Author Options:

over 90W of flyback (back EMF) energy? how to dissipate it? Answered

Continuing my work on the flyback driver, I am investigating methods and practices to protect my MOSFETs from overvoltage conditions. I have measured peaks of as much as 350V across a 250V MOSFET while drawing an arc, and it does not matter what I do it seems like I simply cannot limit the back EMF transients to <250V. I am really impressed by the rugged nature of these MOSFETs, although I have popped quite a few of them. Seeing that they are being bombarded with almost 400V transients, it is a wonder they survive as long as they do.

When I added a medium sized reverse biased zener diode in series with a 10 ohm 1 watt resistor, both went up in smoke, especially the resistor! The zener diode survived. I also tried a much larger screw mounted 30V zener diode w/o the resistor and it reached much over 100 degrees C after only a few seconds and I had very little output on the secondary. Judging by how quickly a 10 watt resistor heats up I assume that the back EMF very low impedance and has to be over 90W average!!! I'll need to try to use the integration feature on my scope to measure the average power dissipation accurately. In none of the configurations involving a resistor did I see the voltage clamped any lower than before. :(


What's the current in the primary doing ? You should have a sawtooth, with no shoot through. The back EMF is high because there is nowhere for the energy to go.

I'm not sure if it is due to more parasitics than expected or that I am actually not operating in a purely flyback mode of operation, but Vds on my scope does not look anything at all what is shown in this application note.


Instead, it appears that I get a 400ish volt transient (it's pretty squared up too) for about 500ns just shortly after another 100ns 400v transient, and lots of ringing after that as the voltage stabilized to the supply voltage. I tried to get a screenshot of the scope but I could not get it to work. This is while an arc is being drawn. Surprisingly when there is no arc being pulled the gross overvoltage condition vanishes.


The waveform does not seem to change too much if I connect my 10W 10 ohm resistor and reverse biased schottky diode in reverse as a typical snubber network. And still that resistor gets very very warm!


DO you think that the reverse body diode in my MOSFET is clamping the voltage to 350 to 400v, and that using a higher voltage FET will not solve the issue, or will a higher voltage device be an effective workaround to the problem?

Until I see a current waveform, I won't be convinced by anything ;-)

You can see the expected sawtooth wave. Lets see what it is in the unloaded (open circuit) and loaded (delivering an arc), essentially short circuit.

I am not getting anything that looks like a ramp wave on the scope using the current meter, it appeared to be the simalar waveform as Vds. Very chaotic. Investigating that, I did discover the voltage on the gate drive is pretty crazy too, there is some bouncing between switching on and off. The fundamental is still there but I need to do something to get rid of the bouncing.


I replaced the 10W resistor with a 150W 120V lamp, again, did not help the voltage spikes much, but it was at full brilliance when the flyback output was completely shorted! :O Yeah, so far it seems like nothing in the universe will fix the damned problem!


It seems like adding a 0.47uF film capacitor directly across the MOSFET drain and source seems to reduce the voltage spikes to less than 200v, and the waveform became much more rounded. I'll try to measure the inductance of the primary and calculate a closer value to 40 KHz, the resonant frequency of my flyback transformer. Hopefully this will not hamper the audio quality too much.

Do you have a KNOWN primary inductance ?

4 turns of wire around the core of my flyback transformer, with 1 layer of electrical tape to create the gap in the core. That's all I know. That's why I said I'd need to setup an experiment to measure the inductance.

Alright, it seems like I got the circuit to be very robust and reliable now, thanks for the help! Here is a video demonstrating the capabilities! as well as the circuit diagram if you want:


The resistor is dissipating that back EMF 40,000 times a second, I called it a watt because it is energy over time. And the resistors I used get very hot!

How can I measure current with my scope? I only have 5W 0.25 ohm hollow ceramic wirewound resistors but they get way too hot after seconds of use in series.

Make a little current transformer.

Take a small toroid from something like an old PSU, wind a few turns on it: they're your secondary, and will go to your scope. Put a 1K resistor across the turns, so there is a load. There HAS TO BE A LOAD ON A CURRENT TRANSFORMER.

Put a wire carrying the current to be measured THROUGH the core.

