The classic IRFP250N vs FDP33N25? Which one is best for high power SMPS?

I have an idea to start selling some powermax singing arc kits. In the kit I plan to include all the parts including a printed PCB to create an audio modulated flyback driver. I figure to start off with, since I have no clue what demand would be, I want to play it safe and only get enough parts for 10 - 25 quantity, which works out to be about $40, or about $4 a piece, not including custom PCBs. However, at >$2, the IRFP250 is not cheap even in 10's of quantities on Digikey.

Looking to shave off pennies, I found the FDP33N25 for about half the cost of the classic IRFP250N and best part is that it appears to have simalar specs, and in some ways it looks like the FDP33N25 is even better! Bolded parameters are the (I think) prefered values. I do not know much about selecting the best transistor for the job, but I did compare the specs below:

IRFP250N: 85mΩ Rds ON, 30A, 200Vds, 2800pf gate capacitance, TO-247, 0.65 *C/W

FDP33N25: 94mΩ Rds ON, 33A, 250Vds2135pf gate capacitance, TO-220, 0.80 *C/W

One thing I am concerned about power loss and therefore switching speed, so the MOSFET must be driven with a 15KHz to 40KHz square wave with very sharp rise and fall speeds. I was able to reduce power dissipation in my last circuit by simply shortening the wire carrying the signal to the gate, as well as adding a complementary pair of NPN and PNP transistors to the output of the 555 timer. That seemed to lower the drive impedance and switch the mosfet more rapidly, but considering that I still need to use active cooling and a brick sized heatsink, I think I can do better. Any suggestions?

sort by: active | newest | oldest

From my experiments with my induction heater I learned one thing very fast...
If you feed a high frequency signal to the gate and also need extreme currents to be switched you really want a low internal resistance and a strong gate current - or gate transformer.
But single most effective way to get maximum power output and only ver low heat up on the mosfets is to drive the circuit in resonance.
Here the mosfet is switching no loads at all (in the ideal world) and the most power comes out of the tank circuit when the two mosfets are switched over.
In such a config it is only important to have the internal resistance as low as possible and that the mosfets are rated for four times the input voltage.
The way I know you it should be no problem to design a TL494 based circuit with an additional coil to keep the resonance.
Did not check the prices but something like a NTP60N06 or the clones might be a better solution.
And you only need to go outside the audible frequency range for a singing arc, so why go all the way to 40kHz?

-max- (author)  Downunder35m1 year ago

A) The arc is only capable of reproducing high frequencies, Typical audio sample rate is 44.1KHz (sometimes 196KHz), so it would be nice to drive the flyback at those frequencies for good audio fidelity. However, I gave a large range because every transformer is different and the circuit needs to be tuned into resonance with the transformer secondary for maximum voltage output.

B) The transistor you listed does not have price info on digikey, and is only rated for 60V, 4 times my 24V power supply is 96V, so I don't think it will be suitable for this circuit.


You mentioned ZVS switching, but how would I audio modulate that? I have messed with that circuit, and I can get some fat arcs out of my flyback transformer with any heatsinks!! However audio modulating it is a bit more difficult, I did make an attempt by injecting an audio signal into the power rail but it was very quiet. I have also considered modulating the magnetic saturation of the flyback core but never got around to trying that, as maintaining a magnetic feild in a small coil will cause it to heat up significantly.

However, operating in a resonance mode is different from the flyback mode, correct? I like the 555 driver for being able to generate really high voltages, long thin arcs, and loud audio with a relatively small input voltage. So much so that 24V is all it took to cause dielectric destruction of 2 transformers lol! However, maybe I can get the best of both worlds and connect 2 MOSFETs to the ends of a center tapped primary simalar to the popular ZVS circuit and drive the MOSFETs with that TL494???

That last bit was my thinking exactly ;)
The ZVS circuit is powerful but as you noticed not really capable of audio induction.
But if I am not thinking totally wrong adding a second primary winding that carries the audio signal should be enough to change the magnetic field.
Was thinking of 4-8Ohm and driving it directly from a tiny amp.
And I suggested the Mosfets for 12V use as this should be sufficient, but they should work fine in your circuit.
Should be possible to make a highly efficient driver circuit for the flyback that does not require several amps to run.
Those tiny things for the plasma globes work fine for smaller arcs and some already offer a mic input - with a change in sensitivity and maybe a slightly bigger transformer...

-max- (author)  Downunder35m1 year ago

My fear with adding a L3 coil to the flyback to inject an audio signal directly into the core is that the strong dϕ/dt on that core will induce an EMF right back into any device connected to it, potentially destroying it. I wonder if I could do something simalar to what a dimmer switch does, and build a circuit that phase locks with the oscillator and abruptly forces whichever transistor happens to be conducting at the moment to turn off, simalar to how those SCR based dimmer switches work.

There are some nice schematics out there in several variations but not so many with an audio input.
Was thinking of the audio problem and interference, ever considered adding the audio signal to the DC that is switched by the mosfets?

-max- (author)  Downunder35m1 year ago

I don't think it is my MOSFET driver circuit that is limiting how fast the MOSFET switches, as I am achieving a slew rate of about 90 MV/s according to my osciloscope. The waveform is very square in shape. Strangely when I apply power, the waveform on the gate becomes more chaotic, which I guess is expected, I think I need to "shorten the loops." as the wires from the MOSFET are several inches long and stray inductance is playing a role.

-max- (author)  Downunder35m1 year ago

I think that was how my original design worked (effectively a class A amplifier) , then I switched over to adding a secondary winding to the inductor between the center tap and Vcc, and driving that with a class AB amplifier. Both scenarios are incredibly inefficient, as my ZVS driver can pull up to 6A at 12V. Also this requires additional power transistors, which are not cheap.

-max- (author) 1 year ago

Why is it that the IRFP250 has a lower Rds on resistance and a better thermal conductivity (between junction and case) but is rated for less current? I guess the actual on state resistance depends on more factors and the figure given (tested at 10Vgs w/ 18A drain current vs 10A of current for the FDP33N25)

iceng -max-1 year ago

Thanks for the spec sheet pointers. IR iso-drain 19A @ 100`C using hexfet construct vs Fairchild 20.4A @ 100`C unknown construct... While transconductance

are spec games...

Anyway, where it matters they are the much the same.

-max- (author)  iceng1 year ago

Ohhh, "transconductance" big words, me like! Maybe I get to frighten someone away from electronics with that word lol!!! :P

Can you please explain what that means?

I guess I will order a few and see if they are suitable. I have not had much luck with the smaller TO-220 FETs in the past with this project.

iceng -max-1 year ago

Whenever you meet a push/pull mosfet amp or audio power arc singer, after you choose a fet then you then want to test which two devices have the closest CLM

Channel Length Modulation the modern term for transconductance of the old, and now becoming new again, as vacuum tubes for amplifiers are growing and manufactured in Russia.

You would never replace a single dying tube but always the push pull as a closely matched and ballanced set.

-max- (author)  iceng1 year ago

So transconductance and CLM are just fancy terms for being in the linear region???

iceng -max-1 year ago

More like gain change as bias shifts and you want both gain bias slopes to be the same (identical) for best balanced audio.

iceng iceng1 year ago

Actually the linear region is just that. Trans-conductance is the effect a voltage device exhibits as its impedance changes with the applied bias or gate voltage..

A snazzy engineer could use that effect to make a better bipolar oscillator using current fluctuations.

I don't ever remember if it was normal or inverted, but you have to admit that it is desirable to match Transconductance for push pull applications.

-max- (author)  iceng1 year ago

OK, so it is kindof a way of to describe the dynamic resistance as a function of gate voltage, or more specifically the gain at any particular Vgs voltage. It sounds like a representing the linearity of a device.

However, what I am more concerned about is switching losses. I am using either PWM or FM modulation (probably not AM modulation as that requires me to deal with that power transistor in the linear region) to audio modulate the circuit. How can I reduce (what I believe) is mostly switching losses? (because the same transistors used in the ZVS circuit seem to reduce power loss in the MOSFETs a great deal.)

What factors limit how fast a MOSFET can be switched on and off? When I measure the voltage on the gate without power applied across the flyback transformer, I see that the the gate input is very square in appearance. It does not look like I should see much power loss. Even once I apply power to the flyback, I still have sharp rise and fall times, although I see a bit more harmonic junk and a few really high frequency ringing artifacts on the flat parts of the square wave. This might have to do more with how I was probing it than the voltage present, IDK. Any ideas?

iceng -max-1 year ago

Besides driving a gate hard to overcome gate charge time capacitance, keep very short wire lines (Iv seen a 6" ten strand straight wire inductance add a 100v spike at a very fast, too fast, turn off time) and consider litz or many stranded welder wire to force even current distribution in your power runs...

Different manufacturers use different gate geometries which should decide how fast gate rise time travels activate the channel...

Obviously as the driven switching PWM or FM increases fq it will heat up.. Resonance can run cooler but I do not believe you can do music with resonance.

-max- (author)  iceng1 year ago

Yeah, short wires for sure! I noticed that when I went from using a small 6 inch breadboard jumper wire for the gate to a much shorter 2 inch wire, my MOSFET ran noticeably cooler. where before it took only a minute of operation to get to a point where the MOSFET got got enough to melt the soldered wires right off, to only getting very very warm on my CPU heatsink. When it comes to litz wire, wound 10 or so strands of 30AWG wire cut the mustard? I made it in the past with only 6 strands of thicker wire but did not notice a significant difference.

iceng -max-1 year ago

Multi strand wire 40 or more allow insulated wire to use less space in hi-fq xfmrs to use smaller gauge wire because you use the full cross sectional area of a wire rather then the outside edge...

That's why RF does not kill you by flowing on the outside of your skin...

-max- (author)  iceng1 year ago

The skin effect is certainly a strange one! So I guess 10 strands made by hand is not really worth the effort, is it. :( I guess at 20 KHz to a few MHz range it is not too big a deal. I wonder how well a tesla coil secondary would work with litz wire, ummm.... Thanks your your answers iceng, I would have given you best answer award but you never "answered" my question, just reply to my own "answer" lol! I wish the Instructables commenting section is sooo pathetically broken!

-max- (author)  iceng1 year ago

It seems that a few sites and videos are saying that transconductance is kindof a parameter of how "good" the MOSFET is, or it's gain. (Vgs-Vthres)/Id. So would a MOSFET with a better "transconductance" basically turn on "harder" with a lower voltage?

-max- (author)  iceng1 year ago

I noticed the cheaper MOSFET actually has almost half of the output capacitance, and a much smaller miller capacitance. What I don't know is if that would significantly improve switching losses?? I was reading an article about a 4MHz SSTC design, and to achieve ZVS, he added a damped tank circuit to the primary so that the damped ringing would cause the voltage to come back down close to 0 just in time at the same time the MOSFET gets switched on. It was an interesting design! I wonder if my circuit would benifit enough from such a configuration?