New Idea - IGBT Based Induction Water Heater

 New Idea - IGBT Based Induction Water Heater 

The basic components are the inverter, driver, coupling transformer and RLC tank circuit. I'll show you the schematics in a little bit. Let's start with the inverter. This is an electrical device that changes the direction of DC current to AC current. For a high-power unit this must be robust. Above you can see the shielding that is used to protect the mosfet gate drive from any stray EMF. Stray EMF causes noise, which results in high-frequency switching. This leads to overheating and failure of the mosfet.

The high-current traces on the circuit board are underneath. Many layers of copper are used to allow them to carry over 50A of current. You do not want them overheating. Also note the large aluminum water-cooled heat sinks on each side. This is needed to remove the heat generated by the mosfets. I originally used fan-cooling, but to deal with this power I have small pond pumps moving plain water through the aluminum heat sinks. As long as the water is clean there should be no conduction. I also have thin mica pads underneath the mosfets to ensure there is no conduction through the sinks.

This is the schematic for the inverter. The circuit is really not that complicated. An inverted and non-inverted driver moves a 15v voltage high and low to set up an alternating signal in a gate-drive transformer (GDT). This transformer isolated the chips from the mofsets. The diode on the mosfet gate acts to limit spikes and the gate-drive resistor minimizes oscillations. Capacitor C1 removes any DC component. Ideally, you want the fastest rise and fall times on the gate, reducing heat. The resistor slows this down, which seems wrong. However, if the signal is not dampened you will get overshoot and oscillations, which will destroy the mosfet. If you look up "snubber circuit" you will get more information.

Diodes D3 and D4 help protect the mosfets from reverse currents. C1 and C2 provide non-shorted pathways for the current to flow during the switching process. T2 is a current transformer so the driver, which we will talk about next, can get feedback about the current going to the tank.

Wow. That's one big schematic. Well, you can read about a simple, low-power inverter. If you want the big power you need a competent driver. This driver will lock onto the resonant frequency all by itself. As your metal melts it will stay locked to the correct frequency without the need for any adjustment.

If you have ever built a simple induction heater with a PLL chip you will probably recall tuning the frequency as your metal heats. You would watch the waveform move on the oscilloscope. You would keep changing the clocking frequency to maintain that perfect point. No need to do that anymore.

This circuit uses an Arduino microprocessor (uP) to follow the phase difference between the inverter voltage and the tank capacitor. Using this phase it calculates the correct frequency using a C algorithm.

I will walk you through the circuit:

The tank capacitor signal comes in on the left to LM6172. This is a high-speed inverter that converts the signal to a nice, clean square-wave. Got to be clean. This signal is then isolated using the optical isolator, FOD3180. These isolators are key! This signal goes to the PLL through the PCAin input. This is compared to the inverter signal on PCBin, which drives the inverter from VCOout. The Arduino finely controls the PLL clock using a 1024-bit pulse-width modulated signal. The two-stage RC filter converts the PWM signal into a simple analog voltage that goes in at VCOin.

How does the Arduino know what to do? Magic? A good guess? No. It gets the phase-difference information about PCA and PCB from PC1out. R10 and R11 limit the voltage within 5 voltage for the Arduino, and the two-stage RC filter cleans the signal from any noise. We want clean spanking signals because we don't want to pay more money for expensive mosfets after they blow up from noisy inputs.

Take a breath

That was a lot of information. You may be asking yourself do you need such a fancy circuit? The answer is "it depends". If you want a self-tuning circuit then the answer is yes. If you want to manually tune the frequency then the answer is no. You can build a very simple driver with just a 555 timer and use an oscilloscope. You can get a little more sophisticated and add a PLL (phase-lock-loop) chip.

Anyway. Let's continue.

There are a few approaches for this part. If you want a high-power heater you will need to have a capacitor array to handle the current and voltage.

First, you need to determine what operating frequency you will use. Higher frequencies have greater skin effect (less penetration) and are good for smaller objects. Lower frequencies are better for larger objects and have greater penetration. Higher frequencies have greater switching losses, but there is less current going through the tank. I choose a frequency near 70khz and wound up with about 66khz. My capacitor bank 4.4uf and can handle over 300A. My coil is near 1uH. The capacitors are from Illinois Capacitors. They are pulse film capacitors. They are axial lead, self healing metallized polypropylene, high voltage, high current, and high frequency. Mine are 0.22uf/3000vdc. The model number is 224PPA302KS.

I used two copper bus bars and drilled matching holes on each side. I used plumber's solder and fixed the capacitors to the bars. I then ran copper tubing on each side to carry cold water to the coil.

Do not get cheap capacitors. They will self-destruct and you will pay more money than if you did the right thing the first time. Here is another thought. You can buy a Celem capacitor. Although they are expensive, the cost is not that far off from the cost of building a good capacitor array. Trust me, I've been there already.

I've posted a picture of a Celem capacitor for your viewing pleasure.