Calibration is an exercise for the pupil. If its not sensitive enough, add turns.

The current transformer will be the equivalent of AC coupling the scope, correct?

No, it will be transforming current to voltage in a completely isolated manner.

I measured/calculated the inductance to be approximately 20uH with just the flyback transformer core. I could not make an accurate measurement when the secondary coil (rest of the flyback transformer) was included.

It appears that a 1uF or 0.68uF (or more precisely, a 0.75uF) capacitor is needed for resonance with my particular configuration.


However it appears that like in my last flyback driver video, adding this tank capacitor severely impacts audio quality, adds a lot of hiss and attenuates the music (particularly the high end) significantly. My guess is because the LC tank circuit oscillator is underdamped and acts a bit like a low pass filter, and of course plasma speakers are not good with bass.

I am driving the gate with roughly a 10v ~40KHz square wave whose duty cycle is modulated with music. The voltage powering the transformer is about 16v, but I want the circuit to operate between 12v and 30v. Reliability seemed to be fine at 16v even with the huge voltage transients. Also it seems like most failures occurred when adjusting the overall duty cycle to slightly over 50%.

You've got to make sure that at your maximum duty cycle the primary inductance can't saturate, or is on the very edge. I am sure you are saturating the core at high ON duty cycle.

Saturation is definitely a concern! How can I tell if I am saturating the core? Sound quality seems to be best at 25-30% duty cycle, and 50% allows the hottest thickest arcs but with audio distortion. I think this might be a result of saturating the core.


I removed a single turn so now there is only 4 and added a 150nF film capacitor, now I am getting impressive sound quality and the capacitor isn't hampering the audio quality anymore! It is a much smaller value than needed for resonance (at the fundamental), but enough to soak up the high voltage spikes. Vds waveform appears like one big hump followed briefly following by 2 smaller humps. I'll screenshot it for you later. The maximum voltage appears to be only 120v and I can easily start a hot arc at 3 inches! Also it seems like the MOSFET is staying much cooler than it was before, probably because that body diode is not absorbing/dissipating all that back EMF anymore lol!

I believe the hissing was a parasitic effect not directly due to the addition of the capacitor, but some sort of unwanted feedback from the high voltage secondary into the oscillator of some sort, likely into the long audio input lead. Selecting certain value capacitors will cause lots of hizz, but also at a particular arc length. :-/ Very strange, and I don't have a clue to any of the weirdness. But thank you for all the help!

You can tell if you saturate the core with the CT I suggested you make. You'll see current ramping up in the inductor, if though, it saturates, the current will shoot up to the Supply volts/ DC resistance of the winding, because L falls to zero.

Given your drive voltage, and your inductance, calculate roughly how long it takes to reach a saturation flux density in the ferrite. Your switching frequency must be higher than 1/that

What I mean is that a current transformer will not be sensitive to DC, unless there is enough net current to saturate the core I guess.


2 years ago

Consider a trifecta of three flyback outputs linked by three primary transformers linked at 1/3 your FQ... Now that waste energy can be used to pass pre-energize the next stage instead of stressing the one turning off.

Can you give an example circuit of what you mean? Do you mean to add a small choke or inductor in series with the flyback transformer? I added a small random ferrite core between the positive and the primary but this just made the audio very hissy and did not solve the problem.

So are you using a single fet drive on then off or a push pull configuration ?

The first one you mentioned. It is not synchronous.

When one FET is turned off the excess energy should transfer to the other side by magnetic flux.

The diodes, if you use them, will allow a prolonged current to flow as the magnetic field collapses before you fire the other side.

When no diode is present the inductive collapse attempts to raise voltage in an attempt to continue the current you interrupted. This is why diodes are placed across relay coils, to limit inductive kick-back from damaging the circuitry..

You may want to also play with rotating one set of primary leads...

I leave gate drive to your accomplished hands...


When I add diodes in the purely flyback configuration, the result is a 1mm arc instead of 3 inches. :( I have tried the dual primary configuration as well but the problem was using 1 heatsink to dissipate the heat and the thermal bottleneck those sil pads are. Also tried to use mica insulating pad things with thermal paste which was slightly better but the MOSFETs still stopped working once they got hot enough for the wires to melt off.