Royer induction heater and high current power supply

[Mads Barnkob]

I asked the seller for a picture of the insides and that revealed that it is only a huge stack electrolytic capacitors, so it is completely out of interest now, that is simply too expensive for something that is maybe worth 150€ in total.

[RocketScienceSmurf]

Too bad but really nice of the seller to provide the picture. My search continues...

[rikkitikkitavi]

Kamelryttaren, have you considered larger server power supplies for server racks. There are two types in princip , 12V and 48V types. I was thinking about the later.

The good thing about these is that they are

1) high PF- this is important , utilising your AC outlet fuse to the max
2) high efficiency (90+%)
3) for the larger, 48V types floating chassis, ie it is NOT connected to either of the 48V terminals. It is of course connected to PE.
4) compact
5) whiny (as in fan whines) - maybe not so good after all
6) easily found at ebay etc for a reasonable price mostly


They can be paralleled at convenience.  They are designed for parallel operation by current sharing but there is a data bus protocl which  is propietary but for most of them you can find which pins to tie for startup as a standalone unit and run them as standalone supplies via current share resistors as output voltage is trimable a few volts.

They could also be series connected up until some critical insulation voltage but as most of them are compact and hold 400VDC bus with a high-pot aproval I am not worried about that.
I expect to be able to use 3 in series for 150V DC bus at 40A.

I was planning to use some for a induction heater project in the future.

They probably needs a soft start on the induction heater and proper filtration to prevent back emf interfering with the DC supply feedback loop, ie some secondary DC bus cap that can supply the current peaks.

[RocketScienceSmurf]

Thank for the suggestion rikki. Haven't really thought about it but since I have most things apart from the caps I think I will stick with my +200kg three phase psu. I don't like SMPS and I think my 37VDC is already a bit on the high side. I would have preferred something closer to 24-30VDC but this will have to do.

I got another ebay package today with a Hantek 1008 8-channel USB oscilloscope and a 65A current clamp probe so I can keep track of what's going on with the induction heater circuit. I need a few 20:1 attenuators as well.

Btw, you wrote something about soft starting the induction heater. That's a great recipe for smoke and other undesirable special effects.They need to be jump started as brutally as possible. I have managed to get my hands on a pair of Infineon IPT007N06N that I will try and use as a power switch but I haven't really figured out how to water cool it and connect it. I think I will try and solder the thermal pad itself to a copper pipe and use the pipe as conductor as well as heat sink and then just solder the fattest wire I can to the source pins.

[Mads Barnkob]

The server power supplies are handy, cheap and small, if they can withstand the abuse and injected noise from a Royer oscillator, I got no idea, but I do have some of these supplies somewhere, maybe I should smack together a Royer just to test them

There is no doubt that your iron core power supply will be one of the few things left after 3rd world war along with the scorpions that will be sitting on top of it.

Good that you got yourself a oscilloscope, that is one of the best tools in electronics, if you need a cheap attenuator you can just use a ferrite ring core of suitable permability / AL above 4000, wind 20 turns on it and put the wire through you want to measure, this gives you a 1:20 step down.

The IPT007N06N MOSFET looks good on paper in a high current switching power supply, but take a look at figure 4, if you use this as a switch to conduct DC at high current, it will melt really fast, even at 50% duty cycle it will have a temperature rise of 2 Kelvin per Watt dissipated, how much power do you expect to dissipate across its R(DS)on? If you end up with more than 30 Kelvin rise, I think you will be on thin ice in regard to it living long.

Would it not be a better idea to get a large SCR to switch it on with?

[RocketScienceSmurf]

If I assume worst RDS(on) of 0.75mOhm at about 50A current draw that would mean about 0.5W per mosfet and a voltage drop of 0.02V. Water cooling it should be sufficient I think unless I have severely misunderstood something in the datasheet.

I am not familiar with SCR. I tried searching but if you have a good source for info I would appreciate it. Since I have a pretty good grasp on mosfets I just thought that would be the easiest way to go, especially since I intent to use an arduino to control my power supply when it is finished.

