Royer induction heater and high current power supply

[petespaco]

What do you gain with the 100 ohm resistors?

[RocketScienceSmurf]

My assumption was faster rise time for the gate signal. I treated the gate part more or less as a voltage divider with 100 Ohm fixed resistor and the mosfet gate as a dynamic value with a large resistance when not charged. I haven't been able to do any measurements to confirm this as my oscilloscope does not have the required bandwidth but it seem logical.

At the current levels I am running it is crucial to keep the time the mosfets operate in the linear region to an absolute minimum.

[petespaco]

I agree that you don't want the Mosfet to be in the linear region any longer than they have to be.
  But, in my limited experience, as long as you are using 48 volts or so, the original design is pretty good.
that gate ckt appears to be more than  a simple  voltage divider to me.  The gate itself draws almost no current at all.  The apparent "voltage divider" would be the 470 ohms and the 10K ohms (in the original ckt).  So the difference there is 20:1 vs 100:1.  But that's not the end of the story.  when the zener gets forward biased, you now have  36 volts across it.  Now you get some heat.
But there's more, isn't there?
When the commutating diode forward biases, doesn't it essentially ground the gate to turn that Mosfet off?
If so, now you have 48 volts across that 470 ohm (or, in your case, 100 ohm resistor.
Here's how I see the power at that resistor.  Please correct me, someone, if I am wrong:
      Commutating mode (gating off)         
Voltage applied to R      48   48      
            R                  470   100                     
                 I=               0.102    0.48      
            P=                4.90   23.0   Watts   
               
      Gating Charging mode         
Voltage applied to R      36   36      
                         R   470   100      
                        I=     0.076   0.36      
                        P=    2.76   12.96  Watts   
               
Note: divide by two if you want, because each Mosfet is Off               
Half the time.
               

[RocketScienceSmurf]

I think you may be partially right. To my knowledge, once the mosgfet gate is charged it consumes almost no power at all, but to turn it on from an lmost non-charged state it takes quite a bit of power to do that quickly. The capacitive load of the mosfet gates basically presents a dead short initially.

You are right about the power dissipation from both the zener and my power resistors, hence the much beefier components. My resistor is a 20W TO-220 resistor that I actually used water cooling for and my zener diodes are quite big ones rated at 8W.

I'm not sure how you calculated 36V across the zener. Won't the resistor make sure the voltage never exceeds the zener voltage?

[Mads Barnkob]

I think you should really get some measurements done instead of just assuming that switching faster is better. It is a general rule of power electronics that you shall try to switch the current as fast as slowly possible. Meaning that too slow you have higher losses, but too fast and you have excessive ringing.

I do recall that when going up in voltage, someone else used 680 Ohm resistors instead, exactly to lessen the heat dissipation.

[RocketScienceSmurf]

Excellent point. I never thought about the possibility of ringing. To measure that I do however need a better oscilloscope.

I think I will focus on making a better assembly of the whole thing. On closer inspection I realized I melted the solder on the pcb via connecting all the caps, the mosfets and the standoffs for the working coil. Since I didn't have the board flat, the solder flowed to one side of the board leaving the via much thinner in one section causing the via to break and the last connected capacitor leaving the pcb with violent force since the solder was melted.

What bandwidth would you recommend for measuring the gate signals? I don't think I will ever use it at frequencies above 1MHz.

[petespaco]

A quote:  "I'm not sure how you calculated 36V across the zener. Won't the resistor make sure the voltage never exceeds the zener voltage?"
Answer: No.

It's not 36 volts across the zener.  The zener clamps the gate to 12 volts, right?  If so, the resistor sees the difference between the supply and 12 volts.
eg:  48 (if that's what you are using) - 12 = 36.

By the way, I don't see any ringing to speak of using the "standard" values.
Regarding frequency:  I doubt that you will be running the tank at any higher than about 150 Khz unless you start removing capacitance and/or making coils shorter than about 6 turns or smaller in diameter than about 1" ID.
My little DSO-112 pocket "scope can see what's going on pretty well, but I use my ancient Techtronix 422 15 Mhz scope most of the time.  It was very conservatively rated, though and still has good response at 30 Mhz. (From back in my 27 Mhz radio control days.)

[RocketScienceSmurf]

when the zener gets forward biased, you now have  36 volts across it.  Now you get some heat.

I don't understand what you mean by this. I thought that the resistor in combination with the zener diode made sure that the resistor never sees more than supply voltage (about 40V in my case) and that the zener never sees more than the zener voltage no matter what.

When I wrote 1MHz I meant frequencies for other possible future projects, not induction related. This was just to help decide what scope to maybe buy in the future, but I think I will borrow a 2 channel Hantek from my brother instead just to check the gate drive signal.

[petespaco]

Sorry at about this:
"But that's not the end of the story.  when the zener gets forward biased, you now have  36 volts across it.  Now you get some heat."
should have read: "But that's not the end of the story.  when the zener gets forward biased, you now have  36 volts across the 470 ohm resistor.  Now you get some heat.".
  I guess I had my mind stuck on your change to 100 ohms.

[JardCrocker]

Hi...i am a new user here. I want to knowwhat kind of switches will you be using to handle all that power ? I used to push 30A constantly through a set of irfp250's which didnt last very long. After pushing the voltage higher than 40v I had to move up to IRFP460's and as a last resort I switched to IGBT's.

pcb assembly

[RocketScienceSmurf]

My plan is to use a pair of Infineon IPT007N06N rated at 300A each. The current just when the circuit is closed doesn't seem too bad, but when it starts to oscillate the current draw is insane at first and it slowly stabilizes at a more reasonable level.

