Tuesday, September 5, 2023

Pulse Height Analyzer (PHA) option for Eberline ASP-1

 When it comes to those big, beige, analog metering system, heavy metal and clunky radiation detectors from the 80s... I am a huge fan of ... not Ludlum Model 3 but the Eberline ASP-1.  I have 3 such units and they are really versatile! 

For anyone interested I made a brief review and comparison between Ludlum Model 3 and ASP-1.

Eberline ASP-1 was and still is a fantastic meter for its time! In the name "Analog Smart Portable" (ASP), the "Smart" stands for "Microcontroller" - an 8-bit Intel 80C31 is in the heart of the meter, driving an 8-bit DAC (AD7524) which on the other hand drives the Analog Metering System. Two old-school 74HC157 multiplexers read the state of various config DIP-switches and the main rotary Range switch.

Reviewing the Service Manual for the meter years ago, I noticed that Eberline designed a Pulse Height Analyzer (PHA) module for it but for all these years I was never able to find, or even see a picture of one in existence. I am not even sure if this option was ever manufactured / sold for ASP-1 but I have seen dozens of ASP-1 units and not a single one had this module installed.

The Pulse Height Analyzer allows the meter to count pulses only when they have a specific energy range (of course, when the meter is equipped with an energy discriminating detector (scintillator)).

The PHA creates a "window" which has an adjustable width and can also be moved up and down the energy range so the counter can count only pulses produced due to gamma rays from a specific isotope - for example it can count only Cs-137 gamma at 662keV while ignoring any other energies, making it very selective at what activity is measured.

From a practical point, I can set it up for a specific U or Th decay chain isotope and being able to differentiate in the field if I am dealing with U or Th for example (that is if I don't have my Raysid with me)

The PHA module is interfaced via a "PHA Option" socket on the main board of ASP-1. This socket must be populated with a special shorting connector when the PHA Option is not installed and so was the case with all of my 3 ASP-1 units, until now...

After I was unable to procure the ASP-1 PHA Option for years, I got tired of looking for it and decided to make my own modules.

The PHA Option socket with the Shorting Connector removed.

The same PHA Option but for a different meter - the Eberline ESP-2. 
The schematics are virtually identical - only board outline and component topology appear to be different from the one, specific to ASP-1.

Fortunately, the Service Manual includes full documentation and schematics for this board.

This is the schematics for the PHA board published in the Eberline ASP-1 technical manual.
There are no exotic components and everything for this circuit is readily available on DigiKey. 
The circuit itself is fairly simple, using comparators and voltage dividers for reference to determine the energy "window". There is also a Dual BCD up-counter alongside a bunch of other passive components. 
The board outline is very specific as this board needs to fit between existing connectors and components located on the top side of the main board. 
The R32 Trimmer-potentiometer controls the width of the energy "window", and the slide switch allows to turn off PHA counting mode and to GROSS count all pulses. 

I started by re-creating the schematics in AutoCAD Eagle.

 I decided to keep the old-school DIP-14/DIP-16 versions for the ICs but replaced all other components with their SMD versions, using 1206 footprint for quick and easy soldering. 

Since it is a fixed size board there was no gain of going SMD for the ICs and their SMD versions were actually more expensive. 

I kept the original component names as listed in the technical manual - this makes component installation and troubleshooting very easy.

The PCB Layout created in AutoCAD Eagle.
Because of the location of the board, right behind the analog metering system, I realized that this board can also conveniently host LED backlighting for the metering system (the stock backlighting uses 2 incandescent light bulbs). 
I utilized the unused pin 2 on the Option header to provide the switched ground to the LED lighting and a small modification of the main board ties Pin 2 to the light switch. The LED Anodes are connected by jumper wire to the +VBAT

Top Layer of the completed PHA Option Board. Some additional milling of the board outline was needed to make it fit between the components.

Bottom PCB layer with the male DIP-16 header for interface with ASP-1

The PHA Board installed in the Option socket. The Blue Pot is the "Window" trimmer, and the red slide switch changes the mode between "GROSS" and "PHA" counting.

The PHA Option board sits nicely behind the Analog Metering System.

I have 3 completed PHA Option for ASP-1 modules left, available for purchase at $50/each + shipping.

Saturday, July 15, 2023

Gamma Dog hardware - PCB v4.0

I finished the design of PCB v4.0 for the Gamma Dog and OSH Park service came through again with excellent manufacturing quality boards.

