Transformers With Tertiary Windings - John Dunn, Consultant, Ambertec, P.E., P.C.

We were trying to make 100 kV power supplies using the push-pull class-C amplifier technique and we got into trouble with our high voltage transformers.

With a x20 voltage multiplier on the transformer's secondary side, reflected capacitance into the primary winding was very large. Even with using 0.003 inch gapping on our EC70 transformer cores, the resonance between the primary winding's magnetization inductance and that very large reflected capacitance was at far too low a frequency.

Trying to raise the resonant frequency by lowering the magnetization inductance by further increasing the gap between the cores also increased the windings' leakage inductances beyond tolerance.  

 

The remedy was to remove the gapping altogether from the main transformer thereby raising its magnetization inductance value way up, but also lowering its leakage inductance values way down. Then we added a tertiary winding which was connected to a completely separate inductor that was wound on a 0.003" gapped 3622-sized pot core.   

 
The net magnetization of this arrangement was now that of the inductance of the separate pot-core coil as reflected to the main transformer's primary winding and appearing in parallel with that very much increased magnetization inductance value mentioned before.

Transformer resonance was now in the 25 kHz area under the control of a proper value of magnetization inductance, just like we wanted it to be, without having to pay the penalty of excessive leakage inductances. 

It worked beautifully.

Comments

[Joel Koltner]

Slick, that's quite clever!

[david pacholok]

I wonder how you were measuring leakage inductance, John. My experience says that leakage inductance(absolute L, not as a percentage) depends upon the flux linkage between the pri and sec, that is to say the physical space between the geometeric center of each winding. The bigger the physical space the greater the leakage, as more lines can "leak out" when the sec is loaded and opposing flux is generated by the secondary current.

In test after test, Lsubl measurements were the same with or without the core, if we are talking absolute Lsubl, not expressing it as a PERCENTAGE of Pri L.

As an aside, having built many HV power supplies up to 150 KV and 3KW DC for charging of pulse disch caps, I learned the hard way to choose a topology that provides sine wave current to the HV xfmr. This not only reduces parasitic resonance issues with xfmr/varactor effect of HV diodes, but seems to be a lot easier on the diodes. (less extra commutations due to ringing)
Suitable circuits are the push-pull or full bridge current-fed sine-wave converter, or some form of series resonant converter or LLC converter.

Another trick is to beef up the insulation petween pri and sec to 1/2 the DC Vout and feed the center of the Cockroft-Walton mult. This will improve open-loop regulation of the HV out by a factor of 4 with the same size HV caps, which get big and expensive at higher powers. Of course this will increase Lsubl and is just another engineering tradeoff.

[John Dunn]

Leakage inductance would manifest itself as ringing on the drive transistors' Vce swings. The Vce swings would go to a plateau on each half cycle that was set by the DC of the voltage multiplier scaled down by the transformer turns ratio, but the Vce overshoots would become more extreme as leakage inductance got larger. The ringing frequency versus the net tuning capacitance allowed leakage inductance estimates to be made using the resonance equation. This was not a precise leakage inductance measurement by any means, but it was qualitatively revealing. It was also the limiting factor for how high in voltage a power supply like this could be made to go. The upper limit had been only 50 kVDC and leakage inductance had made even that quite difficult. This trick allowed making 100 kVDC easy. Sadly, we never did try to see how high in voltage this trick might have allowed the high voltage DC to have gone.

[Harry Dellamano]

John Dunn, I agree with David and believe you have totally missed the point on this one.
Leakage inductance is a function of the windings and will not change with different gapping. You have a capacitive load that cannot be driven by a voltage source push pull topology. As David points out, a LLC topology will efficienly use the leakage inductance in a resonate manner and reduce secondary losses.
What is your 100KV load and Vcc? A proper LLC design will be more efficient, have less switching noise with a smaller transformer.

Harry