HPGe Detector, Part I: Repair

A high-purity germanium (HPGe) detector is the ultimate instrument for energy spectrometry of gamma radiation.  For the nuclear hobbyist, an HPGe opens a window into a fascinating realm of  home-accessible, low-intensity nuclear reactions that are obscured by background in other detectors lacking the superlative resolution.  Weak alpha sources available without a specific NRC license can be used to detectably excite (a,n) and (a,p) reactions attended by emission of gamma rays from product nuclei.  Radioactivities induced at the fractional Bq level by weak (a,n) or DD fusion neutron sources can be identified.  The downsides of HPGe detector ownership are obvious to most amateur scientists who have considered them: they’re fragile, consume liquid nitrogen, and—perhaps most significantly—require multidisciplinary knowledge to return to operation.

I was kicked into these uncharted waters when Taylor Wilson sent me an older 2″ Ortec coaxial HPGe detector in unknown condition, and I hesitantly began an effort toward making it work.  Right away I knew Lady Luck hadn’t smiled on me: the input FET was blown.  As I detail in the gallery below, I replaced it with a $2 Japanese audio FET, rigged a vacuum pumping scheme for the Dewar, adjusted the preamplifier, and—voila!—the thing works now, ultimately providing about 1.7 keV FWHM at 662 keV.  From my limited experience repairing an HPGe detector I can’t generalize too much, but perhaps other amateur nukeheads will find encouragement in the success story documented here.

Gallery 1: Teardown and Repairs (click any image for larger captioned version)

 

The following steps comprised my path to a working detector.  Additional details for some procedures can be found at TRIUMF’s website.  To make these repairs, you need an oscilloscope, an MCA, an electrometer, some NIM-standard electronics, and a high vacuum system.

1. Demount and test the HV filter.  Jon Rosenstiel has found the filters to be a weak point in his repair experience.  Not only will blown resistors and capacitors in the filter prevent the detector from operating, they can cause failure of the input FET.  Make sure the filter’s through resistance is a stable high value (200 MΩ in my model).  These filters do not appear easy to replicate or repair, so if yours is bad, you can pretty much count on spending $500 for a new one.  Nice to know up front before getting too involved in the project!
2. Test the detector’s preamplifier.  With low-voltage power applied from a NIM bin (but no HV bias), monitor the preamp output for noise on an oscilloscope or MCA.  At room temperature, there will be lots of thermal noise if the FET is alive.  If you’re lucky and your FET checks out, skip the next two steps.
3. Replace the FET.  You can either pay hundreds of dollars for a new one specially culled by the manufacturer…or you can take a little pot luck on a $2 off-the-shelf part.  For relevant noise and capacitance information on specific commercial FETs, Amptek’s note here is a must-read.  (I initially tried a pair of 2SK152s in parallel, having made questionable assumptions about the crystal capacitance.  Later, when I tried a single 2SK152 transistor, I did not obtain a measurable difference in system noise.)  Take apart the detector head and break the main vacuum o-ring seal on the detector cap.  Solder in the transistor(s) using no flux.  Use a clip lead to ground the crystal HV electrode during this procedure to protect the FET.
4. Check for high voltage clearance between the cap and the crystal package.  Sometimes there is a thin (0.01″) plastic spacer sheet interposed between–check it for burns or holes.  Damaged plastic sheets may be replaced with the plastic from a clear binder cover (from an office supply store), carefully washed and dried.
5. Evacuate the Dewar.  Even if the FET is OK, Dewars tend to go soft over time…and that puts the FET in jeopardy because of low-pressure HV breakdown.  Preemptive attention to the vacuum may even be warranted.  You can buy an evacuation attachment from the manufacturer for hundreds of dollars, or you can drill a hole in the Dewar wall (carefully! slowly!) with a standard jobber drill and epoxy a vacuum fitting through it like I did.  Whatever you do, make damn sure the vacuum is good (< 10 mtorr) and will stay good.  Whether to continuously pump or seal off is your choice, but I do the former.
6. Remount the HV filter and preamp components.  Supply power to the preamp (but not the HV bias!).  If you have an Ortec detector, adjust the preamp charge loop per these instructions.  Failure encountered in this procedure probably indicates a blown FET, but I am told the hybrid ICs on the Ortec preamps go bad sometimes too.  Leave the cover off the preamp; the charge loop procedure (and PZ procedure) will have to be revisited once the detector is cold.
7. Obtain liquid nitrogen.  Pricing in small quantities is ~$1.20 / liter, so don’t get ripped off by opportunistic asshats at the welding shop who smell teh noob.  Some dealers freak if they see you driving a Dewar around in your passenger car.  If you do take a Dewar in your car, make sure it is strapped in so it can’t spill, and roll the windows all the way down for ventilation.  My 30-liter supply Dewar weighs 83 lb full and sits very nicely in the back seat of a sedan.
8. Wait several hours after filling the detector Dewar for the detector to be operable.  You can observe the decrease in thermal noise from the preamp output as the detector and FET cool down, and you can periodically readjust the charge loop circuit to track 0 millivolts per the above instructions as the temperature drops.  This adjustment will stabilize when the FET gets cold.  In my system, the process takes just under 1.5 hours.  I recommend waiting several hours before applying bias.
9. Give it a try: Turn on a variable HV bias supply set at 0V initially.  Use an oscilloscope or MCA to monitor the preamp output.  Approach a radioactive source to the detector head.  Counts should appear even with bias at 0V due to the photovoltaic effect.  Raise the bias to ~100V.  Noise should decrease dramatically.  Keep pushing the voltage while collecting spectra from your favorite gamma source.  At this point, hopefully you’re witnessing your new toy’s sick resolution.

Gallery 2: Testing and Initial Operation (click any image for larger captioned version)