The improvements in the schematics and layout are not huge but this version incorporates the progress in the development, which continued after PCB v3.0.

Gamma Dog Main PCB v4.0

Bottom layer with a Marie Curie quote.

Changes from PCB version 3.0:

  • changed the placement of the DAC board and EEPROM board connectors. The new version of the 12-bit DAC daughterboard is slightly larger due to additional QT connectors and it wouldn't fit in the old footprint. 
  • added HV voltmeter circuitry for measuring and adjusting the HV PMT Bias in the field - the measurement range is 0 - 1255V (max). The High-Voltage reading is displayed when entering the menu system's diagnostic screen and will allow for field adjustments when changing different detectors. The circuit provides HV PS status data for the Self-Diagnostics / Health-Check during startup. The accuracy is better than the on-screen resolution of 1V.
  • added the I2C digital potentiometer for the Digital Audio Volume Control to the main board (MSOP-10 package) - with the previous version it was an in-line add-on board as it was developed after fabrication of PCB v3.0
  • added support for the LED in the GREEN/DOWN button - this LED serves as "Charging" (flashing) and "Charge complete!" (solid) indicator during battery charging.
  • added an SMD jumper to configure the power source for the Audio Power Amplifier module - options now are either regulated 3.3V from the MCU board or direct battery power (4.1V max) (affording the highest possible audio volume).
  • re-organized (separated) and moved the DOWN Button and Audio Amplifier connectors - the GREEN/DOWN button now has a dedicated 3-pin connector (previously, shared it with the Audio Power Amplifier), and the connector is placed closer to the button's location.
  • removed all of the unnecessary component footprints, related to the old analog circuit for setting Minimum Pulse Height Threshold.
  • Improved the matching between the motherboard's pads and the corresponding HV PS module output and GND pads for easier installation of the piggybacked PS module.
  • Various small changes in component placement, trace routing and overall layout optimizations.

The PCB with SMDs, sockets, headers and connectors installed. Visible on the right is the HV voltmeter's VD string of precision resistors. 
The input impedance of the voltmeter is >1 GOhm to reduce any voltage drop caused by the measurement circuit.

DAC, EEPROM, Comparator and HV Power Supply + Pulse Amplifier 

The nRF52840 MCU module along with various sensors and RFI Shielding Can around the HV circuit

Latching Relay module and RFI shielding cover installed.

Completed, calibrated and tested V4 boards.

Firmware version is now v4.5, supporting the new hardware features and changing how unit-specific data is handled. 
All of the "Identity data", unique to each Gamma Dog unit, such as calibration offsets for various voltage dividers, pulse amplifier DC bias value, exact DAC reference voltage, serial number, etc is now stored in a dedicated "Read-Only" area of the EEPROM and not hard-coded in the firmware as before. 
Each unit is flashed with its "identity" during manufacturing, removing the need for unit-specific firmware code and simplifying firmware updates. 

The PMT Bias reading shown when entering the menu system. 
The value is constantly read and displayed at 4Hz until a button is pressed. 

Thursday, May 11, 2023

Modifying Scionix-Holland 38B57/1.5M-E1 Scintillating Detector

 Commercial scintillating detectors are usuallu very expensive - hundreds, often thousands of dollars for a good size NaI(Tl) detector. Such detectors are not always affordable for amateurs, even on the secondhand market but for many applications they are the optimal, and sometimes the only solution.

There is a hidden, and often overlooked and underestimated gem though, made by a leading in the scintillator business Dutch company - the Scionix-Holland 38B57. This detector is nearly impossible to beat when it comes to value and it is pretty much "the best bang for the buck", delivering an incredible performance for its very low cost on the used parts market. The detector was manufactured as an OEM part about 10-15 years ago and not available for purchase as "new" but there are plenty of salvaged units out there, offered by various Internet sellers.

The 38B57 is a "classic" NaI(Tl) detector - the crystal is 38mm by 57mm (1.5" x 2.25"), surrounded by reflective powder, coupled with a 38mm Hamamatsu R980 10-stage head-on PMT and mounted in an Aluminum + Stainless Steel tube enclosure.

38B57 is employing an integrated design - the NaI(Tl) crystal is not encapsulated in its own aluminum canister, but it is directly interfaced (glued) to the PMT's Head-On photocathode window and then both, PMT's front part and the scintillating crystal are sealed together in an air-tight aluminum can. The assembly is wrapped with the Mu-metal magnetic / electrostatic shielding and together with the VD PCB is housed in a stainless-steel tube with an aluminum end-cap. 