I made some quick tests today with the scope and the current clamp. It works really well but I need to get my big power supply up and running. I took the old UPS battery just to get the heater to oscillate and it works but the gate voltage is just under 7V so they don't even fully "open". The gate waveform looks really bad as well. I was hoping for  something close to a square wave but it more resembles a saw tooth wave.

[Mads Barnkob]

You are right, I was being pessimistic, even at 100A through them you do not have more than 7 Watt dissipated.

If duty cycle D=0.5 has a 2K/W rating, I think we have to be pessimistic again and say that DC is around 10K/W, so you need to cool a 70 degree temperature rise, so water cooling is maybe a good idea, but maybe a large passive heat sink / forced air cooling is enough, you better test this before building a large water cooling setup

I did not have any good SCR resources in mind, just a fast googling it was something like this I had in mind: http://www.bristolwatch.com/ele/triacs.htm

Be careful with such low voltages, below 10V and you are in the MOSFET kill-zone 

[afk]

On the topic, I'm having some questions regarding water cooling. Some sites warned me about tap water having minerals (mainly sodium, calcium and potassium ones) that make water more conductible, and this can cause electrolysis of water and these minerals, which then can corrode the copper tube. Is it true? I'm asking because distilling water is a time-consuming process and waste a lot of gas/electricity for heating. Furthermore the tank current is alternative at a few tens of kHz so the electrolytic effect shouldn't exist (or to be a problem). I'm thinking of taking advantage of the rice cooker in my house, though. If tap water can be used for around thirty minutes then that should be okay for my application.

[RocketScienceSmurf]

You are right, I was being pessimistic, even at 100A through them you do not have more than 7 Watt dissipated.

If duty cycle D=0.5 has a 2K/W rating, I think we have to be pessimistic again and say that DC is around 10K/W, so you need to cool a 70 degree temperature rise, so water cooling is maybe a good idea, but maybe a large passive heat sink / forced air cooling is enough, you better test this before building a large water cooling setup

I did not have any good SCR resources in mind, just a fast googling it was something like this I had in mind: http://www.bristolwatch.com/ele/triacs.htm

Be careful with such low voltages, below 10V and you are in the MOSFET kill-zone 

My water cooling setup for these mosfets will be very simple and nothing fancy at all. Will post pics when I have it soldered together, but it will probably be a while since I can not find any affordable caps

[RocketScienceSmurf]

Regarding the dc bus capacitors, is there any advantage of the first configuration over the second one in this image?

I have purchased caps now. Proper film capacitors but I couldn't afford to reach more than 1900uF. I'm hoping it will be enough to start experimenting at least. I have tried to figure out the resulting ripple voltage but I don't know how to calculate this so I will just measure it when everything is up and running.

I got 5 of these for a good price: http://www.kemet.com/Lists/ProductCatalog/Attachments/123/F3303_C4DE.pdf

They are rated 100 Amps (10kHz 25C), will they tolerate more or less at 300Hz?

Now I have something fun to do on my upcoming vacation

[edit] Receied the caps and copper rails yesterday so today i drilled holes and test mounted the capacitors. I still haven't drilled the holes for connections but I will wait until I have decided how to mount everything. Hopefully I can do some test with my power supply soon since the caps where the last items missing.

[RocketScienceSmurf]

It's alive!

I finally got the power supply together (but not completely finished) and fired it up. Great success and no smoke

Tomorrow I will bring the IH and connect it to make some early measurements

[RocketScienceSmurf]

Yesterday I connected everything and it is working great!

I made some triggered measurements with my Hantek 1008 oscilloscope and it looks ok but I am a bit puzzled about the gate waveform. Despite around 40V power supply voltage it does not look very "square-y" but I'm beginning to think that my oscilloscope knowledge or the actual oscilloscope itself may be at least part of the reason. Could it be that the bandwidth is a bit too low for what I am trying to do? Despite that it was very interesting an educational.

At first I did a few runs with nothing inside the coil. There was a large current spike at the beginning but it dropped quickly. Startup current at around 40V was roughly 20A.

