Tuesday, October 26, 2021

Gamma Dog - The Ultimate Radioactive Rock / Mineral Finder!

It was time to put my newly built large NaI(Tl) scintillator to work. 

A friend of mine and an avid REE mineral collector Charles Young designed sometime ago an instrument intended specifically for finding radioactive rocks. The design evolved from his initial concept to become a pretty clever and viable solution. Charles and I have been talking for awhile, exchanging thoughts and ideas what such instrument should be and what features are needed. Finally, Charles managed to do something I've been thinking about for years but never really had the time to dive into! His work was truly an inspiration for me and eventually precipitated as the "Gamma Dog project".

With the core R&D work done by Charles, we decide to collaborate in the design efforts and improve both, the hardware and software side in order for this instrument to become the Ultimate tool for the Radioactive Rockhound.

Currently, there are no such specialized prospecting tools on the market. The glory days of Uranium and REE prospecting are long gone! Consumer radiation detection is reduced mainly to inexpensive Geiger counters and dosimeters

Geiger Counters are more or less useless due to their lack of effciency / gamma sensitivity and commercial scintillators require that your eyes are "glued" to the meter's scale or listen to a divided click-rate and trying to figure out if the rate changes "by ear" which is a pretty strenuous activity by itself. 

We wanted something which is easy to use, very sensitive and allows the user to focus on the environment / terrain and not on the instrument itself. Something really intuitive to use which makes the activity of radioactive rock hunting more enjoyable than simply starring at a meter waiting for the needle to jump. 

Meet the γDog! 

The instrument's housing is a 14.5" long 4" diameter ABS plastic tube with two end-caps, a few handles and control panel on the back cap. Weight is ~5.5 Lbs with 63 x 63mm NaI(Tl) crystal. 

I think currently the outside look is a bit ominous and I'll probably need to dress it a bit with some decals for the sake of the public. After all, you press a red button on a big, black cylindrical object - it beeps and the first thing you see on the display is the radiation trefoil - I can see how an ignorant or less intelligent person could hyperventilate seeing this.

The 3 major internal components are the scintillating detector, electronics package which includes an advanced microcontroller, pulse amplifier and HV power supply, along with other boards and components and the 6600 mAh LiPo battery (there is also a coaxial cable interconnect).
The scintillating detector is heavily padded with closed-cell rubber foam ("Neoprene") to protect the delicate crystal and PMT from any mechanical shocks and vibrations. The unit can accommodate different LiPo battery sizes - usually 2x18650 cells in parallel (4400 mAh) or 3x18650 cells in parallel (6600 mAh). Bottom cap is solvent-welded and it is water-tight. The top cap is removable (3xSS Screws) and hosts the entire Electronics module and control panel.

Currently, I am prototyping the latest version (v3) of the Gamma Dog. I designed the PCB with AutoCAD Eagle and the fabrication was done through the OSHPark service - the PCBs came out absolutely superb with a beautiful purple solder-mask! Highly recommend OSHPark for small projects - quality, pricing and ordering process are simply fantastic!
The MCU board and latching relay board are stacked on the motherboard headers. Visible between the headers are the DAC and EEPROM boards.

The new hardware version expands v2 with a second control button for an improved UI, 4K EEPROM for storing settings and internal data and a 12-bit DAC for digitally adjusting the minimum pulse height threshold.

There is an RFI can for shielding the HV circuit which is attached to the HV board (not yet installed in this picture).
There are 6 connectors on the motherboard - one female BNC for connecting the scintillator and 5 other for power, buttons, display, speaker and LEDs (4 are on the top side) - not counting the IPX to SMA coaxial connection.
Not pictured here is the latching relay shield which plugs into the top MCU headers. 

A quick demonstration of the "γDog v2 PLUS" and the UI.
(Youtube link)

Hardware Features:

Most of the instrument's housing is occupied by the padded scintillator (see the diagram above). There is also a battery compartment and the entire electronics assembly is located in the top end-cap. A fixed metal handle with a soft nylon strap handle and a a couple of special finger pulls / attachment points facilitate easy handling and maneuvering of the instrument while searching and digging. The enclosure is sealed and water-proof up to the control panel so it can be partly submersed if needed. The control panel is water-resistant and can withstand occasional splash of mud, water or light rain and it is easily protected with a transparent plastic "shower cap" and a rubber band when weather conditions deteriorate severely. There is 1/4" thick Plexiglass bezel protecting the display.

