Email John
Last update 09/20/2009

A 500 1200 Watt Induction Heater - Update

since 09/16/09
The heater presented on the previous page is known as the parallel resonant LC tuned architecture.  The tank consists of a capacitor and inductor in parallel.  The resulting high impedance, high voltage circuit is matched to the low impedance driver by an L-C network.  This is explained in great detail by Richie. This is a simple architecture but after you work with it awhile, you basically discover that, in the words of Tim Williams, it sucks.

The problem is that the L-C match is only good for one frequency and load condition.  For all other conditions it presents a reactive load to the inverter driver.  That means that the inverter has to supply reactive power (VARs) to the circuit.  The practical real world result is that the transistors run much hotter than necessary and the unit is quite inefficient at other than the design conditions.

The other common architecture used in industry is the transformer coupled, series resonant design.  In this design, the work coil and capacitor are in series and are fed by a transformer whose primary is connected to the inverter.  This architecture presents the inverter with a mostly resistive load under all conditions when the tank is in tune.  This is very good.  The real world practical result is that this little heater's input power jumped from about 500 watts to over 1200 and most of that power ends up in the load instead of heating the transistors and matching network.

I originally shied away from this architecture because I knew that typical industrial heater transformers (heating stations) are big and bulky and heavy. I ignored the fact that as the frequency rises, magnetic parts miniaturize.  After a chat with Tim Williams, I decided to give it another shot. Zounds!

If you build this heater, this architecture is a must-do.  Forget the parallel resonant design.  Old history.  Now let's take a look at some details.




Here is a photo of the new tank.  Doesn't look much different, does it?  A capacitor at the bottom and the work coil standing up, apparently hooked in parallel with the capacitor.  The difference is that stack of ferrite cores hanging off one leg that has several turns of wire wound around it.  That is the transformer.  The primary is several turns of fairly light wire and the secondary is the leg of the work coil that forms a single turn secondary.


If we draw the tank circuit a little differently, then the architecture becomes obvious.  The capacitor and work coil are in series and are fed by the transformer secondary, that length of the coil leg that passes through the core.

Ferrite is not a good material to use at this frequency (300-400khz). It is lossy and has a fairly low Q relative to the proper material, powdered iron.  In other words, it gets HOT. Here is a photo of the tank using two Amidon Formula 15 T157-15 powdered iron cores.  The primary is 15 turns of teflon coated #16 gauge wire.  I recommend going with another couple of turns until you're sure of your FET cooling system. 15 turns loads the FETs right up to their maximum rating. 

There is a considerable amount of heat conducted from the work coil to the transformer so if you don't have teflon coated wire, I recommend inverter grade magnet wire. Lacking that, any wire with >100 deg C rating will do.

With the frequency carefully tuned to resonance, the unit will crank out a solid 1300 watts with 15 turns.  I don't recommend running at that level because it approaches the transistors' absolute maximum ratings.  Adding another turn or two to the transformer primary limits the power to something in the 1000-1200 watt range, yet is safe for the FETs.


Connection is simple. Simply solder the two wires coming from the transformer to the PCB's output terminals.  That's all there is to it.

The inductor that you see in the above photo has a 2 inch inside diameter.  This is a good all-round size for general purpose heating.  For neon, it will heat anything from a large cold cathode electrode to a tiny 8mm unit.  It will also fire a 10 mm getter, though with a little effort.  The original 1" ID coil is better for getter firing.

The 2" ID coil can be wound by hand without making any kind of jig.  I found a can of butane fuel that was the correct diameter.  I simply filled the copper tubing with sand and then rolled the tubing around the can.  2" OD PVC pipe would be another good coil form.

Finally, notice that the work coil is discolored from heat but that the discoloration stops just outside the coils themselves.  That is because the work coil induction-heats itself!  One of my planned experiments is to silver plate a coil to about 1 skin depth and see what difference that makes.  I suspect that it will make a LOT of difference.

The difference in operation is amazing.  The FETs stay cool and calm.  The tuning is very smooth and operation is stable as the load varies.  The photo shows the test board wired up on my bench.  The tiny heat sinks are good for up to about 600 watts.  After that I direct a small muffin fan on them.

One word of warning.  Keep your FETs cool.  That is, not much above room temperature.  A power FET's Rdson (resistance from drain to source when turned on) rises with temperature.  That means that the FET dissipates more power the hotter it gets. That can quickly lead to a positive feedback, runaway condition.  I've popped more than one FET by ignoring my own advice!  Remember, a cool FET is a happy FET!



Finally, to show the awesome power of this heater, this is a photo of a cold cathode neon electrode that has been slagged.  Yep, the heater can melt iron!