The primary current across a whole bang is the only tuning evidence a loaded QCW gives you, and few builders photograph it. It is also coarser than it looks. It shows the current running away at the top of the ramp and points at which pole the driver is on, but it will not tell you how far to move the tap.
We have not switched this machine on, so there is no capture of ours here, and there is no published one to borrow. The builders cited below describe what their coils did, but none of them prints an envelope next to the tuning that produced it. What carries over to your coil is the mechanism, so read your own envelope against that and not against a shape described in words.
Why the envelope is the only place to look
Most of a QCW's tuning can be done cold. Dead time and phase lead are arithmetic, and both apply at the start of the ramp, where no arc is needed. Detuning is different. The rule is to detune by as much as the arc detunes you, and an unloaded coil has no arc. That check has to be made while the coil is making sparks, and the primary current is what a running coil will show you.
The frequency page says the same and sends the reader here. It does not claim that the envelope reads out the detuning rule, and for a long time this article did claim that. In practice the envelope reads out something coarser.
Three things set the shape, and only one of them is tuning
- The bus climbs several-fold across the ramp. That alone carries the current up with it, so a QCW envelope is never flat and a flat one is not the target. The MMC page rules a flat top out as unphysical for this machine for the same reason.
- The resistance the bridge sees climbs while the arc is short. More load on the secondary means more reflected resistance and less current for the same bridge voltage. This pulls against the bus and stops the envelope from simply following it.
- Then the resistance stops climbing. It rises with arc length, turns over and falls again. The curve is on the primary's page: the turnover sits at 0.88 m at
k= 0.40 and at 1.25 m at 0.50. This coil's 0.421 lies between them, so between 41 and 17 per cent of a 1.5 m ramp is spent past it. Past the turnover the load falls away and the current has to climb to put the same power in.
This means a tail that turns up is already there on a coil with nothing wrong with it. At the couplings a QCW runs at, 0.30 to 0.50, the turnover falls inside a 1.5 m ramp, and the tail by itself is not a tuning fault. Earlier versions of this article got that wrong. The curve comes from a model and has not been measured: the closed form behind it has never been fitted to a built coil, so treat the turnover as a shape and not as a length to the centimetre.
What it does read out
The envelope can tell you which pole the driver is on. The two accounts of the mechanism disagree about most things, but they agree about this.
Anders Mikkelsen's account is quoted in full on the poles page. On the lower pole the tank draws more power as the arc adds capacitance, which makes it hard to ramp the power up slowly, and in his words "most people prefer to run QCWs on the upper pole as it gives a more linear response of power vs bridge voltage". Note the QCW in that sentence. The same page has the lower pole as the ordinary choice on a DRSSTC, which has no ramp to break. On the envelope it looks like current getting away from you at the top of the ramp.
Our own arithmetic, on the same page, disagrees about the sign. At a fixed bus it gives negative feedback on both poles: the reflected resistance rises through the ramp and the current holds itself back, with the self-limiting about 3.7 times weaker on the lower pole. He describes a runaway, the model a slack brake. The poles page says which to doubt first. He has built the coils, and the model has one calibrated constant in it.
Both accounts bend the envelope the same way, steeper on the lower pole than on the upper. They differ on how much steeper, and a photograph cannot tell you which is right. If the tail is steep, suspect the pole first. The envelope will not give you a number for it.
There is a cheaper check of the same thing that does not use the envelope. The buck page points out that during the wick, with the arc barely alight, the two poles are furthest apart, 202 kHz on this coil, so the frequency in the first millisecond shows plainly which one the driver captured. Later in the ramp both have come down and closed up, and the frequency no longer separates them. If the envelope makes you suspect the pole, confirm it there.
What it cannot do, which is most of what it was brought in for
- It cannot give you the detuning percentage. The turns arithmetic is on the primary's page, and its advice is to cut the taps generous in both directions, because you do not know the number until you have measured the detuning. The envelope does not measure it.
- It cannot tell detuning far enough from detuning too far. Both leave the system coming into tune across the ramp, and both give a weakly rising envelope. Telling them apart would need the current level rather than the shape, and a level means nothing without a reference current for the arc you got.
- It cannot be read at the start of the bang. The first stretch is the wick, a bus shelf set as a duty in the ramp generator. The envelope does not begin at zero, and the height it begins at is a setting rather than a symptom.
An earlier version of this article had a list of shapes and fixes: add primary turns if the envelope curls up at the tail, take turns off if it is high and flat. Both were wrong. The first treats the impedance turnover as a tuning fault. The second describes the same envelope that the section above calls correct, so one observation got two opposite verdicts.
Two ways to misread the picture
What the shape costs the bank
The shape of the envelope decides how hot the tank runs far more than its peak does. The two figures are with the MMC: a linear rise gives I_pk/sqrt(6), 0.408 of the peak, and a quadratic one I_pk/sqrt(10), 0.316.
0.408 / 0.316 = 1.29 RMS current
1.29 squared = 1.67 dissipation, which goes as the squareMoving a coil from the second shape to the first is 1.29 times the RMS current and 1.67 times the heat in the same parts [derived, from those two figures]. Both shapes are ramps. The flat case is left out on purpose: the same page rules a flat top out as unphysical here, and an MMC overload figure computed on one has already been withdrawn once.
Where the mechanism comes from
Gao Guangyan sets out three ways a conventional feedback driver can be tuned, and what each does as the spark grows, on his QCW 1.5 page. Primary at the secondary excites both poles and notches the current at the beat. Primary below the secondary drives the lower pole, and the coil starts out of tune and goes further out as the spark loads it. Primary above drives the upper pole, and as the secondary falls the upper pole follows it down and the system comes closer to tune, which he calls the best of the three and still not right. His own trick goes further, tuning the primary below the secondary while forcing the upper pole, so that the coil becomes "more and more in tune as the spark grows".
He gives numbers beside that: five feet of spark at 110 A on QCW 1.5, against about 200 A for the same length on Steve Ward's QCW 1 under conventional upper-pole tuning. The comparison of the two machines is his, not ours.
No one has published an envelope read against a known tuning on one coil. Until someone does, what is here is a mechanism and not a calibration.
The nearest published pictures are not of a QCW. They are from Steve Ward's DRSSTC-1, a bang about a hundred microseconds long rather than a ramp, but on one coil at one scope setting with only the load changing between frames. That is the comparison still missing for a QCW.
The rings it is measured on are on the department diagram, and there is a piece on them. The trajectory the tuning is set against is on the frequency page.