  •  Large (63x63mm NaI(Tl) crystal) Scintillating detector - this is the detector I previously built, but almost any NaI(Tl) / CsI (Tl) / CsI (Na) scintillating detector can be used or even Bicron BC412 Plastic scintillators. Requirements are: a voltage divider impedance of 50M-120M, common signal / PMT HV bias line and voltage up 1000V - Ludlum 44-2 probe for example is a possible "off-the-shelf" solution.
My large 63 x 63 mm NaI(Tl) detector is the "heart" of the γDog and covers the "traditional" for NaI(Tl) detectors Gamma rays energy range of 25keV to 3000 keV. The larger the crystal is, the more efficient and sensitive the instrument is going to be - this is critical when looking for radioactive specimens "in the wild", which can have very weak emissions due to the shielding effects of soil and rocks and/or low N.O.R.M. content. Smaller size detectors or scintillating plastic detectors on the other hand are better when searching in tailings piles, mine dumps or inside mines, where the radioactivity is abundant to begin with.
  •  Adjustable, regulated and filtered, boosted High-Voltage Power Supply for the PMT Bias, positive polarity, outputting in the 600V-1000V range with 4V input. The low-voltage supply rails of the HV PS is controlled with a latching relay for a reduced power consumption.

γDog's High-Voltage Power Supply and Pulse Shaping Amplifier.
  •  Pre-amplifier with adjustable gain ratio and adjustable pulse shape - rise/decay time-constants as well as decoupling capacitor and load resistor are all hosted on the HV board as well/
  •  Adjustable minimum pulse level detection threshold using a comparator on the input of the MCU to "digitize" the scintillator pulses for counting.
The scintillator pulses as they are amplified (yellow trace) are "digitized" by the comparator (blue trace) when they exceed the preset voltage threshold (white cursor line on the Y axis).

Nordic nRF52840 System on a Chip is the "brains" of γDog.

  •  Nordic nRF52840 SoC (installed on the Adafruit Feather Sense Board) - ARM Cortex M4F processor, 1MB Flash and 256K SRAM, 21 GPIO, 6 x 12-bit ADCs, up to 12 PWM outputs and an array of environmental sensors - Accelerometer, Temperature, Humidity, Magnetometer, Barometric pressure, etc. as well as Bluetooth LE connectivity.
  •  Amplified (Class D amplifier) panel-mounted 1W speaker with adjustable volume
  •  3 backlit buttons (Power, Squelch, Display mode) with indicator LEDs
  •  High-contrast, ultra-low power SHARP Memory eInk graphics display (144 x 168 pixels)
  •  Rugged Micro-USB port (/w protective cap) for battery charging and firmware updates. Use-while-charging with an external 5V USB battery pack is also possible. 
  •  Built-in LiPo battery charge controller with 200mA charge current
  •  6600 mAh / 3.7V LiPo Battery (or 3x 18650 cells in parallel) which provides up to 24 hours of continuous run time. Smaller capacity battery can also be used - 2x18650 (4400 mAh) and will still provide a full day of continous operation.
  • BLE or Low Energy Bluetooth connectivity for outboard data processing and storage (on a Smartphone or PC)
  • Configurable BT power levels

Software Features:

There are a couple of novel features which make this instrument different than anything else out there - the count-rate based variable frequency tone feedback coupled with an adjustable level squelch control. Both features are aimed at easy detection and location of radioactive sources. 

Listening to slow clicks while trying to adjust to the background count rate in order to detect any changes when radiation is detected is way too strenuous on the brain - changes in the pitch of a tone on the other hand are picked up immediately and very easy to follow. The squelch control on the other hand makes the instrument "vocal" only when needed, near a radioactive source above background level.

