Fusor Update - Liam David

Current images of fusor efforts, components, etc. Try to continuously update from your name, a current photo using edit function. Title post with your name once only. Change image and text as needed. See first posting for details.
User avatar
Liam David
Posts: 509
Joined: Sat Jan 25, 2014 5:30 pm
Real name: Liam David
Location: Arizona

Fusor Update - Liam David

Post by Liam David »

I've been working steadily on my fusor for the past several months and have a few noteworthy updates to share.

DAQ / Control System
I made a new control box that has hardwired connections for voltage, current, and MFC deuterium flow. The signals are passed through 100ft of 15-core, stranded, shielded wire running alongside a fiber optic cable. Fiber to ethernet converters on either end connect to a computer and an Arduino Ethernet shield to allow for data collection. The Arduino drives four ADS1115 16-bit ADC modules that monitor pressure and neutron production rate, and soon will also measure deuterium flow, and several other variables. My Spellman power supplies are also connected over ethernet for voltage and current monitoring. Values are periodically sent through the fiber and a custom Matlab GUI parses, displays, and saves the values. There are still several bugs to work out, most notably that the Arduino is EMI-sensitive. I also replaced the needle valve with a 15sccm MFC which gives superior (and remote!) pressure control with the turbo throttled. The voltage, current, and deuterium flow are hard-wire-controlled through the cable and a control box. I can now comfortably sit up to 100ft away from the device, should I desire.

Version 3 of the data acquisition system
Version 3 of the data acquisition system

Don't look, it's pretty ugly...
Don't look, it's pretty ugly...

15-conductor and fiber umbilical cable, 100ft
15-conductor and fiber umbilical cable, 100ft

MFC
MFC

Control box
Control box

Power Supply
I obtained a second Spellman DXM70N600 power supply off Ebay in working order. With the help of some engineers from Spellman, I replaced the DXM firmware with SLM firmware. This allows the supplies to regulate against voltage and current setpoints rather than shutting off upon exceeding them. They work in parallel as a simple master-slave system with both voltage programming lines connected in parallel and the master current monitor driving the slave current programming line. The setpoints are controlled through the aforementioned 100ft cable. I 3D printed custom connectors for both supplies and potted them with RTV silicone. The outsides of the connectors accept o-rings and are packed with silicone grease to prevent arcing. So far I've had no issues at >50kV. (Edit: One connector arced along the length of the cable and now has to be replaced. Voids in the silicone are to blame.) The HV cable is standard RG-58U with the shield pulled ~8" back from the ends. Each supply connects to separate 200k resistors and 100kV diode strings before joining at the HV feedthrough. This configuration allows me to run the system at up to 16mA, but I usually keep it below ~8mA to reduce heating. I only push it higher for short runs measured in the tens of seconds. The grid and chamber get very hot very quickly at high currents which cause the neutron production rate to plummet (wall/grid loading, I presume).

Not proud of the floating HV components, but it works ok
Not proud of the floating HV components, but it works ok

3D printed connector
3D printed connector

Feedthrough
I made a couple changes to the vacuum end of the HV feethrough. The ceramic washer that was designed to shield the quartz from direct plasma bombardment has sputtered full of material and now arcs profusely at anything above 25kV. I retracted the quartz and added an alumina stalk for the last few cm. More arcing then forced me to forego all ceramic and leave the stalk bare. Like I mentioned in the dedicated design thread, this new version works flawlessly.

New feedthrough
New feedthrough

Gamma Spectroscopy
I made a 3" NaI(Tl) detector using a Photonis PMT and crystal from Russia. The housing is 3D printed PLA. I use an old Ludlum Model 12 set to 1100V as the HV supply and for simple count-rate measurements. Background is ~20kCPM. My poor-man's MCA, if you ignore the $500 scope (Siglent SDS1104X-E), is as follows: I tapped into the the model 12 circuit directly behind the HV capacitor and passed the signal through the housing with a BNC connector. This connects to an oscilloscope which then interfaces with a computer via Ethernet. With the scope set to 100-200ms/div, a Matlab script queries the scope for the waveform data. The 14Mpts worth of data are then filtered in-software, peaks are located and binned, and then a histogram is displayed. I get about 7% FWHM resolution for 137Cs and can detect x-rays down to ~15kV before the noise floor interferes. It easily picks up x-rays from the fusor, and I've confirmed the calibration using 241Am, 176Lu, and background K40.

Gamma spectrum with oscilloscope MCA
Gamma spectrum with oscilloscope MCA

Custom 85x60mm NaI(Tl) probe
Custom 85x60mm NaI(Tl) probe

X-Rays
I picked up a CD V-742 dosimeter from a Hamfest a while back and have been placing it at the fusor viewport behind the lead shielding. It read zero when I got it but is now at 60 roentgen. Yikes. Keep those shields up.

Dosimeter
Dosimeter

Breaking the Mega Mark
I am reasonably confident that I hit 1e6 n/s the other day while pulsing the system at higher currents. Since the system heats up very quickly and neutron rates drop, I was limited to ~5s runs. I verified that my neutron detector, which is simply a SNM-17 tube surrounded by paraffin coupled to a Ludlum Model 12, is immune to fusor noise using the standard moderator-removal test. I waited for the needle to stabilize for a few seconds before recording the CPM. The detector axis is just 13.5cm from the poissor, so given the anisotropy from beam-target fusion and the neutron-emission resulting from a single-beam grid, the results should be taken with a grain of salt. The three most impressive runs are as follows:

28kV, 12mA, 13.3mtorr, 100kCPM, 3.4e5n/s, Q=1.13E-09
37kV, 14mA, 17.0mtorr, 200kCPM, 6.7e5n/s, Q=1.49E-09
37kV, 11mA, 14.3mtorr, 300kCPM, 1.0e6n/s, Q=2.86E-09
33kV, 11mA, 17.0mtorr, 350kCPM, 1.2e6n/s, Q=3.66E-09
43kV, 15mA, 17.0mtorr, 300kCPM, 1.0e6n/s, Q=1.84E-09

I found that the neutron production rate was heavily dependent on grid temperature, which is why the 43kV and 37kV runs were less successful. I did not give the system adequate time to cool. Also, the rotation of the grid and symmetry with respect to the 6-way conflat arms was critical. I did my best to align the grid holes with the chamber axes. On these runs only a single bright beam formed along the long chamber axis (extra conflat crosses, while the other axis has blanks), although a very faint orthogonal beam is visible.

