Induction furnace

Awemason has it right.

Basically there are two topologies when dealing with induction power supplies.

#1) Variable DC Buss, with the inverter running at (or very close to) the tank's natural resonant frequency. - This is the on that Awemason has described.
Basically in this mode, the power (or voltage applied to the tank) is controlled by the amount of DC Buss voltage achieved.

#2) Fixed DC Buss and "Swept Frequency Inverter". This style usually has diodes in the rectifier (unless the design engineer wanted to use SCR's for the shutdown mode). The inverter will start at a frequency far different that the natural tank frequency. Then as power is dialed up, the frequency of the inverter is moved closer to the tank frequency, thus producing power more of the time in the tank, and thereby increasing the average power produced in the load.

Some manufacturers have used a combination of the two topologies.

Some induction units are series tuned tank circuits, some are parallel.

Then there is always the famed "Crowbar" induction units. Which usually work until the day that any one component fails, then ALL components fail at once. This type of machine is easy to troubleshoot. Step 1, replace all components, Step 2 turn the machine back on. Step 3, if Step 2 failed repeat Step 1, then replace the cabinet, then proceed to Step 2. :D

That is pretty much how I thought it would work. I am planning to do pretty much the same thing, at least in overall concept. I preferred to use a high frequency buck regulator instead of using SCRs to control the main dc supply. It is far easier to control one IGBT than six SCRs !! But for really high power, the SCRS make a lot more sense. For relatively low power, the buck regulator is a lot simpler.

I will run six diodes in an ordinary bridge rectifier, and then PWM the dc output of that at 40 KHz with an IGBT. That will make my dc feed choke much smaller. Your feed choke is forced to operate at only 300 Hz, and it must be an enormous iron cored monster. This one will need to have a powdered iron core, but in physical size it will be fairly small for the power.

Using SCRs in the H bridge unfortunately limits the operating frequency to below 10Khz. My H bridge will use a commutation diode in series with an IGBT in each leg. The operating mode is really identical, except IGBTs allow a much higher operating frequency. The very efficient zero voltage resonant mode switching function is exactly the same in either circuit.

I feel I am on the right track with this, just getting my hands on some of the rather special parts is proving to be a hurdle, but the electronics design for it should be fairly straightforward. At least I hope so.

Quote:

---

Originally Posted by Shawn D

Awemason has it right.

Basically there are two topologies when dealing with induction power supplies.

#1) Variable DC Buss, with the inverter running at (or very close to) the tank's natural resonant frequency. - This is the on that Awemason has described.
Basically in this mode, the power (or voltage applied to the tank) is controlled by the amount of DC Buss voltage achieved.

---

That is what I figured. I see the variable dc bus topology as having far fewer problems, being much easier to control, and generally being more robust.

While sweeping the frequency off resonance may be very ingenious, I can see that building a control system from scratch without any prior knowledge or experience to fall back on, would be a lot more of a challenge. There are quite enough problems to solve as it is, without seeking more.

I'm not understanding the IGBT and Diode series setup. Do you have a schematic of your inverter?

Yes I do have a proposed schematic. But I am very reluctant to post completely untried ideas that others may be tempted to copy.

The basic idea is based on the conventional well known self commutating SCR full bridge converter, where the tank voltage reverses and turns off the pair of SCRs that are conducting.

The tank voltage itself turns off the SCRs, so the SCRs can never become a dead short across the tank, even if the gate triggering goes completely nuts.

IGBTs have a few problems, in that the collector cannot swing negative of the emitter more than a very few volts without destroying the device. Also under fault conditions, it may be possible for a pair of upper, or a pair of lower IGBTs to conduct together shorting out across the tank, with disastrous results.

The cure is to place a diode in series with the collector of each IGBT. That makes each IGBT/diode combination reverse blocking, (just like an SCR). If the tank tries to force the collector negative, the diode just turns off.

In fact it would be quite possible with this circuit to turn on all four IGBTs in the bridge simultaneously. No damage will occur, because two of the diodes will ensure that the particular pair IGBTs will be protected from reverse polarity from the tank.