The integrated design keeps the cost down, but it means that this detector is not really serviceable beyond its voltage divider circuit / PCB - crystal and PMT cannot be decoupled from each-other and replaced without the complete destruction of the detector - this is one of the down sides of such design.

38B57 on the other hand, is a really a high-quality, spectroscopy grade detector and the lack of an additional glass window in front of the crystal improves the resolution by reducing photon refraction and/or reflection which would normally occur with the extra glass window of an encapsulated crystal.

These detectors are part of the Exploranium GR-135 RIID device and hundreds such units are being decommissioned all the time by various US Government agencies (Border Patrol, Cost Guard, etc. ) and sold to equipment recyclers / salvagers.

These detectors will often show up on eBay for as little as $80/ a piece (at the time of this writing, but prices do change) and sometimes for even less, making "good size NaI(Tl) Scintillator for under $100" possible!

The 38B57 detector located in its black, shock-proof rubber protector, inside an Exploranium GR-135 Radioisotope Identifier unit. The white connectors on top are how the detector is connected to the electronics - the left one is the temperature sensor for compensation and the right one is for power to the detector.

Obviously, if the PMT or crystal are damaged because the unit was mistreated or accidentally dropped, the detector goes in the garbage bin but if they were treated well and are in a good, working condition, the detector can provide excellent post-service life as a Gamma Scintillator (Counting or Gamma Spectroscopy probe) - I routinely measure the FWHM resolution to be better than 7% (@662keV) for the 38B57 detectors. This makes it an excellent choice for those who need a scintillator probe, are just beginning and want to try Gamma Spectroscopy and are on a budget. 

I, actually started my Gamma Spectroscopy experiments years ago with such detector.

Unmodified, freshly decommissioned detectors.
I cut off the connectors in order to remove the detector without damaging the protective rubber booth.

As the detectors are removed from the Exloranium GR-135 units and sold on eBay, they are not directly useable - they have a custom voltage divider circuit with transistors and diodes in the last stages, intended for use with the GR-135 hardware and must be modified with a "standard" voltage-divider circuit to get the best performance for both, linearity and resolution.  Even the original connectors and the way they are powered is specific to the GR-135 unit. People have tried to use them without any modification, but the results are not great, and linearity is very poor. 

The simple and easy to do modification brings it to a completely new level and one will be rewarded with a very capable detector once it is done.

The modification process consists of removing the original voltage divider, installing a "classic" VD circuit with appropriate impedance and mounting a coaxial connector on the housing.

This modification is not difficult but requires basic electronics, soldering and mechanical skills and one should be comfortable, working with SMD components in order to perform the procedure.


Modification

Funny enough, the most difficult part of the modification process is opening and removing the rear aluminum cap of the detector.

The cap is glued very well, with 2 different types of adhesives (including a special conductive adhesive) and one must use a heat-gun, an utility knife, flathead screwdriver and some patience to take the cap off. 
Fortunately, there are no heat sensitive components in the very back of the housing, but heating must be done quickly before the heat creeps down the housing, towards the NaI(Tl) crystal. 
TIP: Holding the detector with a moist paper towel can provide additional cooling and heatsinking effect to the crystal housing while performing this procedure.

Enemy #3 of these inorganic scintillating crystals are rapid temperature changes which can cause the crystal to crack. (Enemy #1 is moisture and Enemy #2 is mechanical shock)

It requires quite a bit of heat for the adhesive to fail and let the cap go. 

Inserting the blade of the utility knife between the edge of the tube and the cap while hot, allows for the cap to be pried off - this action must be carried out repeatedly at different spots around the perimeter of the cap until it comes off.

If the cap doesn't budge initially, just reheat quickly to a higher temperature, while monitoring the temperature of the crystal housing and once the cap is open, slowly cool down the top, heated edge, of the stainless-steel tube.

The aluminum cap can retain heat, so the process is as follows - heat up the cap, then quickly put down the heat-gun and try to pry it off with the utility knife, then repeat as necessary and change position around the perimeter of the cap.

Once the utility knife blade widens the gap enough to slip in a flat-head screwdriver between the edge and the rim of the cap, things become easy as twisting the screwdriver applies quite a bit of force to pry the cap open.

Some caps will come off quickly and easily, but others will have excessive amount of glue and can be "tough cookies".