Gamma Dog is running on over 2000 lines of code (firmware V3.5)

  •  Audio is main method of feedback via the variable frequency tone and alert beeps.
  •  Super-simple and responsive 1 button UI for Squelch level adjustments with click, double-click and long-press actions (all with audible feedback) to execute various actions such as Sql Level Adjustment, Auto-Set, etc. 
  •  2 user-selectable display modes - Rate and Histogram. Modes can be selected at startup or changed later, during normal operation, with 2 consecutive long-press actions within 3 seconds 
  • In the Rate mode, single-click of the Squelch button will increase the squelch level and double-click will decrease the level. Audio frequency will increase 10Hz for each 10 CPS and displays shows numerical count rate.
  •  In the Histogram mode - display shows 140 seconds scrolling count-rate histogram and double-click will toggle between constantly open Squelch and Normal Squelch Control. In addition, in this mode the audio is generated with smaller (5Hz) steps per 10 CPS. This mode can be useful when searching in places with many high-activity sources like mine dumps / tailings. etc. as very high rates will keep the audio frequency half of the rate - i.e. 4000 CPS will result in ~2kHz tone when squelch level is set really low.
  •   2 indicator LEDs - Green Squelch button LED flashes when the number of counts reaches 2x the squelch setting counts and also indicates the startup sequence point for Histogram mode change, Red Power button LED indicates sleep status / HV power off (when blinking) and normal operation (solid glow)
  •   Graphics display shows Firmware version, Internal Diagnostic results and alerts, Current Count Rate in CPS (refreshed every second, resolution 10 CPS) or a count-rate rolling histogram, battery level as voltage and percentage of capacity (refreshed every minute), Current Squelch Level Setting, Count Overflow and Low Battery Alerts.
One of my contributions to the project was the development of the display interface and the firmware code for it, including the UI. γDog PLUS uses 1.3" Ultra-Low Power SHARP Memory display (144 x 168 pixels, same kind as the one used in the Pebble Watch). The display is a very high-contrast eInk type, super-easy to read in direct sunlight, recessed in the front panel for mechanical protection and it shows useful information improving on the ergonomics and the interaction with the instrument.
Data on the display is refreshed every second for the Count Rate readout an every minute for Battery-related information.

  •   Internal diagnostic routine reports over-temperature, water intrusion (by measuring the internal air humidity level in the enclosure) and various levels of the battery condition - both visually, on the display and with a series of beeps.
  •   Audible and on-screen "critical battery level" reminder - every minute when voltage drops bellow 10%
  •   On-screen Charging / Charge Complete Status Indicator
  •   An Advanced Accelerometer-based PMT / Battery saver will shut down the High-Voltage Power Supply and the pulse pre-amplifier after 10 minutes of inactivity (no instrument movement) during normal operation or after 1 minute while charging the battery overnight. The normal operation resumes within 1 second of the instrument detecting a physical movement (for an example - picking it up). During normal operation the red power LED glows solid and blinks when the instrument is in Sleep Mode. I designed this feature with the use of a latching relay to further save power. Going in and out of sleep mode is accompanied by an audible feedback (descending or ascending tone chirp). Current draw drops from 70 mA to 10 mA when the HV supply is powered down. The DAC is also powered down during sleep. There is also selectable "No Sleep" mode.
  •   Auto-Squelch Set will sample and establish the squelch trigger level at the current sampled background radiation level during startup or when it is activated by the user with a long-press of the Squelch button.
  •   Tone frequency generation is set always to begin at the current squelch level and frequency always starts low when the squelch opens, regardless of the current rate or range of operation - this allows for the instrument to operate in a comfortable for the human ear frequency range regardless of the current measured count-rate. In other words, the lower end of the audio tone frequency range is dynamically adjusted to begin at Squelch level. 
  •   Histogram Mode displays scrolling and auto-ranging relative count-rate histogram

The scrolling count-rate histogram displays the count-rate history over the last 140 seconds. The histogram is dynamically normalized to the count-rate range over this period of time with the minimum count-rate overt the past 140 seconds on the bottom.

Typical background rate histogram (130 CPS to 200 CPS). 
The histogram is constantly updated in background at 1 sec intervals and always available for the last 140 sec. It displays also Min, Max and Average Rate.

