Introduction
I came a long way, since when I first started to build nuclear instrumentation. For projects like a Geiger Counters there are plenty of example projects out there to rebuild. But Gamma Spectrometer projects are rare if they are self build and so I build one my own with lots of experimentation and some parts of which I’m documenting here. I won’t be documenting the high voltage power supply. This can potentially kill you and if you know what you are doing, you can safely build one by yourself or you know how to use a bought one. And if you don’t know what you are doing, you better buy a pre-build spectrometer.
Always handle radium watches, thorium lantern mantels, uranium glass etc with care. They are radioactive and you should never underestimate them. Just because they are radioactive products on the open market freely available and therefore might give you a false sense of safety, doesn’t mean they are automatically safe. They are as safe, depending on how you use them. So store them far away from where you normally are at your place and so only you can access them. Preferably in shielded containers and always use them with caution, when taking spectra.
Twitter: @Pico_Ampere
Getting the Signal (negative HV)
The easiest way to get the signal with that precious spectral information out of the photomultiplier(PMT) is a transimpedance amplifier(TIA). This circuit works only for negative PMT supply voltages.
The circuit is powered via USB and filtered. The negative rail is made using a ICL7660 with a filtered output. The signal is conditioned with a TL084. A similar OP-Amp can be used. R7, R8, C8 can be omitted, if the pole–zero correction has no effect.
The anode of the PMT is directly connected to the TIA input, with two antiparallel diodes at its input for protection. The TIA converts the current pulses from the PMT’s anode into voltage pulses, which are further amplified with a selectable amplification and the send through two high pass filters to get rid of noise and to prevent DC voltage from the sound-card microphone input to enter the OP_Amps output. One of the outputs goes into the microphone input of a sound-card and one is passed onto a BNC connector. A better solution for a signal filter would be a simple CR-RC shaper circuit, but the noise of the system is already good enough, so without it the performance won’t be degraded. But it could happen that with other detectors (LaBr), different applications (x-ray spectroscopy) or a high sensitivity detector (Large scintillator –> pileup) a CR-RC shaper is needed.
This image shows how to connect a PMT for a negative supply voltage. The schematic shown is for a R580, but the principle is valid for any tube. Check the respective datasheet for the pin-out. Here N°14 is the cathode and N°7 is the anode. The rest are dynodes, or secondary anodes. Some tubes also come with a additional grid or modulator input. Just treat that like another dynode. The standard resistance is 1M 1/4W or whatever you picked. Pick one where you can find easily a multiple from the standard resistance, given by the ratios in the datasheet. The capacitors are 47nF 400V. The resistance ratios between dynodes can vary according to the datasheets. The FEU85 between pins six and eight has a ratio of 1.5 from the standard resistance. This means 1.5M for a standard resistor of 1M. A ratio of 0.5 from eight to ground. This means 0.5M for a standard resistor of 1M. The anode is the output and connected directly to the TIA’s input.
For example, here you see what it looks like in a XP2202 datasheet:
Getting the signal (positive HV)
Most of the things said for negative HV are valid here too. What changes is of course the detector circuit and the connections to the PMT base and that your signal isn’t DC but AC coupled. Generally I prefer positive HV for spectroscopy. It doesn’t really matter what HV polarity you use in that case. For other applications it definitely matters, but that doesn’t fit in here.
The circuit used here is a CSA. Charge sensitive amplifier. It’s gain is roughly 1/Feedback_C . It cant be made too small, because than you would loose information, since it can’t store the charge coming from the gamma pulse. Also it may cause oscillation. The circuit shown has a secondary amplification stage for that reason. You can use any op-amp fast enough, just use always the same or a faster one for the second stage. Otherwise you will get a distorted pulse, like I got when I combined the LTC6228 with a AD8055 in the second stage. You could perfectly see that the latter one was too slow, because the nice gamma pulse was looking more like a saw-thooth signal.
The divider in between stage just gives more flexibility in adjusting the output signal level. Also offset can be added to the final signal output.
The schematic is from my first completely DIY gamma spectrometer. This includes the MCA and PC software. It is work in progress and once done, I will post it on my website.
The PMT connection diagram is from Hamamatsu. This is how you would configure your PMT for positive HV. First three resistor of the divider have a capacitor in parallel. 10-47nF 400V. Normally you have PMT’s with 8-10 dynodes so just adapt. PMT datasheet should give you the resistor ratios for the divider.
For the anode the resistor is 100kOhm and the signal decoupling capacitor is 10nF 3kV. The high voltage rating is for safety. The PMT runs normally on 500V-1kV.