Single beam
Single beam

I realize that the voltages are rather low for this kind of output, and perhaps the higher currents compensate, but I have not found any sources of noise. I was able to achieve rates between 3e5 n/s and 1e6n/s for two days until an unrelated issue forced me to shut down and open the chamber. Due to the semi-repeatability, I doubt it's simply Poisson noise, and detector pileup isn't an issue at these count rates. I should have moved the detector and verified an inverse-square relation, but such is high voltage safety and hindsight.


Chamber Reconfiguration
My 6-way chamber was nestled between other conflat components which made shielding difficult and the entire thing difficult to access. I reconfigured the vacuum system, put the chamber above the frame, and subset the rest of the parts below. Now 5 of the 6 sides are available for the feedthrough and viewport, leaving 3 blanked-off for whatever comes next.

Chamber above and vacuum system below, HV feedthrough off for maintenance
Chamber above and vacuum system below, HV feedthrough off for maintenance

I took the opportunity to replace the viewport and sacrificial glass. The latter is chipped and sputtered brown, and since I previously used the viewport without protection, it has degraded as well. Since I didn't properly mount the sacrificial glass, it rested on the bottom of the tube and led to the following:

Offset section of glass protected from sputtering
Offset section of glass protected from sputtering

Chipped sacrificial glass
Chipped sacrificial glass

The new viewport is better protected by a new glass disc, which is now held in place using expandable metal sleeving.

New viewport
New viewport

Indium Activation
After reconfiguring the chamber, I activated a piece of indium foil. It was placed between two paraffin blocks at one of the end caps and irradiated for 1-2 minutes at 1e5-3e5 n/s. Count rates on a 2" pancake were between 180 and 300 CPM and quickly decayed to background. I haven't yet taken a gamma spectrum nor irradiated the foil for longer due to chamber heating. The chamber reaches well over 100C after just a couple minutes, and I prefer to keep my viewport intact.

Indium foil
Indium foil

Some miscellaneous pictures:

Neutron detector. Wires bridge analog meter for neutron reading by DAQ.
Neutron detector. Wires bridge analog meter for neutron reading by DAQ.

Pressure gauge for monitoring remaining deuterium
Pressure gauge for monitoring remaining deuterium

Feedthrough broken by my idiocy, not heat or HV.
Feedthrough broken by my idiocy, not heat or HV.
User avatar
Liam David
Posts: 509
Joined: Sat Jan 25, 2014 5:30 pm
Real name: Liam David
Location: Arizona

Re: Fusor Update - Liam David

Post by Liam David »

Solenoid valve to protect turbos in case of power failure
Solenoid valve to protect turbos in case of power failure

And some plasma pictures:
Attachments
IMG_20200118_114257.jpg
droidcam-20200724-111146.jpg
droidcam-20200720-103904.jpg
droidcam-20200515-105748.jpg
droidcam-20200120-153339.jpg
droidcam-20200120-153339.jpg (13.3 KiB) Viewed 16203 times
User avatar
Mark Rowley
Posts: 908
Joined: Sun Dec 29, 2002 12:20 am
Real name: Mark Rowley
Location: Sacramento California
Contact:

Re: Fusor Update - Liam David

Post by Mark Rowley »

Wow! Excellent report Liam.
The remote operating arrangement is quite something and almost unheard of.

I have also noticed the same drop in numbers as the chamber heats up. Active cooling is something I’ve been toying around with. Forced air seems to help but chilled water in a copper tubing “jacket” may be the best bet.

Congrats on the mega mark and indium activation!

Mark Rowley
User avatar
Richard Hull
Moderator
Posts: 14976
Joined: Fri Jun 15, 2001 9:44 am
Real name: Richard Hull

Re: Fusor Update - Liam David

Post by Richard Hull »

Great work and a great report. Very detailed.

Richard Hull
Progress may have been a good thing once, but it just went on too long. - Yogi Berra
Fusion is the energy of the future....and it always will be
The more complex the idea put forward by the poor amateur, the more likely it will never see embodiment
John Futter
Posts: 1848
Joined: Wed Apr 21, 2004 10:29 pm
Real name: John Futter
Contact:

Re: Fusor Update - Liam David

Post by John Futter »

Liam
looking at your droidcam photo i see the hallmark of radiation exciting the camera plenty of white / colored pixels
remote is good when you see that amount of radiation interaction--you do not want a suntan from this process
keep up the good work
User avatar
Liam David
Posts: 509
Joined: Sat Jan 25, 2014 5:30 pm
Real name: Liam David
Location: Arizona

Re: Fusor Update - Liam David

Post by Liam David »

Thanks for all the compliments!

There is a fan blowing directly onto the chamber which helps by a few degrees, but it's far from enough. It cools stuff down between runs more than anything. I've started thinking about water cooling as well, perhaps some 1/8" copper or aluminum tubing around the arms. There's just so little space between the flanges, especially with the nuts and bolt stick-out. Some 2.75" crosses like the one Jim Kovalchick uses have longer arms which would be nice.

