Ryan Ginter's Power Supply

This forum is for specialized infomation important to the construction and safe operation of the high voltage electrical supplies and related circuitry needed for fusor operation.
Post Reply
Ryan Ginter
Posts: 51
Joined: Fri Nov 25, 2022 9:25 am
Real name: Ryan Ginter

Ryan Ginter's Power Supply

Post by Ryan Ginter »

This will be a very long post, my intention is to detail my learning process in an attempt to expose any misconceptions I may have come upon. I'm not asking for the right answers here, but if the way I'm going about any of this is fundamentally wrong I would appreciate it if someone would call it out.

With my vacuum system nearing completion, it's about time I got to work on a power supply. My prior experience with high voltage electronics is limited to a nst driven tesla coil I made almost 10 years ago. My intent is to design the supply in a manner that allows easy modification, such that it can be improved over time as my understanding of high frequency switching supplies progresses.

What are the initial design requirements for the first iteration of this supply?
  • A simple design that can be produced at home without the need for specialized equipment.
  • An output voltage that can be throttled using a variac, allowing me to control the fusor from a distance.
  • To avoid the use of expensive components, allowing for cheap and rapid replacement should damage occur.
  • Two 5kV outputs relative to the center-tap, 180 degrees out of phase to allow the driving of a full wave Cockroft-Walton multiplier

For this reason, the first version of the power supply will be driven using a commercial ZVS. Using this type of driver does present a few disadvantages. Primarily, it lacks deep throttling for voltage and has no method of constant voltage or current control.

The ZVS I'm using was selected for its low cost. Going into the project I was certain it would be destroyed in my testing. Despite this, it has managed through all tests thus far, not bad for $20. The listed rating was 30V at up to 20A. I've been able to drive the unit a few volts above this so far without issue.

Should the ZVS driver fail, or prove incapable of the power needed for a fusor, I will replace it by making my own.

What voltage will the primary coil be driven at?

Assuming it will be driven at its rated voltage of 30V

Vout(pk-pk) = Vin*pi
Vout(pk-pk) = 30V*pi = 94.2V(pk-pk)


Being a half-bridge driver, we only expect to see half this potential.

Vout(pk-pk) = 47.1V(pk)

So the bus voltage for the primary of the transformer will be 47.1V(pk)*0.707 or 33.3V(rms).

The first transformer I constructed was made for the sole purpose of testing. Being new to high frequency transformers, I wanted to probe its behavior. Regardless, I still wanted the voltage output to be similar to that of the final transformer, so I made the secondary using 1,760 turns of 26 AWG magnet wire.

My 3d printer was used to produce the bobbin. There were 8 segments in total, each of which was given 220 turns of wire. While a 3D printer may technically count as specialized equipment, they are rather commonplace today and there are a variety of ways a similar coil could be made.
Winding First Transformer.jpg
The resulting Secondary was placed into a UY30 Ferrite core.
First Transformer (2).png
Next, I needed to determine the minimum number of turns for the primary coil.

Most of the calculation is straightforward, however using a ZVS driver I won't have any control over frequency. I needed to guess a range and then verify experimentally. Had I made my own ZVS it would have been possible to calculate the frequency, but the driver I purchased did not list the values of its components. I suppose I could have measured the values, but I didn't consider this worth the time.

With a UY30 core I get a cross-sectional area of 5.98sqcm. I'll calculate with a possible frequency range of 30 kHz to 80 kHz.

N(turns) = Vin*10/delta B*Ae*F*2

N(turns) = 33.3*10/0.32*5.98*80*2
N(turns) = 1.09

or

N(turns) = 33.3*10/ 0.32*5.98*30*2
N(turns) = 2.87


So I considered 3 turns the minimum for my testing.

I was now prepared to test the transformer with different arrangements of primary coil. The first primary tested was the 5+5, as I thought this would cause the lowest level of flux in the core(it was actually the highest). It was here that I discovered the Secondary coil had been wound wrong.