High-rate event (2560 CPS) (specimen of Uraninite), followed by drop to background level. 
The displayed histogram is scaled to show the maximum rate peak which on the other hand "pushes down" the background level.
  •   Response to count rate changes with a pre-set hysteresis which prevents the squelch from opening during short rate fluctuations.
  •   Excessive Count Rate / Overflow alert above 10 000 CPS (600K CPM) - the maximum rate is not really limited but the count accuracy will drop when excessively high rates are detected.
  •   Compensated count accuracy for overall error in counting of 0.1% (electronics only)
Test of the hardware and the software counting algorithm. Feeding the input with pulses from a function generator shows that the counting accuracy is spot-on! Maximum error is 10 CPS.

Test Setup for counting accuracy and pre-amp alignment.
 Maximum count rate is up to 17K CPS (~1M CPM) for short pulses (~50 uS). Average pulse length of 100uS will allow count rate of 10K CPS (600K CPM).

   Instrument Usage

Sensitivity comparison between γDog and a Geiger Counter - there is no contest!

(Youtube link)

Charles Young showing his field use of the standard model Gamma Dog.
(no display and indicator LEDs but otherwise almost identical)
(Youtube link)

Prospecting for Radioactives in New Mexico - Charles Young, August 2021

(Youtube link)

Prospecting for Radioactives in Southern New Mexico - Charles Young, November 2021
(Youtube link)

UPDATE: development of Version 3 hardware and firmware has been completed. The UI is now using 2 buttons - it was specifically designed to improve the user-interactions and ergonomics if display output is also present. 

The UI can now adjust the Minimum Pulse Height Threshold. This parameter along with another 8 configurable parameters are saved in the unit's EEPROM. Adjustments of the Pulse Height Threshold are done using a 12-bit Digital-To-Analog Converter (DAC) and displayed in millivolts (accuracy is better than 1mV) while the device is running in normal counting mode - this way the result of the adjustment can be seen immediately. There are UI coarse and fine steps while changing this value.

Additionally, there are 10 items in the Configuration Menu as well as a "Reset to Defaults" feature which re-writes the EEPROM with the default values.

The Version 3 hardware installed in the front panel cap.
Only 2 connections are needed to the body of the instrument - coaxial cable with BNC connectors to the detector and a connection to the LiPo Battery pack.

Current draw for v3 is 68mA during the normal operation mode with Squelched audio and only 10 mA during Sleep mode.

The Blue button is the Squelch Control (UP/+), the Green button is Display Mode and Squelch Control (DOWN/-) and the Yellow button is the power ON/OFF switch. The user now can toggle and latch the squelch control open/close by a double-click of the Green button. Squelch Auto-Set is activated by a Long-Press of the Blue button.

Version 2+ (/w one button UI) on the left and Version 3+ (/w two buttons UI) on the right.

Bluetooth connectivity 
Gamma Dog is equipped with a Low Energy Bluetooth connectivity (BLE).

Bluetooth can be enabled/disabled from the configuration menu. On startup, Gamma Dog will automatically start advertising, waiting for connection.

5 levels of BT Tx Power can be set from -20dBm for level 1 up to +8dBm for Level 5.
This menu item is context-sensitive and will show up in the menu only if BT is enabled.

There are two different BT icons, indicating in total 4 different states of the link - whether there is a currently active connection or not and whether BT data streaming is on or off.
 Double-click of the Squelch button toggles the data logging on/off and the BT icon blinks showing Data logging is in progress.
Data logging is automatically suspended during Sleep Mode and resumes on wakeup.

 A string of parameters is broadcasted over BT every second and includes various internal parameters, detected current Count Rate and Squelch Level. Each data set includes a rolling counter as well.
 Currently, the data can be received with the free Bluefruit Connect App in UART mode or Plotter mode. Commands can be send back to the Gamma Dog to toggle on/off the logging mode and to stamp "Waypoints" with GPS coordinates. 
There are plans for development of a companion Cell Phone App which will store the received data while stamping each set with the current GPS coordinates and a time-stamp. Parsing this data and converting it to Google Earth KML/KMZ format would be next logical step. Overlay over the terrain showing the path with current rate and "waypoints" of found minerals will be a pretty powerful and useful prospecting tool.

The Bluefruit Connect plotter mode shows Count Rate vs. Time Graph (orange) and Squelch Level changes (blue).

P.S. This is old but here is an article in Atlas Obscura on radioactive mineral collecting (shameless self-promotion:  I am quoted a couple of times and there are pictures of minerals currently in my own collection).

