I spent over a year researching the Adams Pulsed Electric Motor Generator, built several working models, gave a presentation on the subject at the 2021 Energy Science & Technology Conference (ESTC), and even used MyHeritage.com to track down one of the inventor’s daughters in the hope that she would be able to give me the last bits of information I needed to replicate Robert Adams’ impressive results.
Unfortunately, I hit a lot of dead ends, got discourages, and spent most of this year working on Tesla Turbines instead.
However, at this year’s conference, Paul Babcock and Mike Clarke gave two very inspiring presentations and hinted at new avenues to explore, like using state-of-the-art silicon carbide or gallium nitride MOSFETs for superior performance, which fully rekindled my enthusiasm for pulse motors!

I already have a design for a new and improved pulse motor in my head, but first I wanted to write this article for anyone who is experimenting with a Bedini SG, Adams Motor, Newman Motor, or any other pulse motor generator, and needs some inspiration on how to push their project to the next level of performance.
I’ll cover each and every part of a pulse motor generator system and how to maximize its potential, as well as how difficult and costly it is to implement.
So let’s go!
Table of Contents
- Intro: What are we optimizing for?
- Switching: SiC or GaN FETs
- Gate driver: use one!
- Diodes: SiC all the way
- Connections: low impedance
- Coils: lots of copper
- Air gap: 1 – 1.25 mm
- Magnets: stronger is NOT better
- Batteries: bigger & higher voltage
- Duty cycle: 25-35%
- Generator: on the shaft
- Mechanical construction: precision is key
- Closing thoughts
Intro: What are we optimizing for?
In Paul Babcock’s presentation, we hear him say that higher voltage is better, and faster MOSFETs are better, but that raises the question “why”?
To answer that, we have to first determine what it is that we are attempting to optimize.
It seems to me that there are 3 main factors to optimize in a pulse motor generator system:
- Charge recovery from the flyback
- Generator output
- ΔV/Δt, or rate of change of voltage over time
I’ll briefly touch on each of those, as they are crucially important to understand before we can look into ways of maximizing them.
UPDATE 28-01-2023
Recently I received a document written by Julian Perry, who was able to achieve a CoP of 38 (!!) using a solid state coil pulsing system, essentially a pulse motor without a rotor.
He got to this incredible result by tweaking factors like pulse frequency, flyback voltage, duty cycle, and discharge % before swapping.
He found that each battery, especially if they have a different chemistry and/or capacity, has a different optimal pulse frequency and flyback voltage, and so the chance that your pulse motor happens to run at the exact ideal frequency of your particular battery is close to zero, and so your results will most likely disappoint.
UNLESS you reverse the process, and start by finding the optimal frequency, flyback voltage etc. for your battery, and only then design a motor that can run at that frequency.
This was an eye-opener for me, and I believe this is the reason why hardly anyone building pulse motors ever sees any CoP > 1 results, so I urge you to read his work in detail to fully grasp the consequences.
Charge recovery
When we pulse our drive coils, a magnetic field is set up around them.
When the pulse stops, this magnetic field collapses in on itself in a fraction of time, resulting in a high-voltage flyback spike.
We want to try to maximize the energy stored in the magnetic field, and then subsequently recover as much of this collapsing field energy as possible, so we can use it to charge a battery, charge a capacitor bank, or even reroute it to add additional torque.
Generator output
Both Paul Babcock and Peter Lindemann mentioned that the excess energy in the system mainly exists as mechanical energy, and so to extract it, a generator will have to be hooked up to the system.
However, pulse motors are not the strongest, so we’ll have to find ways to maximize the torque of the system in order to be able to run a generator off of it.
ΔV/Δt
Now this last one might seem a little out of place and requires a bit more explanation.
What does rate of change of voltage over time, or ΔV/Δt, have to do with any of this? Everything it seems…
This is what Robert Adams said:
There are a number of various methods of harnessing considerable aetheric energy from these machines. In this respect I strongly urge you to study Tesla radiant energy.
Robert Adams, ADAMS SPECIAL RELEASE OF INFORMATION FOR THOSE EXPERIMENTING WITH MY MOTOR GENERATOR TECHNOLOGIES (2001)
This leads us to a commonly shared, but hard to verify story of how Tesla supposedly started thinking about “radiant energy”.
A curious anomaly occurred in the very first instant of throwing the power switch at the DC generating station: Purple/blue colored spikes radiated in all directions along the axis of the power lines for just a moment. In addition, a stinging, ray-like shocking sensation was felt by those who stood near the transmission lines.
Tesla realized almost immediately that electrons were not responsible for such a phenomena because the blue spike phenomena ceased as soon as the current started flowing in the lines. Something else was happening just before the electrons had a chance to move along the wire.
The phenomena only exhibited itself in the first moment of switch closure, before the electrons could begin moving. There seemed to be a “bunching” or “choking” effect at play, but only briefly. Once the electrons began their movement within the wire, all would return to normal. What was this strange energy that was trying to liberate itself so forcefully at the moment of switch closure?.
https://sites.google.com/site/johnbediniresearcharchive/radiant-electricity
So something odd happens the instant a switch closes and the high-voltage DC of the generating station rushes into the transmission line.
This sounds a lot like the “impedance phenomena” Tesla experimented with in his hairpin circuit.
Robert Adams also refers to Wattless, currentless, electron-less power:
Due to the subliminal velocity of the aetheric energy gaseous flow over the surface of the windings, the current becomes “lost” and, therefore, left behind.
Robert Adams. “On The Phenomena Of Wattless (Currentless) Power.” Applied Modern 20th Century Aether Science (1999)
The higher the voltage and the shorter the time period it is delivered in (ΔV/Δt), the higher the impedance, which then seems to result in the curious situation where energy flows over the surface of the conductors at incredible speeds, but no electrons have actually started moving yet, as the pulse hits a brick wall of impedance, like slapping your flat hand on the water at full force.
I’ll share one last comment from Robert Adams on the subject:
There are at least two possible methods of causing a finite delay time thereby retarding current flow within the stator element, in addition to impulsing the source – one, of using doped winding wire (at present a difficult one) and – two, in designing the stators by incorporating sacred geometry, i.e., PI and PHI involving the Golden Ratio.
Robert Adams. “Translocating Potential Gradient To The Motor.” Applied Modern 20th Century Aether Science (1999)
If we somehow manage to achieve this delayed current flow, this means we’ll use way less input power.
Thus it appears that a high ΔV/Δt is essential to unlocking some of these unique pulse motor phenomena, and knowing this, it all of a sudden starts making sense when Paul Babcock suggests to increase your input voltage from 12 to at least 48V, and to get the fastest MOSFETs possible…
Switching: SiC or GaN FETs
I forgot where I read it, but apparently Jean-Louis Naudin once was at a conference in Australia with his Adams Motor prototype, and Robert Adams himself and his wife were there as well.
When Naudin asked Adams about what he should do to improve his motor, Adams had one quick look and said: “use a physical commutator”.
This sentiment is also echoed by Robert Adams in some tips he gave in 1993:
Should you experience any difficulty in designing and constructing the tapered disc contactor (machining, etc.) then use electronic switching, i.e. photo, Hall effect, or inductor effect, with switching current transistor, etc.
…the input current to supply the electronic switching will raise the total input quite steeply. The point to be made here is that in using electronic switching, in a larger machine, the degree of loss due to this use of electronic switching is negligible.
However, for those who are seeking greater efficiency figures, it is advised to stay with the tapered disc contactor method.
Robert Adams
So it seems he said to use a physical commutator for small devices, so you don’t incur extra losses due to the power required to switch a transistor.
The original Adams Motor Manual does include some transistor circuits, which use a TIP32B or 2N3055.

The famous pulse motor circuit diagram from John Bedini’s original 1984 book also contains several 2N3055 transistors.

We know that in his later years, John Bedini preferred to use MJL21194 transistors, often found in audio equipment.
So why can’t we use those transistors? It worked for them, right?
That was my attitude for a long time, until I realized what it was that we are trying to optimize (ΔV/Δt), and then realized that modern day silicon carbide (SiC) and gallium nitride (GaN) FETs are far superior to anything Adams, Bedini, or Newman had access to in their days.
Remember, they started working on this stuff in the 70’s and 80’s; that’s over 40 years ago!
Like Paul Babcock said, we’re living in a golden age of electronic switching right now, and we should make the most out of it!
These new types of FETs are breathtakingly fast, can handle incredible voltages and currents, and are very energy efficient.
They are however not cheap, at about $20 to $30 a piece, but will be well worth the investment.
From my research so far, it seems GaN FETs are a little easier to control, as most SiC FETs require a gate voltage of 20V and a low of -5V, making it harder to find a proper gate driver and complicating the circuit.
Most GaN FETs I found can be driven from 10V and don’t require negative gate bias.
I’m currently looking into using a TP65H035G4WSQA 650V GaN FET for my next motor project, so lets compare the numbers to a power MOSFET I commonly see in pulse motor circuits, the IRF840LC:
IRF840LC | TP65H035G4WSQA | ||
Voltage | 500 | 650 | V |
Peak amps | 28 | 240 | A |
On-resistance | 850 | 32 | mΩ |
Gate Charge | 39 | 22 | nC |
Rise time | 40* | 10 | ns |
Fall time | 30* | 10 | ns |
Diode reverse recovery time | 490 | 59 | ns |
The GaN FET is the clear and obvious winner on all fronts!
If you want to go even faster, surface mount GaN FETs like the GS66502B take the cake, but I have no idea how to get those soldered onto a circuit board :-p
Gate driver: use one!
Now it’s one thing to have a FET capable of fast switching, but you also need a way to turn it on that fast, and for that you need to dump as much current as possible into the gate capacitance in the shortest amount of time, and the very best way to do this is to use a fast gate driver IC with a high current rating.
I’m currently looking at the UCC27524P by Texas Instruments, which is a 5 Amp high-speed, low-side gate driver with incredible 7 ns rise and 6 ns fall times!
A gate driver essentially boosts a small logic signal from for example a Hall sensor or optical sensor to the higher voltage and current required to fully turn on the FET as fast as possible, thereby maximizing switching speed and minimizing heat generation and thus losses.
It also offers protection to the rest of the circuit, as a switching MOSFET can cause a back-current from the gate back to the driving circuit, which these drivers are designed to handle.
These drivers are also pretty cheap, but they can be a little tricky to implement due to their many pins, and the fact that they sometimes come as surface mount parts, although you can buy through-hole adapters for some of these packages.
However, if you spent $20 on a GaN FET, you should really implement a good gate driver as well and stop switching the gate with a basic transistor.
Diodes: SiC all the way
One of the simplest, and cheapest things you can do to significantly increase the amount of energy captured from the flyback is by upgrading your flyback diode.
Many Bedini SG circuit diagrams online, even some in the Complete Advanced Handbook written by Peter Lindemann, contain basic 1N4007 diodes.

Now these diodes are commonly used for 50/60Hz mains rectification, not for high frequency operation, and while their 1000V rating is generally sufficient to handle the flyback spikes, their 1500 ns reverse recovery time is abysmal!
In comparison, modern silicon carbide diodes like the C4D02120A are so incredibly fast that their reverse recovery time is usually not even stated, or else stated as 0 ns, resulting in, quoting the datasheet, “essentially no switching losses”.
A flyback spike contains a certain amount of energy, and this energy will be dissipated over a certain amount of time.
So the slower your diode, the more of the spike’s energy will have already been dissipated as heat before your diode finally turns on, and so less energy will be left over to recover to charge your capacitor bank or charge battery.
This is why an ultra-fast flyback diode is essential, and since it’s a drop-in replacement that costs just $2.50, you have no excuse not to implement it right away!
Connections: low impedance
Now this inductive flyback spike is high-voltage (~300-900V), exists only very briefly, and so has an incredibly high ΔV/Δt.
This means it will be very difficult for this energy to pass through a conductor due to the skin-effect, “which causes alternating currents to flow closer to the surface, rather than through the centre of a conductor, effectively increasing the resistance”.
To counteract this, make sure that all your connections that will receive higher-voltage pulses are short and have as much surface area as possible.
So way thicker wire than the expected amps would require, multiple insulated strands of thin wire (Litz wire), or flat copper strips.
This means the connections between your batteries, between your batteries and the drive coils, between your coils and your MOSFET to ground, and especially the connections to and from your flyback diode!
Coils: lots of copper
No one gets excited about the prospect of re-winding their coils. as it’s a mind-numbing, time-consuming task, but please hear me out 🙂
First, the goal of the drive coils is to act as an electromagnet, either pushing or pulling the rotor magnets, or simply demagnetizing the core, as in the Adams design.
So how can we maximize the coil’s magnetic field strength?
The Basic Electricity Naval handbook tells us there are 4 main factors:
- The amount of current flow through the coil
- The number of turns of conductor
- The ratio of the coil’s length to its width
- The type of material in the core
We want to expend as little energy as possible, so increasing current flow like in #1 is not ideal, but lets explore the other options.
Number of turns & copper content: size matters!
Joseph Newman, brilliant inventor of the Newman motor turned crazy person, writes in his book:
For a given energy input into varying materials of identical volume, the generated strength of the magnetic field varies drastically! …I therefore instantly knew that the strength of the magnetic field had to be a result of the nature of the atoms comprising the material and not a result of the electrical energy input.
Joseph Newman
This indeed makes sense, as we know that for the same amount of current input, the magnetic field strength of a coil increases with the number of turns, which is why Magneto Motive Force (MMF) is measured in Ampere-Turns.
Newman says that “the current input simply acts as as a catalyst” for the coil’s copper atoms to align and set up a magnetic field.
He then describes two coils with identical resistance and identical internal diameter, but one uses 40-gauge (0.08 mm) wire and the other 5-gauge (4.6 mm) wire, the first weighing a mere 0.02993 lbs (133 g) and the latter one a whopping 335,469 lbs (16.77 tons)!!

When 100V is put on these coils, they both draw about 95 mA of current, or 9.5 W of power, but the second coil’s magnetic field is 1905 times larger, and the energy stored in the field is 8 million times greater!
Holy crap!
Reading this was eye-opening for me, and so to get more power out in the flyback for the same amount of power in, simply use a lot more turns of thicker wire!
Robert Adams seems to have reached a similar conclusion, as he always mentioned how important a high-resistance coil was, which generally means more turns, and so more copper.
The cardinal mistake being made here is that most of these experimenters are concerned about I²R losses! If you are seeking high/super performance with these powerful magnets [neodymium], then discard all concerns in relation to Ohms Law, for in the Adams technologies Ohms Law becomes a non-entity.
Instead of expecting results of a high order with stators of very low resistance, such as under 10ohms, increase the total series electrical resistance instead to 72ohms.
Robert Adams
By knowing the resistance values Adams targeted and by analyzing the coil dimensions on pictures taken of his original machines, it seems like Adams potentially used 0.35 mm wire on his smaller models, and up to 0.5 mm wire for his multi-kilowatt models.
Much thicker wire was obviously used by Newman on his machine, but you’re also going to want to keep the coils in proportion to the rotor magnets, which is why a larger motor with larger magnets and larger coils will most likely lead to much better results.
If Newman’s theory is true, you could literally just pulse a large coil of wire, and get a ton of flyback in return, without having to add the rotor, which is exactly what Tim Harwood from the Adams CD Motor Project tried to do with his Air Core Generator, in which he tried to replicate the effects of the Adams Motor in the simplest way possible.
This is what he told me in an email:
If you pulse an air cored coil at 2.2khz at 2% duty you get the same physics.
Slam that in @ 360v , in 120v increments, and it starts to get interesting!
Tim Harwood
So maximize copper content of your motor coils, which is in line with Paul Babcock suggesting to forget about the single coil Bedini motors, and instead have as many coils as you have magnets on your rotor.
Coil dimensions
To be honest, I was surprised to see the Basic Electricity handbook mention length to width ratio as a way to tweak the magnetic field strength.
However, when I looked into it, I soon stumbled upon the Brooks Coil, which says that the dimensions that maximize the inductance and minimizes losses of an air core coil are as follows:
A Brooks coil is an inductor wound in such a manner as to produce maximum inductance for a given length of wire. Specifically, it is a coil wound on a circular spool of diameter 2c. The length of the winding is c. The thickness of the winding is also c! Note the winding has a square cross section. The total OD of the coil is 4c. “c” is the Brooks constant.

Now this coil shape might or might not work for your setup, as there are other factors to take into account besides maximizing induction.
For example, to get the most torque out of your machine, you’ll want to approximately match the size of your coil face to the size of your rotor magnets, or else a lot of magnetic field lines miss the magnets and therefore won’t assist in making the rotor spin.
So that decides your coil diameter, but how long should your coil be then?
Well, there’s something called the paperclip test, where you place a paperclip on a flat surface with a ruler next to it.
You then slowly move a rotor magnet closer to the paperclip, and note the distance at which the paperclip flies towards the magnet.
That distance should be the length of your coil, according to this test.
But then again, if you want maximum flyback to recharge your batteries, you might have to wind a coil much larger than that…
Finally, there is the fact that when you put a current into your coil, one side of the coil becomes the South pole, and the other side the North pole.
However, in nearly all pulse motors I’ve ever seen, the back of the coil, so one of the poles that you “paid for” with current, doesn’t get used at all!

There seem to be 2 main ways to utilize both sides of your coil:
- Wind the coil on a U or C shaped core, like in the image above
- Sandwich the coil between two rotor sections, each with magnets of a different polarity
This should give you a significant torque boost for the same amount of input power, but you’ll obviously also need to have a rotor with both North and South facing magnets, like the advanced rotor from Robert Adams pictured below, on which I’m basing my next design.

As you can see, the ideal coil dimensions are unfortunately not clear-cut, but hopefully this section gave you some things to consider.
Core material
Bedini and Adams used soft iron cores in their drive coils, and DIY researchers are using all types of core materials.
Using a soft iron core has several perks:
- It greatly increases the magnetic field strength
- It keeps the magnetic field more focused
- The rotor magnets will be attracted to it
However, in his presentation, Paul Babcock recommended using air core coils instead, and Joseph Newman’s motor also used an air core coil.
So who’s right?
I’m not sure yet, but cores tend to be expensive and hard to source, so if Paul Babcock and Joseph Newman had success with air core coils, I think that is the cheapest and easiest way to start.
You can then always progress to trying various types of cores, because as soon as you use a core, you’re going to have to deal with eddy currents, hysteresis, saturation, and other complex variables, making it much harder to get it right.
And if you do use an iron core like Adams, then think of following his instructions to have the core diameter be no larger than 75% of the rotor magnet diameter, so for a 20 mm magnet use a 15 mm core.
Other: bifilar windings
There have been countless people emailing me or commenting on my videos saying I should use bifilar windings on my drive coils.
John Bedini even went as far as having, I believe, 16 separate windings on a single coil!
Peter Lindemann once told me that this was because Bedini noticed each wire had its own “radiant element”, so he could capture one spike per winding, which greatly complicated the circuit, but apparently it was worth it.
I haven’t tried this out myself, but the Newman and Adams motors both used simple single layer windings and supposedly were able to achieve similar results, so to me this is a very low priority option you should only consider after trying all the other things in this article.
Air gap: 1 – 1.25 mm
In 1955, E.B. Moullin performed an experiment where he used a variable AC power supply to energize an electromagnet, and then measured the input current, as well as the induced voltage in a search coil placed on the bridging member.

This induced voltage was a proxy for the amount of magnetic flux across the air gap, as magnetic flux determines the amount of mechanical power available.
He gradually increased the air gap, then adjusted the power supply so that the voltage measured in the search coil remained constant, and then noted the amount of input current.
What he found was a big surprise!

The straight line is the amount of current that is expected by today’s theory to be needed to maintain the same mechanical power across an air gap in a magnetic circuit.
However, we see that in reality the amount of current slopes DOWN and is in fact a lot LOWER than theory predicts.
So where is this additional energy coming from? The original experimenter didn’t know, but Adams believed this is proof that energy is entering the machine from the environment.
And you see that the 1.25mm gap Adams used is just in the positive region where excess energy starts to show up.

In the above drawing from 1993, you see two magnets, one of which is the rotor magnet, and the other one is the drive coil, which acts like an electromagnet.
In the accompanying text he explains that the energy from the surrounding space is gated by the magnets into the air gap.
Now whether this is true or not, Robert Adams built countless prototypes over several decades, and found the 1 – 1.25 mm gap to give the best results, and so this might be worth implementing in your own setup.
Magnets: stronger is NOT better
The rotor magnets can be of any type, but Adams said that “there will be much disappointment to a lot of people out there to learn that magnet energy product does not govern efficiency in any way whatsoever.”
Stronger magnets require more input power to achieve the same amount of rotation, and so efficiency remains the same, although torque increases for the same size machine.
So since we already learned that size seems to matter, and we don’t want to be handling large neodymium magnets as you might literally lose a hand, I highly recommend to stick to ferrite / ceramic magnets.
If you do decide to use neodymium magnets, heed the following warnings:
The cardinal mistake being made here is that most of these experimenters are concerned about I²R losses! If you are seeking high/super performance with these powerful magnets [neodymium], then discard all concerns in relation to Ohms Law, for in the Adams technologies Ohms Law becomes a non-entity.
Instead of expecting results of a high order with stators of very low resistance, such as under 10ohms, increase the total series electrical resistance instead to 72 ohms and instead of expecting spectacular results using these powerful magnets with only 12 – 24 volts, increase the voltage to a minimum of 120v.
Upon having done this you must give attention to other important factors, i.e., stator to magnet air gap should be 1 – 1.25mm… Reduce the face area of stators to 75% of the magnet face.
Robert Adams, ADAMS SPECIAL RELEASE OF INFORMATION FOR THOSE EXPERIMENTING WITH MY MOTOR GENERATOR TECHNOLOGIES (2001)
Many people decide to use ultra-powerful neodymium magnets, but then still try to power it with a wimpy 12V, which means they have to dramatically increase their stator to magnet gap, which then makes the motor way less efficient, less powerful, and they then miss out on the “magic” that Adams believes happens if you use a small 1 – 1.25mm air gap.
So unless you’re ready for 120V+ drive circuits (at least if you’re using iron cores), it seems better to stay away from neodymium magnets.
Batteries: bigger & higher voltage
One of the main things I took away from Paul’s presentation, is that we will never achieve anything remarkable using a small 12V battery.
We need to both significantly increase battery voltage, which will help our ΔV/Δt, and also increase battery size, as larger batteries have lower impedance, and are therefore easier to charge.
Paul suggests a minimum of 48V, as that is also still a relatively safe operating voltage, but the higher the voltage the better (just be very careful).
Adams also went up to 120 and even 240V, and Tim Harwood also told me that the really big effects start to happen between 120 and 360V.
Rick Friedrich explains that larger physical battery size will be able to capture more energy.
Now this all requires a significant investment in batteries, and the higher voltage most likely means you’d have to redesign your entire circuit as well, so it’s not a small step to take, but apparently needed if you want to see any serious results.
For my next motor, I’m looking at using 8x 12V 35aH flooded lead-acid batteries, 4 for running, 4 for charging, for a total of 48V, which will set me back about $500… ouch…
Note: Julian Perry told me he didn’t find a relation between battery voltage and CoP, and that using a higher voltage battery on the same coil would even result in lower CoP due to more current consumption, so higher voltage should probably go hand in hand with larger, higher resistance coils, which is what Adams also did.
Chemistry
In an old Bedini interview we hear him mention that LiFePO4 batteries are the future, and it’s true that these batteries have some interesting characteristics, like a smaller form factor and a very low impedance.
However, they are also about 5x the price of a lead-acid battery with the same capacity!!
Paul Babcock and Mike Clarke successfully use lead-acid, and Julian achieved CoP > 1 with both lead-acid and LiFePO4 batteries, but achieved his highest CoP values with 18Ah LiFePO4’s.
So I think lead-acid would be the economical place to start, and once you see success with those and find a jar of money somewhere, then you can consider giving LiFePO4 a try.
Swapping
In my original Adams Motor design, I had only one battery which was running the motor, and which I also tried to recharge from the flyback at the same time, as I tried to keep the circuit as simple as possible.
Now the thing is that a battery can only really do one of those at a time, and if you do what I did, the ions will just be sloshing around, and the battery won’t be charged efficiently.
Note: battery swapping is essential. I can’t stress that enough.
Paul Babcock
As you can see in the quote above, it seems like you’re going to need a second battery bank which will be charging while the first bank is running the motor.
Paul then has a circuit that will swap the run and charge battery around in the circuit every 3 minutes or so, but unfortunately he did not want to disclose his circuit during the presentation.
There is also a battery switching board for sale by Energy Bat Labs for a whopping $600, which actually uses 3 battery banks, where 1 bank is “resting, and switches based on voltage levels instead of time.

If you search really well, you might come across some circuit diagrams like the one above, which I haven’t tested myself.
But if you want to simply switch batteries after a certain period of time, like in Paul Babcock’s situation, the following video has a potential answer for you!
It uses a 12V 4-Pole Double-Throw latching relay, the Panasonic SP4-DC12V, driven by an Arduino for the timing and a tiny circuit to determine whether to SET or to RESET the relay.
Another key thing to consider when using relays is to make sure you turn the motor OFF while the relay switches, or else current will still be flowing, thus creating arcs inside the relay, which will eventually cause the contacts to burn out.
I turn the motor off by utilizing the ENABLE pin on my MOSFET gate driver IC.
In my latest video I show how I implemented battery swapping:
You can find the schematics, wiring diagram, and Arduino code for the battery swapper in the description of the video.
Caps
A small addition to the battery story is that you might want to add 500-1000uF of capacitance right after the run batteries in your circuit, as they will be able to provide energy to your pulses much faster than the batteries alone could muster.
It also seems that once you have a system that charges faster than it runs down, there might be a point where batteries could be replaced by capacitors altogether, but I don’t know of anyone who has reached that point yet.
Duty cycle: 25-35%
In a pulse motor you obviously pulse the coils, but for how long?
Adams recommends a sweet spot for the duty cycle of 25-35%, with the lower range being more energy efficient, while the higher end gives you some more torque.
Going higher than 35% will waste unnecessary power.

One easy way to achieve this is by using a photo-interruptor and a timing disk attached to the shaft.
Generator: on the shaft
In Adams’ manual, he shows a motor generator system where there are 2 drive coils and 4 generator coils placed around the periphery, which I replicated, and which wasn’t very effective at all at generating power.
However, in the British Adams / Aspden patent they write the following:
If the object is to generate electricity by operating in generator mode then one could design a machine having additional windings on the stator for delivering electrical power output. However, it seems preferable to regard the machine as a motor and maximize its efficiency in that capacity whilst using a mechanical coupling to an alternator of conventional design for the electrical power generation function.
R. Adams & H. Aspden, UK Patent GB 2,282,708, Electrical Motor Generator (1995)
Paul Babcock recommended the same thing: use the pulse motor as a motor, and attach a conventional generator to the shaft for electrical power output instead of fabricating your own, probably very inefficient, generator coils.
Mechanical construction: precision is key
When it comes to pulse motors, everyone seems hung up on the electronics, but then I see some motors made out of wood sawed and drilled by hand, a bend curtain rod as a shaft, and cheap bearings full of grease…
Such a build just isn’t going to work.
I understand that not everyone has the tools, skills, and funds to build a high-quality machine, but the reality is that the mechanical aspects of these motors need to be spot on, and made of high quality, precision made parts, or else the losses will be tremendous and you’ll never achieve the results you’re looking for.
Bearings
When it comes to bearings, we want as little friction as possible (so no thick grease inside), and we want to make sure they are non-magnetic.
The bearing type of choice for me has been full-ceramic bearings, which are very low friction, non-magnetic, waterproof, need no grease or other lubricants, and promise a much longer lifespan compared to conventional ball-bearings.

They are usually more expensive, unless you get then off of AliExpress.
I can highly recommend the brand Fushi, which sells through AliExpress, and who make precision ceramic bearings in many sizes for competitive prices.
Shaft
Your shaft also needs to be very straight and also non-magnetic, so either 316 stainless steel or aluminium.
I tried ordering stainless steel shafts in the past, but they ended up being a different type of steel and so were actually magnetic.
Mike Clarke had the same thing happen to him, and he only found out after he epoxied his expensive rotor to the shaft!!
That’s why I’m mainly using hardened aluminium precision shafts, and when you go up to 20mm and up, a hollow shaft might be a great option, as it is lighter and stronger than a solid shaft.
Balanced rotor
During all my research into pulse motors, I’ve hardly ever heard anyone discuss the importance of balancing the rotor, which is a crucial step in the construction of any type of rotating machinery.
It was observed that average power loss due to unbalance is of 0.11watt/gm.mm unbalance. This can amount to 2.75 kw if there is unbalance of 50 gm at a radius of 500 mm.
M. Bulsara, Energy loss due to unbalance in rotor–shaft system (2016)
The above quote from a 2016 research paper shows that just 50 grams of unbalance on a 1 meter diameter rotor can result in 2.75 kilowatt (!!) of power loss due to vibration, which is a huge amount!
And the faster a rotor spins, the more important a good balance is, as the centrifugal forces grow tremendously with speed, excarcerbating the vibrations, and pulse motors tend to go FAST!
At certain speeds, your system will hit a resonance point, where the vibrations can become truly destructive.
Unfortunately, proper balancing is not exactly super easy, as you’ll have to deal with both static imbalance and dynamic imbalance, which are very clearly explained in the video below.
Dynamic imbalance becomes more important the wider your rotor is, so luckily for most pulse motors out there with relatively flat rotor disks you can focus primarily on balancing your rotor statically, which is fairly easy to do.
In the video below you see how someone statically balances a motorcycle wheel, mainly to give you an idea about what a simple balancing setup looks like.
Below is a brief description of the static balancing process.
Static balancing was the primary method for balancing before more sophisticated machines became available that could dynamically measure imbalance in rotating parts. Static balancing can be accomplished by placing a part on knife edges [or some low-friction bearings]; the part will usually rotate until the heaviest side rests at the bottom. Material can either be removed from the heavy side, or added to the opposite side to try and correct the imbalance. This process is repeated until the part can rest at any angle without turning.
https://xyobalancer.com/static-balancing/
Static balancing is more suited for non-critical machines where a rough balancing process is adequate e.g. car wheels are usually statically balanced and weights are added to the wheel rim to compensate for imbalances. High speed rotating equipment will typically have to be dynamically balanced, or else excessive vibration could lead to failure and/or damage.
As you can read above, a dynamically balanced rotor is ideal, but even just a static balance will already reduce your power loss due to vibration significantly, and will put you way ahead of 99% of the other pulse motor builders.
Since I’m planning on building a fairly wide and powerful rotor, I’m considering to let a professional balancing company balance it for me, for which I was quoted €300… ouch again.
Rotor weight
In Mike Clarke’s presentation he describes how he casted his rotor from epoxy, and that it’s so heavy you need two adult men to lift it.
This is not an ideal situation in my eyes, as more mass means more energy required to make it spin, so any non-functional weight should be kept to a minimum, like the inside of the rotor.
Robert Adams used a lightweight HDPE foam filler in his rotors to keep them as light as possible.

I plan to 3D print my rotor, and use thicker walls, but a mere 20% infill to keep the structural weight down as much as possible.
Closing thoughts
A lot of pulse motor design considerations were discussed in this article, some easier and cheaper than others, but all important in their own way.
At the very least, I hope this article gave you some food for thought on how to take your pulse motor to the next level.
And if I said anything stupid or if I missed some important points, please let me know in the comments below so I can amend the article.
Finally, be sure to check out the Paul Babcock and Mike Clarke ESTC presentations!