January 20th, 2021 by Steve Foster

What do you do when you’ve built a radiation detector? Why, you build another one, of course! I wrote a while ago about how my son and I had built a cheap electron (β-particle to you) detector using a low-cost design from CERN. Well, CERN actually have designs for two cheap detectors, the other one being an alpha particle detector. So, we decided to build that too.

The alpha detector uses many of the same components as the electron detector, which is good. We had discovered that you cannot buy single quantity numbers of many components. You can only buy resistors in multiple of 10, for instance. Even the circuit board for the detector had a minimum order of three, so building another detector was a good way to use up surplus components.

Building the alpha detector is only slightly more difficult than building the electron detector, but there were other factors that made a two hour job drag on for 8 weeks. My son had returned to school and had homework to contend with, and now seems to have a wide circle of on-line game-playing school mates competing for his attention. Also, I had returned to study myself by starting an MSc. course in Astrophysics. It was only when the schools and universities closed down for the Christmas break that we were able to complete the assembly.

(Note the detector diode – minus glass cover – on the far left. The marks left by the pliers are plainly visible)

Snags and niggles

Unlike with the electron detector, there were a few minor problems in building the circuit board. First of all, one of the components in the KitSpace parts manifest was wrong and we were sent the wrong resistor for one component. (Don’t panic! This is now fixed on the site.) Then we discovered that one of the holes on our circuit board hadn’t been drilled out properly. This was easy to fix with a large darning needle. The biggest problem was the surface-mounted 40MΩ resistor: this thing is tiny! If you sneezed or breathed too hard, it would vanish into the distance and take forever to find. (I recommend you work over a sheet of white paper when positioning this part.) Despite all this, we soon had all the parts bar one successfully attached to the board and continuity checked.

The final part to be added was the silicon detector diode itself. This is a tiny, flat silicon chip in a metal can with a glass cover. As even this small glass barrier will block alpha radiation, it has to be removed – carefully, as the devices cost around £10 each. We used small cutting pliers to make 4 small dents in the case around the edge of the metal can. Each time required a gentle, slowly increasing pressure on the plier handles until the glass cracked. After four such cracks, I turned the diode upside down and flicked it with my finger. The pieces of the glass cover fell out easily, and the board was soon complete.

As previously, we mounted the board and its various connectors and switches in an old metal watch box provided by my wife. We connected it to a web oscilloscope, powered it up and I flicked the metal case with my finger. A waveform on the ‘scope meant that everything was looking good.

Finding your alpha source

You can’t test an alpha detector without a source of alpha particles. This is more difficult than finding an electron source, since pretty much everything containing potassium (e.g. a banana) gives off β particles. Uranium is a good source of alpha and surprisingly common. There seemed to be two reasonable options: uranium glass or Fiestaware.

Uranium glass is a decorative glass that was popular from the mid 1800s right through to the mid 1900s and, as the name suggests, it contains uranium. It has a characteristic greenish colour and fluoresces nicely under UV light. I’m pretty sure my parents had a glass bowl made of the stuff when I was a kid. However, the more I read about uranium glass the less confident I was that it was a reliable alpha source. Alpha particles originating in a lump of uranium glass are most likely to be absorbed within the glass itself, and the density of uranium near the edges of the glass is very low. The stuff was just too safe and inactive.

Vintage Fiestaware looked a better bet. It’s a type of domestic crockery popular in the US in the 1930s-1970s because of its deep orange-red colour. That colour is caused by a pottery glaze laced with uranium, so all the radiation comes from a thin layer on the outside of the material. We found a supplier on eBay and sent off for a small sample. Had I known exactly how small that sample would be for £20, we would have been better off buying a whole Fiestaware cup (£29.99 on eBay), taking a hammer to it, and selling on what we didn’t need!

Anyway, we now had a reliable alpha particle source and a mildly amused wife and mother.