Monday, September 20, 2021

Bigger is better or building a larger NaI(Tl)-based scintillating Gamma Spectroscopy detector

In the world of scintillators, larger crystals means more sensitivity - when you have a bigger volume of scintillating material, it can capture more photon interactions and will produce more pulses. High-energy photons sometime can pass thru a crystal without interacting with it and more material in the way means higher chance for interaction.

I've been doing some radioactive rockhounding and sensitivity is everything when it comes to finding that one specimen a foot or so deep in the ground, under rocks, dirt and mud (which btw act as a pretty good shielding). A larger scintillating detector was in order and having excellent results with my first build, I just went right ahead. This new detector is intended for both, Gamma Spectroscopy and Counting mode when prospecting.

The Photomultiplier Tube 

For PMT I wanted a high quality 3" device and Tom from Irad Inc came thru with a very nice, brand new 8-stage Hamamatsu R6233-05 PMT. This device is intended for use in medical equipment (-05 designation) and came with a factory installed PCB which Tom removed before shipping as it didn't serve the GS purpose - the PCB only contains a few components and not a complete voltage-divider.

The device is fairly short (only 123mm) and compact for its 3" diameter size. It has 8 dynode stages which are nested tightly in the "neck" portion. A plastic cap on the back insulates the pins. 
The gain (2.7 x105) is less than R980 but this is to be expected with 2 dynode stages less - still it is impressive to see that the 10-stage R980's gain is 3.7 x105

The Bialkali photocathode has an optimal response at 420nm wavelength of the NaI(Tl) crystal. The Photocathode is a tad more sensitive than R980. Effective diameter is 70mm which works perfect for me since the crystal I was going to use is 63mm and larger photocathode than the crystal is optimal - no photons will be lost leaking.

The factory label - P.H.R. (Peak to Height Ratio) is very good at 7.8% but in reality it is way better than this. The listed PHR is measured with some "standard" for the factory test crystal, with Co-60 and not Cs-137 (hence the lower resolution range) and it is more or less worst case.

The PMT came with the factory PCB removed and the 3 stand-offs cut off. I marked the Cathode pin (K) with a silver dot and Anode pin (P) with a beige dot, To the left of the Anode pin is the Grid pin and to the right of the Anode is D8. On the left side of the Cathode pin is D1 and then clockwise all the way to D8 in sequence.
The identification of K and P is easy - the Cathode and Grid are the only two pins separated by a single(!) blank space (missing pin)

All VD components - 10 resistors, a cap, lead wires and BNC connector were provided as a kit by iRad Inc.
A 20M resistor is connected between the Grid and the Cathode and another 20M between the Grid and D1. Eight 10M resistors are connected between each Dynode (D1 to D8)  and between D8 and P.
The resistors are with precision of 1% tolerance and were all checked with an LCR bridge.
Total impedance of the VD is 120M which is excellent for battery operated HV supplies.

10nF / 2kV filter capacitor is installed between D8 and GND (Cathode effectively). The leads of the resistor are kept as short as possible. The A and K leads are silver-plated Teflon insulated stranded wires inserted in a second Teflon sleeve.

Schematic diagram of the 8-stage Voltage Divider used.

The machined plastic cap came from iRad Inc. as well. 
A BNC connector was installed on the cap and connected with 2 silver-plated Teflon insulated wires. I drilled a small hole on the side of the cap for the grounding lead for the Mu Shield. The hole and wire were sealed on the inside with black RTV sealant after the wire was soldered to the BNC's ground lug.

For anyone interested in PMTs and technical details, Hamamatsu published the most comprehensive document I've seen on the subject - The PMT Handbook.

The NaI(Tl) scintillating crystal 

Larger crystals have better stopping power resulting in overall better efficiency and this affect mostly the high-energy gamma response.

My NaI(Tl) crystal was purchased on eBay and came from Ukraine. The 63mm x 63mm crystals out of Russia and Ukraine seem to be currently very popular on eBay and they tend to produce excellent results.

Date code is Dec 1987, most likely made in USSR.

The crystal is hermetically sealed in an aluminum canister with an optical glass window and surrounded by reflecting material. The size of the crystal is 63mm x 63mm. The outside diameter of the aluminum can is 71.1mm.

