Re: It works! Solid state driver

From: 	RODERICK MAXWELL[SMTP:tank-at-mail.magnolia-dot-net]
Reply To: 	tank-at-mail.magnolia-dot-net
Sent: 	Sunday, December 14, 1997 10:13 AM
To: 	Tesla List
Subject: 	Re: It works! Solid state driver

> On Fri, 12 Dec 1997, Tesla List wrote:
> > It is *essential* in a pushpull design, all else being equal, to
> > *bifilar* wind the primary. The imbalance is often the result of both
> > less than best possible coupling between the primary halves and, more
> > importantly, the difference in copper mass between the two halves if
> > one is wound on top of the other.
> There are other things as well. PWM driver output is not perfectly
> symmetrical. Output paths (layout) are not symmetrical always. Most
> important one is dissimilar switching devices: turn-on or turn-off
> delays etc are usually not equal even for two same brand mosfets. This
> results in dissimilar conduction period which means dc flux saturating
> possibly the core eventually.

> Bifilar winding is essential, Malcom is right on that one. However,
> switching dissimilarities should be considered as well. Usual solution
> (for push-pull) is to use current mode controll which switches the
> devices off as soon as certain current level is reached. That will be
> below huge saturation currents and will therefore protect system.
   I will probably change pwm chips as well. I'm in the process of
looking at some parts made by Telcom Semiconductors Inc. They are
available from Digi-Key. As well as that Hexfred I was told about. 
> For tesla systems this approach may not be too easy to implement if near
> full duty cycle is wanted. The coil tends to draw sinuoidal current and
> using peak current limiting system would switch off usually at around
> peak current of sinuoid that is duty cycle might be reduced to half due
> to current limiting.

   The new pwm will include current limiting.

> Gapping the core to make it able to handle enough dc component as well
> is probably the best solution to try out with push-pull.
 What about U cores used in TV flyback transformers? All you would have
to do is slide a piece of paper between the two halfs.

> Bridge topologies on the other hand are easier in this problem: just put
> a dc-blocking cap in series with the transformer and the problem is
> solved.

    Has anyone tried series zener diodes across the drain/source
terminals to limit flyback voltages? Digi-key has some transient voltage
suppressor zeners rated at 1500watts. I know this is only for a few ms
but they also have a steady state power rating as well. Maybe 4 or five
in series in addition to another clamp that would take the brunt of the
power? These zeners are supposed to have a clamp time of 1 pico second!

                             Frankensteins Helper