This topology of four IGBTs plus four diodes in H bridge, should be hazard free from cross conduction caused by noise or any fault in the gate drive system. In fact the circuit will also work without any allowance for dead time. The commutation diodes ensure smooth turn off as the tank polarity swings through zero voltage, regardless of how the IGBTs are being driven.

Quote:

---

Originally Posted by Shawn D

I'm not understanding the IGBT and Diode series setup. Do you have a schematic of your inverter?

---

O/k here is a link to my PROPOSED driver and phase locked loop. Parts of it have already been built and tested in isolation, but the unmarked components are as yet undetermined.

I will leave this link up until the final tested working schematic is available, and I will post that in all it's glory when it finally becomes available.

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

I think in part of that discussion above you refer to a dead short across the tank whereas you mean across the DC rails perhaps? The problem with series diodes is the energy loss, which is more heat to get rid of. Even with high efficiency avalanche type you are almost doubling the losses already encountered in the switching device itself, be it SCR or IGBT as the forward voltage drop is comparable.

On the swept frequency issue and tuning, it is important to realise that the resonant frequency of the tank circuit varies as the melt proceeds. This is for two reasons: a/ As more charge is added the inductance of the coil changes and b/ As the charge warms up and changes its magnetic properties again the inductance of the coil changes. It is very easy to detect resonance by either having an oscilloscope across the tank while tuning (best method) or a current meter in series with the DC feed to the H bridge which will peak on resonance.

If you observe the oscilloscope waveform while sweeping the drive frequency over a wide range you can easily differentiate between harmonics and sub-harmonics and the true resonance, whereas the simple current meter can easily confuse you unless you start very close to the correct frequency. I spent a considerable time doing this and working out control strategies and mechanisms when I was developing my low power pilot unit.

When I stopped my developments due to buying the 'real thing', rather than throw out my prototype or put it in a cardboard box never to be seen again I put it on ebay thinking that the power semiconductors (IGBTs) and heatsinks would be useful to somebody. Amusingly it was bought by a company making induction heaters to see if I had any ideas that they hadn't thought of, which I very much doubt!!! The pulse width modulation circuitry however was rejigged to form the basis of the motor contoller for a friends electric bicycle and is still in operation to this day <G>

There are two fairly major issues with any H bridge design that could lead to device failures.

The first is turning on either two "top" devices or two "bottom" devices simultaneously. That could potentially place a dead short directly across the tank as described in my previous post. The commutation diodes are required anyway if either IGBTs or MOSFETs are to be used to prevent the possibility of sudden catastrophic device failure. Neither of these devices have any inherent reverse blocking ability by themselves. SCRs act like diodes within themselves and block reverse voltages naturally, so this can never a problem with SCRs.

If circuit operation above 10Khz is a requirement, SCRs are really out of the question, and the extra conduction losses of IGBTs or MOSFETs, plus commutation doides is really unavoidable.

The second problem is what happens if one "top" device, and one "bottom" device on the same side of the H bridge conduct simultaneously. The dreaded evil cross conduction, which effectively places a dead short across the two dc supply rails.

But remember, in this topology the power is coming from a constant current supply, not from a stiff constant voltage source. The dc current feed choke with its comparatively enormous series inductance, prevents any sudden instantaneous rise in current, even with an instantaneous dead shorted load. The control system up stream that sets the constant current will take care of any slow changes, while the feed choke takes care of very sudden changes.

The constant current supply powering the output bridge circuit is an extremely important feature of this design topology. Destructive current spikes are simply impossible in a constant current high impedance circuit. Voltage spikes can certainly occur, but they are far more manageable with some simple clamping diodes.

What actually happens is that the H bridge is always switched precisely at the voltage zero crossings. At resonance the current is exactly ninety degrees out of phase, so at the voltage minimums, the tank current has just reached it's peak and is about to reverse. The H bridge switches just at the current reversal point, so the incoming constant current is always steered through the tank at the correct time, and in the correct direction to add to circulating internal tank energy.

Realize too that zero voltage, means not only zero volts directly across the tank, but effectively it pulls the voltage down across the whole bridge to zero as well. The constant current source is pulled down to almost nothing by the tank circuit and two conducting IGBTs. If you looked at the supply voltage to the bridge circuit it looks like the voltage coming out of a full wave rectifier. The current is constant, but the bridge supply voltage is always changing.

