The primary is a few turns of heavy copper around the base of the secondary. It has no electrical connection to the secondary anywhere; the two are coupled only by the magnetic field.
This article explains what the primary does, why it is a few thick turns, the shapes it comes in, and how it is tuned with a tap. It ends at the one design number you have to choose before winding one, the impedance of the tank.
What it is and why
The bridge pushes current back and forth through those turns at the coil's resonant frequency. The current makes a magnetic field, the field passes through the secondary, and because it is changing it induces a voltage there. Only part of the field reaches the secondary, so a single cycle induces a good deal less than the turns of the two windings would suggest.
This means the coil is not a step-up transformer, and the high voltage on the secondary does not come from a turns ratio. The secondary is a resonator. Each cycle's small push arrives in step with the one before, and the voltage builds up over many cycles instead of being stepped up once. That build-up is covered on the secondary's own page. The primary's job is to put a field through the secondary, over and over, in time.
It is thick because the whole tank current runs through it, and that current is large. You do not pick the size of it; it follows from the impedance of the tank. Builders use copper tube or wide strip. At these frequencies the current flows only near the surface, so the middle of a round wire carries almost nothing. The practical result is that hollow does as well as solid, because what matters is where the copper sits, not how deep the current goes. A bundle of bare parallel wires does not help: one magnetic field surrounds the whole bundle and it behaves as a single conductor. Strands only win when each one is insulated from its neighbours. The figures, including the comparison of tube against solid wire, are on the design page.
It is short because the primary is only half of a circuit. The capacitor in series with it is the other half, and the two together are the tank. The pair rings at
f = 1 / (2 · pi · sqrt(L · C))
and that frequency has to land on the coil's working frequency, because the secondary only builds voltage from pushes that arrive in step.
So turns and nanofarads are one decision, not two. At a chosen frequency their product is fixed, and you only choose how to split it. More turns and less capacitance gives a higher impedance and a smaller current; fewer turns and more capacitance gives the reverse. Everything else on the design page is about what each direction gives up.
"The coil's working frequency" needs care, though. Once the primary and secondary are coupled there is no single secondary frequency to tune to. The pair has two resonances, and which one the driver lands on is something you decide. Aim the tank at the secondary's unloaded frequency and you will be near neither of them. That is explained in there is no secondary frequency, and it is why builders deliberately tune the primary off, in a direction they still argue about.
The primary stands beside the secondary, not around it, because of coupling. Coupling is the fraction of the primary's field that reaches the secondary, and it depends on geometry: the primary's height relative to the secondary, its diameter, and where the secondary sits inside it. More coupling means more of the field is shared. What that gets you, and what it does not, is covered in an article of its own.
What one looks like
There are three shapes in common use. A flat spiral is wound outwards on a disc like a pancake, with the secondary standing in the middle. A helix is wound up a cylinder around the secondary. The cone sits between the two, leaning outwards as it rises.

Two arguments pull against each other in the choice. A cylinder runs at lower current, which is easier on the bridge, the copper and the bank's RMS rating. A cone puts its top turn further down the secondary and lowers the flashover risk. Both are real, so the current decides it: pick the side you can afford to lose. This is worked through on the design page.
Lower is safer because the secondary's potential rises from its earthed base upwards. The part of the winding beside a low primary is the least stressed part of it, so turns kept low face volts instead of kilovolts. That one fact sets the shape, the height, and the reason there is a gap at all.
Those three shapes are the DRSSTC set. Almost no published QCW in this corpus uses any of them. QCW primaries are solenoids, or are interleaved with the secondary, or in one case are the "half-donut" of Gao's QCW 2. The one exception is wound as a cone, and it is on the QCW primary page. On a QCW, then, a primary that sits inside the secondary is normal. There is also a fourth route, which makes a different class of machine: one builder reaches a coupling of 0.71 with a ferrite floor and centre post, using ferrite instead of a difficult primary and paying for it in operating frequency. This also affects the tools. The usual coupling calculator was written for flat pancake primaries, and published QCWs run at three to six times its own recommended figure.
Six to twelve turns is the usual range on a DRSSTC. Our own coil taps between 9.4 and 12.4. Shane Colton calculated six for his and wound seven and a half so there was room to move. An SSTC primary has more turns, eight to nineteen on the builds this corpus has read. The reason is in an SSTC primary is a different part: without a tank to resonate with, the winding has to make up the difference in turns.
The turns are spaced, not wound tight. Adjacent turns sit at nearly the same potential, so the spacing is not there for insulation. It sets the primary's height, and the height relative to the secondary is a large part of the coupling.
You tune it by moving a tap, a clip or bolted block on one turn, so the number of turns in circuit can change without rewinding. Leave plenty of room either side of the tap and wind an extra turn or two, because you will not know the right number until the coil is running. The tap is also the highest voltage point on the winding, which is easy to miss and matters when you mount it.
The grounded ring above the primary is not part of it. That is the strike rail, which catches an arc coming down before it reaches the primary. Its purpose is covered with the things that will kill you, since its real job is protecting the mains side, not the copper. How to earth it, on its own wire and at one end only, is with the RF ground.

Which way round to wind it
It does not matter which way you wind it, but the reason is useful because it tells you what to do after a change. The winding direction sets only the sign of the mutual inductance between primary and secondary. It does not change the size of the coupling, the frequency, or the impedance.
At the driver, flipping the sign on the primary looks the same as flipping it on the feedback transformer, because both invert the same signal. That is why boards come with a phase switch for the current transformer. Eastern Voltage Research says of theirs that it "has the same effect as reversing the wires (or direction of the primary lead) of the feedback current transformer". The two flips cancel, so a primary wound either way runs the same once the feedback phase matches it.
In practice: if you rewind, re-tap or reverse the primary, check the feedback phase again. A coil that started yesterday can refuse to start today. That is a sign error, it takes ten seconds to test by turning the ring over, and no amount of phase lead corrects it; the details are with the two rings. The reasoning here is ours; the quotation is from the board's manual.
The one decision you can make from this article
The number to decide is the impedance, not the inductance. A designer sets the impedance, and the primary current follows from it. That is covered in the primary is an impedance, which is also where turns and tank capacitance get their values.
On our own coil the two halves are 16.96 µH of primary and 12.2 nF of tank. The formula above puts the primary at 349.9 kHz, and sqrt(L/C) gives an impedance of 37.3 Ω [derived, both]. The frequency is a constraint and the impedance is a choice, so the impedance is the one that gets its own article.
What to watch out for
Bringing the two coils too close makes the secondary flash over along its own winding. That sets the ceiling on coupling, and it depends on the winding's geometry, so you cannot calculate it in advance.
The limit is a voltage on that winding, not a coupling number, and one machine on file shows it. A three coil magnifier is wound at about 0.5 between primary and secondary, twice what any two coil machine attempts, and runs fine. In that topology the secondary is not the high voltage coil; the volts appear further along the chain. Tight coupling is only dangerous when the winding it is tight against is the one holding the kilovolts. More in a DRSSTC magnifier is not just a longer coil.
What goes wrong
- The copper gets hot enough to soften. High bangs per second. Give it air and enough cross-section.
- A flashover across the secondary's winding, low down. Racing sparks. Too much coupling, and the ceiling is lower on the lower pole than on the upper.
- The frame draws kilowatts and the bolts melt. The shorted turn above.
- The current is nothing like the calculation. The calculation probably started from turns. The impedance sets the current.
Where next
- The primary is an impedance, the design number.
- Sizing a tank capacitor by RMS current, the part in series with it.
- An SSTC primary is a different part, where the same component is wound the opposite way for the opposite reason.
On the department diagram the primary is the only part drawn touching nothing. Everything else is wired to its neighbours; the primary does its job across a gap.