IntroductionHello everyone!
I'm new to this forum, I'd like to introduce myself first:
I've been an avid follower of Dave's youtube channel for a few years now, his show is an amazing source of information for people in this field. Most of my hardware designer colleagues follow him and praise him! Keep it up Dave!
I'm one of those weird radiation nutjobs, I've worked at a company that designs radiation detectors, and my previous workplace inspired me to design my own ones.
Now granted, I've been an embedded software engineer at that place, and I could never really scratch that
hardware design itch I've always had, so I thought I should keep the stuff fresh that I learned in university, and get a nice hardware project under my belt. It looks nice in a CV to have a hardware design portfolio to go along with your paperwork.
I made a couple G-M counters before, but there's not a whole lot of high-tech stuff going on in there. My next project, that I've been working on for a couple months now is going to be a full-fledged scintillation detector.
My company was very nice to me, they gifted me with a lot of PMTs and scintillation crystals they grew in-house. One of my most prized possessions is this massive Sodium-Iodide crystal with an analog detector shown here:
That's a 2x6 inch NaI crystal used mostly in radiation portal monitors. The resolution was calibrated in-house to be 8.1%, which is rather nice for its size, and more than capable of proper gamma spectroscopy.
I present to you, Project Helios:
My goal is to replace the electronics of this thing with one that I designed from scratch. The first step to get to this would be to have an appropriate divider / pre-amplifier circuit that'll bias the PMT, amplify its output signal and stretch it a little bit.
The "specifications"So let's see what are the key parameters that were fixed from the beginning of the project:
First off, PMTs are expensive devices. I have to work with what I have, so the choice for the PMT is an ET Enterprises 9266B.
The datasheet for the PMT is right here:
https://et-enterprises.com/images/data_sheets/9266B.pdfET is essentially the daughter company of Thorn EMI, so they have a solid name in the industry.
The sad deal with this tube is that it's not in an industry standard 14-pin package, so it's not exactly interchangeable with other, more commonly available PMTs on the market, but otherwise it's a very nice, serviceable tube that gets the job done very well. It doesn't need a whole lot of biasing, it can produce a proper spectrum with a 2x2 inch NaI crystal when biased at around 600 Volts, so the IPC-2221 clearances won't be too unforgiving.
For the biasing scheme, I chose a negative bias supply, so the anode will be at ground potential, and putting a negative high voltage on the photocathode. This degrades some parameters of the tube, notably the dark current noise will be larger than with a positive biasing scheme, but it has the added benefit of being able to use DC coupling. This also allows me to approximate very large dose-rates through anode current measurement, if such a need arises.
The basic idea will be to supply the bias voltage through an SMA connector, and read the signal from another SMA connector. The maximum working voltage an SMA connector has is generally 500V RMS when using a rigid coaxial cable, which is about 700 volts peak. I have seen SMA connectors with lower working voltages, and the RMS-peak notation is sometimes mixed up in the datasheets I've seen, so it's absolutely possible that the manufacturer didn't specify the proper voltage values, but the dielectric withstanding voltage is well over 1000 volts usually, so with a little bit of elbow grease, using SMA connectors will just fit the spec. The worst I'll have is a little bit of leakage current, and that's a risk I'm willing to take to save a lot of space. Actual proper high voltage connectors are bulky and expensive. The other option would be to just directly solder the high voltage cable onto the PCB, but
come on!
The biasing networkFor gamma spectroscopy, one of the more important factors is high pulse linearity. Commonly used resistor chains don't provide me with that. The reason is simple: At high counting rates, the last dynodes in the chain pull a lot of current out of the power supply, and this will throw the other dynode stages off-balance, as the voltages will change during a pulse event. The way to fix this problem, is to replace the resistors at the final dynode stages with transistors. Other than these modifications, the resistor values were chosen according to the manufacturer's recommendation in the datasheet. For more information on this circuit, see 5.1.7 in Hamamatsu's photomultiplier handbook.
The chosen components are thick film resistors and NP0 capacitors as they are the least noisy, least drifty components that I think I can lay my hands on. Several documents advise against using carbon resistors for my application.
Gain stabilizationAutomatic gain stabilization is all the rave these days in scintillation counting. The common devices in the gone-by era used an internal Cs-137 check source, but that degrades the overall signal-to-noise ratio of the detector, especially the sensitivity for Cs-137, and I didn't want to have that, so my stabilization method of choice will be to use an LED pulser. The LED will periodically flash into the PMT, which will produce a stable light pulse for the PMT to adjust its gain to. PMTs are highly sensitive the pulses need to be short and very dim, so the LED's light pulses won't be visible to the naked eye. PWM control doesn't cut it at all, as the PMT is not a human eye. I need very precise control. The LED's color is also an important consideration in the design. Taking a nice look at the PMT datasheet, the sensitivity of the photocathode is dependent on the light's wavelength, so I chose a blue LED for my design.