Once the aluminum rear cap is removed, this is how the detector looks on the inside. 

Next step is to remove the silicone sealant, cables and the cable grommet.

DO NOT try to remove the stainless-steel tube from the bottom, aluminum part of the housing in order to gain better access to the PCB - it is not needed, and any such attempts could lead to the destruction of the detector!

All of the work is carried out through the back opening.

I cut the cables for this picture, but actually the wires should be de-soldered and completely removed. The picture shows the original Voltage Divider, with the transistors in the last stages. It is a tapered VD and the total impedance is fairly low - around 12MOhms.

Most components of the original VD must be removed, and some will be replaced with different values. The only components that stay are the 3 capacitors shown on the picture - everything else, marked with "X" in this picture, must be de-soldered.

The best and fastest way to remove these SMD resistors is using 2 soldering irons equipped with fine tips (I use ETP tips for this task with my Weller stations). This method also carries less chance for PCB damage. Each resistor is heated simultaneously on both sides and picked up by the two soldering iron tips as if tweezers are used. It takes me just a few minutes to remove all of the unnecessary components. Solder wick is used to clean the pads and prepare them for the new resistors. The old soldering flux can be cleaned off with alcohol pads or alcohol-soaked Q-tips.

This picture shows how the PCB should look like after de-soldering the original divider. 3 out of the 4 SMD capacitors (10nF/200V) are left in place. The 4th capacitor on the very left is removed and later a resistor will be installed in this position.

Next step is to install the SMD resistors for the new, "standard" voltage divider. I have discussed choosing resistors for PMT VD in other posts but for general purpose (counting with Eberline or Ludlum meters for example) 10 MOhm resistors are normally used. This is a 10-stage PMT with 2R (20M) between K and Dy1 and R(10M) for all other resistor positions (between the rest of the Dynodes).  The footprint on the PCB requires 1206 package resistors. The total impendence of the VD will be 120M, which causes minimal voltage drop even with weak HV power supplies. 
(For Spectroscopy, a lower value for R should be used - 1M or 2M is generally a good choice.)

I use 10 Mohm / 1/4W / 1206/ 0.1% tolerance resistors - Digikey part # 749-MCA1206MD1005BP500CT-ND - Vishay High Stability chip resistors.

Resistance tolerance is not super-critical as there are already differences in the PMT's Dynode stages to begin with, and 1% tolerance should work just as well. 

All resistors should be installed just as shown on the picture.

(tip: buying these resistors in quantity of 100 pcs from Digikey is more cost-effective, especially if more than one detector is modified as each detector takes 12 resistors)

On the picture above:

A. Resistor is installed in the position of the removed capacitor.

B. Resistor is installed on top of the capacitor and in parallel.

C. Two resistors in series are installed between Dy1 and K. Single 2xR resistor (in this case 20M) can also be used but I found to be more convenient if I use 2 resistors as the distance between the pads allows for this, looks clean and helps if 2R is not a standard value.

This is how the PCB should look like after all of the resistors are installed. The yellow wire is connected to the PMT's Anode (P) pad and supplies both, HV Bias to the PMT and return signal - it returns back the positive pulses generated by the PMT to the external circuit.

The grounding lead wire, soldered on one end, to the detector's housing must be connected to the K pad (PMT's Cathode). The stainless-steel part of the housing acts as electrostatic shield for the PMT. 

After the K lead (black wire) is installed, the grounding wire to the housing is connected at the junction of K-2R. 
If the original wire is not long enough to reach the K pad, it can be extended with a piece of bus wire, as shown.

The two pads circled on the picture must be bridged with a short piece of jump wire. 
This is an important step and should not be omitted !
If the jumper is not installed the detector will not work. 

This is the schematics of how the modified detector should be wired. 

The aluminum cap is drilled in the center and a female BNC connector is installed - I recommend using a good quality connector with Teflon center conductor, like Amphenol UG-625/U. 
Alternatively, a MHV or even SHV connector can be used but there is not much clearance on the inside and fitting a standard SHV bulkhead connector will be rather difficult.

The drill diameter for the hole in the cap is 3/8" for a round connector and if the connector barrel is D-shaped (to prevent rotation), then 11/32" drill bit is used, and the rest is shaped with a set of small round and small flat files, until the connector can just fit through the hole without being able to rotate. Using connectors with D-shaped barrels is the better choice as it locks the barrel in place and prevents the connector from loosening itself when operated.