The x-rays get pretty fierce at >45kV and start to bounce around the lead shields I have in place. Rest assured the camera is followed by lead so I won't get a tan. There's a lead sheet cube with an open top around the main chamber and then a large 2'x4' sheet between me and the whole setup (when I'm operating it in the same room). Unfortunately at these voltages x-rays are reflecting off all kinds of surfaces, so while line-of-sight rays are stopped, I occasionally measure ~1kcpm on a 2" pancake by the operator station. I keep runs at these voltages short, and really need to start taking advantage of the long cable for remote operation.

One thing I can attest to about smaller chambers is the need for symmetry. I was repeatably able to hit the mega mark in the previous configuration, but just a few minor changes in chamber setup, grid positioning, and such have limited me to ~3e5 even at 55kV and 8mA. I've started messing with single-axis grids (made from shaft collars since I have neither a lathe nor a welder) but haven't yet exceeded 3e5. Instead of one beam, two form in an X shape in line with the stalk, like in this picture:

droidcam-20200724-110731.jpg

The pattern on the grid tells the same story.

IMG_20200724_113816.jpg


There are no magnets or protrusions in the chamber--it is simply a blanked-off cross. Changing the grid positioning didn't help nor did careful alignment with the center-line. I'm going to grind down the hexagonal adapter to 1/4" OD like the rest of the stalk to see if that helps. Kind of at a loss about what else it could be. Hopefully I'll get back to the mega mark soon. With a limiting voltage of 70kV and based on Jon's cube fusor work, I should be able to hit at least 5e6 with careful design.

Liam David
User avatar
Jim Kovalchick
Posts: 716
Joined: Wed Apr 13, 2011 8:00 pm
Real name:

Re: Fusor Update - Liam David

Post by Jim Kovalchick »

Liam,
Your work is inspiring. Well done.

My fusor really hits a wall when it heats up. I think the ideal solution would be a water jacket as opposed to coils because the coil contact with the chamber can only be so good. I am not sure yet how I would do a jacket yet but water directly on the chamber shell would be great.

Jim K
User avatar
Liam David
Posts: 509
Joined: Sat Jan 25, 2014 5:30 pm
Real name: Liam David
Location: Arizona

Re: Fusor Update - Liam David

Post by Liam David »

Yeah, I don't think coils will cut it without exceptional contact and high flow rates. I agree a jacket would be ideal, just not sure how to implement that at the moment.

I had one of my 3D printed CA11 connectors (for the Spellman supples) fail a few weeks ago. I was also asked to provide more info about it's construction, so here's some more details.

The connector pictured in my post has been working flawlessly and has withstood 60kV at length. It is 3D printed from PLA with about 20% infill at 0.4mm layer height and potted in neutral cure 737 RTV silicone. Credit where it is due, I was given the 3D model by Chad Ramey and the original author is Peter Schmelcher as discussed in the DXM70N600 thread. Referencing the model (attached below), I used the nose, body, and nut components and discarded the taper parts since they serve no purpose with a small diameter cable. The HV cable is from HVStuff but is no longer sold there. I found it on AliExpress as "150kv 8mm electrostatic spray gun cable" but haven't purchased any more. I kept the insulation on the potted section. The connection between supply and cable is made with standard 2mm bullet connectors pressed into the nose component. O-rings (McMaster #9396K428) sit in grooves on the outside of the connector and the remaining space is packed with silicone vacuum grease. Now I can't remove the connector from the supply due to suction...but it works so no need.

The second connector was identical to the first except I used RG213U coax for the HV cable. I actually potted two cables in the connector to give me access to the floating filament, but I didn't use it. It unfortunately failed at ~55kV by tracking along the cable surface and then arcing to the case. Notably, it did not arc through the connector. I think a trapped air bubble was responsible, but the autopsy wasn't particularly enlightening. I'll make another connector soon with RG213U for the cable, but will ignore the floating filament and use just one cable.

Connector after failure
Connector after failure

No voids found in silicone
No voids found in silicone

Failure mode
Failure mode

Silicone used
Silicone used

Hope that helps.

Liam David
Attachments
RE__Claymount_HV_Connector_STL_files.zip
(5 MiB) Downloaded 416 times
John Futter
Posts: 1848
Joined: Wed Apr 21, 2004 10:29 pm
Real name: John Futter
Contact:

Re: Fusor Update - Liam David

Post by John Futter »

Liam from what i see of your failure the earthy end is far too close to your hot end
extend your printed bit alot further ie x2 and put plenty of silicon stopcock grease over the outside of the cable inner to stop air gaps there allowing corona hence conduction to occur that then carbonises the polyethylene surface
Peter Schmelcher
Posts: 228
Joined: Sat Nov 13, 2010 1:56 am
Real name: Peter Schmelcher

Re: Fusor Update - Liam David

Post by Peter Schmelcher »

Nice work/report.

Some thoughts on the connector.

The commercial CA11 connector had vents and far fewer seals so you could nick the O-rings on the inside to create a vent but alternate the vent positions so they do not line up. If a flash over ever happens it should track and damage the surface of the printed connector.

Consider using a copper plumbing cap as a faraday shield near the pins to control the electric field at the cable end. Drill an interference fit hole(s) and slip the insulated conductor through it. Then electrically connect it to one or more of the banana pins. I would trim away most of the cap to make soldering the pins easier. Might as well polish out any scratches in the copper to reduce field enhancement. FYI, I have not simulated the field improvement just intuition.

I also believe polypropylene and other plastics need to be surface oxidized using a flame for chemical bonding of inks and??? For all I know these days you might find a liquid that does the same thing.

Like your HV vacuum feedthrough.
Cheers
-Peter
User avatar
Liam David
Posts: 509
Joined: Sat Jan 25, 2014 5:30 pm
Real name: Liam David
Location: Arizona

Re: Fusor Update - Liam David

Post by Liam David »

I managed to remove the connector that failed (obviously) but with some undue force. The silicone doesn't stick to the polypropylene as you mentioned, so I'm afraid that the assembly will pull apart. Could probably get the other one out, but it isn't broken so no need to tempt fate. I'll likely keep the o-rings as-is given that shouldn't ever have to remove the connectors less a failure.

