# 3 mosfets in parallel on Induction heater.

#### gophert

##### Well-Known Member
Let me help him with schematics, design and theory of operation...

http://bfy.tw/HRBF

#### unclejed613

##### Well-Known Member
oops my bad... induction heating for metals uses 20-50 khz. induction heating for materials like silicon, and dielectric heating use 6, 13, and 25Mhz

#### gary350

##### Well-Known Member
I finally repaired my broken crow bar. I had to make a different coil that is not round, it is 1/4" wider in 1 direction than the other with 8 turns. I made another round coil then squeezed it a little bit flat in the vise to make it narrower on 1 side and wider on the other side. It takes several minutes to heat metal up to a very dull red color I am surprised this is hot enough to hammer metal into shape. I had to re heat metal over and over many times & finally got the broken off tip hammered out close to what it once was. I had to cool the 8 turn coil with water. The big 3 oz heat sink takes a long time to warm up. I shaped the tip with my electric grinder & sander. It won't win a beauty contest but it is good enough for me as long as it works. Mission accomplished it only took about 4 years. LOL.

Last edited:

#### gary350

##### Well-Known Member
This is the final circuit.

#### large_ghostman

##### Well-Known Member
Have you got a pic of the repaired one?

#### gary350

##### Well-Known Member
Have you got a pic of the repaired one?
This is the repairs end I'm not sure why this straight end got broken the other end with the 30 degree angle bend is the end I use to lift machines. About 3/4" of the tip was broken off. I used the grinder to make the end square then heated it and hammered it out to a chisel point again. I use this to move my Bridgeport mill & lathe when ever we move to a different house. I hope we are never moving again new house in on 1.5 acres & work shop is 30'x50'. I pry the Bridgeport mill up on 1 side then slide 1/2" round solid steel rods 4 ft long under the machine so I can push it out the door to the trailer then haul machines to the new house and unload them 1 by 1. It usually takes 2 hours to move this mill from 1 house to another house. If I had a larger trailer to could move everything in 1 load.

Last edited:

#### tcmtech

##### Banned
Someone mentioned higher voltage I can rectify 120 VAC = about 170 VDC and voltage doubler = 340 VDC. I know I have a few 1500v 5a mosfets & might have some in the 400 or 500 volt range.
For a marginally designed circuit that you are trying to push ~ 2 KW though you're going to need switching devices rated for way beyond 5 amps!

In the better designs commercial units they run switching devices with 3 - 4 or more times over current ratings so if you're running 2000 watts at 160 VDC you will need switching devices capable of at least 35 amps to handle the theoretical ~ 12 - 13 amps RMS current level. Voltage wise 2:1 would be fine.

#### unclejed613

##### Well-Known Member
one thing missed by looking at the data sheet for transistors, is there's actually a maximum load line. if a transistor is rated at 1500V and 5A, doesn't mean it can switch 75kW. there's a SOA (Safe Operating Area) for mosfets, and if you go beyond it, the transistor will fail. this is especially true with a reactive load (inductors and capacitors) because the voltage and current are not in phase. here's a graph from the data sheet for the transistor you are using that shows the SOA.

for a resonant switch, the area under the 10mS pulse width and the Rds-on portions of the curve are the safe operating area (fortunately, MOSFETs don't have a secondary breakdown region to worry about). you can see also that the transistor can exceed the maximum E vs I portion of the curve for very short amounts of time. good design practice, as TCM mentioned uses devices of double (or more) capability. with reactive loads, the amount of "overkill" should be more like 3 or 4 times the expected voltages and currents. those are general "rules of thumb", and will improve the reliability of the circuit.

another approach would be to have the gates of the MOSFETs driven by an oscillator, rather than rely on the self oscillation. even though the self oscillating circuit is simple, you have seen that you can't use parallel devices easily. driving the gates with a fixed frequency signal would allow parallel devices without the problems you were experiencing. you could also use a logic level to switch the gate signal to turn the output of the device on and off. since the loaded frequency was measured as 64khz, that would be a good frequency for the drive signal.

#### tcmtech

##### Banned
driving the gates with a fixed frequency signal would allow parallel devices without the problems you were experiencing. you could also use a logic level to switch the gate signal to turn the output of the device on and off. since the loaded frequency was measured as 64khz, that would be a good frequency for the drive signal.

In a more ideal situation where the workpieces inductive characteristics are known and fairly constant from one to the other I would agree on the fixed frequency design but with something that may see a wide range off work materials I would go with the dual function adjustable driver circuit where both the frequency and its PWM levels can be set individually.