Then I decided to see what happens when you put something inside the coil and try to fire it up. I found a piece if solid iron about 3cm in diameter and placed inside of the coil. Now things were starting to get exciting! Startup current was over 50A and it didn't drop as much as before. The iron piece got quite hot in the few seconds the induction heater worked before one of the mosfets blew up (as expected). Since the induction heater started to oscillate and worked well for a little while before blowing up I am assuming the heat losses is the most likely reason for failure. I will try my IRFP4568 next since they have MUCH lower RDSon the the IRFP250N. So far I am using un-modified ebay induction heaters to learn a bit more about them.

In the scope pictures below I have the current clamp hooked up to channel 1, mosfet gates are connected to channel 2 and 3, and finally channel 4 is the supply voltage connected through a 20:1 attenuator.

[Mads Barnkob]

I got 5 of these for a good price: http://www.kemet.com/Lists/ProductCatalog/Attachments/123/F3303_C4DE.pdf

They are rated 100 Amps (10kHz 25C), will they tolerate more or less at 300Hz?

In high frequency applications where these typically are used, it is the power dissipation that is the limiting factor, since losses are higher at higher frequencies. At lower frequencies it is the allowed ripple voltage that is the dominating factor, but since you are using MKP capacitors, this is not a problem, but watch their temperature rise as there is a large current de-rating factor for temperature.

Yesterday I connected everything and it is working great!

I made some triggered measurements with my Hantek 1008 oscilloscope and it looks ok but I am a bit puzzled about the gate waveform. Despite around 40V power supply voltage it does not look very "square-y" but I'm beginning to think that my oscilloscope knowledge or the actual oscilloscope itself may be at least part of the reason. Could it be that the bandwidth is a bit too low for what I am trying to do? Despite that it was very interesting an educational.

At first I did a few runs with nothing inside the coil. There was a large current spike at the beginning but it dropped quickly. Startup current at around 40V was roughly 20A.

Then I decided to see what happens when you put something inside the coil and try to fire it up. I found a piece if solid iron about 3cm in diameter and placed inside of the coil. Now things were starting to get exciting! Startup current was over 50A and it didn't drop as much as before. The iron piece got quite hot in the few seconds the induction heater worked before one of the mosfets blew up (as expected). Since the induction heater started to oscillate and worked well for a little while before blowing up I am assuming the heat losses is the most likely reason for failure. I will try my IRFP4568 next since they have MUCH lower RDSon the the IRFP250N. So far I am using un-modified ebay induction heaters to learn a bit more about them.

Congratulations on having it all working, at least for how long/short it lasted

You are right about your Hantek 1008 oscilloscope having a too low band width and sample rate. 2.4MSa/s is very low for a oscilloscope, which is why this is advertised for diagnostics on engines, not high frequency inverters. 2.4MSa/s is most likely also only if you used a single channel, so using all four it is even smaller, like my Rigol DS1054Z is a 1GSa/s if using one channel, but using all 4 its down to 250MSa/s.

Was it a loud boom when the MOSFET went? Did you get it on video?

Certainly looking forward to the repaired inverter updates! 

[RocketScienceSmurf]

Yes the mosfet blowing up was louder than I expected. It's very nice to have a power supply that won't sag under heavy loads. I realized afterwards that I should have set my oscilloscope on a negative slope trigger to catch all values when it blew up but perhaps I will do that later on when I get it running again which shouldn't take too long.

I haven't inspected the pcb thoroughly yet but I noticed a very thin connection that blew up and it makes me wonder why they would make such a thin via if it was supposed to handle large currents. I will see if I can figure out what it was supposed to connect. The other, similar board I have does have the exact same configuration so I can not use that as a reference.

[petespaco]

Which "unmodified Ebay induction heater" are you using?  If you are using one of the "1000 watt" models, why would you allow 50 amperes to flow?
50 Amps X 40 Volts = 2000 watts.
  When testing the limits of one of these devices, I would not place  any work in the work coil before turning the power on unless I was already confident that the current flow would be acceptable .  I would turn the power on, confirm that the circuit was oscillating, and THEN slowly insert the work, watching the current, making sure not to exceed a practical limit.

In your pictures of the burned traces:
Those components appear to be gate components, not power handling components.  I think those traces are okay to handle several watts.  I think the components overheated when when the Mosfet failed.  I don't think my gate zeners have ever survived a failed Mosfet.