I like the idea of a copper cap for field control. Now that you mention it, the solder joint and bullet connectors have pretty small radii and aren't that smooth. If my school's Ansys licensing server ever gets fixed, I could simulate the assembly to verify the hunch.

Thanks for the suggestions. I'll update once I've made some progress.

Liam David
User avatar
Liam David
Posts: 509
Joined: Sat Jan 25, 2014 5:30 pm
Real name: Liam David
Location: Arizona

Re: Fusor Update - Liam David

Post by Liam David »

Just some pictures of stuff I've been working on.

Some problems, things in progress, or needed upgrades:
1. Mass flow controller doesn't like EMI which causes pressure spikes and then cuts off deuterium flow. Obviously not good for stability. Need to rework wiring and shielding.
2. Shielding the DAQ and speeding up the software to reduce errors.
3. Adding more data collection variables including chamber temperature, backing pressure, etc..
4. Data post-processing (filtering, interpolation, error detection, etc.) to remove noise and transients. Mostly done but can be improved.
5. Getting a proper MCA so I can do gamma spectroscopy and more interesting neutron activation.
6. Building a spectrometer to measure deuterium purity without risking my RGA.
7. Building a smaller neutron detector that's easy to move. Perhaps a Russian corona tube or BC-720 replica if I can find some ZnS(Ag).
8. Actually calibrating my neutron detector using more than just simulations. Would like to avoid buying a bubble detector. "Calibration" is currently TIER = 9.27*CPM at 15cm.
9. New aluminum end caps for better cooling and to allow for varying the chamber length.
10. Making the feedthrough more precise since symmetry is my limiting factor. The conductor is slightly bent thanks to McMaster's shipping.
11. Making more graphite grids of various sizes. Neutron numbers aren't that impressive yet.
12. Perhaps a chamber upgrade to a 4.5" conflat... Closer to the ID of the cube fusors and stability seems to be improved.
13. Fixing a Ludlum Model 19 micro-R meter. NaI(Tl) crystal is shattered and some of the front end transistors appear to be shot.
14. Cleaning up the cable salad that resides under my fusor.

Cooling the end caps with 40mm water blocks. Limiting factor is the (very) poor thermal conductivity of stainless. Fan still required to keep other arms cool. Neutron numbers have stabilized and there's no noticeable drop as the chamber warms.
Cooling the end caps with 40mm water blocks. Limiting factor is the (very) poor thermal conductivity of stainless. Fan still required to keep other arms cool. Neutron numbers have stabilized and there's no noticeable drop as the chamber warms.

Neutron detector with paraffin moderator re-wrapped and bolted to the frame at a constant distance from the fusor.
Neutron detector with paraffin moderator re-wrapped and bolted to the frame at a constant distance from the fusor.

Feedthrough with new grid. 3D printed parts for alignment. Note bare stalk to grid.
Feedthrough with new grid. 3D printed parts for alignment. Note bare stalk to grid.

New graphite grid 0.75" OD x 0.75" long. Changing ID appears to have little effect on fusion rates. Needed a lot of heating and outgassing but is now stable.
New graphite grid 0.75" OD x 0.75" long. Changing ID appears to have little effect on fusion rates. Needed a lot of heating and outgassing but is now stable.

Patterns on end caps. Note asymmetry between the two. Grid is slightly misaligned.
Patterns on end caps. Note asymmetry between the two. Grid is slightly misaligned.

Misalignment evident because of the two beams on one side.
Misalignment evident because of the two beams on one side.

Fusion plasma. The system can run indefinitely at 5e5. Arcing limits operation to 50kV currently.
Fusion plasma. The system can run indefinitely at 5e5. Arcing limits operation to 50kV currently.

Data acquisition GUI, since upgraded to 10 samples per second. Neutron counts are acquired via an interrupt on an Arduino and integrated in software. Scalar agrees well with ratemeter.
Data acquisition GUI, since upgraded to 10 samples per second. Neutron counts are acquired via an interrupt on an Arduino and integrated in software. Scalar agrees well with ratemeter.

New aluminum water cooled end caps in the works with 1.5" quick connects.
New aluminum water cooled end caps in the works with 1.5" quick connects.

Cleaning the new grid: https://youtu.be/d3YjzriigFY

Crappy quality video of GUI: https://youtu.be/WbZE7FX8KIw


Some plots using data from the DAQ. All are from the same short run where I varied the current limit from 0 to 8.5mA.

vcpn.png

tier vs voltage.png

tier vs power.png
User avatar
Richard Hull
Moderator
Posts: 14976
Joined: Fri Jun 15, 2001 9:44 am
Real name: Richard Hull

Re: Fusor Update - Liam David

Post by Richard Hull »

Great work Liam! I see you are like me....always working off a list to stuff that needs doing. I am working fusor V prior to HEAS in October and have a project of finally using my $500 BC-720 5" scintillator coupled to my Hamamatsu 5-inch PMT in a light tight housing with internal pre-amp. Doubt if I'll have it ready for HEAS but the 3He detector is as great as always.

Keep up the good work. I look forward to you continued fine efforts. I love that out-sized insulator.

Richard Hull
Progress may have been a good thing once, but it just went on too long. - Yogi Berra
Fusion is the energy of the future....and it always will be
The more complex the idea put forward by the poor amateur, the more likely it will never see embodiment
User avatar
Mark Rowley
Posts: 908
Joined: Sun Dec 29, 2002 12:20 am
Real name: Mark Rowley
Location: Sacramento California
Contact:

Re: Fusor Update - Liam David

Post by Mark Rowley »

Impressive work Liam. I like the water cooled end caps and the new versions look like they can handle a much higher flow rate. Fwiw, I’ve found that Jon’s recommendation of using automotive super coolant makes a significant difference. Throw some ice in the coolant and it’s even better.

