High frequency royer oscillator. / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

High frequency royer oscillator.

1 2 3 4 next

Move Thread

LAN_403

Marko

Mon Aug 27 2007, 12:49AM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Hello guys...

I started to dislike class E topology mainly because it's instability and touchiness in TC driver duty. It's very sensitive to load impedance and frequency changes, making it hard to run from secondary feedback. One can use a fixed oscillator but then there are even more things to tune together.

Naturally, I wanted to take feedback from amplifier itself to keep it always in tune. The I put another amplifier in anti parallel.

This is how class E evolves into royer: a single bus inductor is used, and instead of shunt capacitor over each mosfet a single capacitor is used for both of them.
Feedback is taken through two diodes and pullup resistors.

One of main problems with operation in high frequencies are the resistors: in order to charge the gate fast enough they need to be small and dissipate a lot of power.
In Mhz range it gets especially bad, and may need solid state solution which I'm exploring right now.

Other thing is diode recovery time. They need to be really fast and yet have low enough voltage drop to turn the mosfet off.

Tuning is troublesome, and it would be good to have some bigger air-variable caps able to stand the voltage. Some exotic techniques like saturable reactors may also come into game.

I didn't post this into Tesla coils because I also intend to use this oscillator for magnetically-coupled wireless power demo.

First board was not-so-good mockup which served only to test the theory.

Following pics is new oscillator I built today
This one is built with really minimal inductance and is highly compact.

The resistor lash up is only temporary for a test tomorrow.

I hope to create some kind of usable active pullup circuit for high-frequency operation.
Drivers like UCC's should be avoided because they create really lot of delay while having unnecessarily sensitive input.

I never really tried putting any serious power into a resonator, nor tuning it properly.
Lots of work ahead..
(To be done with your support.. I hope I renewed some interest in high-frequency coiling

)

Cheers..

ragnar

Mon Aug 27 2007, 02:43AM

Registered Member #63 Joined: Thu Feb 09 2006, 06:18AM
Location:
Posts: 1425

Hi Marko -- great to see that you've got the most critical part (low-inductance layout) under control. Reducing those gate resistors is frustrating in that you need more current through the diodes to swing it, not to mention burning more power in the resistors, but as you mentioned, there is a solid-state solution.

I've used UCCX732X chips as 'boosters' in a Royer circuit before (i.e. between the feedback diodes and the MOSFET gates), but the difficulty is the frequency limitations that result because of the delay time, etc. It can make the whole circuit run a bit more efficiently though.

Marko

Mon Aug 27 2007, 11:57AM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Hi

I don't want to remain on resistors - I want to try any possible active-pullup technique, while leaving pulldown to diodes (since they are really fast and powerful enough).
If I was to ditch diodes, I would have chosen different topology than this.

Ok, some pics; with 50ohm resistors gates don't look too good even at just 8W input power.
Here it's running at 1Mhz.
Slow turn-on causes slow turnoff in other mosfet and vice versa.

Active pullup is definitely in order before trying anything big.

Marko

Thu Aug 30 2007, 02:59PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

OK, at some point I must have stuck. Designing an active pullup or driver circuit proved much harder than I expected at first. All ideas I had have proven implausible.

Some kind of 'ordinary' active pullup was not expected to work because transistor never turns off due to diode drop. And I proved that in practice.

'totem poles' are out for same reason.

The last two ideas I haven't tried out because I don't have any good P-channel mosfets, but I'm still pessimistic about it because of diode drop. What seemed elegant solution at first has become source of trouble.

I found these small, really fast mosfets

I hoped to sample. They have excellently small delays, and are still strong enough to be used as a gate drive. I would use one as pre-drive and several in parallel to drive the big gate.

But, they are designed with very low threshold voltage (1V) and I fear they won't turn off properly. Diodes I use (UGF8JT) have max forward drop of 1.7V, and all I could hope is that it would be much less in quiescence.

In sketch 4 I tried to fix that by using only 2 mosfets and elevating the first few volts from ground, while still being able to turn on the P-channel. Bad thing is that now I need to use pullup to turn the P-channel off, making it suffer shoot-through and having the diode turn off main mosfet (plus through at least one schottky diode drop).

I hope some of you guys could give me better ideas, because all this looks shaky to me.

Ideally, I would want the main mosfet pulled into negative, and with smallest possible delay.
(Making the delay >50ns or so defeats the purpose, as UCC's would already be more useful).

Sulaiman

Thu Aug 30 2007, 03:07PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

For a high frequency 'Royer' wouldn't it be better to ensure quick turn-off
rather than quick turn-on.
This would prevent/reduce any current 'spikes'
and gives a simple drive circuit .... one diode and one resistor as standard!
You could use a bjt to short Gate-Source (Gate-Emitter for igbt)
but two per switch would be required to maintain phase. (invert & invert again)

Steve Conner

Thu Aug 30 2007, 03:48PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

WaveRider made one that uses feedback coils to drive the gates, but I can't find the link.

uzzors2k

Thu Aug 30 2007, 04:00PM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

Right here.

And that's about all I can contribute to this thread, sorry Marko.

Marko

Thu Aug 30 2007, 04:29PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Steve Conner wrote ...

WaveRider made one that uses feedback coils to drive the gates, but I can't find the link.

How could I forget.

In the end the circuit was a quite odd, used feedback with no clamping and seemed to operate largely in linear mode. Mosfet's (especially 20V-GS) aren't easy to drive in class C like BJT's or vacuum tubes.

Other than that, when used with Tesla coil, he appeared to use no resonant cap at all, and massive flyback voltages from primary leakage inductances were really hurting mosfets.

I'm actually genuinely puzzled how to really use feedback windings a ZVS converter.
I need to take voltage, and not current feedback as far as I know.

I imagine I should use large ferrite transformer or something for that purpose which I don't want to.

If I just brought some coupling loops near the primary the thing would tune out of ZVS as soon as I coupled another resonator to it and blow up.

If I could perfect feedback of that type I could probably just use a bridge instead of push pull. Some really cool possibilities show up, like seriesing switches and running off mains.

For a high frequency 'Royer' wouldn't it be better to ensure quick turn-off
rather than quick turn-on.
This would prevent/reduce any current 'spikes'
and gives a simple drive circuit .... one diode and one resistor as standard!
You could use a bjt to short Gate-Source (Gate-Emitter for igbt)
but two per switch would be required to maintain phase. (invert & invert again)

I'm using 50ohm resistors which are already grossly hot and turn-on is still extremely slow. I want it to happen close to zero voltage, not during half of a cycle

I suspect slow turn-on causes the slow turn-off of another switch and those really bad looking bumps on the falling edge.

I killed the inductance completely down, diodes have recovery time of 25ns and drain voltage is low, I shouldn't have that much ringing by any means.

uzzors2k

Thu Aug 30 2007, 07:10PM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

Have you seen the Skori coils before? They use a single base current feedback winding to drive both mosfets. The sparks he gets from 12V are amazing! I tried building a similar circuit, but I could never get it to start oscillating every time. With a proper start-up circuit I think this is the type of feedback to use. To run it offline one could simply use the mosfets to run a GDT and then a bridge. Once I wrap up the smps project I'll start working on this again.

Marko

Thu Aug 30 2007, 07:21PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Uzzors wrote ...

Have you seen the Skori coils before? They use a single base current feedback winding to drive both mosfets. The sparks he gets from 12V are amazing! I tried building a similar circuit, but I could never get it to start oscillating every time. With a proper start-up circuit I think this is the type of feedback to use. To run it offline one could simply use the mosfets to run a GDT and then a bridge. Once I wrap up the smps project I'll start working on this again.

Sure, but if I used a current transformer hooked to resonator with wires, the thing wouldn't be so wireless anymore?
The point is to keep parallel primary resonance and perfect ZVS.

To run a Tesla resonator the circuit needs tuning which is necessary evil, but not a problem at time.

What I want is real royer circuit to run independently, and be able to use it for wireless powering of stuff.

I see I've gained some interest, so can this gate drive problem be fixed. I can't use resonant /class E gate drive, so what would be best to do next?

Or diode-feedback really just isn't of any merit at higher frequencies...?

1 2 3 4 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob