[awemawson]

Having replaced the crucibles recently in both my furnace bodies I am perhaps TOO intimately aware of their construction which is:

A/ Each turn of the coil has 4 brass bolts whose heads are brazed on leaving the threads pointing outwards in the four compass points

B/ There are four vertical insulators running outside and parallel to the axis of the coils through which the bolts pass to support the coils very rigidly

C/ The four insulators are bolted to the bottom and top plates of the furnace body sandwich fashion.

D/ The inner surface of the turns of the coil is 'grouted' with a plastic thermal insulator or capram

E/ A layer of refractory paper is laid on top of the grout forming a tube whose outer wall is stuck to the grouting

F/ The base of the furnace body is built up with refractory cement within the bottom of this tube to give a platform for the crucible to sit on.

G/ A layer of 'drypack' (unbonded refractory compound) is put on this platform and a crucible 'bedded in' by twisting to make sure it sits level and centred.

H/ Layers of dry pack are poured between the crucible and the tube and rammed very tightly with a slender iron bar until up to within 1/2" of the crucible top, which should align with the top of the top plate of the furnace body.

I/ 'Capram' which is a plastic refractory putty is then formed round the top of the crucible and taking up the last 1/2" of the crucible/tube space.

J/ Small (1/16") holes are made in the capram to ventilate any moisture from the drypack space.

H/ The furnace is then fired at a low long heat to ensure that the drypack is indeed dry and the capram set.


The drypack forms a protective layer that will effectively stop molten metal reaching the coils in the case of a cracked crucible (yes it does work, I've had it happen !!!)

So you see, the crucible becomes an intimate part of the coil assembly

[Geof]

Having replaced the crucibles recently in both my furnace bodies I am perhaps TOO intimately aware of their construction which is:...

Sounds familiar .

[NinerSevenTango]

Warpspeed,

I'll check back here once in awhile and put in a tip or two if I can help you out. I work in the field too, and I even met Shawn a few times, hehe.

On the cooling thing -- go ahead and use water. Even tap water will be all right, the only thing you have to do is use more than two feet of non-conductive hose per 1,000 volts of potential. Coil the hose up into loops and tie-wrap them together. Suspend the hose away from grounded surfaces where it's really close to a high potential. Try to route the water cooling to similar potentials if possible, and in the places you can't, just add a foot or two of extra hose if you're worried about it and forget it. If the hose is non-conductive (important no matter what you cool with) and it's long enough, and the water is tap water quality or better (not too hard), then it will be many years before you have a problem. We cool 15,000 volts DC with superimposed 400 KHz on it this way. I wouldn't bother with glycol, unless the thing is in danger of freezing.

And you can make your own sacrificial anode if you're really worried about it, but it's probably not necessary.

In order for you to develop the power you want, you might find yourself needing a matching transformer at the output unless you can alter the output coil turns to match the voltage. For that reason I recommend you start with more turns than you think you need and put taps on the turns. You will definitely need to be able to tune around with your capacitors to get the load into the range you want. Every time you alter the capacitance or inductance, it changes the voltage/current. And you can't develop full power until you get full current at full voltage. If you end up going too far in either direction with the inductance in order to get the voltage in range, then the Q of the circuit suffers. For multi-turn coils like used in melting, it's common to use an autotransformer that can be kept upstream of the tank.

You might need more filter capacitor than you are anticipating, not to smooth the DC, but to provide the high amplitude, short duration current pulses to your inverter section (appropriate source impedance of the DC supply at the desired frequency). And the series inductor should take into account instantaneous fault current that will be seen at your chopper and upstream when the thing fires out of sequence. Just mentioning it in case you haven't already taken it into consideration.

You might want to learn how to silver-solder if you aren't already set up to do it. Copper is easy to work with, and most if not all of your conductors can be easily made with common refrigeration tubing, with tabs brazed on. You bolt thru the tab, and connect a cooling hose to the end. You can make low impedance buswork by brazing a length of tubing to a strip of copper -- make two of them and separate them with a layer of teflon. It makes a difference in the feed to the tank circuit, and in the tank circuit itself.

There is an induction heating company with a facility in Australia, Inductotherm, which also owns Inductoheat. If you call them up and try to make friends with the service manager or some kindred soul, you might be able to talk them out of some obsolete or cast-off parts from used equipment. Heck, they might even offer you a job once they find out what you are up to! Too bad you don't live near me, I have lots of stuff around.