Using the detector

As we discovered when building the electron detector, the CERN software for it actually is optimized for the alpha detector version. There’s a web oscilloscope, and a python script for recording pulses, plus another python script to process the data files. The platform that we chose to run the software was different to what we’d used before. My 10-year-old Asus mini laptop simply wasn’t up to the job, but a Raspberry Pi 4 model B performed excellently. We equipped it with the same cheap soundcard that we’d used previously. (There’s an important step with this, which is to use the “alsamixer” utility to set the recording level to max otherwise you’ll be underrecording the energy levels of the pulses.)

First step: check the detector works with the web oscillosope. We actually used the detector upside down so that a shard of Fiestaware could easily rest on the diode detector. Here’s what we saw:

It doesn’t look much at this stage, but it confirms both that the detector is working and that our Fiesta ware really is Fiesta ware!

The python data recorder is much more reliable and is suited for collecting large amounts of data, but there are limits. If you’re going to do any sort of series analysis, you’ll probably need to collect data overnight. Many of the pulses will be electron pulses so we’d decided to collect at least 100,000 pulses, expecting about 60,000 of these to be alpha pulses. Our pottery shard gave us around 4 counts per second, so that’s 25,000 seconds or roughly 7 hours of collection time. We left the detector running for about 15 hours and could see that it had collected about 145,000 pulses. Unfortunately, the recorder holds the pulse data for all of these pulses in memory and, when I tried to save the data, the Raspberry Pi (4GB) ran out of memory. The whole collection run was lost! I tinkered with the collector script settings (setting the parameter THL = -1000) to eliminate most of the electron pulses and began another long collection run.

Interpreting the output

Recording the data is only half the story; displaying and interpreting it is a journey of its own. It took us quite a few attempts, with a lot of input from the ever-helpful Oliver to understand the energy spectrum from my pottery shards. There is a data analysis python script on the project wiki and this can be used to produce some really great graphs, though you may have to tinker with the settings a little.

First, let’s look at the first data plot that we produced:

Given that our detector seemed to be working perfectly, I was a little disappointed that our graph (the blue line above) didn’t look anything like I was expecting. I wasn’t expecting it to look like the red line as this is a simulation for a completely different radiation source. What I was expecting was something like the output from the glazed pendant that’s described in the wiki. Instead of a broad peak around 4MeV, we seem to have the tail end of the electron peak (below 1MeV) and then just a broad plateau with no discernible features. What had gone wrong? The answer, as it turned out, was “nothing”. Manipulating the output from the sound card is a just a bit of a dark art. So here is what we learned.

  • The electron peak at low energies is H-U-G-E. If you use the analysis software without modification the it will scale the y axis for this peak, and the detail for the alpha pulses will be almost washed out at the noise level. For our data, I had to set a parameter in the script called “max_y_level” (look around line 490) to a setting of 400. If you try this for yourself, you’ll need to adjust it to whatever works for your data.
  • Not all sound cards based on the same chipset behave the same. Our sound card was based on the same CM108 chip that’s mentioned in the wiki. However, ours seems to have some additional amplification factor in it that was amplifying the alpha particle pulses, and hence their apparent energies, by a factor of almost double compared to the reference ones. There’s another parameter in the script that extends the display plot to the right to show [supposedly] higher energies. I had to set “max_bin_kev” (around line 480) to 16,500.

After these adjustments, we re-ran the analysis and we got this:

(Ignore the MeV scale on the bottom as it’s for a differently calibrated soundcard)

Now, that’s more like it! We have a working detector and working software. You can follow the discussion on getting to this point here.

What next?

So, this is where we are at the time of writing. There’s still a bit of work to do to calibrate the detector properly. It may be that’s some way to adjust the amplification of my sound card back to the same levels as the reference cards and we’ll continue to look for that. Plus, my son and I will look for some more radiation sources like the radon balloon or one of the many igneous rocks that I have acquired over the years as part of a small mineral and fossil collection. I’ll keep you posted.

Posted in blogposts Tagged with: , , , ,

CERN DIY electron detector and connected laptop
September 1st, 2020 by Steve Foster

What do you do when you’re stuck at home with a 16-year-old son during a pandemic lockdown? Well, the first answer is that you build a gaming PC. That was a good project with a lot of incentives to complete, but that also meant it was soon over. So, what next? I spotted a post about an outreach project from CERN to build your own particle detector for under £30. Sounded like fun, so here’s how we got on.