With this circuit, I can achieve this. I can accurately set the LED current with this configuration. I highly recommend everyone to NEVER use voltage control to adjust an LED's light output, because thermal runaway can occur. Current control is a safer option. The circuit consists of an SPI controlled DAC, that drives an improved howland current pump. The circuit can supply about 16 mA of current, which is more than enough to drive a single LED. The solder bridges are for debugging purposes, and for the occasional "in case i mess it up" situations. They don't take much board space, they won't degrade performance, they don't cost extra money, and they're very helpful.
The preamplifierThe signal coming out of the PMT is, even after an amplification ratio in the range of
millions, still very small. And it's a current signal. I need a method to convert current to voltage, and a tiny bit of signal shaping needs to occur. An important thing to note: Even though my design is going to use a NaI crystal, "phoswich", or phosphor sandwich crystals are also a big buzzword in the industry. There are specialized crystal assemblies that allow detection of alpha, beta, gamma particles, as well as cold and hot neutrons, or any combination of these, and discriminate between the pulses. This is done through a method called pulse shape analysis. Basically, different crystals produce differently shaped pulses, and the shape corresponds to the detected particle. Also, since I use an LED pulser, I need to differentiate between live decay events, and LED pulses. The chosen charge-sensitive preamplifier configuration allows me to retain pulse shape information for further processing down the line with a high speed ADC and a DSP/FPGA. The input pulses will be very fast, and very small. I need a high slew-rate op-amp with low bias current to detect the signals with acceptable performance, and rail-to-rail input and output is also a necessity. I chose an AD8038 voltage-feedback op-amp for my preamplifier.
The important part of a scintillation pulse is its pulse height, and decay time. The charge sensitive preamplifier's pulses convert pulse shape information into rise time. All I have to do on the signal processing level, is to measure the input pulse's height and rise time to determine where the signal came from. Neat!
Various extra tidbits for convenienceAs seen in the PMT divider image, I have also included a temperature sensor in the schematic. This is for approximating the LED and crystal temperatures, so that some minimal backup stabilization can occur. I may or may not need this feature, but if I do, it's already there. The chosen temperature sensor also communicates through SPI.
The PCBI decided to go with a 2 layer PCB, and selected JLCPCB to be the manufacturer. They're super cheap. For a PCB of my projected size, it'd cost me a good 4 dollars at the fab for 5 pieces, and they were kind enough to give me a 5 dollar discount as this'll be my first purchase! That's one way to make customers come back!
Since this'll be a high voltage PCB, I have to keep IPC-2221 in mind on top of JLCPCB's own design rules, but at such a low component count and not too high frequencies, I don't think EMI will be an issue.
The board would need to be of a circular shape, and with high voltages being present, I decided to let my artistic mind go loose and tried out the radial grid functionality in Altium and used only curved traces and rounded pads to avoid making any spark gaps:
Dare I say it looks marvelous!
The finished productI sent the order to the fab, and I had it in 4 days. I didn't expect them to ship it so fast, I thought it'd take them weeks to make and send, so that was a very pleasant surprise. The board itself looked very well made, my only critique with it is that the solder mask expansions on the PMT socket's pads are missing. This may or may not be an issue as those pins can have 200 volts between each other.
I immediately grabbed my TS-80 and started to populate the board. I never soldered SMT components before, so it was an interesting challenge. The TSSOP-8 component was a real pain in the backside to solder properly, but everything else was surprisingly easy. After a couple components, you quickly learn the method.
The black solder mask with white silkscreen looks wonderful. Here are some pictures of the fully populated board, missing only the socket and the charge-sensitive preamplifier op-amp:
Getting back a little bit to that negative biasing schemeSomething very important that needs to be done with the PMT is to "dress it up". The PMT wall needs to be at cathode potential, or else some of the electrons will stray aside, possibly knocking into the glass instead of traveling down along the dynode chain. To do this, I bought some electric guitar shielding paint, which is basically just black paint with graphite mixed into it, and then connected it to the cathode with conductive copper tape.
After that, the magnetic shielding needs to be applied. Magnetic fields, for instance the earth's magnetic poles can and will also affect the electron trajectory, essentially turning the PMT into a very expensive compass with the proper circuitry. To mitigate this problem, I applied thick heatshrink tubing over the PMT, and then put some mu-metal shielding foil over that.
And here is the final assembly:
Closing thoughts and further readingIf I inspired anyone to mess around with scintillation detectors or PMTs, I'd like to recommend a couple books to start out:
Radiation detection and measurement from Glenn F. Knoll: The holy grail of radiation detection. I'm thinking about buying it in a hard copy actually. It has a LOT of useful information.
Photomultiplier handbook from HamamatsuPhotomultipliers: principles and applications from PhotonisThank you very much for reading my little "blog" about my passion!
Home
Help
Search
Login
Register