You’ll totally enjoy the MCA once everything is all set. Lots of fun.

Mark Rowley
User avatar
Dennis P Brown
Posts: 3147
Joined: Sun May 20, 2012 10:46 am
Real name: Dennis Brown

Re: Fusor Update - Liam David

Post by Dennis P Brown »

Great reporting and data collection/presentation. Top notch.
User avatar
Liam David
Posts: 509
Joined: Sat Jan 25, 2014 5:30 pm
Real name: Liam David
Location: Arizona

Re: Fusor Update - Liam David

Post by Liam David »

I finally got around to modifying a Ludlum Model 3 to output 5V pulses for each detection event, with the intent of making it a scalar. I tapped one of the schmitt triggers and passed the signal through a BNC connector in the meter face. An Arduino acting as a counter sends each second's total to a computer via serial.

The purpose was to make pretty decay graphs. I've been activating indium for several months but hadn't yet gathered quantitative data. I irradiated a ~1.5"x1.25" piece of indium "foil" for ~10min at 5e5n/s. A few-second transfer to a 2" pancake detector and an hour later, I obtain the following curve:

Indium decay curve (sharp drops at ends of curve are an artifact of binning).
Indium decay curve (sharp drops at ends of curve are an artifact of binning).

The decay of In-116m1 (54min) dominates the counts at later times but the initial count rate is much higher due to other short-lived isotopes including In-116m2 (2s), In-116 (13s), In-114 (72s). I haven't done a proper exponential fit, but by eyeballing the graph the counts appear to drop by half after about an hour. If only I had a gamma spec...

Also, some interesting patterns on the graphite grid:
Notice ring on ID and different texture on ends of OD.
Notice ring on ID and different texture on ends of OD.

Modified Model 3 with Arduino.
Modified Model 3 with Arduino.

Beat up, torn, etc.. piece of indium flattened from 1g ingot
Beat up, torn, etc.. piece of indium flattened from 1g ingot

Paraffin neutron oven, in need of an upgrade. Indium is usually sandwiched between the two larger blocks.
Paraffin neutron oven, in need of an upgrade. Indium is usually sandwiched between the two larger blocks.
User avatar
Dennis P Brown
Posts: 3147
Joined: Sun May 20, 2012 10:46 am
Real name: Dennis Brown

Re: Fusor Update - Liam David

Post by Dennis P Brown »

Again, impressive. I'd be very interested in seeing how to connect to a computer via an interface as you did with that Ludlum; that is something others (I'd think) and certainly I would like to learn how to do. Could you maybe post a thread on this and provide details on both the method and code?
User avatar
Liam David
Posts: 509
Joined: Sat Jan 25, 2014 5:30 pm
Real name: Liam David
Location: Arizona

Re: Fusor Update - Liam David

Post by Liam David »

I have an older Ludlum Model 3 (#73001) with a manual that no longer exists on the internet, so I had to do some oscilloscope probing to find a suitable tap point. Pin 3 on the quad NAND gate CD4093BE is pulled high (10v) for 10us for each detection event. To make it TTL and Arduino compatible, I soldered a resistor divider between pins 3 and 7 (VSS). The signal passes to a BNC connector I installed in the front face.

IMG_20200914_123226.jpg

IMG_20200914_123235.jpg

IMG_20200914_123300.jpg

For newer models most of the components have been changed and are also surface mount. I don't have such a meter, but here are a few possible tap points:
model 3 new.PNG
U8 acts as a comparator/discriminator for the signal coming from the preamplifier. The output is passed to the analog meter stages (not shown) and the audio chain, of which U9 is the first component. U9 is a monostable multivibrator, so it gives a single square pulse for each event. The circuit appears to be 5V so a voltage divider may not be necessary.

The signal is then connected to pin 3 of an Arduino UNO microcontroller set up as a scalar. Pin 3 is interrupt-compatible and each rising edge increments a count variable. For different Arduinos the compatible pins vary, however, so check the datasheet. Once a set time has passed, in my case 1s, the Arduino disables the interrupt, sends the count in that one second block to the computer using the Serial USB port, resets the count, and then reconnects the interrupt. In code this is very simple and takes just a few lines:

Code: Select all

#define INTERRUPT_PIN 3 //For Arduino UNO

uint32_t counts = 0; //32 bits is waaay more than needed, but doesn't hurt anything here
uint16_t period = 1000; // in milliseconds

void setup() {
  attachInterrupt(digitalPinToInterrupt(INTERRUPT_PIN), count, RISING); //Attach the interrupt to pin 3, triggering on a rising edge

  Serial.begin(115200); //Open serial communications at 115200 BAUD. Serial monitor on computer must be set to same BAUD rate
}

void count() {
  counts++;
}
 
void loop() {
  delay(period); //Wait for one second, during which count interrupts can happen
  
  detachInterrupt(digitalPinToInterrupt(INTERRUPT_PIN)); //So a particle event doesn't interrupt the serial transmission
  
  Serial.println(counts); //Send count to the computer
  
  counts = 0; //Reset count for next one second block
  
  attachInterrupt(digitalPinToInterrupt(INTERRUPT_PIN), count, RISING); //Reattach the count interrupt
}
On the receiving on the computer, I just open the serial monitor in the Arduino IDE and have the values accumulate in the console. Getting the data is a simple copy-paste. I will write an automatic MATLAB script at some point to gather the serial data in a more elegant way, but for now this suffices.
User avatar
Liam David
Posts: 509
Joined: Sat Jan 25, 2014 5:30 pm
Real name: Liam David
Location: Arizona

Re: Fusor Update - Liam David

Post by Liam David »

I spent some more time in the workshop today and put together a rudimentary MCA for my gamma spectroscopy setup. A Ludlum Model 12 provides the HV and the probe is a custom 85x60mm NaI(Tl) scintillator mated to a Photonis XP5312 9-stage PMT. A 3D printed housing holds everything together. The model 12 is tapped directly behind the HV capacitor and the signal is sent through a BNC connector installed into the can. I used a simple op-amp inverter and RC filter to smooth and shape the pulses such that peaks last a couple microseconds.