Firstly, a bit about ourselves: I’ve just retired from an IT architecture role at a major UK bank, a brave decision when you still have a child at secondary school and there’s a global pandemic looming. My son, Sam, has just completed his GCSEs and is about to enter the sixth form. I have a degree in Physics and Sam has a keen interest in scientific things. We both suffer from hypermobility of the finger joints, meaning that we have limited fine motor skills (translation: we are both klutzes. Even klutzes can successfully build this detector.) I haven’t picked up a soldering iron in anger for a long time and Sam has only done two or three connections for a school project so we are effectively total noobs to this sort of project.

Anyway, here is a write-up of our experience and lessons learned rfom building the electron detector. If you’re planning on building one yourself then you may find this useful

Buying the components

The great thing about the CERN kit is that you can order most of the parts easily using kitspace. Can you really build it for under €30 (about £27) as claimed? Well, yes and no. The exchange rates have all moved, which doesn’t help. The individual components certainly cost less than €30 but you cannot just buy individual components. You have to buy 10 resistors at a time but they are dirt cheap and you’ll just end up spending 30p instead of 3p for the parts you need. The biggest surprise was that you have to buy a minimum of three circuit boards, but this turned out to be a blessing in disguise. It’s handy to have a spare board if your soldering skills are questionable, plus the alpha detector variant uses the same board so you have an incentive to build that model too. All-in-all, the cost of all the electronic components same to just under £40. If you were buying for a class project – a dozen or so boards – I think the bulk price would be closer to £33/€37 per board.

There are some optional costs too that you may want to take into account. We bought a BNC to audio jack cable for about £8 on Amazon as this was quicker, easier and ultimately cheaper than building one from scratch. We also had to buy some tools that we needed but didn’t have in our existing toolbox. You may prefer to buy a metal case or project box for around £6, however I would recommend using an old sweet tin or similar. We used a presentation case off a Fossil watch that my wife received one Christmas. To mount the board in the case, you’ll need some sort of metal standoff. We bought a box of M3 brass standoffs of assorted lengths with all the necessary screws and bolts for about £11 on Amazon. (120 components when you need only one!) You’ll also need some small lengths of connecting wire – about 30cm in total. (Tip: Amazon Prime is well worth the cost if you need to order all these things as you go along.)

Finally, you are going to need some sort of radioactive source to try out the detector. We bought 100g of analytical quality potassium chloride (KCl) on Amazon for under £5. This is a better price than that of the potassium-rich salt substitutes that are also on Amazon and has a higher proportion of potassium to boot.

Tools that you’ll need

Before you start, make sure that you have all the tools that you’re going to need, Here’s a list of what we found necessary:

  • A soldering iron with a small tip suitable for the components and board
  • A de-soldering pump and braid for when you mess things up
  • Small cutting pliers for tidying up excess wires when you’ve soldered a component
  • Wire strippers for the ends of the connecting wires
  • A small multi-meter for testing joints and checking the board
  • A nail punch to make holes in your old tin box
  • A file to file down the sharp edges on the holes that you’ve just punched
  • Some things to hold the board and components in position while you solder them. We used an old wall tile and a hexagonal pencil.

Constructing the detector

We spent about three elapsed hours over about 3 weeks building the board and maybe another hour building the case and mounting the board into it. Why so long? Well, this project didn’t have the same appeal to a 16 year old as does building a gaming PC. Plus, use of said gaming PC frequently took priority over this project. We decided to solder the components in groups, doing the resistors first, then the capacitors then the semiconductors. The build instructions in the wiki are pretty good and easy to follow.

We tested every joint as we soldered it, making sure to test from another point on the circuit board so as to prove the connection was good. My son did all the soldering on one of the PCBs. I started work on a second board partly as a backup and partly to have something that could form the basis of the alpha detector. (I discovered the hard way that the two variants begin to differ at the third component, R3!)