On another note, the frequency you are using will be just fine for what you are doing, you don't really need lower frequency until you get into larger loads. Generally, higher frequency couples better with non-magnetic materials (if you keep the coil close enough to the load).

What are you planning to use for your IGBT drivers?

--97T--

[Warpspeed]

Thanks 97T, excellent advice exactly what I am seeking. While I have an electronics background, my practical working knowledge of induction heating is just about zero.

Taking your points in order:

I plan to liquid cool not only the tank components but the various IGBTs and diodes as well. Heat dissipation and semiconductor junction temperatures becomes a limiting problem long before these devices reach their full rated maximum current. Cool running devices are also faster with lower losses. Thermal design of the power electronics is a very important consideration. In short, I believe air cooled heat sinks are just not really up to the job. Liquid cooling of the power electronics also allows a very compact assembly which reduces stray parasitic inductance between the switching devices. Anyhow, as you suggest the various voltage differentials can be quite high, and massive coils of insulated flexible pipe are not going to be exactly convenient.

The way I see this is that oil offers superb electrical insulation properties, and completely eliminates any possibility of corrosion. I know I am probably alone in this, but if there prove to be unforeseen problems with oil, I can always go back to water later on. Immersion in oil is very common in very high voltage equipment for both insulation and cooling. While I suspect oil cooling may be relatively unknown in the induction heating industry, it is in common use elsewhere. So I would like to at least try the idea first and see how it goes.

Matching power into the tank is always going to be less of a problem with a semiconductor driver than with a high voltage vacuum tube driver. There are fewer volts and many more amps being switched in a semiconductor driver. I don't anticipate any matching problems directly driving the tank circuit with a fairly high switched constant current.

In my circuit the tank itself forms part of a self resonant oscillator, so it is just not really possible for the tank to go off resonance. The operating frequency can wander all over the place (and it will), but the power driving the tank will always be exactly at the tanks own self resonant frequency. The constant current driven converter topology is extremely tolerant of high reactive power and overload. It is also tolerant of random switching events and cross conduction in the H bridge driver.

The issue of sufficient filter capacitance is extremely important, and I am glad you have raised it. The constant current buck regulator will be pulse width modulated, switching on and off. When it is off, there is no power drawn at all, when it is on, there is an extremely heavy power load from the rectifier. Any series inductance in the up stream rectifier and filter circuit will create massive voltage spikes that would be death to my switching devices.

The trick is to use a low self inductance filter capacitor located right at the switching devices with almost no lead length. It must charge and discharge every PWM cycle and not allow any inductive voltage spikes to appear. It is not just a case of sufficient capacitance, but low enough source impedance at the PWM switching frequency. That capacitor also needs a high rms current rating to prevent overheating or failure. I plan to use a sufficiently rated GTO MKP snubber capacitor that is designed for that type of application. Something like this perhaps:

http://www.schusterusa.com/IMAGES/wimagto.pdf

I am very familiar with silver solder, it is wonderful stuff to work with. My semiconductor heatsinks will be slabs of copper busbar with a copper cooling pipe silver soldered onto it exactly as you describe. I have already run some thermal tests, and the heat sinking capacity of this is just amazing.

This whole exercise fascinates me, and even if I could buy a fully working commercial induction heater for peanuts, I would still enjoy the technical challenge of designing and building my own completely from scratch.

Ha-ha, not looking for a job, I am a gratefully retired power electronics design engineer. But later on I may sniff out some useful contacts at the induction heating companies that you mentioned.

As to which IGBTs to use I have not yet decided. But I will probably use some 600 volt devices in the TO247 plastic packages. I will fabricate my heatsinks so mounting several devices in parallel will be possible. That way I can test at initially low power with single devices to debug the system. Additional devices can then be added in parallel to increase the power capability.

That approach allows me to use low cost readily available IGBTs, as well as try different types. It may take several attempts to optimize switching loss with conduction loss. There is certainly a choice of device types with widely varying characteristics available to try.

The buck regulator running at 40Khz is the most critical and highest stressed part. I had thought I may first try the SKW30N60 from Infineon:

http://www.ortodoxism.ro/datasheets/...SKW30N60_1.pdf

[Warpspeed]

More random thoughts....

Andrew, your tank coil construction appears to be very well thought out, and it has given me quite a few very welcome ideas. I had not considered anchoring individual turns, but thinking about it, the thousands of circulating amps that you have there, must set up some pretty fearsome repulsive forces. I may just embed my tank in castable refractory, at least to begin with. Crude and simple, but probably better than nothing.

