[Phys-L] Re: Foucault pendulum circuit

[John Denker <jsd@AV8N.COM>]

Michael Edmiston wrote:
> I'm guessing the "feedback" John Sohl was talking about is unintentional
> feedback as opposed to designed feedback.  If that is not true, ignore
> the rest of this message.

It is very common to have an intermediate case.  People like to
design "regulated" power supplies, where "regulated" is just a
fancy way of saying there's feedback in the design somewhere.
Up to this point, the feedback is intentional.

The wrinkle is that designs that work just fine facing a resistive
load might not work so well facing a highly reactive load.  This
includes
  1) a "constant voltage regulated supply" facing a big capacitance,
   e.g. a battery charger, or
  2) a "constant current regulated supply" facing a big inductance,
   e.g. a magnet driver.

In such cases, an otherwise-well-behaved regulation circuit will go
unstable.  John S. said things went "very unstable" and until I hear
otherwise I will assume he meant what he said.

Note that examples (1) and (2) above are mirror images with respect to
the transformation
  current --> voltage
  voltage --> current
  capacitor --> inductor
  inductor --> capacitor
  series --> parallel
  etc.
which can be considered a _contact transformation_ within the meaning of
classical mechanics (e.g. Goldstein).

> If you try to square-pulse a coil

Square voltage, or square current?

> with a logic circuit, the inductance
> can create voltage spikes that can fry or confuse the logic.  You can
> get the "tail wagging the dog."  Another example of this is running a
> stepper-motor with logic pulses.  You need some components to shunt
> induced emfs so they don't get back into the logic circuit.

That unnecessarily conflates two or three ideas.  Each is a good idea,
but it would be better to consider them separately.
  A) If you apply a step-function of current to an inductor, you can
   easily create a voltage spike big enough to let the magic smoke
   out of your output stage.
  B) If you don't use proper grounding and shielding techniques, you
   can create stray couplings (sometimes called "feedback") between
   your power circuits and your control circuits.

Note that you can easily have problem (A) without problem (B) and vice
versa.  OTOH they are loosely related since high frequencies create
bigger "L I dot" voltage spikes and also create stronger radiative
couplings and mutual-inductance couplings.  In any case: make sure you
don't produce a big I dot unless you really need it.

> Connect the two ends of the coil together with a varistor or with
> back-to-back zener diodes.

That's mostly a good idea, but most of my experience has been with
systems that are big enough that no varistor and no zener on earth
can carry the current by itself.  (Consider for example driving
100 amps into a 25 henry load.)  In such a case, the zener needs
to hook onto a point one or two gain stages upstream of the power
stage.

This section applies only in the case where you need a constant-current
regulated supply for your magnet ... which is sometimes necessary,
but not AFAICT for the Foucault pendulum driver.

Also:  the zener would *never* be needed if the control circuit
could be trusted to send reasonable I(t) commands to the power
stage ... but you need to consider exceptional cases, e.g. what
happens if there is a power failure.  The control logic is crashed,
but the power stage has to keep doing its job for many minutes.
(Do the math:  L I is 2500 amp-henries, and L I dot is only a few
volts.)

>  Another example of this is running a
> stepper-motor with logic pulses.  You need some components to shunt
> induced emfs so they don't get back into the logic circuit.

That's a bit misleading in this context.  People drive square-wave
*voltages* into stepper motors all day every day.  This is standard
operating procedure.  This does not require additional components
such as "varistors or back-to-back zeners".

Voltage drive is what I recommend for the Foucault pendulum ...
unless there is some as-yet-unmentioned special requirement.

Use the physics to your advantage:  If you know the inductance,
voltage, and timescales, you know the current.  All you have to
do is handle that amount of current.  The only halfway tricky thing
is that the current will be bidirectional (and some off-the-shelf
constant-voltage regulated power supply designs can only handle
unidirectional current).  Usually you can solve this problem simply
by adding a resistor, as in the design I suggested in my previous
posting.

Nitpickers will point out that my design could be made more
energy-efficient by the addition of a couple of high current
diodes ... but for this application the efficiency probably
isn't a big issue, and simplicity is a virtue.
_______________________________________________
Phys-L mailing list
Phys-L@electron.physics.buffalo.edu
https://www.physics.buffalo.edu/mailman/listinfo/phys-l