Here are some shots of the board at different stages of construction:

Stage 1

Resisitors from the bottom
Resistors from the top

Stage 2

Capacitors bottom view
Capacitors top view

Stage 3

Completed board, tbottom view
Completed board, top view

Stage 4

The completed detector in its case with switch, battery and connector

Testing the detector

The first thing to do when testing the detector was to make sure that there were no shorts by measuring resistance across the power connectors. (Tick! A satisfactory 9.8 kilo-ohms) and make sure that power was reaching the board. Then we plugged it into a PC and switched it on. We were using the pulse detector JavaScript to check that it was working. I flicked the switch and…nothing. So, we examined the board. We knew we were novice constructors and some of the joints looked potentially to be dry and on one or two others we could not be 100% sure that we had an air gap between adjacent joints so we de-soldered and remade about 3 joints. Again we connected to the PC and we saw the satisfying pulses when the device was turned on and off. We also saw quite a bit of microphony being picked up when the tin was moved or when one of us spoke near it.

Ok, so next step was to try it with a radiation source. I weighed out roughly 10g of KCl into a plastic lid and positioned it under the diodes in the detector. We hadn’t made a radiation window in the lid of the tin and the diodes were about 1.5cm above the surface of the KCl. I switched on the detector and…nothing, not even the on/off pulses. The detector seemed dead. I took the device apart and retested very connection with the multi-meter and made sure all the case connections were good and that the case was properly wired to earth. I checked that power from the battery was reaching all the necessary parts and that the battery power was good. I could find nothing wrong.

It quickly became apparent that there was an intermittent problem. Sometimes, the JavaScript detector would see a pulse when the device was switched on or off and sometimes – usually – there was nothing. Sometimes you would see noise on the web oscilloscope and sometimes not. I was scratching my head as to what the problem was. Eventually, it occurred to me to try actually listening to the audio output, a trivial thing to do under Windows. BINGO! I could hear white noise when the device was switched on and, just audible over the noise, a soft click every 20-30 seconds. Clearly the detector was working and therefore what I was facing was a software problem.

The importance of sound cards, and repurposing yet again an ancient laptop

I have to say that Oliver Keller was an enormous help while I was trying to get the detector working and I cannot stress enough how important it is to listen to his advice about audio connections and sound cards. The problem seems to be that sound cards (or chips) in modern PCs are too clever for their own good. They are designed to optimise microphone input for Skype and Zoom and that means removing the noise and extraneous clicks that are exactly the signal that the detector puts out.

Modern technology was assuming the detector signal was unwanted noise so I needed some primitive technology. I have an old notebook laptop, an Asus EeePC 901. I bought this about 10 years ago and it has been the subject of several hobbyist projects over the years. I upgraded it’s memory to 2GB. I replaced it’s laughably small 4GB SSD with a 32GB one. I upgraded it from Windows XP Home, via several versions of Ubuntu, to Windows 10 Home. You may be surprised that such a machine can run Windows 10 and MS Office but it can do so, just not well. I backed it up (virtualized it for VMWare) and installed a fresh copy of Lubuntu. I fired up the web oscilloscope under FireFox and now I have a working detector:

You can quite clearly see a detector pulse in the screenshot. The noise sometimes peaks and can give a false reading but these are fairly easy to spot. The false pulses are one sided and lack the rebound peak of a true pulse. From around 25g of KCl, I’m seeing two or three pulses per minute in the detector

So what comes next?

OK, now that the detector is working, what comes next? Our first plan is to try out the python script to make sure that we can see and correctly analyze the energy spectrum from the KCl. Step three will be to try various things to reduce the electrical noise. I have a suspicion that the tin leaks a little light into its interior but that can be remedied with black tape. Step three will be trying out various samples from my small mineral collection in the detector; there’s bound to be something mildly radioactive in there. Step four will be a bit of fun. You may have heard the old adage that even bananas are radioactive? well, we plan to prove it by measuring the beta decay rate from a lump of delicious, potassium-rich banana! I’ll keep you posted on how we get on.


Posted in blogposts Tagged with: , , , ,