Gate resistor values for Mazzilli ZVS / High Voltage / Forums

Forums

4hv.org :: Forums :: High Voltage

« Previous topic | Next topic »

Gate resistor values for Mazzilli ZVS

Move Thread

LAN_403

nestor

Mon Jan 29 2018, 01:09PM

Registered Member #3331 Joined: Mon Oct 18 2010, 07:07PM
Location:
Posts: 4

Hi everybody,
even though this isn't about high voltage, I've decided to place this question here, because this circuit is often used to generate high voltage. But maybe it should be moved to "general electronics".
I've spent the last couple of days with fruitless attempts to figure out how the gate resistors in a Mazzilli-ZVS circuit have to be dimensioned, given the other system parameters (operating frequency, supply voltage, gate capacitance of the mosfet/igbt). All my attempts failed, and in an attempt to understand what's going on, I turned to a reference system that seems to work fine, namely Mads Barnkobs indutction heater:

http://kaizerpowerelectronics.dk/general-electronics/royer-induction-heater/

Turned out I have no idea how this system manages to work! Let me explain my problem: this is a ZVS-system, but the switches have to turn substantial currents on and off. So it should be most important that they spend only a small fraction of the cycle in a region where they are not fully turned on. The working frequency of the reference system is about 100 kHz, so the lentgth of one cycle is 10^(-5) seconds, which means that the length of one half cycle (which is the relevant time for us, because its the time one mosfet/igbt is on) is 5 * 10^(-6) seconds. Therefore, the gate of the mosfet should be charged to at least 10 V in a small fraction of that time. But this is not what seems to happen in this system. The effective gate capacitance of the used mosfet (IRFP250N) when charging it to 10 V is 12.3 nF according to the datasheet (Qg = 123 nC @ 10 V). With R = 470 Ohm, Cg = 12.3 nF, Uin = 30 V, Ugate = 10 V, and using the equation

t = -ln(1 - Ugate/Uin)*R*Cg

to calculate the time required to charge the gate to 10V, I arrive at a charging time of approx. t = 2.34 * 10^(-6) seconds, which is almost half of the on-time of one mosfet. According to my "understanding", this system should go up in smoke and not work at all.
The conclusion of this is: I simply do not understand how to find the proper gate resistance, and I am stuck. My aim is to design a Mazzilli-ZVS-oscillator for a target frequency of about 100kHz, using the IGBT IRGP50B60.

hen918

Mon Jan 29 2018, 08:14PM

Registered Member #11591 Joined: Wed Mar 20 2013, 08:20PM
Location: UK
Posts: 556

MOSFETs don't turn on linearly. The time in which it turns on is much smaller than the time to charge the capacitor.

I loaded up LTSpice to investigate, and the screen shot below should help with your understanding. The red trace is the output voltage, the blue trace is the current through the 470 ohm charging resistor for M1, the green trace is the current going into the base of the MOSFET. The flattish region on the on-slope of the blue trace is where the MOSFET goes from mostly off to mostly on, and vice versa on the right of the trace obviously. With the values used on the link you gave, that transition took 1.6 microseconds. Which is in the same order of magnitude as your estimation, which is good! but the circuit oscillates at around 10 kHz.

The other thing to note is that 100kHz is ridiculously fast for the Mazzilli. I have never managed to operate one over about 30 kHz. and I would think that IGBTs would have too high a current tail, especially with the high impedance gate charge.

I redid the simulation to test at 100 kHz, using a 0.5 uF capacitor. I was able to get fairly good results (it worked) but only below 90 W output power. above that you would get explosions.

Attempting to simulate IGBTs failed, even at 10 kHz. I was able to get it oscillating, but several components would have exploded in real life, and above 60 W things stopped working all together.

nestor

Tue Jan 30 2018, 01:37PM

Registered Member #3331 Joined: Mon Oct 18 2010, 07:07PM
Location:
Posts: 4

Hey Hen918,
thank you for your reply, it is much appreciated. It is very interesting that you had difficulties running this circuit above 30 kHz, because the system I've linked to (built with the mosfet IRFP250N) seems to operate happily at 106 kHz with a power throughput in the kilowatt range.

hen918

Tue Jan 30 2018, 03:46PM

Registered Member #11591 Joined: Wed Mar 20 2013, 08:20PM
Location: UK
Posts: 556

Using IRFP260 I can get a max power of about 500W at 100kHz before things become shaky. The switching time of the IRFP260 is about 0.6 us with a 470 ohm resistor. i.e. significantly better than the one I simulated before, which explains the improvement. IGBTs still no-go though :(

nestor

Wed Jan 31 2018, 01:04PM

Registered Member #3331 Joined: Mon Oct 18 2010, 07:07PM
Location:
Posts: 4

That makes a lot of sense, so there seems to be good agreement between theoretical considerations and practical results.
With respect to igbts, I still think it can be done. Apparently, this circuit was originaly developed by V. Mazzilli to drive a Tesla coil, and the original design seemed to use igbts. Just have a look at this reference:

However, I no longer think the IGBTs I originally planned to use (IRGP50B60) are good for this. The reason being their relatively high Vce in combination with their relatively low threshold voltage, wich could make it difficult for them to turn off properly. A better IGBT for this application might be the 40N65FL2, because it has very low Vce and a pretty high threshold voltage, so even by adding a generous diode drop, they should switch off cleanly. Apparantly, this fellow

has used them to good succes in an mazzilli-zvs based induction heater.

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