PMT input left, output to ADC right.
PMT input left, output to ADC right.

The ADC is a "Particle Photon" microcontroller that is intended to be used as an Internet of Things platform. Its core clock is 120MHz so it's far, far faster than an Arduino. The other main attraction is its 16-bit SAR ADC. I disabled the wifi which hogs processing power and simply connected the RC output of the circuit to pin A0. Unfortunately the usual analogRead() cannot achieve the necessary sample rates because of overhead and various default ADC settings it uses. I haven't measured the performance, but I'm guessing it's a few 10s of kSa/s at best. Since each sample takes ~100us and the peaks are only ~2us long, the problem is apparent.

The solution is to interface with the ADC directly, and interleave sampling with peak processing to reduce overall dead time. I set the ADC sampling period to just 3 clock cycles and and adjusted several other settings that I found in code examples online. The sampling and binning algorithm, in order of execution, is as follows:

1. Initialize an array of 2048 values representing the spectrum.
2. Initialize a buffer of 256 values that store the ADC data as they arrive.
3. Create two variables storing the current peak height and its index in the buffer.
4. For each index in the buffer, start an ADC conversion.
5. While the conversion is in progress, the CPU is free. Compare the previous ADC sample in the buffer to the current max peak height. If it's greater, the max peak height is this value.
6. Retrieve the current ADC reading.
7. Repeat 4-6 until the buffer is full.
8. If the peak height is above the threshold, in my case 1500/65535, and it does not occur within 10 indices of the start or end of the buffer (to mostly avoid catching rising or falling slopes), divide it by 32 and increment that index of the spectrum by one.

Every 1000 counts the spectrum is sent to the computer via serial. Dead time from processing and such is <5% so the majority of peaks are captured. It is important to note that this algorithm only captures a single peak (the maximum) per buffer, so obscuration of lower-energy peaks is possible but really only an issue at very high (>500kCPM) count rates. The most impressive metric, however, is that I can achieve an astonishing 900kSa/s at 32 bits. This means I can accurately capture the peak pulse heights which last ~2us. If I configure the microcontroller to periodically send the buffer instead of the spectrum, I can see each pulse:

Concatenation of multiple buffers showing peaks
Concatenation of multiple buffers showing peaks

I don't have many gamma sources, but I do have americium. The spectrum is a little rough, but I get ~25% FWHM for the 59keV gamma and something around 5-6% for background peaks around 600keV. The 26keV peak is "there" too, but the op-amp circuit smears it out since the pulse height is so small. Background peaks like K40 are there, as well as a couple others I am unsure about.

Am241 spectrum
Am241 spectrum

Some tweaking is still to be done, but I think this poor-man's MCA, costing some $20, should suffice for measuring indium and other activation products. Getting an indium spectrum is next on the list.

Here's the code, as a backup for myself and interested people:

Code: Select all

#include "Particle.h"

SYSTEM_MODE(MANUAL);
SYSTEM_THREAD(ENABLED);

const uint16_t BUFFER_SIZE = 256;
uint16_t BUFFER[BUFFER_SIZE];

const uint16_t NCHANNELS = 2048;
uint32_t CHANNELS[NCHANNELS];

const uint16_t DIVISOR = 32;

uint32_t COUNTS = 0;

void setup() {
	WiFi.off();
	pinMode(D0, OUTPUT);
	
	setADCSampleTime(ADC_SampleTime_3Cycles);
	analogRead(A0);
	
	ADCx_Configuration(ADC2);
	configure_ADC_channel(ADC2,A0);
	
	Serial.begin(230400);
	waitFor(Serial.isConnected, 2000);
}

void loop() {
	while(true) {
		ADC_SoftwareStartConv(ADC2);
		while(ADC_GetFlagStatus(ADC2, ADC_FLAG_EOC) == RESET);
		BUFFER[0] = ADC_GetConversionValue(ADC2);
		uint16_t maxVal = BUFFER[0];
		uint16_t maxIdx = 0;
		
		#pragma unroll
		for(int i=1;i<BUFFER_SIZE;i++) {
//			pinSetFast(D0);
      
			ADC_SoftwareStartConv(ADC2);
			
			if(BUFFER[i-1]>maxVal) {
				maxVal = BUFFER[i-1];
				maxIdx = i-1;
			}
			
			while(ADC_GetFlagStatus(ADC2, ADC_FLAG_EOC) == RESET);
			BUFFER[i] = ADC_GetConversionValue(ADC2);
     	
//      pinResetFast(D0);
		}
		
		if(maxVal>1500 && maxIdx>10 && maxIdx<BUFFER_SIZE-10) {
			CHANNELS[maxVal/DIVISOR]++;
			COUNTS++;
			
			if(COUNTS%1000==0) {
				sendSpectrum();
//          sendBuffer();
			}
		}
	}
}

void sendSpectrum() {
	for(uint16_t i=0;i<NCHANNELS;i++) {
		Serial.print(CHANNELS[i]);
		Serial.print(" ");
	}
	
	Serial.println();
}

void sendBuffer() {
	for(uint16_t i=0;i<BUFFER_SIZE;i++) {
		Serial.print(BUFFER[i]);
		Serial.print(" ");
	}
	
