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.

previous 1 2 3 4 next

Move Thread

LAN_403

WaveRider

Mon Sept 03 2007, 08:52PM

Registered Member #29 Joined: Fri Feb 03 2006, 09:00AM
Location: Hasselt, Belgium
Posts: 500

Without more details on what your measuring, I hesitate to comment in detail. ZVS is, as you know, difficult to maintain under widely varying load conditions. Controlling the feedback will only get you so far.

If Q is low without any loading, it means power is being lost in

1. resonator
2. Driving circuit
3. or in parasitic coupling/radiation

We would need to know more (resonant freq., coupling to driver, feedback ckts, etc.)

Cheers!

Sulaiman

Tue Sept 04 2007, 12:41PM

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

To work at all reasonably a zvs needs a "Q" of at least 2.2 (forgotten reference)
I would consider Q=3 low, Q=5 ok and Q=10 overkill (unless you want a 'good' sinewave)

The Vpk = PI x Vdc is true!, I worked it out !
It's because the center-tap is PI/2 x Vdc due to peak/average value of a sinewave

"Q" refers to VAR/Watts (VAR in the primary capacitor)/(output power)

P.S. For the royer type that uses diodes & resistors I found that it's best to apply the gate-drive power before the main power, if possible. Makes for a very reliable startup.

The dc inductor also needs some consideration;
not only must you consider the peak current (core saturation)
you also need to consider ac flux density with frequency (core heating)

If you don't get Vpk = PI x Vdc (approximately) then you should investigate the problem.

Marko

Tue Sept 04 2007, 02:44PM

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

I'm so ashamed to receive such a great help and advice from you guys and not have time to work on this

I hope to do something about this this weekend at least, as I'm occupied by another (still confidential) project.

The Vpk = PI x Vdc is true!, I worked it out !
It's because the center-tap is PI/2 x Vdc due to peak/average value of a sinewave

And there must be some mathematical proof of it..? I really don't understand it.

What blocks the voltage from rising further with current source?

To work at all reasonably a zvs needs a "Q" of at least 2.2 (forgotten reference)
I would consider Q=3 low, Q=5 ok and Q=10 overkill (unless you want a 'good' sinewave)

I couldn't get a good clue of unloaded Q for this circuit, basically because of very low resistance of the wire I used, but it could be in over 100. With load it varies widely, but as I think it will never drop too low with air core. I only for sure know that characteristic impedance which is 20 ohms (3uH, 7.5nf, 1Mhz) if it means anything to you RF guys.

The dc inductor also needs some consideration;
not only must you consider the peak current (core saturation)
you also need to consider ac flux density with frequency (core heating)

If you don't get Vpk = PI x Vdc (approximately) then you should investigate the problem.

Never in this circuit I noticed the voltage to be much different from pi. I usually calculate the maximum inductance value for highest current I expect before it saturates. (B = uNI/l),

but this time I simply use 80uH inductor which is really an overkill current-wise, but in overall large variety ov inductances worked with circuit practically unaffected.

I could use some tips on this point.

Sulaiman

Tue Sept 04 2007, 08:44PM

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

I'll have to sit down and do the math again,
but the 'secret' to the math is - the current in the inductor is constant (almost)
so at the center-tap
the integral of volts.time below Vdc MUST equal the integral of volts.time above Vdc

In practice the current in the inductor is some dc value plus ripple
The change in inductor current is equal to integral of volts.time above (or below) Vdc
The value of the inductor does not (theoretically) affect the dc current.

You do not need to know the unloaded Q of the primary inductor//capacitor
only the primary VAR.
Example;
Vdc = 12 V, F= 25 kHz, Lp = 170 uH
so Cp = 238 nF and Vpri = 12 x pi / sqrt(2) = 26.66 Vrms
current in primary capacitor = 26.66 x 2 x pi x 25E3 x 238E-9 = 1 Arms
so the primary VAR = 26.66 Vrms x 1 Arms = 26.66 VAR

So the maximum output power is 26.66 Var / 2.2 = 12 Watts

In practice the dc inductor has resistance, so does the primary inductor
and the transistors drop some voltage
so you will not get 26.66 Vrms.

(my mouse has started double-clicking so I hope I haven't made a dumb mistake)

Marko

Wed Sept 12 2007, 08:08PM

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

OK, I'm dying of curiosity now.

This are stats of the circuit.

I'll have to sit down and do the math again,
but the 'secret' to the math is - the current in the inductor is constant (almost)
so at the center-tap
the integral of volts.time below Vdc MUST equal the integral of volts.time above Vdc

First are LC voltage and drain voltage of a single mosfet.
Even though circuit may not be in it's best, I measured the voltage to be almost exactly 40 volts.
I measured supply voltage to be 12,6V which *pi gives 39,6V!

This has proved constant for all parallel resonant royer circuits. I would appreciate a lot an explanation for this!

I returned to resistors for gate drive. 3 paralell 100ohm resistors do their job, although not too well, and are hot (5W of dissipation). Inefficient, un1337, but works.

Still the circuit worked and I had a lot of fun with it.

Now note the bad looking humps and apparent shoot through condition. This may be ringing, reverse recovery current of diodes, or both, but is anyway bad. The fact that I can't pull gates negative messes things up additionally.

In practice the current in the inductor is some dc value plus ripple
The change in inductor current is equal to integral of volts.time above (or below) Vdc
The value of the inductor does not (theoretically) affect the dc current.

Now take a look at inductor current. I attribute the uneven-ness to fact that inductances aren't perfectly balanced. The current varies inverse-proportionally to DC link inductor size.
Larger inductor makes current more smooth.

From other side, changing the inductor value didn't seem to affect weakly coupled output at all, so, is there anything I should look after in inductor design except bigger = better?

For the end, Bad news.

I blew two mosfets and still don't have much clue why. First gone when I messed with gate drive transformer. The transformer worked, giving some badly shaped sinewave-like waveform (at least it was going negative), until one solder point disconnected. I foolishly wasn't using clamp diodes and I think it blew the gate out. I never got pics of those.

I decided that transformer is a bad thing, too much leakage inductance no matter what I do and not worth of effort.

When I restored the circuit it worked happily with up to 70V input, like 50W of power and I had a lot of fun with it. Mosfets were hot, but not too hot and continuously operable.

At one point, I turned off the LV supply without disconnecting the 70V supply. When I turned it back on, a mosfet instantly blew - I have no clue what happened,
I'm so frustrated!

This circuit is a mosfet killer machine like my previous class E work.

Now I'm going from beginning with small (and free) mosfets, and examine some new possibilities.

I think I had a great idea - I would use separate isolated power supply between gates and FB diodes to drive the gates negative. This would also allow whole bunch of new possibilities, like true active pullup and preamplifier circuits, and *cheap* and *fast* feedback diodes like stack of series 1N4148's!

This would actually give a big new chance for the cirucit. If it's the only gate drive, then it's definitely still worth. I badly need to know your opinion on this guys... I hope I've done at least enough homework!

1189627693 89 FT30472 Isolated Pulldown Sources

Sulaiman

Thu Sept 13 2007, 08:11PM

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

Marko, I suspect that the capacitances within the mosfets are causing 'non-ideal' waveforms,
at lower frequencies the waveform on the drain shows the 'plateau' symetrically
at the beginning and end of each half-cycle, if you 'scope both drains the plateau of the mosfet turning off ends just before the plateau of the mosfet that's turning on.

I'm expecting some SUP75N06 mosfets in the post soon
I fancy trying a zvs in the MHz range, I'll let you know how I get on.
At the moment I have no useful comments because
you're working at a higher frequency with zvs than I ever have, so any new info is appreciated.

P.S. no main power without gate drive power

Marko

Fri Sept 14 2007, 02:34PM

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

Hi Sulaiman

I wouldn't use SUP75N06's for >1Mhz operation - they are quite slow, and I already considered 450's too slow for the job.

I actually thought to use small things like IRF730 in paralell which are very fast. For highest frequencies I even plan to use IRFU410 or d2n52's - they are small but insanely fast, and I have a bunch of them free from CFL bulbs.

Some of those small ones may even be good for a small hard switched design with direct feedback as d2n52's have 30V gate rating.

P.S. no main power without gate drive power

Doh, I did that for ages and never had a mosfet blow up with this circuit. I guess I'l have to use a relay to turn the main power on.

Tom540

Thu Oct 04 2007, 08:26PM

Banned on 3/17/2009.
Registered Member #487 Joined: Sun Jul 09 2006, 01:22AM
Location:
Posts: 617

Hey Marko,

i just realized you started this topic before I did. I remember the guys from MIT had good results with 1MHz but I'm thinking perhaps its because they only tried frequencies above that? I wonder how efficient a lower frequency would be? I was thinking of giving this a try at 100KHz instead. I had the same idea you did about using the ucc's in a royer circuit or some sort of hybrid as well when I found this post. This is driving me nutts. I'd rather not screw with mosfets in linear mode and I'd like to drive mine off wall voltage so that I don't need some monstrous transformer.

Have you had a chance to evolve this any further?\

-Tom

Marko

Fri Oct 05 2007, 12:17PM

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

Hi Tom

I did, but didn't post anything due to lack of progress and general lack of interest in the thread.

For a constant shared inductance maximum power you can transfer will be proportional to your frequency, so higher is better, to a point.

UCC's are a bit slow and suffer from high dissipation as frequency goes into Mhz range.

I believed that using diodes to pull the gate down directly which is way faster, but there is a problem of diode voltage drop. I devised some more ideas regarding that but never really tried them out properly.

I'd rather not screw with mosfets in linear mode and I'd like to drive mine off wall voltage so that I don't need some monstrous transformer.

With royer I basically said goodbye to mains voltage, as you would need 1200V switches for that.

There is although bridge ZVS topology which may allow use of 600V switches and much better performance in overall; all of that yet needs to be heavily researched.

Marko

Sat Nov 10 2007, 02:20PM

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

If I can ask some simple questions, since i'm stuck;

- I was told that royer will not be operable if Q is too low.
So that means, I should not use too large value of L over C if my series resistance/loading is too big. (Since it's parallel RLC Q is Rp*sqrt(C/L) )

But some other things are counter intuitive about that. Even after doing some simplest possible workbench simulations I actually got completely lost...

Q is ratio of stored and wasted energy, so how does it just come that huge C bridged with piece of wire is better in this case?

And if I want maximum ampere turns from the coil for wireless power, does this mean I should use *low* Q, more precisely large L and small cap? My resistance increases proportionally with number of turns, L with square but Q is dependent on sqrt of L so I seem to get nothing no matter what arrangement I use.

So it looks like I could use a single fat loop of wire and a huge cap which is completely counter-intuitive for me...

How do I get most of circulating ampere turns vs. power input into the parallel system? Small Q? High Q? I'm miserable...

previous 1 2 3 4 next

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