The crystal looks pristine!
There is a lot of junk sold on eBay - cracked crystals, failed seals, yellowing or cloudy crystals etc.
One need to very carefully choose what they are buying - some of the garbage crystals I've seen are not even good for counting yet sellers ask hundreds of dollars. If you see any blemishes or imperfections these could be signs of trouble and better to stay away! A failed seal will let moisture in and will destroy the crystal over time - NaI is an extremely hygroscopic substance.
Fortunately, my crystal looks perfect as if it was made yesterday - not 30 years ago.  
Even when the crystal visually looks great it doesn't mean one will get the best resolution tho!

The Size-adapter / Centering Collar

The outer size of the R6233 PMT is listed as 76 +/-0.8 mm and the outer size of the crystal canister is 71.1mm so there is a net difference of ~5mm.
I wanted to center the crystal to the photocathode and have some reliable way to keep it centered while taping the assembly together so I decided to design a centering collar.

The centering / size-adapter was designed with Autodesk TinkerCAD.

The adapter was printed with conventional 3D printing technology. 
The bottom portion for the crystal was originally very tight - there was a small "shrinkage" of the print in this area - about 0.5 mm total so I had to sand it until the adapter can slide freely around the crystal.

"Dry fit" of the three components.

The purpose of the "fingers" on the adaptor is to be able to tape it or glue it to the PMT and secure it reliably.

During the assembly, the adapter should fit tightly around the PMT but it should allow the crystal canister to move freely ("float")

The inside area where the transition from one ID size to the other ID takes place is tapered for the glass edge of the PMT. The adapter will center the optical window of the crystal canister exactly to the Photocathode area, while providing both, mechanical stability and light-proofing of the interface area.

Optical Assembly

An edge reflector was applied around the front edge of the glass. On this PMT, the photocathode wraps around, to the side for a few millimeters and a reflector in this area will bring back photons otherwise escaping thru the glass edge. I used pure white PVC tape around the edge and then the whole PMT was light-proofed with the cap taped securely as well.

For optical interface medium, I used crystal-clear silicone fluid with viscosity of 100K cSt. It is sold as Differential fluid for high-end RC cars. Consistency is much thicker than honey but flows and spreads nicely and it is absolutely transparent. 
I started by applying an air-bubble-free drop of the fluid in the center of the window and pressing the PMT firmly on top.
I worked the PMT with large circular movements while pressing downwards for a couple of minutes, overlapping the edge as much as 1/3 the diameter of the window, until any excess fluid squeezes out from the edge and only a super-thin smooth layer is left. This process shouldn't be rushed and excessive force can damage the PMT and/or crystal.
I did notice about 0.2% improvement in FWHM resolution vs. the usual optical silicone grease, probably due to eliminating any possible air pockets / bubbles. 
I also used a 60K cSt fluid in the past with excellent results.
The purpose of the fluid is to eliminate glass-to-air and air-to-glass transitions and reduce the refraction and reflection from the surface of the glass - this makes for an optimal interface between the crystal and the photocathode.
Every time a photon reaches a glass-to-air surface it can be either reflected back or refracted in some direction. The optical fluid substitutes the air with almost the same refractive index medium as the glass, thus reducing light scattering and loss.

I placed the collar around the crystal before I applied the silicone fluid and mated the two glass surfaces. When I was satisfied with the quality of the interface, I simply pulled up the collar while adjusting the PMT's position and pressing it down to the crystal canister.

The entire optical assembly was taped together using electrical tape (3M 88 tape). I started with 8 pieces of tape stretched as tight as possible along the length of the assembly from the crystal's cylinder to the top of the PMT. This "pulls down" the PMT to the crystal with a slight pressure. The assembly was then taped all around starting from bottom to top while still applying pressure between the canister and the PMT.

Mu-Metal Shielding

Electrons just emitted by the photocathode can easily be deflected by an external magnetic field and miss the grid and the dynode cascade. The effect of magnetic field deflection of the electron's trajectory is worst when the field is parallel to the photocathode. Earth's own magnetic field, while weak, it is fully sufficient to cause a deflection for these "freshly emitted" low-energy electrons. Earth's magnetic field is even enough to cause deflections in the dynode stage so good magnetic shielding must be provided for the PMT.