[Mads Barnkob]

Hi Jard and welcome to the forum If you have the time, please make a new thread about your induction heater and show us what you have built.

Kamelryttarn, it has been a while since your last update, is life keeping you too busy for making progress?

[RocketScienceSmurf]

Kamelryttarn, it has been a while since your last update, is life keeping you too busy for making progress?

Yes, I'm afraid so. I know what I would like to try next, but I can't seem to find the time and money for it. Probably my next attempt at this project will be in a few months.

I saw a nice Celem cap on ebay the other day but it got sold pretty quickly. It as a 5uF 350B: http://www.celem.com/Item.aspx?ItemId=33&cid=11&hmenu=29

[Mads Barnkob]

If you are interested, we can find a reasonable price for one of these https://www.ebay.co.uk/itm/232565478339 . I have not sold a single one yet, so apparently noone needs these at 60£ a piece.

It would be better that they found new homes and got used instead of just being on the shelf in my lab

[RocketScienceSmurf]

If you are interested, we can find a reasonable price for one of these https://www.ebay.co.uk/itm/232565478339 . I have not sold a single one yet, so apparently noone needs these at 60£ a piece.

It would be better that they found new homes and got used instead of just being on the shelf in my lab

Thanks Mads! Do you think they would survive a frequency closer to 40-50kHz?

[Mads Barnkob]

If you are interested, we can find a reasonable price for one of these https://www.ebay.co.uk/itm/232565478339 . I have not sold a single one yet, so apparently noone needs these at 60£ a piece.

It would be better that they found new homes and got used instead of just being on the shelf in my lab

Thanks Mads! Do you think they would survive a frequency closer to 40-50kHz?

Yes, you would just have to derate the current, I am not sure by how much, but looking at a similar capacitor type from Vishay http://www.vishay.com/docs/13057/phawo-ks.pdf

Where current ratings are given for a 1500V capacitor at 3000, 5000, 8000 and 10000 Hz, I extrapolated the derating further for the yellow marked numbers, these are just a guess.



So that would make my NWL rated at 680 A, should be derated to something like 400 Arms at 40 kHz.

[RocketScienceSmurf]

I have thought a lot about how to connect the mosfets, providing both cooling and electric connection at the same time. Since I blow the mosfets up all the time they must be easy and quick to replace and I came up with this. The idea is co create 4 of the blocks made of brass and use threaded rods, insulating nylon washers/bushings and springs to clamp everything together. (The mock up TO-247 components are one mosfet and one power resistor)

In the middle I will put somewhere around 144 pcs of 47nF ceramic SMD caps giving me a total of 6,7uF of water cooled ceramic capacitor bank.

To provide good electrical contact between the surface and the components I plan to use small pieces of 0,3mm thick tin foil.

What do you think about it?

[flyrod]

What do you think about it?

I think it will work.  I did something similar:



FETs are screwed directly to water cooled copper blocks with ceramic caps sandwiched in between.

[RocketScienceSmurf]

There hasn't been much progress this summer as I have had other things to to but I have thought a lot about the project and I still want to get back to it. I bought some thick, soft tin toil which I hope will give each and every SMD cap ok contact. In order for the small SMD caps to stand upright and to make it easier to mount them I made a small frame that I intend 3d print in some plastic. Don't know if I need to use some special kind of filament but I will start with regular PLA and see what happens. The frame holds 144 pcs of 1210 SMD caps.

[Max]

I'm a little late to the party, but hopefully not too late 

Yes, you would just have to derate the current, I am not sure by how much, but looking at a similar capacitor type from Vishay http://www.vishay.com/docs/13057/phawo-ks.pdf

Where current ratings are given for a 1500V capacitor at 3000, 5000, 8000 and 10000 Hz, I extrapolated the derating further for the yellow marked numbers, these are just a guess.



So that would make my NWL rated at 680 A, should be derated to something like 400 Arms at 40 kHz.

Actually I don't think that these values are that useful; they come from different capacitors with different capacities. Vishay doesn't indicate it, but a quick calculation based on Urms, Irms and the frequency gives the following values (sorted by voltage):

Urms [V]    Irms [A]      f [Hz]     C [uF] 250 1680 600 1783 250 2640 1000 1681 250 2640 3000 560 250 2560 5000 326 250 2400 8000 191 250 2280 10000 145 600 1267 150 2241 600 1700 300 1503 1500 2640 3000 93 1500 2560 5000 54 1500 2333 8000 31 1500 2267 10000 24 2450 2000 1000 130 3000 667 150 236 3000 1333 300 236 3000 2000 600 177 As you can see, your diagram is based on capacitors with 24uF to 93uF. I'd like to point to another reference. Celem (manufactor of induction heating capacitors) indicates voltage and current deratings vs frequency for their capacitors. Take the example of a 6.3uF 600V 600A capacitor. Here's their diagram: (their other capacitors look quite similar) As far as I understand it, you simply have a maximum voltage and a maximum current which must not be exceeded. To do so, you have to derate the other value respectively. This would mean, that as long as you keep the voltage low enough so that the current does not exceed the maximum rating, you can go as high in frequency as you want (and vice versa for lower frequencies). Now this seems quite simple and there're two situations where I think that this rule is too simple: First, as visible on the diagram from Celem, around the point of maximum power both current and voltage are derated and I'm not quite sure why. Also this does not concern all of their capacitors. Secondly with frequencies much higher than the nominal frequency I could imagine that the dielectric losses rise, too. Which means you'd have to decrease the current. However, I couldn't find such an example at Celem, where the maximum current always remains constant over the whole upper frequency range. Kind regards, Max