The yellow and black wires (silver-plated stranded wire with Teflon insulation) are soldered to the BNC connector. These wires are about 1" long (but could be a bit shorter) and are carefully bent and routed not to touch the board or components when the cap is closed.

The original cable opening is sealed with a piece of self-adhesive copper tape and a length of Kapton tape on top. The housing should be completely light-proof and air-tight. RTV sealant around the BNC connector (on the inside) can be applied before the connector nut is tightened, to seal it as well. 
The aluminum rear cap is glued back with hot-melt glue to the stainless-steel housing.

I also seal the seam between the aluminum part of the housing and the stainless-steel tube with a strip of Kapton tape - just for "good measure".

The modified 38B57 detectors - completed and tested, ready to be installed in Gamma Dogs. After the modification, these detectors can be directly connected to counters such as Eberline ASP-1 or most Ludlum counters. They will certainly outperform Ludlum 44-2 probe for example. 

Here is a Gamma Spectroscopy plot done with one of the modified detectors.
The FWHM resolution for 662keV (1uCi of Cs-137 source disk) is 6.9%. The detector was running on 575V and connected to a Gamma Spectacular GS-USB-PRO. (The second peak from the left is XRF coming off the lead castle - the peak is suppressed due to the graded shielding).
These detectors output ~110 CPS (6600 CPM) for the Natural radiation background at my location when unshielded (~0.1 uSv/h).


"Gamma Spectroscopy Only" Use / 12MOhm Total Impedance VD

If the detector is to be used for Gamma Spectroscopy only, with GS-USB-Pro or a Lab Grade PS driver providing "stiff" HV Bias, lower impedance VD will result in better stability, better SNR and even faster response. 
The 12MOhm VD on the other hand pulls more current and it is too low for portable, battery operated, meters - it will cause a significant voltage drop and an increased battery usage.

The original VD can be modified by removing the active components and only some of the resistors while keeping all of the existing 1MOhm resistors - this is really simple and logical, but I decided to provide the pictures anyways in case somebody wants to follow this guide as a step-by-step.


Only the marked with "X" components should be removed, keeping all existing 1Mohm resistors in place.


This is how the PCB should look like after de-soldering the unnecessary components.

Additional 5x 1MOhm resistors are needed (I used the ones salvaged from other units) to complete the 12MOhm Voltage Divider.

The rest of the modification is just as outlined above.

Tuesday, April 4, 2023

Building a Scintillator, using CsI(Na) crystal and Hamamatsu R6231 PMT

 This scintillator build is not much different than the others except for using a CsI(Na) crystal and 2" / 51mm PMT.

 I understand that people use these posts as a guide when building their own scintillators so I decided to document it as it might provide additional information, which I might have missed in previous posts.

Using Gamma Spectacular PCB for R6231. This PCB is designed for both, single wire (HV Bias + Signal) and 2 wire (HV Bias and a separate Signal line) interface. I populated the PCB for the Single wire interface using R value of 5.6M for the voltage divider. Had to improvise a bit with the 2R resistors between K-G and G-D1 as there was only one footprint per resistor but there are no 11.2M resistors.
Total resistance of the VD is 67.2M.

Voltage-Divider impedance is not a super-critical parameter, but it is important to consider based on the application. For example, with battery-powered counters and meters where the HV Bias supply can't take much load, high-impedance is preferred as the lower current will reduce the voltage drop. Some PS will drop the output voltage with impedance as low as 60M. In these cases, a total impedance of 120M will work well. High-impedance divider on the other hand will have lower SNR (Signal-to-Noise Ratio) and linearity could suffer as well - for Gamma Spectroscopy with benchtop / lab-grade power supplies, VD impedance of 12M will work great.
For this detector I went "in the middle of the road" with VD around 70M.

Machined PMT rear cap, completed PCB and the PMT ready for the final assembly.

The PMT is a 2" Hamamatsu R6231 - the little brother of R6233. The only difference between the two is really the size - all other specs are the same.
R6233 is one of the best all-around PMTs - I've built more than a dozen detectors with it and absolutely love it - the R6231 should be just as good!

VD PCB installed on the back of the PMT with the two silver-plated, Teflon insulated lead wires. 

Installing the MIL-grade BNC connector (Amphenol UG-625 B/U). 
There is a wire fed thru a small hole in the cap for grounding the electrostatic shield. Heat shrink tubing is added for extra insulation of the Anode lead and both wires are coiled into "springs" and away from each-other before closing the PMT cap so they don't press on the board and stay away from the components.