Digitizing the signal and MCA Software BeqMoni
A sound card is used to digitize the signal. But not any sound card can be used. You have to try some. Some integrated sound cards in PCs and laptops can be of insufficient quality. Also don’t bother with cheap Chinese USB sound cards. The don’t work for this application, I tried some and all are unusable for that purpose. I only found that a SoundBlaster Play 3 is of a decent enough quality that it can be used for spectroscopy. The sound card can be read with a free software like BeqMoni. I highly recommend you use this software, since it gives good results and has also calibration possibility, but it is not so easy to use. If you want quick and dirty spectra, use Thereminio. But you might get strange artifacts in the spectrum and measurements errors.
In order to use BeqMoni you first need to set up a detector. Under Tools -> Edit Device Configuration you make a new device. Then in the “Device Specific” tab you make the necessary inputs: 192khz sampling rate, 24bit per sample. Now comes the experimenting. Depending on your setup and detector, the input level, LLD, ULD and shape threshold need to be adjusted. Under “Standart Pulse Acqusition” you also need the adjust the sample width and peak position and also the LLD and ULD values there. If all is set you click start. You’ll see pulses appearing and they should be in the middle of the little screen and have a decent shape, like in the pictures. No worries if the shape isn’t really the same, as in the example oscilloscope picture. Some under-swing or other slight deviations are no problem. See the second picture below. This pulse shape is fine and gives good spectra. Wait for the software to have acquired 10k pulses. It will stop automatically.
Below are the settings of one of my detectors. This is a sowjet/russian 30x40mm NaI(TI) with a R9420. The PMT is way to good for this crystal. FWHM is the width of the pulse at half height, divided by the energy this peak is at. The lower this percentage, the better. The smaller the FWHM, the smaller the peaks, the more distinct peaks you can see. Ideally there would just be a straight line for each energy peak.
The R9420 can achieve a FWHM(resolution) of 3% at least. Therefor the crystal determines the FWHM of the whole detector assembly. It is around 8-9%. The Hamamatsu PMT was switched for a more fitting FEU85, which is the cheapest PMT I would use, if I had no money for something else. It still gives around 9-10% resolution and there are sockets available. Don’t solder directly onto the tube pins. It might damage the PMT and makes it difficult to change the socket circuit.
In order to do energy calibration with BeqMoni, go to the “Energy Calibration” tab first, then to “Multipoint Calibration”. Click “Pick up channel” and select a peak by clicking on one side of it and pulling the mouse to the other and then led go. You should see that the peak gets marked. Again click “Pick up channel” and do the peak marking two more times. In the “Energy” column of the marked peaks type the energy they should have, then click “Execute calibration” and energy calibration is done. You now can identify isotopes with this specific detector. Also if you want you ca save the calibration to the device configuration of this detector, so you don’t need to do calibration every time.
Analyzing Gamma Spectra with InterSpec
BeqMoni and also Thereminio can export the taken spectrum as a .csv file. With Thereminio you need to delete the heading that gets put there. Otherwise Interspec wont open the .csv file. Interspec lets you do energy calibration on the spectra, measure the FWHM at a specific energy and much more. It’s also available as a Android app.
To do decent energy calibration you select a minimum of three peaks by double clicking under one in the spectrum. You might need to adjust the peak area manually. To do that hover around the sides of the peak and some “grab points” will appear. Adjust for a good coverage of the peak. Another way to mark peaks is to hold CRTL(STRG) and mark the area where the peak is and after mouse release, you will be asked how many peaks you want to mark.
Should there be only one visible peak, say like in Cs137, you just mark it and under “Energy Calibration” you just select the linear therm for calibration on the right, after you added the nuclide and peak energy. But if you also have Kr85, you can’t distinguish that peak from a Cs137 easily. But if you use another spectrum with Interspec before that has known peaks, and then switch to the one with the unknown peak, Interspec asks if you want to keep calibration of the previous one. Click that and you will know if it’s Cs137 or Kr85 in this particular case.
Shouldn’t the peaks appear where you want them, you can hover over the ends of the marked parts of the spectra and adjust manually. After that, follow the “Energy Calibration” steps.
Then under “Energy Calibration” you type the nuclide you expect the peak to be from, and under “Photopeak” you type in the actual energy of that peak. After you are done click “Convert to Polynomial”, then “Fit Coeffs” and your energy calibration is done. By howering over a marked peak you can see the FWHM and other data.
Fast calibration can be done by holding CTRL+ALT(STRG+ALT) and the mouse and moving it around. This adjusts the spectrum at the energy where you clicked.
Newer version of the software allows you to combine channels, so it doesn’t lag as much. The reason for the lag before combining channels is that the software was made with HPGE crystals spectra in mind, not spectra from NaI or CsI. This option is found under “Energy Calibration”. You enter a number meaning how many channels you want to combine to one.
Building the detector
The best place to find PMT’s and scintillation crystals is Ebay. But be aware, not any scintillation crystal can be used for spectroscopy. Look for NaI(TI) crystals. You want white and clear crystals with the glass attached. But they are very rare now. You can also buy some from China, new. I can confirm they are of a very good quality and worth the money. The left crystal is a brand new one from China. To it’s right, there is a old yellowed Sowjet/Russian one. This crystal you don’t want to use, because it’s damaged by water. Therefor the yellowness. I have good clear ones from Russia/Ukraine. One 30x40mm crystal has a resolution of 8.5%. The new one from China reaches the achievable limit of sodium iodide crystals.