Zero voltage switching means exactly that. When the IGBTs switch, there is almost no voltage anywhere around either the tank or bridge. Switching losses are almost zero, and fairly slow devices can be used.

Off resonance, the tank still switches at the voltage minimums, but now the circulating tank current will have moved fairly far from its ideal ninety degrees phase. This will be particularly true with such a high Q tank circuit. The phase change in tank current either side of resonance will be fairly dramatic.

What happens is that the H bridge still switches at the voltage crossover through zero, but now the constant supply current will partly aid, and partly oppose the circulating tank current throughout the ac tank cycle.

This is an extremely important feature of this particular circuit topology. It can work safely and continuously into an extremely reactive off resonance load, with either leading or lagging power factor without any additional extra circuit stress. It can also work continuously into a dead short, as it must when the circuit starts up initially.

The phase locked loop will always track the tank resonant frequency, or at least attempt to do so. But for safe and reliable operation, the power circuitry absolutely must be able to take care of itself under worst case transient conditions.

Yes indeed, tank resonance will certainly be strongly effected by the metal load, and especially ferrous materials below the curie temperature. From my reading I believe +/- 10% frequency change is not unrealistic. But I have yet to experience any of this this for myself.

Circuit losses are unavoidable, but the extra voltage drop of one diode is not that significant in a high voltage circuit. Assuming 440v supply and 34 amps (15KW) if the diode drops one volt at 35 amps that is only 35 watts per diode additional loss. 70 watts total for two diodes (only two are working at any instant), just under 0.5%. That is not terribly serious. One solution to this problem is to run the whole circuit at a much higher dc supply voltage. While quite possible, it is not something I really wish to explore at this stage.

Warpspeed,
I like it. I hope you will let us know how things progress.

Determining tank frequency can be easily achieved with a 9V battery and an oscilloscope. Just make sure that you disconnect the tank from the inverter.
Momentarily connect the 9V battery across the tank. You will notice a sinusoidal diminishing waveform across the tank. This will be at Resonance.
I have found this to be very useful when traveling around the country, since a 9V battery is much smaller and lighter than carrying a function generator. :rainfro: This method also gives you a feel for the Q of the tank. The rate of diminishing is directly related to the Q, the higher the Q the faster the waveform amplitude will diminish.

Shawn, that is absolutely brilliant!

I have never before worked on anything of such low impedance level that had such a high Q. By my rough calculation it should have an unloaded Q of around maybe 260 which is absolutely huge. Still waiting patiently to get my 21uF tank capacitor, until that arrives I feel the whole project is really going nowhere.

Actually obtaining the parts for this has been a real emotional roller coaster.

I bid for a beautiful 10KVA 240v/120v three phase autotransformer on e-bay. I put in a bid about a minute before the close, but was beaten in the last five seconds. I finally recovered from the disappointment of that, to be contacted by the seller about a week later. He said he had three more identical transformers, and was I still interested ? I picked one up for LESS than my e-bay bid. He just wanted to get rid of them and planned to sell the remaining two for scrap copper!!!

I will tell you about the long and tortuous saga of the tank tuning capacitor later, if and when it finally arrives. So far I have paid twice for the capacitor, and three times for the UPS freight costs half way around the world. It is starting to become a very expensive component. But without it, I am sunk.

Until I actually have all the large major parts here in front of me, I cannot really begin to plan a mechanical layout and start to actually assemble things. That is why there are yet no pictures, there is really nothing interesting to show. The design really revolves around the parts I can get, particularly the mains transformer and tank capacitor.

Operating voltage, operating current, power level, and frequency are determined more by the parts I can actually obtain, rather than what I may prefer to have.

Anyhow, although progress may be rather slow, I will certainly keep everyone fully informed as to progress.

I'd like to use an induction heater to solder some small brass parts we use in our products, so I'm following this closely.

I would guess that this design will scale down quite nicely.

One thing I'm missing though is that I don't see the computer. I would probably put a micro in the system with some LED or LCD displays and some buttons. I would want to set the operating parameters that way.