	Serial.println();
	Serial.println();
}

uint16_t readADCx(ADC_TypeDef* ADCx){
  ADC_SoftwareStartConv(ADC2);
  while(ADC_GetFlagStatus(ADC2, ADC_FLAG_EOC) == RESET);
  return ADC_GetConversionValue(ADC2);
}

void configure_ADC_channel(ADC_TypeDef* ADCx, uint16_t pin){
  STM32_Pin_Info* PIN_MAP = HAL_Pin_Map();

  HAL_Pin_Mode(pin, AN_INPUT);
  ADC_RegularChannelConfig(ADCx, PIN_MAP[pin].adc_channel, 1, ADC_SampleTime_3Cycles); //4.4 µs ???
}

void ADCx_Configuration(ADC_TypeDef* ADCx) {
  ADC_CommonInitTypeDef ADC_CommonInitStructure;
  ADC_InitTypeDef ADC_InitStructure;
	
  RCC_APB2PeriphClockCmd(RCC_APB2Periph_ADC1, ENABLE);
  RCC_APB2PeriphClockCmd(RCC_APB2Periph_ADC2, ENABLE);
  RCC_APB2PeriphClockCmd(RCC_APB2Periph_ADC3, ENABLE);
  ADC_CommonInitStructure.ADC_Mode = ADC_Mode_Independent;
  
  ADC_CommonInitStructure.ADC_Prescaler = ADC_Prescaler_Div2;
  
  ADC_CommonInitStructure.ADC_DMAAccessMode = ADC_DMAAccessMode_Disabled;
  
  ADC_CommonInitStructure.ADC_TwoSamplingDelay = ADC_TwoSamplingDelay_5Cycles;
  ADC_CommonInit(&ADC_CommonInitStructure);
  
  ADC_InitStructure.ADC_Resolution = ADC_Resolution_12b;
  ADC_InitStructure.ADC_ScanConvMode = DISABLE;
  ADC_InitStructure.ADC_ContinuousConvMode = DISABLE;
  ADC_InitStructure.ADC_ExternalTrigConvEdge = ADC_ExternalTrigConvEdge_None;
  ADC_InitStructure.ADC_DataAlign = ADC_DataAlign_Right;
  ADC_InitStructure.ADC_NbrOfConversion = 1;
  ADC_Init(ADCx, &ADC_InitStructure);
  
  ADC_RegularChannelConfig(ADCx, ADC_Channel_14, 1, ADC_SampleTime_3Cycles);
  
  ADC_Cmd(ADCx, ENABLE);
  // Start ADCx software conversion
  //ADC_SoftwareStartConv(ADCx); 
}
User avatar
Liam David
Posts: 509
Joined: Sat Jan 25, 2014 5:30 pm
Real name: Liam David
Location: Arizona

Re: Fusor Update - Liam David

Post by Liam David »

I finally got around to 3D printing and potting a new connector for the 2nd power supply, so max current is now 17mA. So far the connector's been fine to 50kV. Each power supply has its own 200k ballast resistor and protective diode string.

Potted connector before o-rings were installed and the whole thing covered in silicone grease.
Potted connector before o-rings were installed and the whole thing covered in silicone grease.

Since I'm using a linear grid, the corresponding endcaps get very hot during operation. Water cooling helps, but at 600-700W input the conductivity of stainless and flow rate aren't high enough and the chamber arms reach some 100C (making better endcaps is still on the list). In a previous post I showed the blue patterns that formed on the endcaps. These are now mostly gone, seemingly ablated by the extra current and heat and leaving bare metal again. I don't think it's simply a thin-film of non-interfering thickness since the graphite grids I use just turn the rest of the chamber black.

Endcaps almost clean again
Endcaps almost clean again
Other endcap
Other endcap

Removing the latest grid, which is graphite and more polished than those I've previously used, revealed a thin ring bisecting the ID. This is the same effect noted by a couple others using linear grids.

Ring in grid
Ring in grid

The graphite tends to outgas a lot, especially when it starts to glow, and the fusion rates plummet as the pressure rises and deuterium purity drops. My best runs are with the turbo minimally throttled and flowing a lot of gas. Baking the graphite with a blowtorch mostly solves outgassing but turns it glassy, crumbly, and brittle. Next step is to use a stainless grid just made from some tubing. My best run so far is ~8e5n/s for a few seconds before outgassing became a problem. 5e5 is sustainable indefinitely.

A few interesting thing of note:

I haven't identified the datasets (from my DAQ) which correspond to this effect, but I noticed a decrease in grid heating even with a 2x increase in power. At 45kV and 17mA the grid didn't start glowing even after several minutes. Once I decreased the voltage to 35kV at 17mA, the grid began glowing after a dozen or so seconds. This effect seems to be extremely sensitive to symmetry. Since my feedthrough is a quartz tube, I can rotate and vertically move the grid 0-360deg and +/-2cm. After messing with the positioning since the feedthrough stalk is slightly bent and needs upgrading, the effect disappeared in later runs. I haven't yet attempted replication since I need to improve the feedthrough symmetry first. I have very little experience with ion optics, but I'm inclined to believe that the increased voltage led to fewer ion-grid collisions caused by space-charge deflection. Also, the thermal angular momentum which contributes to defocusing is proportionally smaller because of the increased ion velocity. While I didn't pay explicit attention to the gauge, the drop in voltage must have been accompanied by a pressure increase, so background scattering rose as well. Bunch of ideas, don't know how much credibility they have.

