8 kW Induction Cooktop Teardown and Reverse engineering

[Mads Barnkob]

Teardown of a 8 kW induction cooktop, supplied from 2 single phase mains and neutral. Modular construction with two inverter modules and changeable coils.

/>
The IGBTs are Toshiba GT60J323 with a mere 120A pulsed rating, but despite that, lets explore reverse engineering what looks like a fairly common half-bridge layout and see with how little external circuit we can turn this into a Tesla coil inverter



Enjoy the teardown video while I map out the circuit and drivers!

[Avenger]

If this is an ordinary half-bridge, and the operating frequency is enough, then I think it can be used in a simple SSTS. The power will obviously be less than in the induction cooker

[klugesmith]

Thanks for sharing.   We look forward to the circuit details.   
Do you think the front end (AC main to DC bus) includes a power factor correction function?
I've seen that in an industrial high voltage PS from Spellman, but never in a consumer product.   
Maybe it's time, as governments begin to ban gas stoves and say "try induction, you'll like it."

[davekni]

Do you think the front end (AC main to DC bus) includes a power factor correction function?

I believe there are regulations in Europe (and I think US too) requiring appliances of that power level to meet power factor lower limits.  I also believe that the common way to achieve high power factor for such heating appliances is to have heating power track line voltage squared.  In other words, no bulk caps.  Half-bridge Vbus tracks rectified line voltage, so current remains proportional to voltage.

[Twospoons]

Its also much cheaper to let the power fluctuate at 2x mains frequency, than to build a high power PFC. For heating it doesn't matter if the power level oscillates.
Can you imagine the cost of the bulk caps alone for an 8kW PFC?

[klugesmith]

Bet you're right about just letting the DC bus look like rectified sinusoid. Mads's video points out small-ish (4 uF) 400V capacitors, guessing they are connected to the DC rails close to each half bridge.

It would limit the musical opportunities in SSTC service, because any interrupter waveform would be mixed with the rectified line-frequency sine.

There's a similar power factor issue with microwave ovens, even the half-century-old design with one capacitor and one diode for voltage doubling.  Vkk at the magnetron, and consequent microwave power, fluctuate at line frequency (not 2x).  Who knows the power factor at wall plug, or wants to look it up, or measure or simulate it?

[rikkitikkitavi]

Do you think the front end (AC main to DC bus) includes a power factor correction function?

I believe there are regulations in Europe (and I think US too) requiring appliances of that power level to meet power factor lower limits.  I also believe that the common way to achieve high power factor for such heating appliances is to have heating power track line voltage squared.  In other words, no bulk caps.  Half-bridge Vbus tracks rectified line voltage, so current remains proportional to voltage.

That is correct. All high powered appliances using induction heating is designed this way. The caps used are only for supplying the switch bridge with energy. It also has the advantage that when the transistors blow to short, which does happen occaisionally there is not a lot of energy stored that is dumped into these during shorting, and also pose a higher risk during maintainence. Still there is energy to wreck havoc on PCB , switchers and commonly the rectifier bridge (I know, I have repaired a few) but limited to these. Instead if there where a 100J charged 1000uF cap bank the damage would be considerabel to PCB and rest of circuit.

Small electronic transformers for the now (EU atleast) fast disappearing low voltage halogen bulbs use a similar topology. A half bridge driving the power transformer but no caps and hence it is a square wave modulated with a sinus.

Most common problem with these during installation is that in the nordic countries atleast the stoves where only 2-3 phase with PE, no neutral.
So when connecting a new induction top it is feed 400V instead of 230V. Boom it goes.

[klugesmith]

Small electronic transformers for the now (EU atleast) fast disappearing low voltage halogen bulbs use a similar topology. A half bridge driving the power transformer but no caps and hence it is a square wave modulated with a sinus.

Low voltage halogen bulbs with HF transformers are new to me.  Are these built in to the lamp base, as with ballasts in screw base CFL lamps? How many watts? How big?

Most common problem with these during installation is that in the nordic countries atleast the stoves where only 2-3 phase with PE, no neutral.
So when connecting a new induction top it is feed 400V instead of 230V. Boom it goes.

Boo hoo, no free IGBTs there.  Maybe free capacitors and Litz wire.

Now what happens if we feed rectified, unfiltered power to a simple Mazilli self-oscillating ZVS driver?  I will guess that  line voltage zero crossings are a challenge, even with separate DC power to gates.  Can "oscillation" stop, each time resonant tank voltage gets too small to make gates switch, and re-start each time Vdd ramps back up in the next half-cycle of AC power?

[rikkitikkitavi]

As for the halogen transformers , good ol´Danyks has some schematics with the IR2153 half bridge driver.

https://danyk.cz/elhatr_en.html

 There is self oscillating variants too , that does not seems to have difficulties with zero crossing etc. I have rarely seen one blow. The ones I have dismantled all where some kind of variant of this, never seen one with IR2153 IRL, or other IC, but I guess the chinese prefered spending a few cents less and use  a coupla transistors. But always seems that there is more transistors IRL on the boards than in the online schematics one finds by Google . Never really bothered to reverse engineer a schematic either from these.

Good ol´Rod Elliot has some (damn, that is a living , working archive, I remember building a P3A amplifier in my 20ies)

https://sound-au.com/lamps/elect-trans.html

But there is a lot about these mre simple ones. I did put a schottky bridge on one and a low-ESR caps and voila, cheap DC voltage. But yes, it did blow the cap eventually. That is when I learned about RMS current rating and ESR gives power loss...

[davekni]

Now what happens if we feed rectified, unfiltered power to a simple Mazilli self-oscillating ZVS driver?  I will guess that  line voltage zero crossings are a challenge, even with separate DC power to gates.  Can "oscillation" stop, each time resonant tank voltage gets too small to make gates switch, and re-start each time Vdd ramps back up in the next half-cycle of AC power?

My ZVS-driven Jacob's ladder project originally had no bulk cap:
https://highvoltageforum.net/index.php?topic=831.msg5491#msg5491
Between remaining tank circuit energy and remaining energy in small snubber caps, oscillation did continue sufficiently to pick up again on next half-cycle.
However, I still decided to add a small bulk cap.  Reason was to keep arc hot enough to avoid extinguishing between half-cycles.  Size was a compromise between keeping arc hot and keeping power factor from getting too low.

[Mads Barnkob]

It turned out as expected, just a regular half-bridge with all the bells and whistles. Regular bridge rectifier and passive filtering. And as its guessed by others, induction cooktops uses quasi-resonant mode in phase with mains current.

Theoretically I could just smack 230VAC on this sucker and inject a driving signal on the HI/LO inputs from TTL logic, its properly more a question about driving power when going from 20-25kHz into ten times that. So the driver section should really just be replaced by a regular GDT setup and no worries to be had. But for the sake of MINIMAL addition of more components, I still think it will fail at those frequencies.

On the backside of the PCB, there was however a high-side driver hiding itself!



It took a while to trace all the tracks, a lot goes under components as well. SMD diode/Capacitor values are best guess, as its impossible to measure them in-circuit.

[klugesmith]

Thank you for a well-drawn and useful schematic.

We see that driver IC L6388 is obsolete and no longer manufactured.  Digikey suggested one alternative which appears to be pin compatible: DGD2304.  Available in single unit quantities.   US $0.57 each in quantity of 1000.

[Mads Barnkob]

I will try to connect a SSTC driver with antenna feedback, to drive the half-bridge. But since there is two individual half-bridges and I want to modify as little as possble, could I just connect the two outputs in parallel to the same primary coil?

I don't think there should be any large issues with that, since its sharing the same rectifier and mains input, driving would be from a 4 output GDT, so that is also in sync.

Did I miss anything that would make it smoke?

[rikkitikkitavi]

What happens if the bridges are out of phase? I mean not 50Hz out of phase, but de synchronized when it comes to the switching?

[davekni]

Did I miss anything that would make it smoke?

Probably work fine.  A few possible issues that you probably already considered:

1) I'm presuming this SSTC design already has an HC14 receiving antenna output, and that HC14 output will be wired to both L6388 inputs.  If it were an SSTC version with antenna going directly to driver chip inputs (as I've seen a couple times here), the two L6388 chips might have slightly different logic input thresholds so switch at different times.

2) Presume L6388 has internal logic and delays to enforce no simultaneous IGBT conduction and dead time for switching.  With paralleled half-bridges, any noise or other extraneous logic transitions on one chip and not the other will cause cross-conduction.

3) Delay matching of the two L6388 chips and associated IGBTs need to be well within dead-time.  Probably no issue, especially since there's a good chance that chips came from same production lot.

4) The induction cooktop is likely designed to maintain ZCS conditions, at least most of the time.  GT60J323 spec lists it for resonant inverter applications.  Switching delays are listed for resistive load rather than inductive.  SOA is more limited than for most hard-switching IGBTs.  SSTCs typically are closer to hard-switching (hard turn-off only).  Your SSTC primary current may be low enough that none of this is an issue.  Might be less of an issue for a small DRSSTC with proper phase lead.

5) SSTC operation is likely at higher frequency than induction cooktop, up to a factor of 10 perhaps.  Even at low duty cycle, does the +15V supply have enough bypass capacitance to supply 10x (or whatever the frequency ratio is) current for the length of each enable pulse?

6) Similar question for half-bridge bus power, can line input handle the pulses of current draw, or are you adding bulk capacitance?

BTW, I've been wondering how easy it would be to use an induction cooktop circuit for my eddy current levitation box (if it ever needs repair).  Use is closer to induction cooking (similar frequency and continuous operation), except load Q is much higher for this use:

/>
Have fun with your experiment!

[romy]

1) most of this is above my head. i understand igbt-cookers have no smoothing stage and thus no power factor correction (v/i in phase). they work with a "modulated-sine-wave" where the sinus gets chopped up (sometimes even only after a half-bridge rectifier) as well as on/of operation. is this correct?

2) will it be diffcult getting rid of the change-in-permeability sensing circuit?

[klugesmith]

Regarding item 1, yes the power factor at wall plug connection is close to unity.
Input current is roughly sinusoidal, in phase with mains voltage.
In this case, achieved simply by accepting rectified sine as the "DC" bus voltage.

Don't confuse that with "power factor" in the induction coil and associated capacitor.
The high frequency circuit will have a substantial phase shift between voltage and current.
Volt-amp product much greater than the real power. 
Maybe the factor is less than 10, because the circuit isn't energized without a load. I didn't see if Mads's report includes any measurements of powered circuit.