Also, I would probably close the phase lock loop with the microprocessor. But, hey, I'm mostly a digital guy.

Ken

Ken, it is the concept, and the circuit topology that is important, and understanding exactly how it is all supposed to work. Scaling it down should be no problem.

My aim is to design a very simple, robust, no frills circuit that others can copy, or at least gain some useful ideas from. That is if I can actually get this working reliably.

Why a microprocessor ? All I plan to have is a big round knob to control the tank voltage and a mechanical amp meter to show the output power level.

Realize that a microprocessor and all it's I/O will be susceptible to electrical noise pickup, and any unexpected malfunction of either software or hardware could cause sudden catastrophic failure of the power devices.

If the current is rising somewhere at 200 Amps per microsecond, servicing an interrupt is simply not going to save you in time !!!!!!!!!!!

Quote:

---

Originally Posted by lerman

I'd like to use an induction heater to solder some small brass parts we use in our products, so I'm following this closely.

I would guess that this design will scale down quite nicely.

One thing I'm missing though is that I don't see the computer. I would probably put a micro in the system with some LED or LCD displays and some buttons. I would want to set the operating parameters that way.

Also, I would probably close the phase lock loop with the microprocessor. But, hey, I'm mostly a digital guy.

Ken

---

Ken,

Unless you are running a dedicated micro that is not running an operating system I think you will cripple the phase locked loop in terms of response time. Even then I think it adds complications. I reckon that having a micro doing supervisory and display functions should be ok though.

AWEM

O/k guys., latest update on progress with the induction heater.

I have finally received my tank tuning capacitor, in fact two of them.

This has all been a rather interesting if not a stressful exercise. It is a rather long and involved story, but Celem of Israel initially sent me the wrong value capacitor, which I promptly sent back. Only it was lost somewhere in transit. After much frantic worry they agreed to send me another one, trusting that I was not in fact some con man trying to pull a swifty. They were really nice about the whole thing, and I decided to order and pay for a SECOND tank capacitor. I have now received both, so the whole project can now proceed in earnest.

A second capacitor allows me the possibility to explore a lower operating frequency in the future, have a spare, or try various other things. So it is not really an extravagant waste.

I have also now received six pulse transformers, another fairly important component of the design. These pulse transformers are rated at 400 volt/microseconds with a 100nS rise time and 2.2mH inductance, with a 1:1 ratio . They allow me to drop fairly low in operating frequency if required, (either by design or mishap), as well as having a particularly fast edged square wave response. These were supplied by Schurter, part number ITRA-0239-D502.

http://www.schurterinc.com/pdf/english/typ_IT.pdf

I have finally bitten the bullet, and decided to switch to oil cooling instead of water cooling, and completely avoid any potential ionic corrosion problems. First I though I would use an oil pump salvaged from a Mazda rotary engine that I already had here in my junk pile. But it pissed oil everywhere, which does not matter when it is completely enclosed inside an engine sump. To seal it properly would be a lot more work than I really want to do at this stage. My next attempt will be with a power steering pump off something or other, which may also conveniently have it's own small oil reservoir. I am guessing it might handle the required gallon a minute or so, of flow, powered by something small like a sewing machine motor, which I also do not yet have. I have my evil eye on a candidate pump on e-bay, so we shall see how it goes.

Meanwhile I have been doing some design work on the constant current supply and control system that will feed power to the the H bridge. All the proprietary switching power supply control chips that I have looked at so far, (and tested), have fairly significant problems and/or limitations of one sort or another for this application. So I have decided to use discrete "normal" chips to do the job instead. More on this later.

You might want to take a look at zone pumps for hydronic heating systems.

Grainger has some 1/25 HP pumps for under a hundred dollars. Some advantages of these pumps include:

* Designed to be used at over 120F.
* Designed for long life with continuous operation.
* Designed to be leak free.

They are NOT designed for use with oil.

Ken

You may find a power steering pump has internal valves which restrict the flow and make it too difficult to drive with a small motor. Look for a regular hydraulic pump; you should be able to get by with a used one that leaks too much for hydraulic use but still has good flow at low pressure.

