Simple induction heater

[T3sl4co1l]

By popular request, a thread dedicated to this item:
https://www.seventransistorlabs.com/Images/Breadboarded_Induction_Test.pdf

Which if you wanted something stand-alone, just use a, like, 555, with pin 3 tied to pins 6+2 with a variable resistance, and 6+2 to GND with a cap, leave 7 open, 1 GND, 4+8 VCC.  Uh might want to adjust the supply voltage, 15V is pushing it for a 555, and nearly for the gate driver, so maybe reduce coil turns and run at 10-12V for instance.

This was just an amusement I'd used to tin whole pieces of copper clad PCB stock.  Basically spread some rosin (I've been using actual straight-up fresh pine rosin for this, why waste the good stuff right) on the board, get a blob of solder going, and using a steel spatula, push it around until the board is covered.  Or if you're feeling fancy, you could spread solder paste.  (For which you could use a reflow oven instead, of course.)  Freshen up the rosin if needed, tweak driven frequency or height over coil as needed to keep the power right (do keep in mind, you're doing this ENTIRELY by hand, no automatic power control at all; the bottom side WILL blister and burn if you leave it too long!).  It's quite efficient too, under 50W seems to be enough to hold a fair area (say 2x3cm or more) at soldering temp.  Which really shouldn't be surprising comparing to the surface area of a 40W soldering iron for example, which gets pretty toasty if left alone at full power (oh man, it's been so very many years, but I actually used to use those, the $10 RadioShack irons ).  It's just that the board itself is your iron, mwahaha!

Actual circuit operation is pretty simple, set frequency near (somewhat above) resonance (you'll need some indicator to show that, maybe as simple as an LED+R across the coil, of course doing this at the bench I've got the scope clipped into it), and get it on the right side so that power is about as needed.  Very hand-wavey and intuitive, just use it like a kind of weird torch, or too-small hotplate.

Note that switching loss increases when below resonance (capacitive load), but this thing is running slow enough and not doing a whole lot to begin with, that that's not a big deal: I didn't even have heatsinks on the transistors for this.  Anyway, it's low enough power that there isn't really much consequence of failure; maybe you let out some magic smoke, maybe you blow the fuse on your power supply (do fuse it, or set current limit to say 5A or so).

Note also that, as you bring "heavy" loads close to the coil, inductance drops, so resonant frequency rises.  This can be used to effectively increase power output on approach, if you've got the setpoints and spacing dialed in right.

"Heavy" here can be defined as, blocking magnetic fields.  So, thick non-ferrous metals.  (That includes nonmagnetic stainless!)  At ~20kHz, 1-2 oz copper is just thick enough that it blocks only most of the field, which means the top layer is also being heated some, as well as anything on top (say a thick blob of solder; or that metal spatula, if you hold it flat to the surface, particularly if it's mild steel (magnetic)).



And no, this won't do anything for you like... make the perfect cup of coffee, say.  At best, it would be an effective warmer.

Which, entertaining that thought:
Scaled up to a few hundred watts, you can make a cup of boiling water in the not-too-egregious time span of 5-10 minutes.  (My microwave takes 2:30 to boil, and it's depositing somewhere around 500-800W.)  Anyway, you need a susceptor down in the cup itself, which works out fine if your cup is metal (and not double-layered!), but for anything else you'll need, like, a ring or plate down in there to receive the power.  And at that power level (1kW+), I would strongly recommend a more full-featured, and above all, safe, circuit, not a toy like this one.  It's enough power that the plate, out of water, will get red hot!  Given a European mains circuit and a power supply to match, you would be able to do a cup in a minute or two.  Faster than that, basically requires not only much more electricity (>3kW), but such high power levels that you may simply boil water off the susceptor in the first place (critical heat flux), and then you have... problems.

Tim

[sandalcandal]

Nice simple half-bridge driver circuit.

I think for people wanting to simply dump some inductive heating into a load, a Mazilli/Royer type circuit is likely to be much more effective and requires simpler components (no gate driver or PWM IC). I remember in the past experimenting with 555 half-bridge driven circuits and the original Mazilli circuit albeit for making arcs with CRT flybacks instead of induction heating. The Mazilli circuit produced significantly more power with significantly less heating of the FETs, which isn't too surprising given its ZVS nature. Power can be controlled by varying the voltage of the input power source.

Edit: I guess a problem with Mazilli drivers is that it will always try to drive the output inductor/transformer at resonance so without a load, the tank current/voltage can get excessive. Some sort of manual momentary control for turning on the switching might be a good idea. It is still quite easy to accidently overstress the system unlike a fixed frequency driver such as Tim's so that's an inherit danger of a self-oscillating circuit unless you add some sort of load detection control which will probably make it less than "simple".