97T, more on the impedance matching issue. I thought all this through initially very carefully, as an old radio transmitter tech, I am fairly familiar with these types of problems. Matching a radio transmitter into an antenna is not that much different to induction heating.

It is mainly a case of thinking in terms of circulating tank energy, and tank voltage required to get it. In other words selecting the right combination of inductance, and capacitance to work at the required frequency, and at the required operating tank voltage. All the factors are interrelated.

My design figures are for a 21uF tank tuning capacitor, and 3uH tank coil, to resonate at 20 Khz, and each will have a reactance of around 0.38 ohms.

Tank voltage will be limited by a closed loop control system to 300v rms maximum. Peak tank voltage works out to be 424v, and I plan to power it from a 440v total dc supply voltage. Full load current to be about 34 amps giving 15Kw.

300 rms tank volts will produce a circulating current of 789 amps when Xl and Xc both equal 0.38 ohms at resonance. Circulating power will therefore be 236.7 KVA. If load power is 15 KW then the fully loaded Q will end up somewhere around 15.8 Those figures all tie in fairly well with the maximum ratings of the tank capacitor. (400v, 900A, 250KVA)

I feel fairly comfortable with those figures, and do not see the need for any additional impedance matching. The driver should couple power very well into the tank, and the tank has sufficient Q to couple well into the work. If anything, tank Q is probably a little high, but I can live with that.

If I am on the wrong track here with this, I would really appreciate some guidance.....

[NinerSevenTango]

Warpspeed,

I still think non-conductive rubber hose with clean water will be the cheapest and easiest way to get it done, because even with oil you will need to buy it (not hard to find). But, at least oil doesn't evaporate like water!

There's a company in France, Celes, that makes power supplies using a circuit such as you describe. But it's been awhile since I've seen one.

There are lots of ways to skin the cat, but for relatively small size like the one you are building, the approach you are taking seems to be a good trade-off between cost and complexity.

In larger sizes running to 100's of KW, some companies use an SCR controller to drive a standard 2:1 input transformer (for running off 480V mains). The secondary is rectified, and this gives a variable voltage DC bus well below the maximum ratings of the output devices. The strategy is to run gate pulses to the IGBT's at all times, putting out maximum frequency when the unit is at idle. As the input power is ramped up from zero, the frequency sweeps down until crossover is detected, at which point it is locked onto the resonant frequency. The things get a special series inductor on the DC bus, and fairly robust filter capacitance. The input transformer is expensive, but it makes up for a lot of engineering time, makes life easier on the devices, and its leakage impedance comes in handy when a device fails. Anyway, that's one company's approach to reliable starting and reasonable fault tolerance.

Later on, I'll take a glance at a few others and let you know what some of the other popular (or peculiar as it may be) approaches are.

One thing every inverter I've ever seen has in common is whenever devices or banks of devices are run in parallel, there are inductances placed in the circuit to allow for differences in device behavior. They are usually called 'balancing coils', and they allow devices to share the current. Without something like this, differences in device performance would leave some devices taking the lion's share of the current. From different manufacturers, I've seen specially wound coils, tubing trombones, and ferrite rings with cables strung through them used for this.

What are you planning to drive the gates of the IGBT's with?

--97T--

[awemawson]

Be very fussy when specifying your 'rubber hose'.

Neoprene is the prefered material as it degrades far less than natural rubber based compounds. Most commercial 'rubber hose' be it rubber or neoprene, contain carbon as an additive, which makes the hose slightly conductive. There is a grade of hose called 'brewers hose' intended for the food and drink industry that has no carbon added and is thus an excellent insulator.

The drive leads to my furnace bodies comprise 4 lengths of 3/4" bore Brewers Hose each with 35mm sq bare welding cable threaded through them and crimped onto a special end fitting.

If you have ever tried singlehanded threading 20 foot lengths of 35mm uninsulated fine braided copper welding cable down a length of 3/4" hose you'll appreciate just how much fun it is !!!!!

[NinerSevenTango]

Andrew,

Thanks for the correction on my careless misnomer. You are correct about the rubber and carbon. The clear plastic stuff (don't remember the compound) is bad, too.

Goodyear makes an excellent non-conductive hose, as do several others.

--97T--

[Warpspeed]

I am still not clear on understanding why non conductive hose is so vitally important. Surely a few microamps or milliamps of leakage current through the hose is not in itself going to be a problem ? Unless it somehow degrades the hose material over time?? Why is this small leakage current a problem, I still do not understand this.