I have several different pumps here already that I have looked at, thought about, and some I have tested. Originally I had planned to use a Holley "blue" electric fuel pump. This is a small positive displacement vane pump, with an internal pressure relief valve. Depending on voltage, it pumped a measured half gallon per minute of water through my heat exchanger. It would be ideal in many respects for a water based induction heater coolant system. But it does use exposed steel internal parts, so the coolant would need to be of the corrosion inhibiting type.

http://www.holley.com/12-802-1.asp

Other pumps I have here were considered, included a hydronic hot water circulating pump, that I had used previously in a solar domestic hot water heating project. It simply did not develop sufficient pressure to pump enough flow through my heat exchanger which has roughly sixty 3/8" tubes, each just over a foot long all connected in series. Ditto with a marine fresh water pump and a washing machine pump. I even tested the coolant pump from my lathe. These centrifugal pumps have vastly more open flow than I need, but insufficient available pressure unless driven really fast with a grossly overpowered motor.

One possible candidate for the job might be an industrial spray pump. These use a wobble plate driving multi diaphragms, can develop at least 40psi and have roughly about the right flow. But I don't have one here to test, and they may not like oil. But may be worth a passing thought for for a water based system in a moderate to medium sized induction heater. Temperature is not a problem, my radiator cools the water down to only a few degrees above ambient. So a plastic bodied pump should be perfectly o/k.

Oil has roughly the same specific heat as glycol, (about half that of water), and much higher viscosity than water. So I need around twice the flow at a greater developed pressure. Oil really requires a gear or vane pump of some type. A small hydraulic pump, engine oil pump, or power steering pump would seem ideal. I ran an engine oil pump in my lathe experimentally, and it needed about 700 rpm to do the required job, but it leaked horribly. Thinking about that a bit more, the pump could be enclosed inside a sump or oil reservoir where leakage does not matter, and driven from a vertical shaft.

I don't have any hydraulic pumps here to try, it is not something I have ever played around with. Looking on e-bay, hydraulic pumps seem to be rather expensive, but would certainly otherwise be ideal.

I have previously dismantled a power steering pump, and know pretty much what to expect inside. They usually use a multi vane positive displacement rotor, considerably larger than the Holley fuel pump, so should easily develop enough flow at probably only a very few hundred Rpm. Steering pumps also normally run quite happily at several hundred psi of pressure, limited by an internal pressure relief valve. I can easily change the spring, and see no reason why it cannot work well at a much lower outlet pressure with comparatively unrestricted open flow. Being pulley driven makes driving and speed changing easy, and having it's own built in small oil reservoir is a bonus. And it will be designed not to leak.

Don't forget when comparing oil to glycol filled systems, that the glycol is actually an aqueous solution with about 20% glycol & 80% water.

Oil will obviously be inflamable so the slightest leakage in a hot spot may cause problems fast ! (But so does water !!!!)

AWEM

I realize that a greater flow will be required with oil to carry away the heat, but I don't see that as being a very major issue.

Yes Andrew, I can well imagine what might happen if a bit of molten iron gets splashed onto the very thin copper tank coil. Oil or water cooling, the situation would be rather serious.

I have thought all this through, and plan to keep the oil volume to an absolute minimum beyond that required to fill the system. One little sqirt somewhere, and with luck the pump will then run dry fairly quickly. While an oil leak somewhere will create a horrible mess, it would be nowhere near as destructive (or dangerous) to the electrical circuitry as a water leak would be.

Another idea might be to encase the tank coil in crust of castable refractory, allowing for some slight movement of the copper coil within the refractory. That should provide both electrical and thermal insulation, and some mechanical protection of the tank coil. Is this normal practice, or is there some reason why this may not be a good idea ?

Quote:

---

Originally Posted by Warpspeed

....Another idea might be to encase the tank coil in crust of castable refractory, allowing for some slight movement of the copper coil within the refractory....

---

This prompts the question:

How do you intend to protect your coil if it is not encased in a refractory covering?

On the commercial equipment I used the coil was embedded in refractory material and there was no way molten material could come into contact with it, splashing or otherwise.