Pete Stanaitis
---------------

[RocketScienceSmurf]

The application I intend to use the induction heater for requires my to have the work piece already in the coil when I start it up so I must make sure it can start to oscillate, which, as it turns out, it really can. Even at startup current of 50A! The heater I am using now is this one: http://www.ebay.com/itm/322519608639

And yes, I am fully aware that I am running it at least 100% over it's rated specs (that are probably optimistic to start with). At this stage I am not interested in seeing if it runs and oscillates. I know it does. I am trying to figure out it's weak points and how to address them. My goal is to get my own induction heater up to about 2kW, preferably using only one pair of mosfets and to have enough cooling for it to run continuously.

Another potential problem I might run into is that the work piece won't couple enough energy (not sure how to explain it). I now know that a solid piece of iron will mean a fairly large transfer of energy, but I will use a thin walled pipe and I have absolutely no idea if that will work or be enough. I will try and find a regular steel pipe to try with and see what happens, but for my finished application I will have to get hold of some martensitic stainless steel pipe which has turned out to be almost impossible.

[edit] the burned pcb vias were actually the source connection for both mosfets. I will bypass them with some serious and also hardwire the source leg of both mosfets to the ground connection.

[Anders Mikkelsen]

Nice and solid construction, it will take more than a line spike to kill that supply!

The rule of thumb of using capacitors with a ripple current rating of twice the load current doesn't sound right, so I did some simulations in LTSpice. With 1600 uF on the DC bus, 153 A RMS load current and an average output voltage of 38 V (with 5.7 Vpp of ripple), the capacitor sees about 7.5 A RMS of ripple. Using the Rifa PEH200 catalog as a reference, a 400 V 1500 uF electrolytic capacitor would have a sufficient ripple current rating.

[RocketScienceSmurf]

Nice and solid construction, it will take more than a line spike to kill that supply!

The rule of thumb of using capacitors with a ripple current rating of twice the load current doesn't sound right, so I did some simulations in LTSpice. With 1600 uF on the DC bus, 153 A RMS load current and an average output voltage of 38 V (with 5.7 Vpp of ripple), the capacitor sees about 7.5 A RMS of ripple. Using the Rifa PEH200 catalog as a reference, a 400 V 1500 uF electrolytic capacitor would have a sufficient ripple current rating.

Thank you for the information about ripple currents and caps. Since I had no way of measuring beforehand and not enough experience with large ripple currents I had to rely on the internet as a source of information. The only actual downside of my current capacitor is the capacitance. Current handling and ESR are great!

[Mads Barnkob]

The only actual downside of my current capacitor is the capacitance. Current handling and ESR are great!

From Kizmo's experiences with using a 6000 uF DC link capacitor bank in his BiggerDR DRSSTC, he saw huge improvements in the peak current delivery time, compared to electrolytics, that actually indicated that he needed less bus capacitance when using film capacitors, to get the same performance in the power electronics.

[RocketScienceSmurf]

Did a new test run today. Replaced the broken IRFP250 with IRFP4568, zeners replaced with Russian surplus zeners rated for 8W and delivering about 13V gate voltage. I had done some testing with the zeners and the new 100 Ohm thick film power resistors and realized they got quite hot so I placed the resistors on the opposite side of each water cooling block for the mosfets. Hard to tell if it was the new mosfets, the water cooling or something else but I could run it at almost 40A for quite a while before one of the tank caps left the PCB with a loud bang and the everything just broke down. It's quite interesting to witness the violent fail of this circuit. I have broken stuff before but this is something else. Some pcb vias and components just seize to exist.

On closer post breakdown inspection I could see that the pcb trace connecting all the caps to the coil had burned off, but that was not all. The drain leg of one of the transistors had evaporated! It's not in focus but it's somewhat visible on a few of the pictures.

I am not sure of what actually failed this time, but water cooling is a must for the circuit when abusing it like I do. I did not cool the work coil which might have contributed to the circuit failing. The hose I used for the water block was a bit too large for the copper tubing in the coil so I would have had water everywhere. I think I will have to make the coil of a slightly larger diameter copper tubing so I can properly transfer the heat. I will make some measurements on the zeners do see I they survived but I think they are actually ok, and if they are not I still have a handful left.