No arcing occurred after bringing the two outer wires together. I must admit I was having difficulty visualizing the coil orientation from the 2D images on the transformer FAQ. Because of this, I focused on the prior description of the center-tap being a wire connected to the middle of a continuous coil. From this I concluded that if the entire coil were wound in the same direction, the phasing would be correct. Unfortunately, the method I used to construct the coil caused the phasing to be reversed.

Rather than pass the wire through gaps in a bobbin, I had thought it would be better to have tunnels printed into the dividing segments to pass the wire through. The result of this was the need to wind the bobbin segments individually and then solder the terminals together. The wire would start at the bottom of a segment, wind a coil to the top, and then pass through an angled tunnel leading to the bottom of the next segment. The issue however, was the center two segments of the coil had the bottom connected to the bottom. This caused the winding direction to be reversed, despite the bobbin always rotating the same way.

Primary Testing

Arcs could still be drawn between either outer wire and the center-tap, so I continued with testing. A single loop of wire was wrapped around the ferrite core and connected to my oscilloscope, allowing measurement of the results.

It should be noted that the ZVS DC input voltage reading was of questionable accuracy, as I was only going off the voltage display on my cheap lab bench power supply. The oscilloscope reading was also not particularly accurate, as it was old, and the trace didn't focus particularly well. As such, the following measurements were rounded at two figures for voltage and three for frequency. The period was measured both with an arc drawn and unloaded.

5+5 Primary, 31V DC input, 34.5kHz (9.26kHz no load), Theoretically 9.74Vpk, measured 9.5Vpk.
4+4 Primary, 31V DC input, 45.5kHz (10.5kHz no load), Theoretically 12.2Vpk, measured 12Vpk.
3+3 Primary, 30V DC input, 62.5kHz (25kHz no load), Theoretically 15.7Vpk, measured 16Vpk.

An image of the Oscilloscope measuring the 3+3 Primary
The X-axis was set to 10 microseconds/division. I measured voltage and frequency on separate runs, so the trace is not centered on the Y-axis in this image. The scaling was also not ideal for reading voltage.
3 Turn Primary Measurment.png
From this I could see theory matched closely with measurement, thus I can estimate the V/turn of any coil I build. Assuming an ideal transformer with 1,760 turns, the approximate voltage output for 33.3V(rms) input would be the following.

5+5 Primary, 16,600Vpk or 11,700Vrms
4+4 Primary, 20,700Vpk or 14,700Vrms
3+3 Primary, 27,600Vpk or 19,500Vrms


Of course the impedance and leakage inductance of any real transformer will cause the actual voltage to be lower than this. The voltage output of the transformer I built would also have been halved due to the phasing error.

With frequency measurements now in hand, I can look back to the transformer equation and find the core saturation during each test.

5+5 Primary, 5 turns = 330/deltaB*5.98*34.5*2, deltaB = 0.16
4+4 Primary, 4 turns = 330/deltaB*5.98*45.5*2, deltaB = 0.15
3+3 Primary, 3 turns = 330/deltaB*5.98*62.5*2, deltaB = 0.15


Interestingly, in all configurations it seems to be reaching about 1/4 saturation. This likely isn't ideal, but it leaves plenty of room to increase the drive voltage in the future. At the very least I shouldn't have to worry about anything getting hot in this setup.

While I did not have the equipment on hand to take quantitative measurements of the secondary output, some qualitative observations were made.

Arcs would start when the wires were brought within a few mm from each other, they were about 0.5cm thick and mostly white in color, with slight tints of blue and yellow. The behavior would be described as halfway between spark like and flame like. Arcs extinguished after being drawn more than 4" apart. At a current draw of 10A the arcs rose about 2" in the center. 10A was the highest input current tested, as that was the limit of my power supply.
First Transformer Arc.png
I noticed the current draw dropped below unloaded levels when the secondary was shorted, is this some sort of self-ballasting from leakage inductance?

Second Attempt

Disappointed with the phasing issue of the first coil, I set about constructing a new secondary. As with the first attempt this secondary was only intended for testing use. Rather than testing the primary, this coil would be used to test the construction method for a completed transformer.

The new coil was again wound with 1,760 turns divided across 8 segments, only this time it was printed as two separate 4 segment bobbins. I was aware that constructing it in this way made it vulnerable to internal arcing, but it was for testing purposes only. The primary of the transformer used a 5+5 turn coil, giving an ideal secondary output voltage of 11.7kV(rms).

I had intended for the final version of the transformer to be submerged in oil, and this two part bobbin would certainly also require oil to prevent arcing, but I had not yet finished the container for the transformer. I had wanted to test the transformer, so I told myself I wouldn't drive the ZVS input past 15V, but upon seeing the output arcs I fell into the temptation and increased the voltage to 20V. It was at this point the insulation on the transformer windings broke down.

The arc started from the coil segments near the edge on both ends, shorting through the ferrite core. Luckily, the damage only occurred to the outermost windings. I was able to repair it by coating the outside in resin. The arc had traveled 8.5mm distance per side, for a total of 17mm through air. The sound of corona discharge could be heard before the arc. This most likely allowed for a buildup of ions in the air, reducing its dielectric properties.
Damaged Transformer.jpg
After making the repairs I finished building the container and assembled the transformer.
Finished Transformer.jpg
Making a new power supply

Having reached the limits of my benchtop power supply, it was time to begin construction of the variable rectified mains supply.

I had acquired a Flux-O-Tran resonant transformer manufactured by Kepco Inc. The transformer had been pulled from its original supply, so there was no capacitor attached. The eBay listing stated the input was 115V and the output was 30v at up to 30A. It also stated there was a second pair of output wires with an untested voltage.

Upon receiving the transformer, I started by measuring the resistance of all wires, both to determine what wires had internal connections and what their intended use was. The only connections were between gray/gray(0.6ohm), yellow/black(0.2ohm), and red/green(2.4ohm). Normally this would indicate red/green to be the mains side, but the seller had listed the gray/gray wires as 115V.

I next used my LCR meter to measure inductance. Gray/gray read 60mH, black/yellow read 2mH, and red/green read 936mH. This allowed me to estimate the ratio of the transformer. Gray/gray to black/yellow was a step-down ratio of 5.5, and gray/gray to red/green gave a step-up ratio of 4.
Next, I attached the gray wires from the transformer to the output of a variac, probing the wires at various voltages.

Output measurement of Black/Yellow

10.02Vin, 1.99Vout
60.02Vin, 12.05Vout
146.2Vin, 27.71Vout


This gave a step-down ratio of about 5, rather close to the ratio of measured inductance. At 146.2V of input the red/green wires had an output of 570V, definitely something I need to be careful with. Thankfully they are by themselves on the back side of the transformer, so I'll be insulating them and attaching a cover to prevent contact.

The full rectifier circuit was made as follows
The output of the variac was passed through a 6A breaker, and then connected to the input of the step-down transformer. The output of the transformer was connected to a 150A rated full bridge rectifier. The output of the rectifier passed through 2 50v, 1000uf capacitors in parallel. This ran through a 75mV shunt and a volt/amp meter was connected to monitor the output.
Rectified Mains Supply.jpg
Measurements of rectifier input/output

8.25Vrms in, 10.0Vdc out
15.4Vrms in, 20.0Vdc out
22.1Vrms in, 30.0Vdc out
27.99Vrms in, 39.2Vdc out


Theoretically, this should have given me 11.7, 21.8, 31.2, and 39.5 volts at the output, but the forward drop of the diodes along with any other inefficiencies gave the reduced values of the measurements.

This gave me a variable dc power supply capable of 0-40V.

My first attempt at using the new supply to drive my ZVS resulted in a large voltage drop occurring around 10A of output, preventing me from exceeding around 250W of input power. There were two magnetic shunts between the primary and secondary of the step-down transformer. After removing them with a hammer and punch, the voltage drop was greatly reduced.

The high voltage transformer was now capable of drawing arcs nearly a foot in length. The arcs were clearly much hotter than those of the previous transformer. To provide a sense of scale, the electrodes in this image are a 2.75" and 6" conflat gasket.
High Voltage Arc.png
The volt/amp meter indicated 33.0V at 16.5A just before the arc extinguished, giving an input power of 500W. It should be noted that the variac was turned all the way up, so the dc input should have been 40V. This shows that there is still a significant voltage drop from the rectifier supply under load.

I tested the system by holding a drawn arc at 300W of input power for a duration of 5 minutes. I would have tested it at higher input power, but the length of arc determined the power draw and arcs were much harder to maintain at longer lengths. After the five-minute test I checked the temperature of all components, with the exception of the full bridge rectifier, all components were cool to the touch.

This concludes all of the testing I have conducted to this point. I will take current and voltage measurements of the secondary output in the future, but for now I intend to move ahead to version three of my transformer.

The following are issues with version two that I would like to address going forward
  • The maximum power draw I was able to achieve was 500W, though this would be sufficient for proving fusion through neutron detection, the real output power will be lower than the input. For this reason, I would like to increase power draw on the next transformer
  • Since fusion requires the power supply to operate reliably for long durations, I will need to reduce the risk of internal arcing
  • While I succeeded in preventing leaks from the project box, the wires are wicking out the oil and creating a mess
  • The transformer is too difficult and time consuming to wind due to the tunnels
  • The transformer has internal solder joints that increase resistance and the likelihood of arcing
  • The transformer is difficult to remove from the project box
  • The transformer frequency is too low, this will increase impedance in the multiplier, reducing the capacity for current to flow through it. This will result in a reduction of voltage to the fusor electrodes
  • It is possible for the ferrite cores to shift their position, causing leakage inductance and frequency to vary
The following changes are how I plan to address these issues.
  • The current flowing through the ZVS driver is determined by impedance. Reducing the number of turns on the primary will cause a reduction to both resistance and inductive reactance. I would also like to wind the primary with a larger gauge wire to further reduce resistance. It will need to be made from litz wire however, due to the skin effect
  • To prevent arcing, I will be increasing the spacing between the wires and the core. As the primary turns are being reduced, the secondary turns can also be reduced while maintaining the same voltage. As a result, there will be a larger volume available for insulation
  • In an attempt to make swapping components on the transformer less time consuming, I will be designing a bobbin for the primary coil. In this way the primary can be added or removed without having to wind any coils. Additionally, I will be attaching the wires of the primary and secondary coils to the passthrough of the project box using washers and nuts. This will allow them to be removed from the box without having to drain the oil
  • To prevent oil wicking through the wires, I will be passing a copper tube through the cable glands instead of a wire. The copper tube will have a wire soldered to the side facing out of the box, and a bolt to the side facing in. The space between the bolt and wire will be filled with solder to prevent oil from passing through. The primary or secondary coil leads will be attached to the bolts with a nut, and the washer the leads are soldered to will contact the copper pipe
  • The tunnels through the bobbin sections will be replaced with open gaps, allowing the whole coil to be made from one continuous length of wire
  • A plastic stand will be connected inside the box. It will apply pressure to ensure the core halves stay in contact
I will be going with a 3+3 turn primary instead of 5+5, this higher frequency will allow lower impedance on the multiplier, thus higher output power. The secondary will have 960 turns. This will provide two leads of opposite phase, each with 5.3kV of potential with the center-tap.

Making litz wire

25 strands of 38 AWG magnet wire were strung out across a distance of 10ft. The wire was pulled to equal length and placed in a drill. The wire was turned by the drill for 45 seconds, then released and allowed to relieve tension. Four wires total were wound in this fashion, and these four wires were then attached to the drill and used to make a single wire containing 100 strands. A total of 4 wires like this were then made and wound together again, creating a single cable with 400 individual strands.
Making Litz 1.jpg
Litz Wire 2.jpg
New Bobbin

The New bobbins were designed to be much easier to wind. They also allow for both the primary and secondary coil to be removed simply by pulling the core apart. The secondary has extra insulation from the core, both under the coil and along the sides. There is still some risk of arcing from the secondary to the primary coil, but the primary bobbin centers the coil over the lowest voltage area of the secondary. The gap from primary wire to secondary wire is approximately 20mm, so an arc would need to travel around 40mm through oil to cause damage. It should be noted that the top windings of the secondary coil end 5mm short of the dividers.
New Transformer.png

I am hopeful that my new transformer will reach the full 20A rating of the ZVS driver. Once this is complete, I will move on to the construction of my Cockroft-Walton multiplier.
User avatar
Richard Hull
Moderator
Posts: 14991
Joined: Fri Jun 15, 2001 9:44 am
Real name: Richard Hull

Re: Ryan Ginter's Power Supply

Post by Richard Hull »

First rate work and a fabulous and complete report on this very important topic that is often a point of trouble for so many folks looking to home build a suitable high voltage supply. As noted: 500 watts is a start at a suitable supply. One kilowatt is more like a real design goal with 1.5KW being a good reserve capability so often seen at the first strike of a plasma in a fusor. Most fusors, when working well, rarely need more than 750 watts in running mode. It is that reserve that keeps supplies from blowing up in the effort of getting started. We have seen the frustration of those using $$$$ commercial supplies of just barely enough rating. which almost always include over current protection, shut down at the critical moment of plasma ignition. Over building for current reserve at any voltage in the desiderata.

It has long been known that a high frequency drive system at a suitable high voltage and current capability was the key to utilize and form a workable voltage multiplier system. This reduces the need for large, expensive, high energy storage capacitors demanded in normal mains systems employing voltage multipliers.

I hope this type of report will aid others looking to build their own fusor, fusion capable power supply. The beauty here, is that this posting starts at the beginning of the beginning of all such efforts, rather than rush to a finished product. It is how all such complex tasks must be undertaken and that is in an appropriate number of baby steps from beginning to end of any such endeavor.

Thanks for the effort to assist others in a shared experience at your end of this long road. Baby step #1.

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
Ryan Ginter
Posts: 51
Joined: Fri Nov 25, 2022 9:25 am
Real name: Ryan Ginter

Re: Ryan Ginter's Power Supply

Post by Ryan Ginter »

As Richard has stated, I now intend to construct my own ZVS driver capable of 750W sustained output with a reserve of 1,500W. I plan to rewind my step-down transformer to provide the new driver with 50Vdc at the input. However, I will not begin working on the new driver until all the other components of the supply have been tested.

The Cockroft-Walton multiplier I plan to build will have three stages. Using the commercial driver and version three of my transformer, the ideal voltage output would be about 45kV. Using the same transformer with my self-made driver would then provide an ideal output of around 75kV. Naturally, the real outputs will be lower than these values. The highest peak voltage across the transformer would be 15kV with the commercial driver and 25kV with my own, so I will need to keep that in mind to avoid internal arcing.

I will be using the commercial ZVS to test the voltage and current meters I plan to build. By pumping the chamber into a deep vacuum, I will be able to operate the fusor as an X-ray tube. This will allow measurement of both current and voltage without the risk of destroying anything from striking a plasma.

It is only once I've been satisfied with the performance of my metering equipment and the grounding of the supply that I will modify the step-down transformer and replace the ZVS.

Thanks for the kind words Richard. All people living today were born upon the shoulders of giants. Thanks to their contributions, anyone can go about achieving rather remarkable things, but I believe there is great worth in understanding how the giants came to stand so tall.
Post Reply

Return to “High Voltage - Fusor Input Power (& FAQs)”