[Prev][Next][Index][Thread]
RE: Self-tuning Tesla coil
> 1. A computer generates a frequency which is amplified to the point where
> I can drive the primary with it.
>
> 2. I detect the magnitude of a received radio signal with my computer
> (e.g. signal strength monitor in the frequency range I'm generating)
>
> 3. Change the frequency so that I get maximum output.
>
> This sounds like something that has probably been tried before. I was
> wondering if anyone had any insights into this--especially people who have
> experience with using computers with Tesla coils and haven't blown up many
> computers...
Here's an idea - drive a step function (i.e. a hard, as close to
instantaneous as possible, bounce-free switch from one voltage level to
another) into the coil. Sample the output, which will give you its impulse
response. Inverse FFT this and find the biggest peak - there's your resonant
frequency. Assuming that the coil's resonant frequency doesn't change much
with voltage, you could probably do this at very low voltage (<1 volt or so)
so you don't fry your equipment.
Version 2 of this idea would be to use an audio sweep generator and an
oscilloscope. Set the sweep to sweep through the frequency range you're
interested in, and the waveform to sine. Drive the coil via a resistor (1k
should be about right) to limit the current. Connect an oscilloscope's Y
axis to the coil end of the resistor, and the X-axis to the linear sweep
output on the sig gen. You will then be able to see the coil's resonant
frequency(s) as peaks on the scope's screen.
Version 3 of this idea is maybe a bit less accurate. Wind an extra few turns
of wire around the coil to sense the field it's generating. Feed the signal
from this extra secondary to the input of a high gain amplifier, set up to
invert the signal (180 degrees phase shift). Feed the output of the
amplifier into the coil. You may need to add an RC network or two to avoid
settling on the wrong harmonic, but this should make the coil self-oscillate
at its resonant frequency. If you have a sufficiently powerful, high voltage
amplifier, it might just be possible to run the coil like this.
Alternatively, you could do it at low voltage and use a scope or a frequency
meter to read the resonant frequency back. This idea is a variant of a
crystal oscillator circuit, as commonly used in digital clocks and
computers, except that a resonant coil/transformer is used instead of the
usual crystal.
Version 4 of this idea needs a bit more work, but might lend itself to
running at full voltage. Use a sense winding as described in 3, but take the
output into a current limiting resistor, then an resistor
ladder/potentiometer, then a pair of back-to-back Zeners to limit the
voltage swing. Buffer this with an op-amp configured for unity gain, then
full-wave rectify and smooth the output. This gives you a voltage which will
directly read the amount of power the coil is producing, in theory anyway.
Differentiate this voltage, so you get a +'ve signal when it is going up,
and a -ve signal when it is going down. Put that signal to one side for a
moment. Set up a voltage controlled oscillator to drive the power amplifier
stage that is delivering the voltage to the coil. The control signal to the
VCO should be derived from an integrator, so that the frequency is either
'going up' or 'going down' slowly. Control this from a flip flop. Set up the
'going up/going down' voltage so that the direction the frequency is going
is reversed whenever a down slope is detected. You may need a few time
constants to avoid oscillation or over-controlling the frequency, but this
is essentially an all-analogue approach which is guaranteed to home in on
the right frequency. It will drift a little around the centre frequency, but
it's the difference between frying a $1000 PC and $20 worth of chips and
discrete components.
I've not tried any of these. The real question is not so much figuring out
how to generate the right resonant frequency, it is more a case of knowing
how feasible it would be to build a power amplifier capable of driving the
kind of load that a Tesla coil represents. A bank of parallel MOSFETS or
IGBTs in full bridge mode running from +/- 60V rails is probably feasible -
you can deliver a *lot* of current this way. So long as you didn't drop
below mains frequency, it may be possible to then step this up using an NST
or even a pig, but it is anybody's guess how they would behave under those
conditions. A few hundred hz is probably achievable, but as for KHz or MHz,
I have serious doubts. Quite apart from the NST not necessarily dealing with
high frequencies well, there is the problem that big fets will switch way
above audio frequencies, but it is worth remembering that as the current
goes up, it takes them longer to switch because their effective gate
capacitance goes up. This means you have to drive the FETs from pretty hefty
transistors, but more importantly the higher the frequency, the hotter they
get. This is exactly the same phenomenon, although at a very different
voltage/current end of the scale, as the tendency for modern CPUs to consume
significant amounts of power whilst needing non-trivial cooling.
Finally, correct me if I'm wrong, but I was under the impression that Tesla
coils aren't really used as coils at all in the typical circuit - they are
really used as unterminated transmission lines, which allow a wavefront to
bounce backwards and forwards inside them until the voltage builds up enough
to cause a breakdown via an arc or corona discharge. This makes them comb
filters, rather than a more typical LC tuned circuit, yes? As a result, a
sine might well only allow them to deliver a fraction of their potential.
What you really need is a composite wave form containing all the frequencies
that the comb filter resonates at - a step function will do this, but maybe
what you really want is a steep, sharp edged square wave centred on the
lowest resonant frequency. Better still would be a ramp (sawtooth) waveform
which has all harmonics, not just the odd ones, although I have no idea how
you'd manage to generate one reliably at the kinds of voltage/power levels
coilers typically use. I'd guess that sticking with the spark-gap type
sharp-edged, step function impulses and simply going for more voltage and
current is an easier solution than getting fancy - the 'use a bigger hammer'
approach might be best after all.
Sarah