Please note that the Celem is water-cooled. Whether you use the Celem or make your own you need to water-cool these units. I use the same pump to cool the capacitor as I use to cool the working-coil.

If you have been paying close attention you should be asking how does one drive the RLC tank? I jumped earlier from the invert to the tank without mentioning how they are connected.

The connection is through the coupling transformer. I got mine from Magnetics, Inc. The part number is ZP48613TC. Adams Magnetics is also a good choice for buying these ferrite toroids.

The one on the left uses 12g wire. This is good if your input current is under 20A. The wire will overheat and burn up if it is more. For the real high power you need to buy or make Litz wire. I made mine by braiding 64 strands of 23g magnet wire. This stuff could withstand 50A current with no problem.

The inverter, which I showed you earlier, takes high-voltage DC and chops it into alternating high/low values. This alternating square-wave goes through this coupling transformer via the mosfet switches and the DC coupling capacitors on the inverter. The copper tubing from the tank capacitor runs through this making it the single-turn secondary of a transformer. This, in turn, allows the stepped-down voltage to flow through the tank capacitor and work coil (the LC tank).One question I've gotten is how do you make such a nicely curved work coil? The answer is sand. The sand will prevent the tubing from collapsing on itself during the bending process.

Take 3/8" copper refrigerator tubing and fill it with clean play sand. Cap one end with some tape before doing this and cap the other after filling with the sand. Plant a study pole of the proper diameter into the ground. Fix a sufficient length of tubing for your connection and slowly bend the tubing around the pole. Once you get one turn around the rest is easy. Continue to wrap the tubing until you get the number of desired turns (4-6 is usually good). The second free end so it is lined up with the first. This will make your connections to the capacitor easy.

Now, remove the caps and take an air compressor to blow out the sand. Did I mention to do this outside?

Please not that the copper tubing also serves the function of carrying water for cooling. This water circulates through the capacitor tank (or Celem) and through the working coil. The work coil generates a lot of heat from the current. Even if you use ceramic insulation on the inside of the coil (to trap the heat from escaping) you will still have extremely high temperatures on the workpiece heating the coil. I would start out with a large bucket of ice-water and reduce this to hot water in a short time. Point: have plenty of ice to keep this cool.

Above is an overview of the 3kw unit. It has a simple PLL driver, an inverter, coupling transformer and tank. The video demonstrates the 12kw unit at work. The main difference is that it has a microprocessor controlled driver and larger mosfets and heat sinks. The 3kw unit runs off of 120vac; The 12kw unit uses 240vac.

 New Idea - IGBT Based Induction Water Heater 

The basic components are the inverter, driver, coupling transformer and RLC tank circuit. I'll show you the schematics in a little bit. Let's start with the inverter. This is an electrical device that changes the direction of DC current to AC current. For a high-power unit this must be robust. Above you can see the shielding that is used to protect the mosfet gate drive from any stray EMF. Stray EMF causes noise, which results in high-frequency switching. This leads to overheating and failure of the mosfet.

The high-current traces on the circuit board are underneath. Many layers of copper are used to allow them to carry over 50A of current. You do not want them overheating. Also note the large aluminum water-cooled heat sinks on each side. This is needed to remove the heat generated by the mosfets. I originally used fan-cooling, but to deal with this power I have small pond pumps moving plain water through the aluminum heat sinks. As long as the water is clean there should be no conduction. I also have thin mica pads underneath the mosfets to ensure there is no conduction through the sinks.

This is the schematic for the inverter. The circuit is really not that complicated. An inverted and non-inverted driver moves a 15v voltage high and low to set up an alternating signal in a gate-drive transformer (GDT). This transformer isolated the chips from the mofsets. The diode on the mosfet gate acts to limit spikes and the gate-drive resistor minimizes oscillations. Capacitor C1 removes any DC component. Ideally, you want the fastest rise and fall times on the gate, reducing heat. The resistor slows this down, which seems wrong. However, if the signal is not dampened you will get overshoot and oscillations, which will destroy the mosfet. If you look up "snubber circuit" you will get more information.

Diodes D3 and D4 help protect the mosfets from reverse currents. C1 and C2 provide non-shorted pathways for the current to flow during the switching process. T2 is a current transformer so the driver, which we will talk about next, can get feedback about the current going to the tank.

Wow. That's one big schematic. Well, you can read about a simple, low-power inverter. If you want the big power you need a competent driver. This driver will lock onto the resonant frequency all by itself. As your metal melts it will stay locked to the correct frequency without the need for any adjustment.

If you have ever built a simple induction heater with a PLL chip you will probably recall tuning the frequency as your metal heats. You would watch the waveform move on the oscilloscope. You would keep changing the clocking frequency to maintain that perfect point. No need to do that anymore.

This circuit uses an Arduino microprocessor (uP) to follow the phase difference between the inverter voltage and the tank capacitor. Using this phase it calculates the correct frequency using a C algorithm.

I will walk you through the circuit:

The tank capacitor signal comes in on the left to LM6172. This is a high-speed inverter that converts the signal to a nice, clean square-wave. Got to be clean. This signal is then isolated using the optical isolator, FOD3180. These isolators are key! This signal goes to the PLL through the PCAin input. This is compared to the inverter signal on PCBin, which drives the inverter from VCOout. The Arduino finely controls the PLL clock using a 1024-bit pulse-width modulated signal. The two-stage RC filter converts the PWM signal into a simple analog voltage that goes in at VCOin.

How does the Arduino know what to do? Magic? A good guess? No. It gets the phase-difference information about PCA and PCB from PC1out. R10 and R11 limit the voltage within 5 voltage for the Arduino, and the two-stage RC filter cleans the signal from any noise. We want clean spanking signals because we don't want to pay more money for expensive mosfets after they blow up from noisy inputs.

Take a breath

That was a lot of information. You may be asking yourself do you need such a fancy circuit? The answer is "it depends". If you want a self-tuning circuit then the answer is yes. If you want to manually tune the frequency then the answer is no. You can build a very simple driver with just a 555 timer and use an oscilloscope. You can get a little more sophisticated and add a PLL (phase-lock-loop) chip.

Anyway. Let's continue.

There are a few approaches for this part. If you want a high-power heater you will need to have a capacitor array to handle the current and voltage.

First, you need to determine what operating frequency you will use. Higher frequencies have greater skin effect (less penetration) and are good for smaller objects. Lower frequencies are better for larger objects and have greater penetration. Higher frequencies have greater switching losses, but there is less current going through the tank. I choose a frequency near 70khz and wound up with about 66khz. My capacitor bank 4.4uf and can handle over 300A. My coil is near 1uH. The capacitors are from Illinois Capacitors. They are pulse film capacitors. They are axial lead, self healing metallized polypropylene, high voltage, high current, and high frequency. Mine are 0.22uf/3000vdc. The model number is 224PPA302KS.

I used two copper bus bars and drilled matching holes on each side. I used plumber's solder and fixed the capacitors to the bars. I then ran copper tubing on each side to carry cold water to the coil.

Do not get cheap capacitors. They will self-destruct and you will pay more money than if you did the right thing the first time. Here is another thought. You can buy a Celem capacitor. Although they are expensive, the cost is not that far off from the cost of building a good capacitor array. Trust me, I've been there already.

I've posted a picture of a Celem capacitor for your viewing pleasure.

Please note that the Celem is water-cooled. Whether you use the Celem or make your own you need to water-cool these units. I use the same pump to cool the capacitor as I use to cool the working-coil.

If you have been paying close attention you should be asking how does one drive the RLC tank? I jumped earlier from the invert to the tank without mentioning how they are connected.

The connection is through the coupling transformer. I got mine from Magnetics, Inc. The part number is ZP48613TC. Adams Magnetics is also a good choice for buying these ferrite toroids.

The one on the left uses 12g wire. This is good if your input current is under 20A. The wire will overheat and burn up if it is more. For the real high power you need to buy or make Litz wire. I made mine by braiding 64 strands of 23g magnet wire. This stuff could withstand 50A current with no problem.

The inverter, which I showed you earlier, takes high-voltage DC and chops it into alternating high/low values. This alternating square-wave goes through this coupling transformer via the mosfet switches and the DC coupling capacitors on the inverter. The copper tubing from the tank capacitor runs through this making it the single-turn secondary of a transformer. This, in turn, allows the stepped-down voltage to flow through the tank capacitor and work coil (the LC tank).One question I've gotten is how do you make such a nicely curved work coil? The answer is sand. The sand will prevent the tubing from collapsing on itself during the bending process.

Take 3/8" copper refrigerator tubing and fill it with clean play sand. Cap one end with some tape before doing this and cap the other after filling with the sand. Plant a study pole of the proper diameter into the ground. Fix a sufficient length of tubing for your connection and slowly bend the tubing around the pole. Once you get one turn around the rest is easy. Continue to wrap the tubing until you get the number of desired turns (4-6 is usually good). The second free end so it is lined up with the first. This will make your connections to the capacitor easy.

Now, remove the caps and take an air compressor to blow out the sand. Did I mention to do this outside?

Please not that the copper tubing also serves the function of carrying water for cooling. This water circulates through the capacitor tank (or Celem) and through the working coil. The work coil generates a lot of heat from the current. Even if you use ceramic insulation on the inside of the coil (to trap the heat from escaping) you will still have extremely high temperatures on the workpiece heating the coil. I would start out with a large bucket of ice-water and reduce this to hot water in a short time. Point: have plenty of ice to keep this cool.

Above is an overview of the 3kw unit. It has a simple PLL driver, an inverter, coupling transformer and tank. The video demonstrates the 12kw unit at work. The main difference is that it has a microprocessor controlled driver and larger mosfets and heat sinks. The 3kw unit runs off of 120vac; The 12kw unit uses 240vac.