for a high power system say 5kW what I was looking at for a long time is
1) IGBT phase control bridge, use phase the adjust power from a H-bridge to a coupling transformer to the tank circuit
2) water cooling (a bitch with these power levels and voltages)
Not a bitch, just a formality. Helps greatly with power density.
Cooling plates with embedded pipes are off-the-shelf, it's pretty good actually. Pair that with an automotive radiator and fan, a small diaphragm pump probably, and a reservoir, and you're set for life.
3) use isolated drivers with individual power transformers to control the H-bridge (iso5500 or beefier), using all its capabilities like DESAT protection etc, make sure the transformers give clean power and the circuits dont interrupt eachother with pulses or cause ground bounce etc
4) analog circuit that handles the PWM frequency and phase shift
5) digital control of analog circuit setpoints (LTC timerblox might be good for this, it is what my control system prototype was going to use) (i.e. weighs and scales the analog part)
6) welded steel enclosure with test points so you don't mess around with high energy shit
3. DC-DC modules are easier, but line frequency transformers with low capacitance (typically split bobbin type) work, too.
5. Just use the canonical analog-digital control scheme: analog takes care of itself, setpoints set by DAC. Operating point read by ADC. MCU does whatever heavy lifting is needed (e.g., calculating output power?).
6. Aluminum can be better, depending on how well you control the fields inside. Even that can be bad, if you put too much field into it. The tabletop industrial unit I designed, was forced into such a path by the management... turns out putting a work coil half the size of the enclosure, inside the enclosure, turns it into a pizza oven. Had to line the enclosure with ferrite plates, in the end.
This is considerably easier to deal with, in a series or parallel resonant network (of which, series resonant is easiest with conventional semiconductors). An LLC ("series tuned parallel resonant") network only has that problem because you need to put that first L
somewhere...Anyway, pipes close together, or better yet laminated bus bar, is the way to go. This never got hot on the front panel (aluminum),
and I ran it at 5kW (Q ~ 15).
What I have not figured out that well is the feedback to control tuning. Apparantly if you put stainless or aluminum into the coil it can decrease inductance and cause excessive power problems. Most of the ones online ignore feedback and over specify everything and require manual tuning of frequency.
Yes, a close-fitting load can halve (or worse) the coil inductance, while dropping the Q only modestly (for the case of copper or aluminum; for stainless, the Q drop is appreciable of course). A steel load can double it (or more), until it passes Curie temp, at which point it flips to the other side (dropping).
So, frequency control is mandatory for any real work.
A typical use case is forging, where work is constantly being placed into, and removed from, the coil. The supply can be simmered or stopped inbetween, but sooner or later you'll leave it on too long (or turn it on too soon), and be operating at full power into a high Q load. With no limiting, poof, goodbye inverter.
With frequency control and voltage and current limiting, you can set the supply for full output into a steel load, and let it limit automatically when no load is present. Typically you might have a Q of 5-10 with work inserted, and 30-50 without; at the same coil voltage or current, that's a ~5x difference. If it's set for 10kW into the work, it'll simmer at only 2kW into the coil -- it throttle down automatically, tracking the load.
You could even add a feature to monitor the ratio of voltage and current to inverter output, i.e., the load resistance, and throttle it down even further when no work is detected.
This is how induction cooktops work: the coil is pinged from time to time, and if a load is detected, it can move into operation.
Or for geeky purposes, you can measure the V, I and F at the inverter output, which is therefore equivalent to the L, C and R of the tank. It's a network analyzer, solving for a simple RLC load, operating at full power.
Getting information back from the tank after the isolation transformer seems to be a bitch. You can't use CT or rogawski coils because of noise on current switching,
In a series resonant circuit, current sense is exactly what you want. CTs work very nicely. The current ripple does not have sharp transients, at least for any sane build.
Here's an example from my archives,
As measured on a Tek 475, so not that you'd necessarily see tiny squiggles, but they're really just not very important.
You can still see a little bit, the current waveform is broadened at time divs 3 and 8. That's the actual ringing here. It's not much because the inverter voltage waveform is heavily snubbed (hence the trapezoidal waveform), and it's easily filtered in any case.
A shielded CT goes a long way, too. Basically, anything but a Triad CST206-1A, which is absolutely the worst CT I've ever used...
so you want to use a voltage signal. I don't like the idea of a voltage transformer controlling the frequency. I also don't know enough about the high power transformer to be confident about its behavior or if it will have some kind of inrush problem etc being a doughnut. The idea would be some kind of phased lock loop.
You still want to sense voltage, but not because of tank parameters so much as to limit voltage safely, lest you burn through the capacitors. Which you've more or less spent $100-300 on at this point, whether because you've spent that time soldering together an array, or bought a proper Celem or other name brand.
With voltage, you can of course compute R, L and C from V, I, phase and F.
For synchronization purposes, I've been tempted to use a depletion mode MOSFET current limiter, to sense the high voltage directly with minimal phase shift. Should be very effective at lower frequencies, but probably not as good in the MHz.
In any case, it's just signal analysis, with varying levels and impedances of signal, depending on how you sense it. Dynamic range limits apply, so don't expect to track a super weak resonance. It's a pain trying to make an output transformer suitable for a huge range of impedances anyway. A modest operating range is fine (say 30:1 coil resistance?), which keeps the signals from being gross.
What kind of metal research can you do with a induction heater? I don't know if I even see the point in using one for brazing because a kiln would work better in most cases and the shop ones are pretty cheap now for losing bolts etc.
I was going to make a high frequency ("high" is relative; it would be 1-2MHz) model suitable for soldering pipe, brazing stuff, etc.; it would be an all-electronic torch substitute.
With an intuitive control showing tuning, and inverter capacity enough for a generous tuning range, it should be pretty easy to use. The main trouble is changing out coils, and making coils in the first place.
The other thing I was going to do was.....
teslacoil must have some great uses for these things right? i got fired up about induction heaters like once every two years for the last 12 years or so but I always run a bit short on the application 
...My original motivation, back in the days I did home foundry. I would've done iron and steel with induction. Pretty easy metallurgy, no worries about contamination and such (clay graphite crucible, BTW), nothing picky.
I still have a number of my foundry supplies around, but nowhere to use them, so also no motivation to work on a medium frequency (10-50kHz) heater (the one pictured above).
Tim
Home
Help
Search
Login
Register
