This is a page in progress!!

This is my best approximation of the schematic for the popular 12 - 48 VDC 1000 Watt ZVS Induction Heater
(currently being sold on Ebay and Amazon). This is the one that has 6 each, 0.33 mfd capacitors.
Some of the notes are from the original author. Although I have thoroughly traced out this circuit,
I take no responsibility for the accuracy of this drawing.
You must verify it for yourself.

Here are the main points that I want to cover:

1. The Glowing Color

2. Chinglish Instructions

3. Steel's temperature stops increasing

4. Frequency Too High

5.1. Why it breaks, what breaks, and how to fix it

Pete's 1000 Watt ZVS Induction Heater Troubleshooting Guide

5.2. How I am attempting fix them

Some of my ZVS Induction Heating Toys

Update December 3, 2016:
Since the crucible's coupling caused such high current to flow in the system, I made a smaller graphite crucible from a 1 inch diameter graphite rod that I had in stock, so I could see what difference poorer coupling might make.

The larger Crucible is the one that comes with the induction heater.
I made the small one by simply boring a 5/8" hole in a one inch rod graphite rod.

Note the difference in diameter of these two work coils.
I already did understand that one needs to fit the work coil to the job,
but I was about to find out just how sensitive this particular ZVS system is
to a change in coil size.

December 8, 2016:
I repaired two of my boards. The general approach is to replace all 4 diodes and both Mosfets, plus anything else that is obviously wrong.
I'd call that "shotgunning", but that's the way it is for now. becuase, even though I attempt test or at least ohm out the parts that I remove, I am not sure if I have stressed something or not.

Here's the test setup that I used just to see if they worked:

Using a 12 volt garden tractor Battery, Note One Amp at idle.     Output Across Work Coil. Just about 100 Khz

Neither one worked after the parts replacement.
On one of them, I had re-installed a used Mosfet that had tested okay. when I hit that board with 12 volts, it pegged my 10 amp meter.
I will replace that Mosfet and try again.
The other one did absolutely nothing when I fired it up. Upon close examination, I found the the heavy trace between the Positive power input and the "chokes" had actually blown right off the board for about a half inch. That area had been hidden by the toroid above it. I Replaced the missing piece of copper with some de-soldering braid and now it works.
An interesting note:
I decided to 'scope Gate-to-ground for both Mosfets. I was surprised to see that there is measurable difference in the two signals:

The upper signals are Gate to Ground. The lower signals are work coil to drain.
Q1 on the left and Q2 on the right.
Not sure what to make of this. It almost HAS to be a difference in the characteristics of the two Mosfets.
Now I suppose I will have to make the same measurements on a couple of other boards.

6. Operation and Testing

January 12, 2017:
Larger Picture of my 4th ZVS Induction Heater setup
Well, I just got a new test setup built, one where I can use either 24 or 36 volts to supply the ZVS heater PC Board. It worked well enough for me to shoot a youtube video that is good enough to present:
First Youtube Video test of my fourth setup
I twisted the power leads from the batteries as much as I could to cancel out stray inductance, attempting to clean up the gate drive signals.
That my have worked, since the Mosfets stayed pretty cool during this test. But I still think I need to 'scope the gates.
Now I will have to start playing around with various sizes of work coils.
Or, I could begin to work on reducing the operating frequency.
Or----?

January 16, 2017, 'scoping the Mosfet Gates:
This is a report on the the Gate drive signals from my latest setup. In my Jan 12 post, above, I explained that I had twisted the power leads.
It appears that the twisting made a BIG improvement. Here is the result:

Gates with 24 volts applied to the circuit, with circuit "idling"at about 2 amps.


Gates with 24 volts applied to the circuit, with circuit loaded to 8 amps.


Gates with 24 volts applied to the circuit, with circuit loaded to 22 amps.