The Mu shielding was constructed in two parts. The sheets of Mu-Metal were again sourced from iRad Inc. but they can be obtained from a variety of sources.
The top portion of the Mu shield was wrapped around the neck of the PMT in 2 turns and the bottom "fingers" were folded following the contour of the larger photocathode area.
The grounding lead from the cap was folded and soldered across the Mu-metal collar.

The bottom part of the Mu-shield. About 1.5 turns of mu-metal with "fingers" folded down over the top part "fingers", overlapping and covering all gaps.
This part of the shield extends a bit beyond the photocathode and overlaps the crystal - it is the more important part of the shielding as it covers the area between the Photocathode and the Grid.
If the detector will be exposed to stronger magnetic fields, more shielding should be added especially in this area.

The two pieces of metal are grounded together with a few small spot-soldering joints. I was careful to minimize the heating in order to avoid melting the electrical tape or changing the magnetic properties of the shield.

Final Assembly

A couple of layers of electrical tape were applied over the Mu-metal shielding in order to make it a tight fit around the PMT.
I fabricated 2 sleeves out of closed-cell foam ("Neoprene") - one for the neck of the PMT and another for the crystal canister. 
The foam acts both as mechanical and thermal insulation - it protects from mechanical shocks and rapid temperature changes of the crystal. It is also a good idea to have it for protecting the crystal from a thermal shock when the Heatshrink wrap is applied. Another layer of electrical tape covered the foam sleeves.

Applying the large size heatshrink wrap is a tricky procedure - it has to be heated evenly from all sides so it shrinks equally around the assembly while being careful not to overheat the crystal - a thermal shock can cause it to crack if heating is happening too rapidly. 
I did the entire process with little breaks to allow the temperature to slowly equalize as I was heating it with the heat gun.

The bottom closed-cell foam sleeve was closed off with a thin sheet of the same type foam, protecting the bottom of the aluminum canister.

The completed detector came out quite nice and solid. 
I still might try to fabricate an outer shell aluminum casing for it, just to make it "bulletproof"
The "Fat Man" codename was changed to Da Gamma Bee II / DGB-2525


After waiting sufficient time for the Photocathode to "calm down" from its excited state due to being exposed to daylight during the assembly I started testing the detector first in counting mode just to make sure the VD circuit is all good!

(!) A word of caution - PMT should never be powered up when exposed to ambient light - this will absolutely destroy the photocathode in a matter of a second.

For reference - this is the background count measured with my VD-modified Scionix-Holland 38B57 - 1.5" x 2.25" crystal and R980 PMT (counter is set to x100K range)

The newly built detector shows more than 2 times increase in the background radiation count. The counter is set to the x100K range so the reading is approx. 20K CPM.
The background measured later with PRA was ~270 CPS or 16.2K CPM but this is normal as PRA rejects malformed pulses, not matching the sampled pulse shape so count will be a bit lower as only pulses good for GS are sorted.
Testing was done at 650V

A quick Gamma Spectroscopy plot of Cs-137 and the results are fantastic - a really nice and clean plot with resolution of 6.53% FWHM at 662keV!
Such resolution is more or less as good as it gets for these larger NaI(Tl) crystals.
I used Gamma Spectacular GS-USB-PRO PMT driver set at 650V.
The linearity at this voltage seems very good too - I didn't have to use any Audio Gain factor and all peaks showed up in the right bins indicating that 650V is nearly optimal. In Theremino MCA the Energy equalizer stayed in the same state as before, adjusted after linearity testing and optimization of my other DIY detector. Only the Energy trim had to be adjusted a bit. This is very good sign!

"Fat Man" and "Little Boy" in their carry case.
I built these detectors for the fraction of the cost of a commercial detector, getting the same performance. Interestingly enough the cost of the two is very similar regardless of the size difference.

A new member joins my family of scintillating detectors!
 (L-R) Da Gamma Bee I / DGB-1531 with 40 mm x 80 mm crystal and Hamamatsu R980, Da Gamma Bee II / DGB-2525 with 63 mm x 63 mm crystal and Hamamatsu R6233, modified Scionix-Holland 38B57 with 1.5" x 2.25" crystal and R980 PMT and Gamma Spectacular GS-1525 with 1.5 x 2.5" crystal and Adit PMT.