With the most recent graphite grid (the outgassing one) I observed a remarkable increase in neutron output with a decrease in voltage (at constant current). Immediately after turning the power on the voltage dropped from ~44.3kV to ~42.3kV over about 15 seconds. Over the same interval, the neutron output rose from ~3.7e5 to ~4.6e5 (I use a scalar with an integration time of 3s). The pressure likewise rose from ~19.5mtorr to ~21.5mtorr. After those 15 seconds, however, the neutron output started plummeting. Another 15 seconds later (i.e. t0 +30), the stats were ~40.8kV, ~3.3e5, ~23mtorr. Clearly less voltage makes a difference, but I think a few other variables are at play. Initially the grid outgassed its absorbed deuterium, leading to higher fuel density and thus more beam-background fusion. This offset the voltage-decreased cross-section. After this source became depleted and the grid continued heating, the remaining air in the grid began contaminating the plasma. Notice the changing slope of the pressure rise not long after neutron output dropped. My gauge is a cold cathode which is has gas-dependent readings, and deuterium reads about 2x higher than air, so the decreasing slope checks out. A confounding factor is the increased permeation distance of the air as it outgasses from deeper in the grid, but we might expect this to be offset by the heating. It was glowing red by this point.

Time-series voltage, pressure, TIER. x-axis in 0.1s increments so 120s are shown.
Time-series voltage, pressure, TIER. x-axis in 0.1s increments so 120s are shown.

I intend to do independent voltage, current, and pressure parameter sweeps once I reinstall the upgraded feedthrough and have some respectable symmetry. More progress to come, school permitting.
User avatar
Richard Hull
Moderator
Posts: 14976
Joined: Fri Jun 15, 2001 9:44 am
Real name: Richard Hull

Re: Fusor Update - Liam David

Post by Richard Hull »

Great continuing work on your system. It seems there are always tweeks that we seem to find interesting to test out to better the performance. Good work and continued good luck in your efforts.

Richard Hull
Progress may have been a good thing once, but it just went on too long. - Yogi Berra
Fusion is the energy of the future....and it always will be
The more complex the idea put forward by the poor amateur, the more likely it will never see embodiment
User avatar
Liam David
Posts: 509
Joined: Sat Jan 25, 2014 5:30 pm
Real name: Liam David
Location: Arizona

Re: Fusor Update - Liam David

Post by Liam David »

My 2nd Spellman supply gave up the ghost the other day, so now I'm limited to 8.56mA/600W again. One of the two flyback's secondaries got crispy causing the voltage and current to misbehave, especially in parallel with the other supply. Unfortunately it's very much a proprietary part and I don't have the ability to wind another. I believe it was end-of-life when I bought it and operating the fusor finally pushed it over the edge.

Flyback.jpg
Flyback 2.jpg

After years of use, I was finally forced to clean my chamber. The graphite grids I've been using have a habit of depositing on the walls, and after a particularly hot run the other day the coating began flaking and causing arcing. I scrubbed the inside with fine steel wool, used some vinegar to help loosen things up, rinsed with isopropyl, and baked it in the oven at 200C for a few hours. While not shiny like new, it's a lot better.

Deposition.jpg
Thick deposition flaking after intense heating.
Thick deposition flaking after intense heating.

The graphite grids were outgassing enough to limit long-term stability and letting the chamber sit overnight replenished the reservoir even after a thorough plasma bake. I made one from a 3/4" stainless pipe, with the added bonus of it not being super fragile. It turned nice and blue after heating up and there's a distinct ring on the inside as was also seen by Jon Rosenstiel. The ends also have darker bands.

Grid.jpg

There are two main limiting factors at the moment: symmetry and heating. The asymmetric discoloration on the two endcaps and just eyeballing were enough to confirm that the grid isn't quite coaxial nor centered. I 3D printed a new feedthrough aligner and a temporary jig that mounts to a flange and passes through the grid to improve things. Cooling the endcaps using water blocks is insufficient for one simple reason: stainless steel has terrible thermal conductivity. I'm planning to machine an aluminum chamber based on the new cube designs, and the larger ID will also help with stability at higher voltages.

Grid jig.
Grid jig.

I made a new lead box for the chamber which has helped cut down on x-rays a lot.

It has a lid.
It has a lid.

The data acquisition box was in dire need for an upgrade. Here's the much cleaner and less-prone-to-shorting V2, which has op-amps buffering the inputs/outputs to hide the impedances of the instruments, cables, etc... and a more spacious container. It'll also house the electronics, including full data logging, for a 2nd 3He neutron detector once that arrives. I plan to measure the anisotropy of the neutron output...

Control Box.jpg

More to come after I finish the 2nd neutron detector.
User avatar
Richard Hull
Moderator
Posts: 14976
Joined: Fri Jun 15, 2001 9:44 am
Real name: Richard Hull

Re: Fusor Update - Liam David

Post by Richard Hull »

Nice work, Liam! You have done amazing things in that little cross. I just gave up on my effort in so small a vessel. I look forward to your continuing reports.

Richard Hull
Progress may have been a good thing once, but it just went on too long. - Yogi Berra
Fusion is the energy of the future....and it always will be
The more complex the idea put forward by the poor amateur, the more likely it will never see embodiment
User avatar
Mark Rowley
Posts: 908
Joined: Sun Dec 29, 2002 12:20 am
Real name: Mark Rowley
Location: Sacramento California
Contact:

Re: Fusor Update - Liam David

Post by Mark Rowley »

Looking good Liam. Your lead box is similar to what I built which worked quite well. The remaining issue was leakage through the ceramic HV feedthru. Over 50kV the backscatter from that point was still an annoyance.

I’m sorry to hear about the Spellman demise and am going through a similar situation. I decided to try and wind my own flybacks and so far have had some encouraging results. I’m still waiting on some shipping delays with ordered parts but once done I’ll post more about the results and winding methods. Maybe something similar can be done with your supply.

Mark Rowley
Andrew Seltzman
Posts: 814
Joined: Sun Feb 01, 2004 8:02 pm
Real name: Andrew Seltzman
Contact:

Re: Fusor Update - Liam David

Post by Andrew Seltzman »

I happen to remember that a few some forum members bought spellmans that didn't due to issues on the circuit board, those likely have salvageable flybacks in them or perhaps a good candidate for a circuit board swap.
Andrew Seltzman
www.rtftechnologies.org
Post Reply

Return to “Images du Jour”