There is hardly any change at all in gate drive wave shape at different currents.
Now compare these gate drive signals to the pictures in my December 8 post. You can easily see that the gates are "full on" for a much longer percentage of the time with this setup.
At any rate, those Mosfets sure do stay a LOT cooler this way!
If you look carefully at the waveforms, you can see that the tank frequency slows down a bit as the current goes up. I estimate that the frequency drops from about 100 KHz at "idle" to about 90 KHz at about 25 amps.
But---- back to the wave SHAPES--- it looks like there's one other possibility for the poor wave shapes from Dec. 8. There I was only driving the circuit with a 12 volt garden tractor battery. What if that voltage wasn't enough to supply a full 12 volts to the gate by the time it passed the 470 Ohm resistor? Would that situation have kept the Mosfet in its linear region for a longer part of each cycle? Another thing to test, I guess, but I'm pretty happy with the results right now.
Of course, if you look closely at any of the gate traces, you can see that one Mosfet seems to be turned on for a bit longer than the other one.
Is it a difference in Mosfet characteristics? Variations in the actual recovery speed of the "FAST" diode? Hmmmmm.

January 26, 2017: Actually reading frequency and an interesting 36 Volt observation!
My third youtube Video, Last Attempt to Understand the "Glowing" Phenomenon
This time I have added a cheap (USD$50) oscilloscope to read one gate, but this 'scope also gives direct frequency read outs and that's why I attached it.
Here's how the instrumentation looks right now:

See the little Orange colored device at the upper right? That's the 'scope that reads frequency directly.
Here's a picture of its current reading:

The frequency drops to about 83 KHz with when I insert the 5/16" by 3/8" steel bar into the large work coil that you see here.
Now that I have a way to actually measure frequency, I can start to add Capacitance and Inductance to the resonant tank circuit and know what the results are as I attempt to get steel up to higher and higher temperatures.

Now, HERE'S the latest thing----
I just now added some tie points on the system to make it easier to connect to the gates so I can both read frequency and 'scope both gates on my "real" oscilloscope.
So, Here's what the gates look like at 36 volts:

Note that the gate signal is almost a square wave!!!!
That's great, becuase it means that the Mosfets will turn on FULLY more quickly!!!!!
Which means they won't heat up as much as they would if they turned on slower. To me, this is a BIG deal. If you look back to earlier posts on this page, you can see how poor the gate turn-on waveform is at 12 volts.
You can see that it gets appreciably better at 24 volts.
And now--- it's even BETTER at 36 volts.

February 9, 2017: I finally got to 48 volts and I put up two more videos. See the new video links at the top of the page.

Here's the gate drive story to date:

March 6, 2017: Getting the Resonant Frequency of the system lower and lower
I have read that lower frequencies make the energy penetrate deeper into the work. For ferrous metals, anyway. So, to get the maximum efficiency of energy transfer to ferrous metals (iron and steel, mostly), I need to get the frequency down a lot lower than it is with the "stock" 1000 Watt 12 to 48 volt ZVS Induction Heaters that I am playing with.
See "Part 6" at the top of this page for my video of finally getting the frequency down to the 38 to 40 KHz range. I did it by doubling both the capacitance and the inductance. I doubled the capacitance the same way I did earlier, but I wound a 12 turn copper coil for this test.

But, since I was only using my 12 volt setup for this test, I could only draw about 8 amps of so.
So the next step will be to duplicate that test on the 48 volt air and water cooled system.

March 10, 2017: I recently went a little off-track by performing my first MOT (Microwave Oven Transformer) rewind.
You can see the result as the "Part 7" Video, above.
I started that process so I could build a power supply for this induction heating project. I'm not there yet, but I think I can come pretty close to a 48 volt secondary winding with a single 14 gauge enameled secondary or maybe 3 or 4 16 gauge enameled "bundled" wires.
As I have mentioned earlier, I already have a nice 48 volt, 26 amp commercial power supply, so I don't need to rush into this job. But I DO want to get it done!
I don't know how many different sizes of MOT are out there. I already have one that is larger than the one I rewound for this video. And I will continue to look for larger ones. I used this smaller MOT just to get some experience with using enameled wire.

Recently, a friend asked several good questions about that process. I answered him, but thought I'd add those answers here:
Pete, I just watched your MOT coil winding video with the four strands of wire run together. This was very interesting and helpful to see it done. I really appreciated your charts with the results shown at the different resistances.
It looks like this will be a very strong power supply when finished.
Ans: Thank you. I just got an even bigger transformer to work with, too.
Remember, though, this was just a method of getting some hands-on experience with the MOT and the enameled wire. I don't know if I will ever use that one in a project, but at least I do know what its capabilities are.

Question 1:
- How did you connect the two sections of wire together?
Ans: I bolted the ring terminals of the first and second winding sections together.

Question 2:
Did you scrape off the enamel and then use in-line crimp type connectors?
Ans: Yes and Yes. That particular wire was from a local electric motor repairs shop and of pretty high quality insulation.
After I had thought that I had sanded all the "varnish" off (with 150 grit emery cloth, I found that I still couldn't tin it. So, looking more closely at the "stripped" ends, I saw that I hadn't really gotten all the way through. So that's when I got out the razor blade and went at it again. Then sanded some more. But with 80 grit emery cloth. After all that, the wire tinned just fine.

Question 3:
Did you solder them also?
Ans: Yes, I did. I soldered all four leads together, crimped the terminal onto them and then soldered the terminal to the wires.
I learned this requirement the hard way the first time I added capacitors to the induction heating circuit.

Question 4:
- Do you think the four wires together are working the same as a single larger gauge wire would in this same application?
Ans: In general, yes. However, I had to use 19 gauge wire for this one because the motor shop didn't have anything heavier.
If you look up ampacity tables for copper wire, I think you will find that 19 GA comes in at about 1.9 amps. I originally wanted to emulate 12 GA, at 20 amps, so technically, I am off by a factor of 20/7.6= 2.63. But it didn't seem to be much of an issue. Those tables make some assumptions about heating due to insulation, etc.., so I think there is some room to fudge. (I had wanted 16 Ga, by the way).

Question 5:
- Have you done any winding resistance tests to see if you can calculate the turns ratio, primary to secondary, required to produce 48 volts?
Ans: The math would have to do with the efficiency of the transformer core and some rather esoteric considerations that are currently beyond me.
Besides, I saw one video where the guy simply wound 10 turns on and measured the output voltage. He calculated the volts per turn and went on from there. It worked out pretty close for me, too.

Question 6:
Instead, could you have just unwound part of the existing secondary winding to get to around 48 volts?
Ans: Technically, I think that would work, but once you see how fine that wire is, I think you will abandon that Idea.
Those secondaries produce anywhere from 2000 volts to 4000 volts (like the ancient AMANA RADAR RANGE that I just tore apart).
At 0.7 volts per turn, as an example, there'd be about 2800 turns on the secondary. You'd have to unwind the first 2750 turns or so to get into the range of usefulness, right?

Question 7:
---Or, would that finer wire on the secondary not have produced the amount of amps you wanted?
Ans: Corrrect, It wouldn't come anywhere near handling enough current.
I'd estimate it at 36 ga. or even smaller. That's about 0.21 amps on the ampacity table, about 1% of what I need.

Quesstion/Comment 8:
I was at the Tractor Supply Store in Inver Grove Heights today and saw that they have #4 and #2 gauge multi-stranded wire for making replacement welding leads by the foot. It does however have very thick rubber insulation so the turns will be quire limited.
Ans/Reply: Yes, that thickness of insulation IS the problem if you are trying for higher voltages. But those big wires seem to be exactly what the guys who are making spot welders are using. For a spot welder, I think they use only 2 turns. But I don't think you'd use one of them to weld your car body to the frame.

March 23, 2017:
I have recently moved the "40KHz" parts up to the 48 volt system.
I had previously shot 3 segments for a next video, but I wasn't quite ready to assemble and publish it yet. Today I got my cooling water too hot, so I need to add a fan to the radiator. I also need to baffle the fan I added to cool the extra capacitors because it blows too hard on the work coil, cooling the work inside it.

A day later: I added a bit that video, edited it some more and have decided to publish it. See "Part 9" at the top of this page.
Some of you will find it useful, but it is getting pretty long (currently just under 36 minutes).
Sometimes I think it's good to show things in real time, but I appreciate that it might bore some people, so I have taken my best guess for correct balance.

The good news is that I am still using the same ZVS heater pc board that has been on Setup #4 since ever since I built it up with the switch panel. And that's while I have been abusing it at 24, 36 and now 48 volts. In today's tests, I even had unplanned current excursions up to 35 amps or so with no ill effects.

March 30 and 31, 2017:
On March 30, I updated my setup by:
- adding a water cooling fan to the heat exchanger, with an exit air temperature meter
- adding an air deflector for the fan that cools the added capacitors
- removing the water pump and cooling fans from the main power supply current shunt (so I'd be reading system power more accurately)
        Note: In all past tests, the 2 amps that these components draw have been adding to the "apparent" idling current.
-added a meter to read Mosfet temperature
-and a few other "cleanup" items.

Now I want to see what advantage there is, if any, of running at 40KHz vs. approximately 100 KHz.
So I ran and video taped an extensive test, with this goal in mind. During this test, I retested every metal that I had acquired data for in past tests.
For most metals, I simply checked their initial current draw.
Everything stayed nice and cool, but I wasn't pushing for long periods of high current draw.
At the end of that test, I forged a couple of pieces of steel that had been heated in my setup, just to show that it can be done.
That video is Part 10 at the top of this page.

On March 31, I ran a non-video taped test to see how the new cooling system would respond to continuous higher currents. This test lasted 15 minutes, with an average current draw of 16 amperes at 48 volts, producing a total of 768 watts.
I was very pleased with the results. Everything stayed nice and cool.
On April 1, I reran that test, but I pushing the average current up to about 21 amps for the whole 15 (actually 17) minutes.
Here is a link to the Excel spreadsheet that details March 31 test:
15 Minute Test at 16 Amperes

April 2, 2017:

The focus of this video is on running the system at 1000 watts total work coil current at 20 amps or more for a full 15 minutes (actually about 17 minutes) to get a complete temperature profile for the system.
The weight of the steel that was being heated in the work coil was 290 Grams. or about 10.2 ounces.
It is Video Part 11 at the top of this page.
I suppose the next thing to do is to return this exact setup to a frequency of 78 KHz by removing the second set of capacitors and installing the 6 Turn 2 inch ID work coil. Then I should and rerun the same, identical test so we can determine the value (if any) of 43 KHz operation.

April 15, 2017:

On April 9th, I a test similar to the previous one, but at 87 KHz, so I could compare 87 KHz runs with 43 KHz runs. This time, I did measure the temperature of the work, a thing I should have done in the previous test.
It is Part 12, up at the top of this page.
I severely edited this video but also included my comparison analysis of the two tests and more.
I divided this video into 4 sections, so you can see most of the test itself (if you have the time), or, you can view just the comparisons, my thoughts on future process changes and possible practical uses for this particular Banggood ZVS 1000 Watt 12-48 volt Induction Heater.

Probably the best spreadsheet for power and cooling info

July 8, 2017:
This note relates to heating of the capacitors.
Over the past several months, I have watched a number of videos where the author talks about the capacitors heating up a lot.
I have not had much of a problem with capacitors heating. I wondered why others did.
Well, recently I was preparing a simple, portable demonstration using the basic Banggood board, powered by 2 garden tractor batteries, for a total power supply of 24 volts. I was fan cooling the board, but not water cooling the work coil.
After only a minute or so of operation with various ferrous materials in the work coil, drawing 10 to 20 amperes, I realized that the copper work coil tube had gotten REALLY HOT!! It got so hot, in fact, that the longest end of the work coil, which sits directly over one set of capacitors, was tranfering a LOT of heat to those capacitors.
So my point here is that:
-If you don't water cool the work coil, you will seriously heat up the capacitors closest to the long work coil leg!!!

October 23, 2018:
Changing coils without getting water all over the place:
It has always been a messy job to change work coils. First because of the way they are screwed down to standoffs on the board, but mainly because I have to deal with water running out of the cooling system and onto the workspace or the floor.
So, I added a couple of quarter turn ball valves this way:

I think I wanted to go with flare fittings, but I didn't locate any valves that worked that way, so I went with compression fittings.
That's what you see in the picture.
Maybe you already know this, but copper tubing is sized in two different ways. If you go to the plumbing department in the hardware store (USA) and ask for "quarter inch" copper tube,
you will get tubing with a 1/4 inch OD, but if you go into the Mechanical tubing world, I think "quarter inch" refers to the ID.
Anyway, the quarter inch OD tubing is the size I use for all my coils.
I got those valves at Faucet depot:
Faucet Depot Valves for Work Coils
I think that is the one that I used.
I used compression fittings since the ferrules are smaller than the flared part of flare-type fittings, which would have made it easier to insulate the coils with sleeving.
Or, in some cases, to replace the sleeving if I do something bad to it.
BUT!!! I totally forgot that the nuts have to go on before you put the ferrules on. So, you have them in the way for insulating or re-insulating anyway!
So, I guess it really doesn't make much difference which way you go.

May 22, 2019:
Installing a larger cooling fan
When testing Brass Cartridge Casing annealing recently, I noticed that the capacitors on the board were getting warm, which is nothing new and they were not getting anywhere close to "too hot to touch". However, I did realize that the 90 mm diameter fan that I have been using to cool the circuit board all along was not quite large enough to blow on the 2 outer capacitors. So I bought a 120mm fan and replaced it today.

Here you can see the difference.
I still need to add a bulkhead to separate the board from the work coils so the fan doesn't blow on the work coil which can cool the work.
Adding the bulkhead, as I have done on my 2500 watt unit would also be a good idea to take stress off the board's coil standoffs when changing coils

At some points, I have used a second fan to cool additional capacitors when experimenting with doubling capacitors, too.

May 22, 2019 (Continued):
Adding a frequency counter to this system:
A while ago I added a frequency counter to the 2500 watt system so I can see how coil design affects gate signals.
More recently, I added one to this 1000 watt unit too.
It has been very useful when working on the "annealing" work coils.
Here, in this 2500 Watt Video, I describe the freqquency counter in some detail:
Frequency Counter Explained
That video also talks about other 2500 Watt system improvements, too, but I hope it tells you enough about the counter itself.

July 4, 2019:
Brass Cartridge Casing Annealing
I don't do any cartridge reloading myself, but over the last couple of months I have been testing various work coil designs for that process.
There are many folks out there who are already doing this, many using one of these 1000 Watt ZVS Induction heaters. My testing is mainly focused on making coils that heat the casing quickly, while making sure that the heater's components are not overstressed.
I think I have accomplished that goal, as you can see in my youtube videos- Part 13, 14 and 15 at the top of this page.

August 8, 2019:
Testing a large (10 inch OD spiral or "pancake" coil)
I made this coil just to see if it would have any usefullness in annealing thin sheet metal. See the video in Part 16 in the navigation bar.
This work coil has the highest inductance of any coil I have made so far and runs at a much lower frequency. This lower frequency produces very sharp square wave gate signals, which is a good thing to minimize Mosfet heating.

May 6, 2020:
Notes on the negative effects of increasing operating frequency of the ZVS induction heaters:
---------------------------------------------------------------------------------------------
Many of you are probably tired of hearing me talk about being careful to keep the Mosfets from spending much time in the linear mode.
But here I go again:
We can easily see that the idle current varies when we change work coil designs.
From my own work coil data spreadsheet, I recently noticed that idle current and frequency seem to vary together. When I plotted 9 pairs of idle current versus frequency observations, I could see the relationship easily. I got a correlation of about 0.8. Of course "correlation" does not necessarily equal "causation", but it seems reasonable to me. I see this relationship as another indicator that higher frequencies create more lost power within the ZVS circuit components themselves. This, in turn, leads to more stress on, particularly, the Mosfets and the tank capacitors.

But why do we care, as long as things seem to work okay?
Well, here are a couple of reasons:
1. Assuming that you or your system has a maximum safe current limit, the higher the idle current, the less current there is available for heating the work.
In my case, the maximum sustained current that I will allow on my 1000 watt unit is about 22 amps TOTAL. And, for my 2500 watt unit, it's 50 amps, TOTAL.
I hope that is clear to everyone.

But--- I think this item is even more important:
2. The Basic ZVS power circuit that most of these Chinese ZVS induction heaters use works most reliably at a frequency below about 100 kHz. And, the lower, the better, in my view!!!

So, when you design a new work coil, you need to know whether it will overstress the system or not.
You can quickly check it out by powering the system up and noting the idle current. That current is primarily heating up the Mosfets and the Capacitors. Of course it also heats the work coil, but as long as we are water cooling it, we won't worry about that.
If the idle current is higher than about 8 amps, I think you will be stressing the system more than it was designed to be stressed.
See Chart: