Projects

# Making High Voltage Capacitors

Daniel Koch

Issue 61, August 2022

We follow on from the Wimshurst machine by trying to make our own high-voltage capacitors. Whether it's success or failure, we're sharing it all.

One of the biggest challenges in making a capacitor for our high-voltage projects was not actually knowing what voltages we were, and still are, working with. Because these devices are static generators and not a step-up device, and there is such a vast array of variables that we cannot measure, we cannot calculate the voltage. If we really put our minds to it, we could probably calculate the theoretical maximum value. However, the time and effort required for research, re-learning maths that hasn't been used for years, and then doing the calculations, isn't really justified when we can take a far more 'Maker' approach and just use experimentation, and trial and error.

Experimenting with high voltages might sound like a crazy idea, and it certainly can be. These are effectively large static electric shocks but they are still bigger than any common naturally occurring charges. Safety will be paramount and that includes grounding and isolation. We are working with Direct Current (DC), which sets this apart from many other high-voltage projects like Tesla coils that are encountered online.

# THE MYTH OF THE SKIN EFFECT

The 'skin effect' is a phenomenon where an alternating current (AC) travels close to the outside of a conductor. How much so depends on the frequency. At 50Hz as for a mains supply current, the skin depth of pure copper is around 8.6mm. The skin depth is where the current density falls to the inverse of Euler's number (a common mathematical constant), and in most materials, 98% of the current flows in four units of the skin depth. In a solid copper rod 100mm in diameter, the current will only flow along the outer 34.4mm or so at 50Hz. The same applies for bundles of strands if they are tightly bound and not insulated from each other, though the slight inclusion of free space means it will not be 34.4mm anymore. The higher the frequency, the smaller the skin depth. This is why Litz wire exists for high-frequency conductors - a twisted or woven structure of many small, fine wires electrically insulated from each other. At radio frequencies, this is a more efficient carrier of current than a solid wire or bundle of strands. See our website for the maths for those who are interested.

When watching videos about high-voltage projects, or reading blog posts and the like, it is common to encounter the skin effect. Internet 'experts' regularly rely on it for safety, with the assertion that high-frequency AC electricity travels on the surface of the body. This is incorrect and is not based in any science, but rather is an assumption about or misunderstanding of the meaning of the term 'skin effect'. In a pure conductor, this would be true. However, the human body is far from a pure conductor. Our skin is one of the highest-resistance parts of our bodies, and electricity will always take the path of least resistance. Anatomically, the current will likely flow along muscles and blood vessels, meaning the skin effect will definitely not divert electricity harmlessly away from the heart.

While it is demonstrable that high-frequency AC arcs, for example from some Tesla coils, can contact the human body without visible effects or damage, the skin effect is not the reason. The low currents involved in many cases in these high-voltage step-up arrangements allow the current to dissipate around the relatively high-resistance body and not be strong enough in any one place to affect nerves or muscles. Large Tesla coils are often seen used with a Faraday suit because it is said that at higher currents, burns will occur even if electricity does not cause nerve and muscle issues. While these voltages will indeed cause burns, the scientific literature confirms that the high-frequency current on its own would cause the traditional muscle and nerve damage because these organs become the conductors, providing the current is high enough. Often, it is not. While the skin effect will still apply, the current travels at the surface of these internal conductors, not the outer skin of the human body as a whole.

One of the best sources we found which makes this clear in a succinct way is from ePlasty, a medical journal aimed at the study of tissue and muscle trauma through injuries of all sorts, and many related facets of reconstructive surgery. We have linked to the article, titled 'Conduction of Electric Current to and Through the Human Body: A Review' in the 'Reading and Resources' section.

In addition to misconceptions of the skin effect as a safety system, we are also faced with the challenge that the Wimshurst machine is a DC static charge generator. Many online explorers don't make the distinction, and treat all high-voltage sources as if they were AC.

Even if the skin effect was a valid way of staying safe around, say, a Tesla coil, it wouldn't help at all with the Wimshurst generator, because the skin effect does not apply to DC currents. Don't believe everything you see on the internet! Just because someone hasn't got hurt or expired themselves yet, doesn't mean they won't with a slight change in variables.

By the way, there are still enough online sources which do get this right, but many of the really popular sources which appeal to experimenters and those with less interest in engineering-style videos, present the problematic views we have described here.

# OUR SAFETY PROTOCOLS

While we have seen some people touch one of the SparKit Wimshurst machines, every body is different. The shock still hurts, even if it does not do physiological damage. Additionally, the body's response may not always be so controlled, and rapid movements of the muscles can lead to strain injuries on its own.

Any undiagnosed heart conditions or potentially even epilepsy have been linked to electric shocks (though the epilepsy link is disputed). Regardless, we feel it's not worth the risk. Combined with the fact that we don't really know and cannot measure the voltage, we have decided to take some precautions. Remember, the Wimshurst machine produces a high voltage at a very low current, but we're building capacitors which can store a bigger charge and therefore dump what may be significant currents.

We have chosen to use double wrist-straps, connected to the same grounding point. In addition, we have thick leather gloves of the sort used for rigging (there are thin leather rigger's gloves, too). These will be good insulators if they stay dry. Finally, the use of a grounding probe made from a plastic rod, a wire, and a 1MΩ resistor to dissipate any suspected stored charge in either our capacitors or the Wimshurst machine.

Adding capacitors to a spark generator increases the intensity and brilliance of the spark, but not the length. The capacitor charges with the static voltage generated by the charge collectors of the Wimshurst machine. Normally, the only place this charge can be stored and built up is the metal balls at the ends of the spark gap rods.

The charge collects on the surface of the conductor, so having spheres rather than just the rods increases the charge stored by both surface area, and by eliminating sharp points or hard edges, both of which encourage leakage to the atmosphere by corona discharge. When the charge is strong enough to ionise the air and jump the gap, this charge neutralises as the spark dumps the negative charge of electrons over to the positive deficiency of electrons.

Increasing the size of the spheres increases the charge that can be held, but not the voltage. Adding a capacitor increases the storage capacity much more, and this is why many Wimshurst machines have some form of capacitor attached. The voltage stays the same, being dictated by the electrostatics of the machine, but the amount of charge stored and therefore dumped at once into a spark is increased by the use of a capacitor.

# CAPACITOR THEORY

A capacitor is made from two electrically-conductive plates, separated by an insulator called a 'dielectric'. The simplest capacitor is two plates separated by an air gap. If a potential difference (voltage) exists between the plates, then one plate becomes negatively charged while the other becomes positively charged. The charge stored is a function of the surface area of the plates, and the distance separating them. The closer they are together, the more the electric field can influence the other plate.

However, air is not the best insulator and the higher the voltage, the further away the plates have to be to avoid losing charge across the air gap. It is also hard to make capacitors commercially this way without making them physically too big to be practical in manufacturing and assembly of electronic goods, not to mention the desire for compact goods for the modern market. So, commercial capacitors use layers of a dielectric material like plastic, between layers of plate material to increase the surface area. Alternate layers of plate are connected so that electrically, there are still two opposing plates.

All dielectric materials have a breakdown voltage, at which they no longer insulate effectively. For this reason, all commercially-made capacitors have voltage ratings. Dielectric strength, as it is known, is not the only factor in why capacitors all have voltage ratings, but it is part of the story. Most MKT capacitors on retail sale, for example, have a 100V rating. That means that they cannot even be used on mains, and 250VAC-rated ones have to be sought for that. We certainly just use a few electrolytic caps or greencaps on our Wimshurst machine: They will fail and may explode. Hmm, that's a thought.

In the interests of thoroughness, we decided to find out. We randomly selected a 56nF greencap and wrapped the legs around the terminals of the Wimshurst machine. The legs of the greencap where they enter the body are close enough to be a spark gap so if the capacitor withstands the voltage, we should hear sparks with a bit more energy as the capacitor adds to the spark.

We pushed the button on the machine and then ran away. There was no bang, no flying fragments of capacitor casing, and no sparks either. That implied internal failure as expected (though an explosion of the greencap would have been nice). We turned off the machine, used the grounding rod on both sides of the capacitor, then disconnected it.

Double-checking the discharge with the grounding and shorting tools, we hooked it up to the LCR meter. So much damage had occurred that the dedicated instrument didn't even recognise the component as a capacitor. Instead, the display showed 'Low Resistance and Inductance'. Scrolling down, we found values of 0.4Ω for resistance and 0.1µH for inductance. For comparison, a healthy 56nF greencap was recognised as such with the display showing 'Capacitor, 58.85nF, and scrolled values were not present because the device was satisfied that it had identified a capacitor.

# DIELECTRIC MATERIALS

The easiest dielectric materials to get our hands on are plastics, but research revealed that some papers work well, too. Waxed paper in particular is said to be effective, but the only source of waxed paper we could find readily available is kitchen baking paper and is quite thin. The idea was to try rolling aluminium foil between layers of different plastic and paper dielectric materials, to see which performed best. We would measure the capacitance with the LCR meter, but the only way to test if the device had enough dielectric strength to withstand the voltage of the Wimshurst machine would be to try it. At least with these unsealed constructions, there would be no explosion, only leakage across the dielectric rendering the component ineffective.

# DIELECTRIC EXPERIMENTS

We stated earlier that there is no way of knowing what voltage is being generated by the Wimshurst of any other static generator. However, there is an approximate guide: Many sources state that for a spark to jump 1cm through air, 10,000V is required. This gives us a rough guide as to the voltage generated, by moving the spark gap until a spark no longer occurs. However, it is a rough guide only. Humidity drastically affects the ionisation of air, and indoor air does not, if airflow is not good, have the same composition as it should. In particularly enclosed spaces with limited airflow, lighter gases can diffuse away and heavier gases concentrate. This also has an effect on ionisation, albeit a smaller one.

To determine if our choices of dielectric material had any chance of working, we devised a rudimentary experiment: A sample of each material would be placed in the spark gap of the operating Wimshurst machine, starting at its maximum sustainable spark gap. Then, the spark gap would be closed until the spark was able to penetrate the material. The samples had to be big enough to avoid arcing around the edges as well, when the spark gap was almost closed. The sample would be held with a third hand from the soldering bench, and the spark gap closed slowly with plastic rods. To ensure a consistent test, the batteries were removed and the Wimshurst machine connected to a 3V bench supply. The room was closed and air conditioned until temperature and humidity were consistent. The exact values didn't matter, they just had to be the same.

 Material Thickness (μm) Spark Gap (mm) Builders' Plastic 140 NS Document Wallet 290 NS Document Display Sleeves 50 6.5 Display Sleeves Double Thickness 100 NS Kraft Paper 60 15.5 Copy Paper 160gsm 170 18 Waxed Paper 30 14.3 Contact 110 NS Post Satchel 10 5

There were some surprises in the results. The Kraft paper used was not as thick as we would have liked. What was available from the local office supply store was fairly thin, but this material can be found much thicker and often used for wrapping parcels and the like. Sometimes it can be found in very thick rolls as a floor protector for renovations and construction. The Kraft paper broke down much sooner than online sources suggested, and the lack of thickness is suspected to be the reason.

The other major anomaly was the builders' plastic. This was marked on the package as 200µm, yet measured at 140µm. While we have marked it as NS for No Spark, it did in fact pass a spark. However, we noticed that the spark always seemed to go to the same spots on the sample as we moved the plastic around, and in many places there was no spark even with the spark gap rod ends touching the plastic on both sides. On holding this plastic up to the light, we discovered that, despite being new from the packet, there were many tiny holes in it, or at least exceptionally thin sections of plastic like pin holes. The poorly-made, inconsistently-thick-and-thin sheet is full of places where it is thin enough to break down.

# ROLLED CAPACITOR EXPERIMENTS

Given what we had found, we decided to proceed with only the materials that did not break down under the spark gap. This includes the builders' film despite its issues - it was worth a try. What follows is a series of experiments using aluminium foil and various dielectric material candidates.

The basic principle is the same: A sheet of dielectric on the bottom, a sheet of foil, another dielectric and another foil. Then, the assembly is rolled into a cylinder. Wires are taped onto the foils to provide electrical connection, and the foil secured to the dielectric only at the starting end: As the cylinder is wound, the layers slide as they go around an increasing radius, each slightly bigger than the layer below. Some of our early experiments used no former, but this was hard. For structure, a piece of 15mm pressure pipe left over from our rocket launcher project from Issue 55 was cut into sections and used as a mandrel.

# DOUBLE-THICK DOCUMENT DISPLAY SLEEVES

We will show the build using the first dielectric material in more detail, so you can see all the steps involved. For the other materials, we'll show what differences are relevant if any, and the finished product. One of the best tools we found when working with the dielectrics and in particular, with aluminium foil, is a rotary cutter. This tool looks a bit like a pizza cutting wheel, but is invaluable in cutting thin, fragile materials. In many cases, even the freshest utility knife blade tears or rips the material. We even had trouble with the utility knife tearing aluminium foil while it was held down with a metal ruler.

There must be enough distance between the edges of the foil to prevent arcing. Since the maximum spark gap that the SparKit Wimshurst machine generated for us under ideal conditions was 20mm, we decided that a margin of 25mm should be a reliable way to ensure there are no arcs (however, it was a guess at this point). That meant that the foil could not be closer than 25mm to any edge of the dielectric material. The exception was when we wound the capacitors in landscape with some materials - the end was an unbroken, folded edge representing a complete barrier.

Accordingly, we took two A3 document display sleeves from the folder, and cut off the white spine section. Then, aluminium foil was cut to size using the roller cutter, and laid on top of one sleeve. Note that the foil stops at least 40mm short of the end. As discussed above, the layers slide over each other when rolled and we needed to ensure a minimum 25mm isolation. A piece of silicone-insulated wire was bared and taped to one side of the foil.

Then, the second sleeve was placed on top, and more foil on that. Note that the wire on the second layer of foil was taped at the other side. This is critical, as the silicone wire does not have a high enough dielectric strength to avoid arcing or breakdown if the wires are close to each other. Now, the foil can be taped down to the sleeve at the end where the roll will begin, and then the sleeves taped together. Finally, the sleeves are taped to the mandrel, all with standard plastic adhesive tape (sticky tape in anyone else's language).

After rolling up, the layers were secured with more tape around the outside. This needs to go all the way around, as there is a reasonable amount of tension for some of the more rigid dielectrics. Then, the assembly was connected to the LCR meter for testing. This particular one was 36.93nF, while the second one we made was 37.11nF.

# CONTACT

The same process was repeated to make two capacitors using book-covering contact with its backing paper left in place. The differences here were the increased resistance between the layers, necessitating caution while rolling. The other significant difference was that, because the contact comes on a roll, we could decide on a nominal length for these capacitors. We chose to use sections of foil one metre long, and contact cut longer by the necessary margins. They were 87.16nF and 121.1nF.

# BUILDER'S PLASTIC

These gave us the least confidence after seeing the plastic perform in the spark test but, like the contact, the builders' plastic enabled us to determine the length we wanted to make the capacitors. We again chose one metre, and repeated the process described previously. We found that, like the contact, there is a slightly 'sticky' friction between the layers with builders' plastic. Nonetheless, we wrapped two capacitors of 42.74nF and 43.06nF.

# DOCUMENT WALLETS

This material gave us the most hope. These document wallets are a fairly thick, semi-rigid plastic meant for holding several sheets of paper together for neat filing or transport, similar to a manilla folder. These ones are translucent clear/white, but they come in many colours from office supply stores. The process is familiar by now, but the size of these capacitors is determined by the fixed size of the sheet material. They often come in A4 size, but you can find them in A3 like ours if you are careful.

# LEYDEN JAR THEORY

Leyden jars are often the capacitor of choice for Wimshurst machines and some other high-voltage generators, too. They are a form of capacitor but are often discussed as being a separate category of component. They are a precursor to the capacitor, having been discovered independently by two people: German Lutheran cleric and part-time experimenter Ewald Georg von Kleist in October 1745; and full-time Dutch scientist Pieter van Musschenbroek who hailed from the town of Leiden. This town name has been anglicised to Leyden for the name of the apparatus, although there is more to the discovery story and historians continue to debate the exact course. It appears clever marketing obfuscated the truth for a long time.

Regardless of who discovered it, the early Leyden jar consisted of a glass jar filled with water, or a water and alcohol mixture. There was a cork in the top, with a nail passing through and into the liquid. When held up to a suspended bus bar, the other end of which was connected by a chain to a static generator, the jar became charged. Touching the nail gave the experimenter an electric shock.

The significance of the hand was not understood in these early experiments, or indeed until much, much later. The ions in the water (which is never pure, though distilled water comes close) stored the charge inside the jar, with the nail playing a part. The hand formed a capacitive plate which stored the opposite charge, separated by the dielectric of the glass. It did not take too long before the outside of the jars became wrapped in foil, and the inside of the jars soon followed, dispensing with the water. This is the closest form we have to the current construction of capacitor, and was the peak design of Leyden jar.

The idea is that the negative charge is allowed to accumulate on one layer of foil, while the positive charge accumulates on the other. We are going to experiment with several Leyden jar designs, both foil/water and foil/foil designs. We have two different shapes of glass jar, a pair of glass bottles for the water versions, and just for the sake of it, a pair of plastic jars too.

# LEYDEN JAR CONSTRUCTION

The first Leyden jar attempt was the water-filled version. Water itself is not conductive or ionic, but unless water is very, very purely distilled, it contains dissolved solids. While early Elyden jars used plain water, modern designs often use salt water. The increased availability of ions from the dissolved salts means a greater charge can be stored. Therefore, we chose to use table salt in water as our filling.

The 'jars' of choice are two Ikea one litre carafes with silicone lids. These were chosen because they were to hand, being used around the office on staff members' desks as water bottles. They were filled with salt water, and a metal skewer passed through the lid of each and into the fluid inside. Then, aluminium kitchen foil was folded to the correct height, and wrapped around the bottle with a slight curl under the base. This is for connecting the two later via a strip of foil sitting under them.

Following this, two each of two foil/foil Leyden jars were made. These utilise two different shapes of glass jar from Kmart. We have tried to use materials from chain stores for these builds so that anyone seeking to replicate them can get the same materials. While we generally try to support independent suppliers, especially where the exact design does not matter, these items are often available cheap from Kmart and dollar stores, or for a much higher price (for a higher-quality product) from high-quality homewares stores, with little in between.

The two designs of jar were chosen for having straight walls and necks big enough to get a hand in, for applying the inner layer of foil. The choice between the two designs was for one wider one with lower walls, and a narrower one with taller walls. This affects the amount of foil and the relationship of width to height. The process was the same for both sizes of jar: Foil was cut with the roller cutter to fit the height of the inside of the jar with a good margin to avoid leakage between the plates, and laid against the inside of the jar. Tape secures it in place, while a strip of foil connects across the diameter. Then, the outer foil was cut slightly longer to wrap around the base of the jars. This was for connection, like in the bottle version.

Finally, the lids were drilled with a hole for an electrode to pass through. This took the form of an eyelet with length of ball chain or fine chain cable-tied to the thread. The eyelet sits loose. The chain must be conductive. Some chains, though made of metal, are plastic-coated to maintain shine. Others are metalised or metal-effect plastic and may not conduct at all. The chain is long enough to touch the foil strip on the bottom of the jar, and taped there to avoid it moving when the jars are tipped during handling.

Just because we had them, we also replicated the situation above with some plastic containers we had bought at Kmart. Besides the size and configuration, the construction is the same.

# ARRANGEMENTS

In most Wimshurst machines, two Leyden jars are used in series. However, capacitors in series have a capacitance lower than the smallest individual capacitor in the chain. Capacitors have to be in parallel for them to add up. So then, why two capacitors in series when the object is increasing the charge? There are two possible answers. In the first instance, the voltage across a series of capacitors is divided among them proportionally to their capacitance.

So, if two equal capacitors are in series, the voltage across each is half the voltage applied to the whole series. In reality, no two capacitors are identical and our hand-made versions are far from it. However, the voltage at the spark gap is still vaguely halved, meaning each capacitor needs only to cope with half the stress.

However, while we have not found any literature on the subject, we suspect that the choice of having two Leyden jars is simply a matter of mechanics - the Leyden jars are part of the structure of most Wimshurst machines and hold up the spark gap rods, and also often support the charge collector combs. So, while it may not be electrically necessary to have two capacitors in series, we cannot rule it out. We have no information or electrical theory that suggests a need for two capacitors in series, but we also know that there are plenty of things we don't know! So, we will conduct tests with both single capacitors and pairs of series capacitors, and single Leyden jars along with pairs of series-connected Leyden jars.

In the case of series pairs, the rolled capacitors were connected with wires connecting in series. We have no idea which wire was the inner and outer plate in each case, so the connection is arbitrary. The capacitors were simply laid in a line, the outer ends connected to the spark gap rods, and the inner ends twisted together.

For the Leyden jars, the jars were sat on a strip of foil for connection, and the inner electrodes formed both the positive and negative connection, depending on which side they are on. Because we have no way of knowing which side is the negative charge and which is the positive on the Wimhurst, there was never a way to polarise the capacitor anyway.

# ROLLED CAPACITOR TESTING

The rolled capacitors were tested first. The procedure involved testing the series version first, so that the voltage would be half across each. Then, the single one was tested in case the voltage did some damage. The spark gap was started close, and then moved outward until a spark could no longer be achieved. This is the figure listed in the table.

 MATERIAL CAPACITANCE (nF) SPARK GAP, SINGLE (mm) SPARK GAP, SERIES (mm) Document Sleeve 37.11 / 36.93 7 13 Document Wallet 9.569 / 9.836 23 21 Contact 121.1 / 87.16 Motor Overload, couldn't charge Builders' Plastic 43.06 / 42.74 Leakage Too High

Most of the capacitors exhibited audible crackling faintly as they charged. This could indicate expansion of materials, or it could be leakage. It is hard to tell and we are not getting any closer to investigate. The numbers in the table give an idea of how permeable the dielectric is for each type: The bigger the spark gap, the more resistant the dielectric is. No prizes for guessing which one was going to be the winner! However, while some materials are close in apparent performance, time revealed that some eventually lost performance. This implies a degradation of the material. The document display sleeve was a particular victim of this, while the thick document wallet was rugged.

In early testing during the prototyping stage, we had used a document wallet with its stock seam and the foil slid right into this seam. In hindsight, this was not a good idea. While at a casual glance the seam looks heat-sealed, in fact there are small holes everywhere. Not only did this leak charge, but is also let through an arc strong enough to shatter the plastic.

The table shows that we had trouble getting sparks from some set-ups. The larger value capacitors simply loaded the motors too much, and temperature was increasing. With no capacitor, the SparKit Wimshurst machine runs at around 400mA with a spark going. With capacitors that were too large, current draw crept above 480mA. The large charge stored opposes movement of the discs as they are charged devices, too, with opposing charge on each.

Additionally, leakage was an issue for builders' plastic, which was audibly hissing and smelled strongly of ozone.

# LEYDEN JAR TESTING

Leyden jars were tested in a similar way. The first ones we tested were the water version, followed by the glass jars then the plastic. However, when first trying the water version, with a large spark gap set, we heard lots of crackling which we thought might have been leakage. So, you will see in some photos but not others that the bottles are covered with plastic over the foil, because of all the small sharp edges that foil makes when it crinkles.

We are not sure whether or not this made a significant difference. However, in the end, we covered all our Leyden jars most after the testing was complete, with a layer of book-covering contact to avoid losses to the atmosphere. This seemed to improve not so much the energy of the spark but its reliability.

 MATERIAL CAPACITANCE (pF) SPARK GAP, SINGLE (mm) SPARK GAP, SERIES (mm) Water Bottle 585 / 521.4 6 17 Tall 6.8 / 7.2 17 20 Short 6 / 7 8 16 Plastic 6.2 / 5.8 18 20

The Leyden jars had some surprises in store. In each case, the series pair fared better than the single capacitor despite the lower stored charge. We hypothesise that this is because the higher voltage on a single capacitor caused leakage before a decent voltage could be reached. This is supported by the movement of exposed foil surfaces, sound of crackling, and smell of ozone.

We were also highly surprised at the low capacitance of our builds compared to online statements, but we think this is due to the physical construction and use of loose foil. In many cases, the foil is uneven in its relationship between inner and outer plates. In addition, the many crinkles and points provide leakage paths when charged. Many Leyden jar builds are time-consuming things where foil is bonded to inner and outer layers with glue and rolled on, or even via being painted on as metal paint. Ours were far from that.

# CONCLUSIONS

As it turns out, there is such a thing as too much capacitance in this situation. Under ideal (theoretical) conditions, the capacitor will simply charge up until the voltage across it reaches the ionisation voltage of the spark gap. The more capacitance, the more energy stored and the more energetic (brighter, thicker) the spark.

In reality, these capacitors are far from perfect and have many leakage paths. This means that if the capacitor is too big, the voltage across the spark gap never reaches ionisation voltage because leakage bleeds the charge away as fast as it is being pumped in. In a smaller capacitor, the voltage rises much faster, and the area for leakage is smaller. Therefore, the capacitor aids the spark in this case.

However, we validated the use of a single capacitor and in fact got better results in some cases, likely due to the reduced number of leakage paths (though the increased voltage across the single capacitor would have doubtless meant we would not have halved the leakage) and the increased capacitance over having two capacitors in series.

We found it interesting that a single rolled capacitor often worked well, but a single Leyden jar did not. We are yet to try wrapping the Leyden jars in plastic or contact to reduce leakage to atmosphere and see if that changes the situation.

# LEAKAGE AND CORONA DISCHARGE

We have mentioned leakage quite a lot in this article, but haven't described what it really is in this context. Air is generally considered an insulator in most Maker situations. However, air can carry an electric charge: Electrons can migrate between the atoms in a gas, if it is not inert (which is why Argon is used in MIG and TIG welding to prevent oxidation from the air). When an atom gains or loses an electron, it becomes charged. The charged atoms are called 'ions' whether in a gas, liquid, or solid phase. Salts dissolved in water separate their ionic bonds (that term is not a coincidence) and become ions in an aqueous solution, hence they carry charge and can be used as electrolytes. The process is similar in air, but not so easy to demonstrate.

When a voltage is high enough to cause an electron to leave its atom in air, the air becomes 'ionised' and will more easily conduct a current through it. This only occurs at a voltage that is high enough, hence air being considered an insulator in most Maker situations. However, the static charge built up by the Wimshurst machine is well above the level required for ionisation.

A spark is not the only way for a charge to dissipate. As soon as air begins to ionise, charge can travel through it. The charge can just bleed off into the air, as there is an abundant supply of gas atoms for the charge to interact with and spread. So, even before a spark can occur, a charge can leave an area with excess charge. Sharp points aid in this, as it gives the charge somewhere to focus. In reverse, this is how the pointed charge combs work as collectors on the Wimshurst machine.

The same will occur at many points, including the ends of wire strands twisted around the spark gap rods, the corners and even edges of foil sheets if left in air, and any other shape point. The charge can bleed from anywhere, but smooth round surfaces lose far less than hard edges. It is worth noting that ionisation and charge dissipation occurs at a lower voltage than that required for a spark of any reasonable length.

In some of the photos included last month in the Wimshurst build, we showed purple streamers travelling between the spark gap spheres before a spark occured. This is what rapid ionisation looks like, and is called Corona Discharge. Many of us are sick of the word "Corona", but the group of viruses that includes Covid19 (and the common cold and many others) is so named because of the shape of the spike protein that latches into the right human cells to infect them with virus RNA.

That shape is similar to the discharge pattern from a point source of ionisation, where the charge flows straight up at first, due to heat, then starts to spread in many directions. This shape is called a Corona, derived from both Latin (Corōna) for garland or crown, and Greek (korδnē) for crown or curved object but named for a type of seabird.

Despite the name, Corona discharges happen in many shapes and forms. Most still follow the corona pattern when examined closely, while from afar there is often a mass of such discharges together which make the shape seem like a line, ring, or the like. Sadly, we could not get a good photo of a Corona discharge from our Wimshurst machine, even with crocodile clips with pins attached and pointed at each other, secured to the spark gap rods. The discharge at the voltage and current involved is simply too weak.

While we did use a long exposure and very high ISO with a huge aperture, the exposure still had to be over ten seconds to see anything at all, and by this time, there had been enough path changes that the image just showed a purple ball. Most good photos of Corona discharges are of situations involving much higher voltages and currents, such as high-tension power transmission lines in the many tens of thousands to hundreds of thousands of volts range.

# WHERE TO NEXT

We would like to experiment with bigger capacitors in an oil bath to prevent leakage. However, this is a messy process and also requires more than just oil in a container. To be of any real value, this experiment would need to be partnered with other improvements.

Specifically, we need to source high-voltage wire such as automotive spark plug high-tension leads, to avoid leakage after the wires leave the oil and in particular, when they are close to other parts of the Wimshurst machine. We also need to come up with a lower-leakage way of attaching the wires to the spark gap rods, because twisting wires or using crocodile clips provides many leakage points.

In another direction, we also want to experiment with both the rolled capacitors, and Leyden jars in parallel (with their own kind, not mixed). After verifying that a single capacitor works and that there is no electrical reason (that we have discovered yet) to use capacitors in series besides spreading the voltage out, we would like to see just how bright a spark we can make by using capacitors in parallel. We have to balance this against the increased leakage, and we are aware that this may or may not produce a better spark.