That's basically how the modern inverter welders control things in their power supplies. With the right feedback loops it's easy to make constant current or constant voltage output using little more than frequency and PWM adjustment which in a induction heater would give far better tuning ability and function (shallow to deep heating penetration plus variable power level) over the widest range of materials.

#### gary350

##### Well-Known Member
After reading a lot of information I have several questions. Assume I take my existing circuit and replace the mosfets with IRFP460 500v 20a, also replace the power supply with 170 VDC. I assume the circuit will operate exactly as it do with the same 8T coil and 8 capacitors resonance frequency will still be 102 KHz with no load on the coil? I assume the 4uh choke coil does not need to be replaced either? I may want to test this to learn this myself.

Next use a 4047 Flop flop or IR2153 to drive the mosfets on/off at 102 KHz the circuit will still work perfect?

I was reading where it said, iF I put a load on the coil and frequency drops to 90KHz the IR2153 needs to be adjusted to the same frequency 90KHz. It seems to me that will drove a person nuts because as the metal part enters the work coil frequency slowly changes, there needs to be a frequency meter on the work coil and the IR2153. If IR2153 is never adjusted to match the resonance frequency of the work coil will the induction heater still work?

I am slowly working my way in the direction of the H circuit. I will be a good learning experience and FUN to experiment with all these changes before I go straight to the H circuit.

Here is a circuit with IR2153 driving the mosfets.

Some of the driver circuit drawings show no capacitors connected to the resonance coil?

Last edited:

#### gary350

##### Well-Known Member
Hello Gary, Please read over this article from: instructibles.com, i hope it helps, i built this circuit and it has never failed since: https://www.instructables.com/id/MONSTER-OF-ARC-BY-ZVS/
I have not been able to make multiple mosfets in parallel work, 1 mosfet takes the full load and explodes then they all explode. Several people have shown pictures of mosfets in parallel but my mosfets in parallel never work. I have seen photos of 10 mosfets in parallel.

#### dknguyen

##### Well-Known Member
I have not been able to make multiple mosfets in parallel work, 1 mosfet takes the full load and explodes then they all explode. Several people have shown pictures of mosfets in parallel but my mosfets in parallel never work. I have seen photos of 10 mosfets in parallel.
Gate resistors (for each FET right next to the gate or else it won't dampen ringing)? Gate driver? Flyback suppression? What are you building it on? Layout matters. Dyou have an oscilloscope?

Last edited:

#### Timothy_Egoroff

##### New Member
Hi Gary, if mosfet chips don't work in parallel (with your set up), the other option is using IGBT chips such as IXXK200N65B4 or IXXX200N65B4 this will work in parallel in definitely (theoretically).

#### dknguyen

##### Well-Known Member
Hi Gary, if mosfet chips don't work in parallel (with your set up), the other option is using IGBT chips such as IXXK200N65B4 or IXXX200N65B4 this will work in parallel in definitely (theoretically).
Don't IGBTs suffer from the negative thermal coefficient that BJTs do? (i.e. can't parallel without taking special measures since the one that is carrying the most load gets hotter which decreases it's voltage drop which makes it take even more of the load in a positive feedback loop until it's taking all the load).

EDIT: It seems that IGBTs can have a thermal coefficient that varies with Vge which means if you run it at a Vge that results in a negative thermal coefficient...

#### Timothy_Egoroff

##### New Member
yes, this is called thermal runaway, to combat this, a power resistor with a very small resistance which will prevent over current should be placed in series with the source (or in this case the collector) of each igbt, so now no matter what only a set current will be able to pass through each igbt, (I=V Divided By R)

#### gary350

##### Well-Known Member
I would like to build the H circuit but the IRS2453 is $15 for 1 plus postage in USA but only$1.84 from China it takes a month to arrive. It seems to me the 555 master & slave flip flop circuit can be used with 10% off cycle between each one so they can never both be on at the same time.

#### dknguyen

##### Well-Known Member
I would like to build the H circuit but the IRS2453 is $15 for 1 plus postage in USA but only$1.84 from China it takes a month to arrive. It seems to me the 555 master & slave flip flop circuit can be used with 10% off cycle between each one so they can never both be on at the same time.
For the large MOSFETs (high gate charge) and frequencies I imagine you using, the 260mA gate drive current of the IRS2453 does not seem anywhere near enough