There is a company nearby that specializes in supplying hydraulic hoses and fittings, and they have a pretty good range of hoses for all types of industrial, chemical, and food applications. I will check them out and see what they have.

But being a stubborn bugger, I am still determined to try oil first, at least to do some thermal testing as I have already done with water. Gosh there is still a lot to do with all of this, I still need to get some of the larger components, particularly a common mode choke for the three phase rectifier, and most important of all the high frequency choke for the buck regulator. I have some ideas for that, but have not yet done anything about obtaining parts for it.

Controlling the tank voltage by sweeping the drive frequency is rather interesting, but it would be venturing into a completely unknown area for me. Controlling a PWM buck regulator with a simple proportional integral control loop seems to be far simpler, and is much more familiar territory.

Current sharing of parallel devices may be a problem, or may not. I have yet to discover that. But some balancing inductors would certainly be a practical solution. We shall see.

How do I plan to drive my IGBT gates ? Here is the PROPOSED circuit for the H bridge driver and phase locked loop:

http://i144.photobucket.com/albums/r...-43_edited.jpg

And the circuit for the PROPOSED constant current supply and control system:

http://i144.photobucket.com/albums/r...-55_edited.jpg

And lastly, the power supply:

http://i144.photobucket.com/albums/r...-05_edited.jpg

While small sections of these circuits have mostly been tested in isolation, they do not yet exist in a completed form on a circuit board. It is all fairly basic and straightforward, nothing at all fancy about it.

Any comment or suggestions on my intended design approach would be most welcome. The whole thing is still wide open to possible change.

[awemawson]

If the hoses are at all conductive, and you have a few hundred volts kicking arround you have to insulate them from whatever they come in contact with. If the hoses are non-conductive then you don't.

In the case of my external drive hoses to the furnce body, it may be ME that they come in contact with !!!! (Remember both bodies are dynamic - they tip or invert in one case, so the four hoses are moving.

[Warpspeed]

Ah yes, I see !

Thanks Andrew, I was thinking in terms of hoses completely enclosed inside equipment.

[NinerSevenTango]

A trickle current through the hoses will create carbon tracks which will eventually destroy the hose and contaminate whatever fluid you are using. Lots of people have found out the interesting way.

Edit: By 'interesting', I mean a leak spraying all over everything while the thing is running, hehe.

In the circuit I described earlier, the frequency sweeps down on startup to achieve lock-on to resonant frequency, not to control tank voltage. The tank voltage is controlled by varying the DC bus supply.

Your circuits look elegant and well considered.

Edit: Is gate driver pulse width a consideration?

--97T--

[NinerSevenTango]

One other thing to note about liquid cooling -- I once knew a man who had been severely injured by a pinhole leak. He was leaning against a cabinet when the fine mist from a pinhole leak in an output cable landed on him. The high frequency current found its way through the droplets, then through him to ground and burned him where his shoulder was in contact with the cabinet. It was higher voltage and frequency than this, but at shorter distances the same risks apply.

Sobering thought.

--97T--

[Warpspeed]

I have been wondering about the potential for receiving RF burns off an induction heater. In my now distant youth, I can can still vividly remember receiving RF burns from radio transmitters, and they were particularly unpleasant taking a rather long time to heal.

Not sure if the frequency is high enough to cause a similar effect ?? But I am guessing that it may well be.

Another very good reason to completely enclose the tank coil.

Thank you all for pointing out the hose conductivity problem, it is something I would never have anticipated by myself. One reason for choosing oil cooling, was that I was thinking hose lengths could be kept extremely short, maybe only an inch or two. But leakage current through the hose itself could cause some type of electrochemical deterioration in the hose material. Something like teflon hose should be far better than anything based on natural rubber, that is, if I can actually find some.

Fortunately I have a Danbridge non destructive insulation tester here at home, with which I can measure down to picoamp leakage current, with a dc test voltage continuously adjustable from zero up to 10kV.

So it looks as though I will be entertaining myself with some hose testing.

This whole project is absolutely fascinating, it is becoming by far the most interesting project I have tackled for a very long time.

[awemawson]

I don't pretend in anyway to be an expert in oil cooling (!!!) but there must be some reason that large transformers are cooled with a special oil and not normal car oil. I have a feeling it is perhaps a vegetable oil (though wikipedia disagrees). I know it is hygroscopic and has to be changed regularly to prevent breakdown as I managed to get some once when they were re-oiling a transformer where I then worked to refill my arc welder.

[Warpspeed]

I don't know what they use these days. But back in the dim dark past, transformers and large high voltage capacitors commonly used some special oil that contained PCBs, which turned out to be highly carcinogenic. The EPA become absolutely frantic about the careless disposal of old capacitors and transformers that contain any oil.

But looking at the specific heats of a wide variety of different oils, they all seem to fall within a fairly narrow range. I doubt if there is any particular "special" or "best" type of oil to use specifically for oil cooling. I suppose it is mainly long term chemical compatibility with seals, gaskets, and whatever else is inside a transformer.

As my oil is only going to see some copper pipe, short lengths of flexible hose, and a cast iron oil pump housing, I guess automatic transmission fluid should do the job. It has fairly low viscosity, and it will stand a fair bit of temperature without breaking down. It should also be fairly kind to the pump, because that is what most power steering systems use anyway.

[NinerSevenTango]

About electrical burns: I've never tried it, and I don't know anyone who ever did try 20 KHz. I would guess, based on the difference in the way insulators break down under heavy use, that 20KHz would tend to kill you from inside, same as line voltage. The surface burn is probably less lethal, although very unpleasant.

About insulating oil:

http://www.electricenergyonline.com/...g=8&article=54

[Geof]

I don't pretend in anyway to be an expert in oil cooling (!!!) but there must be some reason that large transformers are cooled with a special oil and not normal car oil....

PCBs, PolyChlorinated Biphenyls and PolyBrominated Biphenyls, were used for transformer cooling because they are non-flammable. Now the oil used is just mineral oil (paraffin oil) I think. Even the PCBs used to break down a bit and had to be periodically cleaned. Especially with very high voltage equipment.

PCBs are now identified as a potent carcinogen which is a lot of nonsense. A large use for PCBs in the past was in floorwax and other polishes. Had they truly been a strong carcinogen this would have been discovered among people using these products on a regular basis. They are carcinogenic, much the same as gasoline (petrol) with constant skin contact. Banning them was a good idea because they disrupt endocrine function in birds and they can do the same in humans. The big problem was they accumulate up the food chain in fat tissue.

Regarding RF burns I think you are completely safe at 20 kHz. You need to be in the mega or giga hertz region; wavelengths down near cellular or molecular dimensions to get localized burns. Medium wave can cause problems with whole body heating rather than localized burns. I was told by a Vet (Animal Doctor version) that this can induce liver failure because the liver is susceptible to heat damage more than other tissues.

Depending on the voltage 20kHz could give very bad surface burns if the voltage was high enough. The skin effect works in humans also, literally skin effect. But this is distinct from RF burns which can be deep tissue depending on the wavelength.

[ImanCarrot]

If I were to stand in a big enough induction furnace, would it heat up the iron in my blood?

[Warpspeed]

Really good info there Geof. I believe then, that just about any low cost mineral based automotive oil should be up to the job. Synthetic oil may be even better?

The only contaminant I see as being a potential problem for us is water. If the oil system is vented to atmosphere, it will breathe as the system is thermally cycled. Normal humidity in the air, over time, would slowly trap condensation inside the oil system, potentially leading to a very slow buildup of water in the oil. That is no big deal. Automotive brake systems suffer from exactly the same problem. Most brake fluids being very slightly hygroscopic will absorb a small amount water, which could then flash into steam in the brake calipers under extreme use.

The solution is rather simple. The oil top up reservoir just needs a big floppy diaphragm seal under the screw cap. A pin hole in the cap vents to atmospheric pressure, but the diaphragm creates a flexible pressure balancing barrier, that isolates the air inside the oil reservoir with outside air. The system can breathe and allow for expansion of the oil, but no water vapor can enter the system. A standard plastic remote mounted brake fluid reservoir with its special cap and seal should work fine for us. That would be ideal, because the oil top up reservoir needs to be located at the highest point of the system to prevent oil drain back from flooding the reservoir.

Another step in this direction would be to use a normal automotive screw in engine oil filter, and fill it up with some silica gel crystals. That would dry the oil, and the filter element would trap any stray particles. It would also prevent the silica gel from circulating around with the oil.

All this is probably far too much trouble to go to. If an induction heater will work with a 100% water based cooling system, a minute amount water in an oil based cooling system is hardly going to be catastrophic. But I still believe that oil may make a far better cooling medium than water, for both electrical and corrosion reasons.