[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: El Supremo



Original poster: "K. C. Herrick by way of Terry Fritz <twftesla-at-qwest-dot-net>" <kchdlh-at-juno-dot-com>

Terry (& all)-

I'll interpose comments (and I wish that my received emails did not
include a whole lot of superfluous "carriage returns"; does everyone
suffer that?)...

On Thu, 12 Sep 2002 17:40:10 -0600 "Tesla list" <tesla-at-pupman-dot-com>
writes:
> Original poster: "Terry Fritz" <twftesla-at-qwest-dot-net>
> 
> Hi Ken,
> 
> I was looking at this all afternoon!! :-))))
> 
> At 09:09 AM 9/12/2002 -0700, you wrote:
> >Terry Fritz's groundbreaking work with the OLTC has given me pause 
> to
> >think about my own design and how it might be improved.  Also 
> constantly
> >in my mind is an appreciation of the ratio of my MOSFET failures to 
> his
> >IGBT failures: it's infinity, not to put too fine a point on it, 
> since
> >dozens upon dozens have failed for me and none have failed for 
> him.
> 
> The big factor there is that I am switching at only 120 Hz.  No high 
> speed
> switching stuff to worry about at all.  
> 
> >
> >Terry's 1-turn-primary design is what I had had in mind in the 
> first
> >place, but my MOSFETs would not drive 1 turn and IGBTs back-then 
> seemed
> >too slow.  But now, I see a way to combine Terry's notion and mine. 
>  I
> >call it, perhaps with some want of humility, El Supremo.  With 
> Terry's
> >indulgence you can find its conceptual schematic at 
> >Http://hot-streamer-dot-com/temp/tch2.gif.  See also the accompanying
> >http://hot-streamer-dot-com/temp/tch2drv.  
> 
> The second URL really should have been this but I fixed it so most 
> browsers
> would still work.

Yeah, I goofed there; hope everyone got around that.
> 
> http://hot-streamer-dot-com/temp/tch2drv.gif
> 
> >
> >Referring to TCH2, all IGBT "1"s conduct during one half-cycle and 
> then
> >all "2"s conduct during the alternate half-cycle.  If you follow 
> the
> >current paths you will find that the four quadrants of capacitors 
> become
> >connected in a "daisy chain" around the primary loop in 
> alternating
> >polarities, thus establishing the requisite alternating magnetic 
> field
> >perpendicular to the image.  
> >Each set of IGBTs is driven by a crossover-controlling circuit like 
> that
> >of TCH2DRV.  
> 
> The big thing "I" see here is that with only a minor rearrangement, 
> you
> have 2400 volts instead of 600 driving the secondary in sort of a 
> Marx
> configuration.  That allows higher frequencies (far less secondary 
> loss),
> and lower loss in the primary!!!  Just what my OLTC needs!!!! 
> :-))))

You lose me here.  Do you mean, charge each of the 4 sections to 600 V
instead of 150?  No problem there, conceptually.  But how does that allow
for higher frequency?
> 
> "I" don't see the great need to switch two circuits at the Fo 
> frequency for
> my OLTC type of coil.

I assume here you mean, switching actively for each half-cycle rather
than passively, thru the diode, as you currently do.  No, you wouldn't do
that for the OLTC-type--but El Supremo would have to incorporate that
since there's no primary resonance to set up the cycling.  The capacitors
are just discharging dc-wise.
> 
> >
> >Note that, during each brief interval between half cycles when no 
> IGBTs
> >are conducting, the "inductive kick" of the primary loop will 
> become
> >clamped, regardless of its polarity, by the "fast diodes", as 
> coupled via
> >the loop capacitors.  No transistor will ever see any voltage 
> greater
> >than twice a capacitor voltage.
> 
> That does work very well!!  The IGBT reverse diodes work perfectly!

Yes, if they're built in, so much the better...
> 
> >
> >Now here is where a major difference with my present design comes 
> in:
> >
> >The fundamental difference between sstc's and spark-gap tc's is 
> that the
> >former generates its spark by application of a constant-amplitude 
> burst
> >of excitation whereas the latter generates an 
> exponentially-diminishing
> >amplitude of excitation.  This means that by far the greatest 
> amount of
> >instantaneous energy going into the spark is delivered, in a 
> spark-gap
> >system, during the first few cycles of excitation.  It does not 
> escape
> >notice that, given the same mains input-power, a spark-gap coil 
> will
> >produce significantly longer sparks than a sstc--even given the
> >relatively large loss of power in the spark gap.  So, it is very 
> likely
> >to be concluded that what's wanted is to cram as much power as 
> possible
> >into those first few cycles.
> 
> Remember the "peak power"!  A sstc runs 100% of the time at 1X the 
> power.

Well..not mine:  Mine is "disruptive" just like yours.  So many ms
sparking, so many ms just sitting there.

> A disruptive coil runs 5% of the time at 20X the power.  20Kv at 500 
> amps
> is 1,000,000 watts of power into the streamer!  You be it "jumps" 
> ;-))  Do
> note the remarkable streamer length increases using the "staccato 
> mode" in
> CW coils.  A very interesting affect that is due to electrostatics 
> but may
> be very important to us and streamer length.  We are nearing the 
> point of
> having very good control over such unexplored factors.

Since both my coil and someone else's spark-gap coil are "disruptive",
and they sit side-by-side, so to speak, and draw much the same power from
the mains, and yet--the other guy's sparks are 30% longer...I have to
conclude that that has something to do with the shape of the power
impulse put into the secondary.  That's why I conclude it's best to
emulate a spark-gap coil, in putting by far the most of the power of each
spark into the first few cycles of that spark's excitation.
> 
> >
> >So what I propose with TCH2 is that the capacitances be markedly 
> smaller
> >that what I have in my present design.  In fact, they are to be 
> small
> >enough to give the desired rate of exponential decline in voltage 
> during
> >each spark event.  Note that the current in all the capacitors is
> >unifirectional, not ac and that the primary is still untuned.  The
> >capacitor voltages just decline during the spark event, they do 
> not
> >change polarity.  And since the IGBTs are to be driven from an 
> external
> >signal source, that source can either be tuned at will to match 
> the
> >secondary's Fr or it may readily be derived directly as a function 
> of the
> >secondary's return-current, as I do presently, to make the system
> >instantaneously self-tuned to the secondary's Fr.
> 
> Self tuning and frequency control is really nice.  But the logistics 
> of
> making it work scare "me".  I'll stick with the plain simple 120BPS 
> tuned
> primary ;-))

But "120 BPS" and "tuning" mean different things:  120 BPS is the
repetition rate of the "tuned" burst of sine waves that excites the
secondary.  What's tuned is the frequency of the pulse-burst, & not the
120 BPS--right?  That's where it's really handy to have that always
"tuned" to the secondary's Fr.  Whenever that secondary is moved to & fro
with respect to grounds, its Fr changes, of course, by a substantial
amount.
> 
> >
> >Most importantly, this scheme acts to completely separate the 
> function of
> >frequency-determination from the function of spark-energy storage. 
> 
> >Change the frequency?  Just do it: the energy available to create a 
> spark
> >won't change.  Increase or decrease the primary's time-constant?  
> Just do
> >it: the frequency won't change. 
> >
> >And notice one more interesting thing:  I show the capacitors as 
> being
> >connected to the primary buses all along their lengths.  I think 
> that it
> >doesn't matter where they are connected since a) their current is
> >unidirectional and b) the current in each of the buses is 
> unidirectional.
> > There exists the same length and locus of conductor regardless of 
> where
> >a capacitor is connected.  The advantage is that a large number of 
> small
> >capacitors may be closely attached to the conductors, both 
> maximizing
> >their overall current-carrying capability and minimizing their 
> lead
> >inductances and the overall ESR.  Also, it's a handy place to put
> >them--all around the periphery of the primary.  
> 
> I worry a little about using caps as a primary inductor element 
> directly.
> Not sure what kinds of odd electrostatics goes on there.  Streamer 
> hits to
> a primary cap (optimistic ;-)) would be a bad thing.

Oh, I wouldn't worry about that at all.  Nothing but good old C, L & R
inside those capacitors!  And as to streamer-hits, with the cap's on the
outside of the primary conductors, the hits will go to the copper rather
than to the capacitors.

> 
> >I envision two
> >1/2"-diameter copper pipes for each segment, spaced vertically 
> perhaps
> >1/4" apart and with the capacitors' leads soldered to their outer
> >peripheries.  Easy to assemble, easy to change the capacitors.

Now, I think more like 1": substantially lower inductance.

> >
> >I show 4 primary segments in TCH2 but could be any even quantity.  
> 
> 
> There is a practical limit.  Probably more based on how much 
> voltage/power
> an IGBT can handle and what kinds of faults it may see.  Perhaps a 
> physical
> limit too.  Just have to work the numbers...
> 
> >Also,
> >the capacitors could be charged in various ways, for example from
> >current-regulating supplies as in my present design or resonantly 
> as in
> >Terry's OLTC design.
> 
> They can be "one" supply :-)))  Think Marx generator.  Simple 
> resistors can
> isolate the section during firing since say 5 ohms is trivial to a 
> 0.01 ohm
> primary loop.  All the current will ignore the 5 ohms...

That's right.

> 
> >
> >Current-sharing amongst the paralleled IGBTs should not be a 
> problem:  
> 
> Watch out for that one!!  Just a little inductance or a late turn on 
> and
> boom!  This is especially true when running them at 3X their peak 
> rating
> :o)))  Try to think about using the cap impedances to force current 
> sharing
> like the OLTC.  That works great!!

You may well be right.  But I puzzled a long time as to how I could use
separate capacitors for each transistor & couldn't see how to do it.
> 
> >A
> >very small emitter resistor developing perhaps 0.2V at peak 
> emitter
> >current should suffice to steer current away from the stronger
> >transistors into the paralleled weaker ones.  
> 
> Yips!!  No!!  That 0.2 ohms is an enormous power drain for a 
> disruptive
> coil!  Think of what it does to the primary Q.  If you have say 0.5 
> ohms of
> primary resistance in a 20Arms primary circuit that is 200 watts of 
> real
> power!!  I would never consider adding any resistance in the 
> primary!

No!, no!  Not 0.2 ohms but 0.2 volts: that's all you'd need to persuade
the other paralleled transistors to take up the load, with their
conducting Vc-e's of maybe 3 V or so. 
> 
> >Such a resistor might well
> >consist merely of the lead connecting each IGBT's emitter to the 
> common
> >point.  
> 
> Watch inductances!  13nH is small until you have 200 amps at 100kHz. 
>  Then
> you have 1.6 volts (reactive) which is enough to cause problems.

Yeah, very likely.  Murphy...
  
> That is
> just what goes on "inside" the IGBT case let along what we solder to 
> it next...
> 
> >Further, avalanche-breakdown should never be a problem since no
> >transistor can see any voltage greater than twice a capacitor 
> voltage
> >under any condition.
> 
> TVS transorbes are wonderful things too just to be really sure!!

Try it first without?...

> 
> >
> >And finally, this is a low-voltage design.  I remember, when I was 
> a kid,
> >that the president of the ARRL was killed by his ham apparatus' 
> high
> >voltage.  That's always stuck in my mind.
> 
> Low voltage is wonderful.  All kinds of conventional tricks can be 
> used
> that are not available to us at 20kV!!  However, if we get higher 
> (2400V)
> firing voltages, things get a little more complex.

Yeah, you're just as dead from 2400V--but not quite as dead as from 150V.

> 
> >
> >It seems to me that this idea has great potential.  I hope I will
> >personally have the energy to look further into it & build it and I 
> urge
> >others to contemplate doing so.  Or else...to contemplate telling 
> me that
> >it won't work.
> 
> Such a system for the OLTC looks very promising!!!!  Switching at 
> only
> 120BPS still but using your four section system could solve many 
> problems!!!
> 
> Cheers,
> 
>         Terry

Ken