Used Harshaw and Bicron crystals and detector assemblies on eBay are quite bad. They have often bad resolution >7.5% (yes, I’m spoiled in that regard) and seem to be defective more than others, so that water enters the crystal and it gets destroyed. So try to avoid them unless the listing shows you a good clear crystal with good casing and also a resolution value @662keV or equivalent, the resolution with Cs137. Most Bicron detectors I saw on eBay had indeed spectra with a resolution in their listing and it was >10%. So useless in my opinion. But if that’s fine for you, go ahead. At least they were cheap.
Plastic scintillators in general can’t be used for spectroscopy, so lets forget them for now.
As with crystals, PMT’s are rare and hard to find. Don’t bother with old Russian FEU tubes. They have a FWHM thats to high. This value determines how narrow a peak is, normally measured with Cs137. It also can be found under the name “Pulse height resolution in datasheets”. Here shown with the XP2202 datasheet as a example.
You want a narrow peak, because that means a high resolution. This way you will see more peaks than with a high FWHM/low resolution, that results in a larger peak and therefore hides smaller peaks.
Try to find head on Hamamatsu PMT’s that fits your crystal dimensions (the PMT has to be slightly bigger than the crystal) or other PMTs that have a confirmed resolution or a high blue light sensitivity and a high PHR value. Since vacuum tubes have rather large production tolerances due to their mechanical nature, PHR values vary quite a bit with tubes and are often printed on them. I found that Hamamatsu PMT are most likely work, than others. I can’t confirm if a high blue light sensitivity and high PHR are a good indication of a high resolution, though it seems so.
They often have resolution values given in the datasheet, so pick one with a high resolution. But these tubes are not cheap, but probably the best and definitely worth it. There are of course other manufacturers out there, but I can’t say anything about them, since I didn’t came across PMT’s made by them. Don’t bother with PMTs that have unknown data or are old. They mostly aren’t any good for spectroscopy. I wouldn’t really go above 8% FWHM. Find PMTs that give you max and min values for resolution and not only typical values. Because you might end up with something not really usable, like it happened to me with my XP2202B.
PHR is another name for resolution. It literally means pulse height resolution. It is normally measured with Co57, but check for which isotope this resolution is given. If it would be Cs137, that would be a bad PMT.
Unfortunately resolution varies with each tube, so the datasheets gives you only a rough idea what that resolution is. This is why you want minimum and maximum values. So you know if its worth risking to buy that PMT, because the FWHM wont get higher than let’s say 7%.
Also pick a PMT where you ca find a socket for or you can improvise something, as long it is something to plug.
Don’t solder directly onto the tube pins. It might damage the PMT and makes it difficult to change the socket circuit.
Here is a list of known good PMT, tested by me:
Hamamatsu: R9420, R878, R7724, R3550, R580, R6232, R6231
Photonis: XP2202B (Tested: 9.5%). Hear you see why “typical” values in datasheets can be misleading.
After that with some silicone grease couple the PMT and the crystal. Use very little, just like you would put thermal grease onto a CPU. You will have a good transition of light from the crystal into the PMT this way. Any unevenness in the surfaces won’t be a problem this way. Then the detector assembly needs to be made absolutely light tight. Otherwise it wont work and the PMT gets damaged. If you have done that and connected the divider to the PMT according to the datasheet and connected the high voltage and signal output correctly, you can start plugging things into your PC and start your software MCA of choice(I really suggest BeqMoni). With a radioactive sample you can start taking spectra. I suggest old radium watches or a thorium lantern mantel.
The detector assemblies in the picture are build by a friend.
High Voltage Supply
The only thing I will say about the HV power supply is that it has to be a clean one, with just a few mVrms. Any noise on the supply will influence the output signal and may degrade your resolution.
Next Level Gamma Spectrometer
The RedPitaya platform offers a App to transform the RedPitaya into a excellent Gamma Spectrometer. Add a high voltage supply, input circuitry and you are good to go. There will be a follow up post, when it is build.
A lot has changed
Time passed and I progressed from sound card based gamma spectrometers to ones with a RedPitaya. Every time the quality of the resulting spectrometer improved.
Recently I learned that Stemlab 125-14 with the revision 1.0 are newer than RedPitayas with the version 1.1. Unfortunately the newer ones have noisier switching regulators.
That prompted the start of a complete DIY gamma spectrometer, including the MCA, based on the STM32 series microcontrollers.
Stay tuned for the post featuring the first prototype. A little teaser in the image below. The first spectrum with the new gamma spectrometer. Perfectly visible radium spectrum from a radium watch.
Twitter: @Pico_Ampere