[T3sl4co1l]

Yeah, with a ZVS, you don't quite get the same control response as fixed-frequency.  Probably no big deal for something like tinning PCBs; and with a variable bench supply, you have all the control you want.  The main reason I chose not to use one that day was, several actually:
1. I didn't have the pull-up inductors handy;
2. The work coil was wound with more turns (needs more voltage; it'll develop over 100V when lightly loaded, near resonance),
3. In part because I don't have huge gobs of capacitors to resonate it with, and the low frequency is important here (skin depth in copper foil!).  I happen to have a few 2.2uF polypropylenes, which do the job well.  (I also have a few polyester and ceramic -- both of which are terrible, giving a maximum Q of 20 or so, and with the ceramic drifting as it heats up.)

Tim

[sandalcandal]

I wonder what a minimalist load detection circuit could be like. Maybe, a feedback/detection coil and circuit that compares the driven coil to the feedback coil?

[T3sl4co1l]

Like to power up under load only, like a cooktop?

It's gotta have some kind of excitation to begin with, so it needs to be powered up enough to oscillate, or drive the load, and some kind of impedance sensing can be done.

For the ZVS, just sensing supply current is enough; easy enough to do with an MCU, voltage and current sense, and a supply controller of some sort (switch, or linear or buck regulator), or even discrete logic if you're into that sort of thing.

For the fixed frequency case, you might want a quadrature mixer, to more or less directly measure the R and X of the coil itself.  Drive should be pulsed in the same way to test for load, then switched to continuous operation when Q is below some threshold (or L above, whatever you wish to do).

Note that the ZVS, as it switches at the resonant frequency, acts effectively as a power mixer, shifting the coil impedance down to baseband -- DC resistance.  Thus we can't measure inductance, but we get Q as resistance compared to a reference value.  (We can measure inductance by frequency instead.)

Tim

[sandalcandal]

Like to power up under load only, like a cooktop?

Similar but more to prevent self-induced damage under a no load condition. [Although I guess that's what cooktops do?]

For the ZVS, just sensing supply current is enough; easy enough to do with an MCU, voltage and current sense, and a supply controller of some sort (switch, or linear or buck regulator), or even discrete logic if you're into that sort of thing.

I was thinking along the lines of something much more analog, minimalist in the sense of using the least sophisticated components as a design challenge rather than minimalist in terms for making the design process as simple as possible. I was imagining using a feedback coil to generate negative feedback into the drive circuitry somehow to attenuate the drive power. Although I didn't think too much about how that would actually be achieved. When no load is applied I guess the voltage in the feedback coil would be increased. Not too sure how that could be used to attenuate power, particularly since we'd prefer to do so by increasing the drive frequency above resonance. Perhaps some analog conditioning to provide a (non constant) phase lead then mixing it into the drive signal? I can't remember the technical name for this sort of control, any RF people?  This shifts the switching point "forward" making it switch sooner each cycle, thereby increasing the switching frequency, this feeds back each cycle up to a point where the feedback amplitude can't drive the switching frequency any higher.

A more "complicated" means of achieving no-load drive attenuation in terms of the maths and modelling but "simpler" and minimalist in that it can be implemented with "primitive" analog components.

Edit: or was it phase lag that's needed for this? I think it was phase lead.

[langwadt]

By popular request, a thread dedicated to this item:
https://www.seventransistorlabs.com/Images/Breadboarded_Induction_Test.pdf

Which if you wanted something stand-alone, just use a, like, 555, with pin 3 tied to pins 6+2 with a variable resistance, and 6+2 to GND with a cap, leave 7 open, 1 GND, 4+8 VCC.  Uh might want to adjust the supply voltage, 15V is pushing it for a 555, and nearly for the gate driver, so maybe reduce coil turns and run at 10-12V for instance.

or an IR2153, it's a "555", half bridge gate driver, and a ~15V zener clamp for supply, the D version even have the bootstrap diode build in


https://www.homemade-circuits.com/wp-content/uploads/2013/09/irs2153appicationnotecircuit.png

[T3sl4co1l]

Similar but more to prevent self-induced damage under a no load condition. [Although I guess that's what cooktops do?]

Ah, sure. One could implement a current or voltage limiting scheme.  Which is what ZVS does: the output characteristic is constant voltage (because again, it's a power mixer, so reflects whatever the input characteristic is), so least power comes at maximum Q.  For the CV-driven series resonant, it has to be either input voltage reduced, or driven frequency raised, to maintain output; basically resonance acts to invert the characteristic (so you get maximum current draw at maximum Q instead).

Frequency control has its own issues, as the dominant pole of the control loop is given by the difference between driven and resonant frequencies -- again, the inverter is a power mixer, and you get the superposition of driven and resonant modes on the tank, which interfere to give apparent amplitude modulation (plus whatever it does to phase, when measured by zero crossings, say).  But that's just a matter of going slow enough (with control compensation) for a known worst case coil Q.

I was thinking along the lines of something much more analog, minimalist in the sense of using the least sophisticated components as a design challenge rather than minimalist in terms for making the design process as simple as possible. I was imagining using a feedback coil to generate negative feedback into the drive circuitry somehow to attenuate the drive power. Although I didn't think too much about how that would actually be achieved. When no load is applied I guess the voltage in the feedback coil would be increased. Not too sure how that could be used to attenuate power, particularly since we'd prefer to do so by increasing the drive frequency above resonance. Perhaps some analog conditioning to provide a (non constant) phase lead then mixing it into the drive signal? I can't remember the technical name for this sort of control, any RF people?  This shifts the switching point "forward" making it switch sooner each cycle, thereby increasing the switching frequency, this feeds back each cycle up to a point where the feedback amplitude can't drive the switching frequency any higher.

A more "complicated" means of achieving no-load drive attenuation in terms of the maths and modelling but "simpler" and minimalist in that it can be implemented with "primitive" analog components.

Edit: or was it phase lag that's needed for this? I think it was phase lead.

Speaking of IR2153, there's an application circuit for it, I think, that uses output voltage to feed back to the timing components, say for an LLC resonant converter.  I think such an arrangement lacks one key feature, that it can't throttle down to zero -- most LLC controllers go to a burst/hysteretic mode at light load, while there's no way to arrange this here.  So it would work for supplies with known minimum loads.  Which is the case here, basically.

A downside is, as a "555", its output duty cycle may suck.  It doesn't have the T flip-flop of, like, TL494 and friends.  Does mean the input is AC (phase) sensitive, so you might in fact use AC feedback -- a recent thread here discussing an ultrasonicator used one of these this way.

Current is best sensed with a transformer, but a shunt could be used as well.  Or a capacitor divider, if you put the cap on the low side.  Voltage, easy enough by resistor divider.

Tim

[langwadt]

Similar but more to prevent self-induced damage under a no load condition. [Although I guess that's what cooktops do?]

Ah, sure. One could implement a current or voltage limiting scheme.  Which is what ZVS does: the output characteristic is constant voltage (because again, it's a power mixer, so reflects whatever the input characteristic is), so least power comes at maximum Q.  For the CV-driven series resonant, it has to be either input voltage reduced, or driven frequency raised, to maintain output; basically resonance acts to invert the characteristic (so you get maximum current draw at maximum Q instead).

Frequency control has its own issues, as the dominant pole of the control loop is given by the difference between driven and resonant frequencies -- again, the inverter is a power mixer, and you get the superposition of driven and resonant modes on the tank, which interfere to give apparent amplitude modulation (plus whatever it does to phase, when measured by zero crossings, say).  But that's just a matter of going slow enough (with control compensation) for a known worst case coil Q.

I was thinking along the lines of something much more analog, minimalist in the sense of using the least sophisticated components as a design challenge rather than minimalist in terms for making the design process as simple as possible. I was imagining using a feedback coil to generate negative feedback into the drive circuitry somehow to attenuate the drive power. Although I didn't think too much about how that would actually be achieved. When no load is applied I guess the voltage in the feedback coil would be increased. Not too sure how that could be used to attenuate power, particularly since we'd prefer to do so by increasing the drive frequency above resonance. Perhaps some analog conditioning to provide a (non constant) phase lead then mixing it into the drive signal? I can't remember the technical name for this sort of control, any RF people?  This shifts the switching point "forward" making it switch sooner each cycle, thereby increasing the switching frequency, this feeds back each cycle up to a point where the feedback amplitude can't drive the switching frequency any higher.

A more "complicated" means of achieving no-load drive attenuation in terms of the maths and modelling but "simpler" and minimalist in that it can be implemented with "primitive" analog components.

Edit: or was it phase lag that's needed for this? I think it was phase lead.

Speaking of IR2153, there's an application circuit for it, I think, that uses output voltage to feed back to the timing components, say for an LLC resonant converter.  I think such an arrangement lacks one key feature, that it can't throttle down to zero -- most LLC controllers go to a burst/hysteretic mode at light load, while there's no way to arrange this here.  So it would work for supplies with known minimum loads.  Which is the case here, basically.

A downside is, as a "555", its output duty cycle may suck.  It doesn't have the T flip-flop of, like, TL494 and friends.  Does mean the input is AC (phase) sensitive, so you might in fact use AC feedback -- a recent thread here discussing an ultrasonicator used one of these this way.

Current is best sensed with a transformer, but a shunt could be used as well.  Or a capacitor divider, if you put the cap on the low side.  Voltage, easy enough by resistor divider.

Tim

the data sheet shows a fet to short ct to ground as shutdown

there's another one variable frequency and 50% duty cycle, https://www.infineon.com/dgdl/dt98-4.pdf?fileId=5546d46254e133b401554de8f0885e14

[SiliconWizard]

I wonder what a minimalist load detection circuit could be like. Maybe, a feedback/detection coil and circuit that compares the driven coil to the feedback coil?

You could also use the same coil and a different driver circuit in parallel, just for this purpose.