The photocathode window was thoroughly cleaned with Acetone and any dust particles were removed using micro-fiber cloth and sticky tape until the glass is absolutely spotless. 


The crystal is a "Soviet Era" 40mm x 40mm CsI(Na). Datecode is June, 1991.

CsI is a higher density (4.51g/cm3) scintillation material which makes it more efficient at detecting gamma (better stopping power and higher Z). 
Its light output is 85% when referenced to NaI(Tl) but one big advantage of CsI(Na) when compared to CsI(Tl) is that the emission peak matches perfectly the response of Bialkali PMT photocathode at 420nm wavelength. 
CsI(Tl) on the other hand is better suited for use with SiPM as its peak is at 550nm and cutoff at 320nm.
CsI(Na) is also much faster scintillator with decay time of 630ns compared to CsI(Tl) at 3.5us which allows for higher rate detection. Not as fast as the NaI(Tl) with 250ns decay.

Comparison of the emission peak wavelength and temperature response of both types CsI materials. 
The light yield is slightly lower with 41 photons/keV Gamma for CsI(Na) compared to 54 photons per keV Gamma for CsI(Tl) but greater than NaI(Tl) with 38 photons/keV Gamma.

My crystal is absolutely pristine - no significant blemishes, no yellowing, no cloudiness. It is crystal-clear (no pun intended).
 The glass of the optical interface window was cleaned in the same manner as the PMT's photocathode.

Due to the small difference in diameters between the crystal canister (45mm) and the PMT (51mm), I added a short sleeve from EVA foam around the window area, which will center the crystal during assembly, preventing it from sliding off-center to the PMT. The size difference is very small (~5mm) and there was no need to 3D print a centering collar.

A drop of high-viscosity (100K cSt) silicone fluid is added as optical interface between the two glass surfaces to minimize reflections and refractions by eliminating the air gap between the two glass windows. The silicone oil's refractive index is around 1.41 which is close to the 1.46 refractive index of the borosilicate glass of the PMT.

PMT and crystal are put together and the silicone fluid interface is distributed evenly between the two glass surfaces with repeated, overlapping, wide, circular motion until it becomes a very thin and even layer. The extremely high viscosity of this layer and surface tension prevents it from "running" and it will stay permanently in place.

Strips of vertical electrical tape are used to pull together both, PMT and Crystal with some tension. The tape is stretched during application, and it exerts constant pressure between the two parts ensuring a firm contact. With this small diameter PMT 4 long strips (which loop under the crystal) are sufficient.
Both, crystal and PMT are then wrapped multiple times, around, with special focus on the interface zone so it becomes one tight package.

After completely wrapping the assembly with multiple layers of electrical tape, I added a sleeve of EVA (Ethylene-Vinyl Acetate) foam to serve as mechanical shock protection and thermal insulation for the assembly. EVA foam is a very dense closed-cell foam and does great job for both applications.
The front face of the crystal is also protected by a disk of EVA foam, glued to the crystal's foam sleeve. 
Needless to say, the assembly is 100% light-proof.

The magnetic / electrostatic shield is added as a two-turns sleeve of special Mu-Metal sheet, spot-soldered closed, and the grounding wire is then soldered to the sleeve. The sleeve overlaps the photocathode area and into the crystal canister by about 5-7 mm.

Second EVA foam layer for even more shock protection of the glass-envelope PMT goes on top of the magnetic shield.

After completing the entire assembly, the final protective layer of heat-shrink tubing is applied. 

This is probably the most critical and dangerous part of the assembly process as overheating the crystal can easily cause it to crack. Crystal temperature should not be increased too rapidly. Heat was applied in short burst (with cooling time between them), which allowed the heat-shrink to heat up rapidly but not to transfer a lot of heat at once to the crystal.

Heat-shrinking is not "a must" but provides a nice finish, serves as an additional protective, abrasion / scratch resistant layer and keeps both components - crystal and PMT firmly together.

Just as expected, the resolution is not bad at all and spectrum looks pretty good! 

Running the detector on 600V PMT Bias. For Cs-137 at 662keV resolution is 6.2% FWHM. 

Theremino's algorithm for automatic estimation of the FWHM resolution seems to be a bit on the conservative side so a more realistic value would be actually around 6.0% - not too shabby for a 30-year-old, Soviet Era crystal I must say.