BPX61 to ensure its case was isolated from the PCB.

Thanks for the link and the feedback!

[rdagger]

I installed a new 9.6V battery and replaced my scope leads with a 30 cm BNC cable plugged directly into the detector. This fixed the noise problem. I'm now getting readings on my scope that look like your samples.

Unfortunately, my doped marble is only emitting about 3 pulses per minute so I need to find a better source. I know your site dissuades the use of smoke detector 241AM disks but I've done some research and it appears relatively safe as long as you don't try to remove the isotope from the disks. The americium is electroplated onto the metal disks and is protected by a thin, gold-metal layer that is very chemically resistant. There have been 2 documented cases of people swallowing the disks without health issues despite over 2 weeks in the GI tract.

Here are my references:
Intake and Subsequent Fate of a Ceramic Particle Containing 2.85 μCi 241Am
Smith, L. R.; Sullivan, P. A.; Laferriere, J.; Cumming, E.; Demis, D.
Health Physics: April 1983 - Volume 44 - Issue 4 - p 329-334

Ingestion of 241Am Sources Intended for Domestic Smoke Detectors
Rundo, J.; Fairman, W. D.; Essling, M.; Huff, D. R.
Health Physics: December 1977 - Volume 33 - Issue 6 - p 561-566

[rdagger]

I found a vintage Kodak 814 camera on eBay for $10 US. I believe the Ektar lens is doped with thorium dioxide. It works with the alpha detector but I switched to the electron detector because I'm getting better results with the ADC.

I still need to work on the ADC. I'm using an ADS1220 but it only has a delta-sigma sampling rate of 2000 SPS (samples per second) which is slow for 100 μs pulses. You mentioned you had good success with the ADS1115 which is 860 SPS. Do you have any strategies for capturing the pulses with an ADC?

[ozel]

nice! if you see browning or a yellow taint while looking through the lens it is very likely thoriated (so-called color centers, due to the radiation). Sometimes those lenses emit even alphas right from the surface of the glass.

For the alpha detector, ~50k SPS are needed for recording the whole pulse (deducing energy spectra).
For the electron detector, it is enough to just trigger on pulses and count them, so much lower rates are fine. In that case, internal ADCs come in handy because they are quite fast and linearity/accuracy doesn't matter for detecting pulse edges.
SAMD21 and SAMD51 on Adafruit/Arduino zero/M1/M4 boards work quite well for counting fast pulses from the electron detector.
Keep in mind that the sampling rate setting usually messes with the ADC input impedance which in turn affects the pulse size seen by the ADC: https://blog.thea.codes/getting-the-most-out-of-the-samd21-adc/.
The oscilloscope input impedance is usually fixed at 1 MOhm but ADCs may have 10 to 100 times less, "eating" into the signal. So at the ADC, pulse amplitudes shrink but the negative edge slope is becoming less steep (=slower), too. I thought the ADS1115 was fast enough to catch the negative electron detector pulse edge reliably (not the whole pulse shape) but my memory might be wrong. It also has a comparator mode, that can generate an interrupt. Maybe I was using that.

[rdagger]

Thanks for the comparator tip, which is very helpful! The ADS1220 doesn’t have a comparator so I’ll try out the ADS1115. I’m also experimenting with the SGTL5000 which is a low power stereo codec chip that has analog input, amplified output and an I2S port.

[ozel]

The SGTL5000 is certainly a great choice for recording whole alpha pulses.

But again, please don't waste time trying to record complete pulse shapes if you use the electron detector. While the pulse size is still proportional to the deposited electron energy, it is usually incomplete because some electrons exit the diode after depositing only a fraction of their energy. You won't get a useful energy spectrum or anything interesting out of the electron pulse shapes (at least when recording from typical gamma check sources, or very mixed particle sources like uranium glass or a thoriated lens). See the discussion about figure 7b in https://www.mdpi.com/1424-8220/19/19/4264.

The ADS1115 comparator is still relying on the ADC, btw., meaning it will be just a bit faster if you use the external interrupt function compared to evaluating the ADC values for certain thresholds in software.

I would be curious to know how exactly it did not work with the ADS1220. My approach would be to connect an oscilloscope in parallel and print out short ADC value arrays via Arduino serial plotter once a certain threshold got underrun.

[rdagger]

I did get the ADS1220 working but I wasn’t 100% confident that I was catching all the pulses. This project has been on the back burner for several weeks, but it felt like the audio count averaged higher than the ADC count. It could be because I wrote the ADS1220 driver and I’m using MicroPython. That’s the reason I like the idea of the comparator to compensate for the interpreted code. The reason I like the SGTL5000 is because it can provide audio feedback without any programming overhead.

Here's a Dungeons & Dragons dice demo using the quantum random number generator powered by an ESP32 running MicroPython.

: