Why does my Royer high voltage generator switch to far higher frequency when loaded.

[martin de jong]

Ive designed a circuit that should efficiently make a 10 Watt 3000V DC signal, but when I connect the output to my Delon voltage doubler, or even to a simple rectifier circuit instead of just a (1Meg) resistor, the Royer circuit stops oscillating at its intended frequency (of about 70 KHz), and suddenly switches over to a much higher frequency (about seven times or 500 KHz), which brings down efficiency greatly to about 15% from about 80% when I simply load it with a resistor.

The strange thing is that I cannot simulate the problem, the circuit below shows the LTSpice simulation of my circuit, but it always simulates oscillating at the lower frequency with good efficiency.



I'm starting to become desperate for a solution, or a theory why the royer oscillates at the higher frequency if the load resistance becomes less than say 100 Meg Ohm, without any load it works, but the moment I load it even slightly the strange effect happens, and efficiency goes down the drain.

Note my real circuit under test now doesn't use the PWM regulator, and I simply put something like 3 or 4 Volt between M3d and Vin, so the PWM circuit doesn't affect the royer.

Current theory is that the transformer somehow becomes saturated, but I don't know how that could happen, the Transformer used is rated 14 Watt, so I'm not using it out of specs. also for the frequency to become seven times higher as it does, the inductance (or capacitance)in the LC loop should become the square lower or 7 x 7 = 49 times lower. The royer still generates a sinusoidal signal when it oscillates at 500 KHz, so I believe its still resonating, only with a 47 times lower LC.

Can anyone shine some light on this problem? Why does the way I load the output coil of the Royer matter for the frequency the royer oscillates? In the simulator I can "tweak" the frequency a bit, but it goes only slightly down, not massively up.

 

[duke37]

Please describe the operation of a Royer oscillator. What little I looked at suggested that the transformer runs in a saturated condition. The circuit I saw had two capacitors for feedback, You do not have these, presumably oscillation is due to the capacity of the diodes.

A TL494 will run at a constant frequency, will not saturate the transformer if run fast enough and has a dead time to stop both fets being on at the same time i.e. no shoot through. Efficiency should therefore be higher than the Royer.

 

[martin de jong]

Its a resonating (resonant LC tank) "resonant royer" which can operate at very high efficiency, it not a power converter, so the TL494 will not be useful. Normally the transformer does NOT go into saturation, like normal royers do.

I had hoped I could find an expert on resonant royers here who could tell me if this is a common problem with resonant royers.

 

[duke37]

You say it is not a power convertor, you have proved this in practise.
To get a decent amount of power out, you will need to drive the gates hard but you have no feedback capacitors, only the capacity of the diodes,
What mechanism is present to turn the fets off?

 

[martin de jong]

LOL.
I meant a power converter as the common device that say converts 12V DC to 110/220V AC @ 60/50 Hz.

The gates are driven ON by the pullup resistors, and are driven OFF by the diodes, the diodes are solely there so energy on the resonant tank circuit isn't lost through the pullup resistors, as the voltage on a branch of the resonant tank circuit (the connection point of the coil and the capacitor) swings above Vin.
If the output of the circuit is simply connected to a resistor, the circuit is efficient in converting power, so it already IS an efficient converter of power, the only problem is that it looses efficiency when it starts operation in a wrong mode where the resonating frequency switches to a frequency at which the transformer cannot operate well. the transformer is rated for a maximum of 100KHz and the resonator starts resonating at 500KHz

I'm trying to find out the mechanism that causes the resonant loop to start resonating at a far higher frequency than its natural resonance frequency. Its still resonating, its not in some other operating mode, (like relying on the coil saturation as a non resonating royer does, which produces a square wave) as the signal is still purely sinusoidal. Meaning that either C or L has reduced dramatically, but the question is WHY?

 

[duke37]

I would try a couple of 1nF caps across the feedback diodes to push the fets a bit harder.
The very low diode capacitance will be more effective at high frequency.

 

[73's de Edd]

Sir martin de jong . . . .


You could do better on your feedback technique.

Consult . . . . .Pages 38 . . . . 39

http://digitalcommons.calpoly.edu/cgi/viewcontent.cgi?article=1266&context=theses

Reading all of the whole data base will relate to several coincident problems that you are experiening.


73's de Edd

 

[martin de jong]

This is great! Just what I was looking for, it will help a lot!
many many thanks....!

Martin de Jong (AKA Mahjongg)

 

[martin de jong]

In the mean time we are making progress, we managed to fix (at least for now) our major problem (startup with the wrong frequency of oscillation) and a few other smaller problems, one of our main concerns is now that the efficiency is still nowhere close to what we want (which is better than 90%) We are also not at the 3KV we would like to reach, but increasing the output voltage is doable, we are now simply regulating to 1600 V.
The lost energy seems to be consumed by the transformer, its the one component that gets hot.
We are still studying the white paper to see if it will help us. Next up is increasing the output voltage while remaining a stable feedback, while not re-introducing the wrong frequency mode problem.
In actuality our feedback (voltage sensing) circuit is a lot more complex than in the simplified LTSpice model shown here.

 

[73's de Edd]

Considerations and unknown attributes . . . . .

That is being a hefty level.

Is your transformer core mass being properly sized to the power level at hand ?.

If ferrite based, is the proper core permeability at a prime frequency being considered.?

If operational resonance is being in the post ultrasonic----low RF spectrum . . . .100Khz---2 Mhz
Has utilization of Litz windings been a due consideration ?

Is the transformer design of:

• Toroid,

• Cup pot core

• Enclosed pot core

. . . or forbid . .

• Ferrite .rod construction ?

73’s de Edd

.

 

[martin de jong]

No, it has to be low cost, low weight, and producible in some quantity so we opted for a commercial 14 Watt capable CCFL transformer, with an E-shape ferrite core. We found that using 10nF caps for C2 and C3 was saturating the core, which what was driving the resonating frequency up massively, so we lowered these to 1nF, and that helped.
Also the choice for ultra fast switching HV diodes also helps, as does doubling C1 to 200nF.
I have now built a frequency compensated test probe, so we can watch the HV signals on an oscilloscope, both for DC and AC (80KHz) signals. we are now generating 1600V DC at 2.5 Watt, and now ramping that up, hoping to reach 3000V DC at 9 Watt.