I found some cheap NaI:Tl scintillation probes on ebay. I wanted to play with a NaI:Tl scintillator for quite some time so I couldn't help myself and bought one. The probe was made by Bicron / Saint-Gobain, these probes are normally used to find out if a truck has loaded any illegal goods by essentially X-Raying the entire truck. The probes are arranged in a Line, every probe represents a single pixle. Every probe is fitted with a pot to trim the pixels intensity.
The probe's enclosure is made from thin stailess steel. The crystal seems to be enclosed with aluminium sheets and is glued in (I'd assume by using epoxy).
These probe was fitted with a quite expensive military-style plug by Lemo. As I didn't feel like paying a lot of money for a matching jack, I chose to remove the connector and replace it with a standard BNC plug. The probe has no additional signal output. Instead the signal is aquired by monitoring the current the probe is drawing. I used a resistor to archieve a voltage drop when the current is rising and a small cap to get rid of the DC offset. This setup is obviously not perfect as the voltage delivered to the Photomultiplier will be dependent on it's current and thus affect the PMT's gain causing distortion in the aquired spectal data.
I still had a few CCFL-inverters laying arround, unfortunately there was no magic smoke left in them. However the transformers were still good so I just added a small Royer-ciruit by using some standard TO92 NPN-transistors (2N3904, BC547 ...). I also added a voltage multiplier as the transformers didn't quite archieve the voltage I needed.
The PMT's power supply has to be regulated in order to archieve a conistent BIAS voltage which is not affected by the tube's counting rate. I wanted to use operational amplifiers to archieve that, but I also wanted the entire circuit to run from a signel 5V DC supply. After spending a few hours investigating "strange" opamp behaviour I decided to search the Internet on weird TL07X-behaviour. I imediately found out that TL07X-opamps are affected by phase-reversal (felt a bit dumb having spent several hours scratching my head). So I decided to get one of these nice (and expensive) opamps (AD822) which worked without further problems.
The circuit is quite simple. A 2.5V reference voltage is generated using a LM336 voltage-reference which is then fed into a simple opamp buffer. The buffer's output is fed into another opamp's inverting input via a 10k trim pot. The HV is measured using a resistive divider and fed into the opamp's non inverting input. The opamp will pull it's output low if the voltage on the inverting input exeeds the voltage on the non-inverting input. This will allow a flow of current into the opamp's "output" driving a PNP-transistor if the measured Voltage is below the desired Voltage set using the trim-pot.
I had a few BD678 PNP-Darlington transistors laying around so I just used one of these.
Using a 5V supply I archieved a maximum Voltage of roughly 1200V-. The PMT used in the Bicron probe needs a bias voltage of about 700V. Due to the probe's internal resistor network this results in a voltage of about 900V being required to operate the probe. The circuit draws about 150mA with the probe connected. This translates to a power use of about 0,6VA, The royer's ripple is still present so the circuit will probably require additional shielding/filters.
I also designed a small PCB (layout.pdf):
The CCFL-Inverter I used for the PCB-version was still working so I also salvaged the original transistors and the cap.
First tests without additional signal conditioning worked out suprisingly well. (Software: Theremino MCA, ADC: Roland Cakewalk UA25-EX):
Uranium ore (N.O.R.M) found in Wittichen / Black Forest: