Re: Recent s.s.t.c work

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: "K. C. Herrick" <kchdlh@xxxxxxx>

Frequency splitting...  Yes, occurring when both primary and 
secondary are resonant (equal Fr's or no) and k is greater than 0.01 
or so.  So...I hark back to my previous s.s.t.c. that had a 
non-resonant primary.  No frequency splitting there.  So why should 
one use a resonant primary in a s.s.t.c?  I measure 15 uH inductance 
in my present one, which yields about 9 ohms of inductive reactance 
at 100 KHz.  With ~300 V rms applied from the H-bridge and with 
relatively negligible primary resistance, the current would be n.g.t. 
~330 A.  Thirty amps of maximum mains current would require a duty 
cycle of n.g.t. 9% which is OK.

So once again...why use a resonant primary?  It would seem that the 
answer is, to get more current in the primary which will yield more 
flux cutting the secondary which will yield greater secondary 
voltage.  Except that you have to deal with that pesky frequency splitting.

I wonder if a better feedback scheme would be to feed back from the 
secondary rather than from the primary.  That way, the operating 
frequency of the system would always be at the Fr of the secondary, 
with the Fr of the primary to be adjusted (presumably lower) so as to 
reach the maximum primary current that could be accomodated as a 
function of duty cycle and allowable mains-current.  Does that sound 
reasonable?

As to the 15 ft of "antenna" connecting my delay-adjusting pot to the 
apparatus:  I plan to put the full 17 V of dc that I have available 
across the pot.  Then, at the coil end, I'll capacitively bypass that 
to ground at the a/d converter and then resistively-divide it down to 
the 0-5 V the a/d requires.  That will yield a) the proper 0-5 V of 
signal and b) less than 1/3 of the EMI that would otherwise be present.

Steve, you're right about the frequency-shifting.  But it is pretty 
small: likely no more than 1-3%, I'd think.  At, say, a 6 us delay, 
3% would yield a 180 ns error.  That, of course, might well fully add 
to the 200 ns resolution-limit that I will see in my 32-stage/5 
MHz-clock shift register.  And if I forget primary resonance and go 
with only a single-resonant system, there will still be that slight 
shift--but then, in the Fr of the secondary.  But still, I should be 
able to tweak the pot for best operation at the full power 
level.  Also, I'd be operating at ~100 KHz and not ~70 so the 
required delay would be less.  And in any case, a significant part of 
the delay will be the turn-on time of the IGBTs which, because of 
Miller effect, will, I think, vary with applied h.v. but not with Fr.

I'm going to see if I can add a secondary in my simulation; but maybe 
not because of element-limitations in the freebie v. of the program.

KCH

[snipped]


> And to Ken:
>
> I simulated a DRSSTC with a 1/2 cycle delay added in line with the 
> feedback.  It is pretty close to ZCS most of the time, but there is 
> a small glitch.  Im not sure if you have been following along with 
> some of the theory about the frequency splitting that occures with 
> the DRSSTC, but this is the problem.  When the primary and secondary 
> are tuned equal, you excite 2 frequencies, one above and the other 
> below the natural Fo.  This results in a beating primary waveform as 
> these 2 frequencies beat.  But, it appears (at least in my 
> simulation) that the actual period of the primary oscillation 
> changes slightly as well.  So you know what this means.... some of 
> your cycles are going to be mis-timed, because the IGBT switching 
> was determined from the *previous* half-cycle.  Overall it seems 
> that using the delay method on average gets better ZCS than without, 
> but you definately have to be careful...
>
> Also, you mention a 15' length of wire running to a potentiometer 
> that would be used to control the amount of delay.  This sounds a 
> bit tricky.  Hope your wiring is very well shielded.  I personally 
> would just mount the pot right on the PCB and adjust it at low power 
> and watch the trend as the power increases.  From that you should be 
> able to guess at where to set the delay.  It would probably require 
> some trial and error, though.
>
> Steve