VapCap DIY Induction Heating : Bits 'n' pieces

[babunja]

Just checked on wikipedia, iQos works by heating the device in which a "cigarette" (a stick of glycerin soaked tobacco ) is inserted, the one using induction is Illuma.
I don't own an iQos, but maybe it can be hijacked to be a heat source for weed, the problem being how to check temperature

yes but I would like to replicate iqos don't buy it to modify it, then it heats up yes but I would like to replicate iqos don't buy it to modify it, then it heats up to 350 degrees which is too much for the weed

 

[Gomaruana]

Adjustable Resonance Induction Heater ( 12 V ) with TTL ( or CMOS ) PWM signal input
( from a microcontroller ,for example )





As the previous circuit ,but with a different MOSFET gate driver
( UCC2720x High Frequency, High-Side and Low-Side Driver )
and a single channel signal inverter
( MC74VHC1GT04 Inverting Buffer / CMOS Logic Level Shifter )
at the Low-side gate input signal ( LI ) in order to invert the High-side gate input signal ( HI ) ,allowing the use of a single input PWM TTL ( or CMOS ) signal .
The two 10K resistors are used to securely drop the signal to ground ( VDC ) ,
when LOW .

http://www.onsemi.com/pub/Collateral/MC74VHC1GT04-D.PDF

http://www.ti.com/lit/ds/symlink/ucc27200.pdf




======================================================================
(...)

UCC2720x Application Information

To effect fast switching of power devices and reduce associated switching power losses, a powerful gate driver is employed between the PWM output of controllers and the gates of the power semiconductor devices. Also, gate drivers are indispensable when it is impossible for the PWM controller to directly drive the gates of the switching devices. With the advent of digital power, this situation is often encountered because the PWM signal from the digital controller is often a 3.3-V logic signal which cannot effectively turn on a power switch. Level shifting circuitry is needed to boost the 3.3-V signal to the gate-drive voltage (such as 12 V) to fully turn on the power device and minimize conduction losses. Traditional buffer drive circuits based on NPN and PNP bipolar transistors in totem-pole arrangement, being emitter follower configurations, prove inadequate with digital power because they lack level-shifting capability. Gate drivers effectively combine both the level-shifting and buffer-drive functions. Gate drivers also find other needs such as minimizing the effect of high-frequency switching noise by locating the high-current driver physically close to the power switch, driving gate-drive transformers and controlling floating power-device gates, reducing power dissipation and thermal stress in controllers by moving gate charge power losses from the controller into the driver. (...)

More info about MOSFET gate drivers and high frequency MOSFET switching :

http://www.ti.com/lit/ml/slua618/slua618.pdf

Also search for " TK5Q65W_application_note_en_20180726%20.pdf "

This is really interesting to me, I also came across this board;

Piernas's Fast ZVS Mazzilli Driver

This modified driver is capable of working at much higher frequency than the original. It's very useful if you are looking for high voltage. This is because, in a regular ZVS driver, if you want higher voltage, you have to feed the driver with higher DC voltage, but this also increases the...

hackaday.io

I'm thinking about doing a soft start by running it at a high frequency when powered on, the amperage should stay low and when the metal is inserted we could gradually increase power by dropping the frequency closer to the resonant frequency.

I'm trying to build a 20v 10a induction heater that is very safe, fast and isn't too hard to make a custom pcb for.

 

[jackrod]

This is really interesting to me, I also came across this board;

I've been getting interested in induction heater design too . After looking into it, the common circuit used on the cheap blue "12V ZVS" etc modules is kind of crazy, right? It makes sense that the MOSFETs get killed commonly, given that there is no gate driver. And if the work coil and capacitors fail to start oscillating, both transistors can get turned on at the same time, and you basically get a dead short. I thought ElectroBoom explained these issues pretty well in this video and article.
https://www.electroboom.com/?p=1198

It does seem like some external control is needed to make the induction heaters safer. I assume it is what is in the ispire wand or other commercial products. This video shows a digitally controlled induction heater, using a really simple delay in a microcontroller.

That way, out of control switching can be controlled. But you could use the same thing to control the drive frequency to whatever you want.

I'm thinking about doing a soft start by running it at a high frequency when powered on, the amperage should stay low and when the metal is inserted we could gradually increase power by dropping the frequency closer to the resonant frequency.

This is a cool idea for detecting the metal. You mean a frequency just slightly off of the natural frequency of the empty coil, and capacitor(s)? When the metal is inserted, the natural frequency should drop. So maybe you would actually want the driven frequency to be below the natural frequency, and then the power would increase as the metal is inserted, which might be easier to detect. Maybe once it reaches a certain point, you can go back to "locking on" to the oscillation, like the regular ZVS.

There are some articles on better ways to do induction heating if you search "parallel resonant inverter" on google. A little bit over my head, I'm not an EE. There are some interesting strategies for avoiding crazy power spikes, varying the power level, etc. But using a real gate driver, adding some protection to keep the whole thing from shorting out, maybe some thermal protection, could be a lot safer for sure. Let us know if you build anything cool!

 

[Gomaruana]

This is a cool idea for detecting the metal. You mean a frequency just slightly off of the natural frequency of the empty coil, and capacitor(s)? When the metal is inserted, the natural frequency should drop. So maybe you would actually want the driven frequency to be below the natural frequency, and then the power would increase as the metal is inserted, which might be easier to detect. Maybe once it reaches a certain point, you can go back to "locking on" to the oscillation, like the regular ZVS.

I was more thinking in the lines of finding a way to avoid the situation where there is too much metal inserted and the circuit can't start oscillating because the demand is too high. What I would do is start the circuit on a frequency way higher than the resonant frequency of the whole with the part inserted, so that there is not much coupling and not much load.

Then we gradually lower that frequency so it comes closer to the resonant frequency of the tank/coil+metal circuit increasing the amperage. Using an amp sensor you can make sure you stay within the safety limits of your power supply. But that was an idea I had, I'm not an EE either, but I am in contact recently with someone who is willing to help me out in making a circuit that is more robust compared to the cheap zvs drivers we all know. If it works well I'm sure he will share it as an open source design, we're both big fans of that way of doing things.

 

[jackrod]

I was more thinking in the lines of finding a way to avoid the situation where there is too much metal inserted and the circuit can't start oscillating because the demand is too high. What I would do is start the circuit on a frequency way higher than the resonant frequency of the whole with the part inserted, so that there is not much coupling and not much load.

Then we gradually lower that frequency so it comes closer to the resonant frequency of the tank/coil+metal circuit increasing the amperage. Using an amp sensor you can make sure you stay within the safety limits of your power supply. But that was an idea I had, I'm not an EE either, but I am in contact recently with someone who is willing to help me out in making a circuit that is more robust compared to the cheap zvs drivers we all know. If it works well I'm sure he will share it as an open source design, we're both big fans of that way of doing things.

Cool! I read up some more on the theory of IH, and I think that you are right to start on a higher frequency. This page explains a lot, including that if you try to drive the tank below its natural frequency, the inverter (2 mosfets in the cheap ZVS) experiences a capacitive load and so Zero-Voltage-Switching is not possible. https://www.richieburnett.co.uk/indheat.html. And this guy on youtube shows what you describe, sweeping down the frequency until resonance is achieved.

. Another video of his explains how this works. Using a phase-locked-loop controller, the freqency of the inverter can be changed until it is in sync with the tank.

. Pretty cool solution. He used a microcontroller, but there are some other solutions that are more analog.

You can't really be too harsh on the cheap ZVS design, since it just locks onto the natural frequency using only the 2 transistors, diodes, and resistors. But the tradeoff is the gates have to be driven solely by what the tank is doing. I made a simulation model to help understand how it even works.

The two main problems I see are:

Startup is finicky (though it does seem to work most of the time, practically)
The oscillation is started when one transistor (M1 or M2) is turned off, and the other one is turned on. The cross-connection (gate1 to tank2 through D2, gate2 to tank1 through D1) gives a lot of positive feedback here. If one transistor is on (gate high), then it can sink current from the choke inductor (yellow in the cheap ZVS, L2, L3) and tank to ground, as well as discharge the gate of the other transistor, turning it off. But the supply voltage, and its rate of change, can modify this behavior a lot.

Particularly, the main issue is the transient which happens between the time both transistors start conducting, and the steady oscillation, where the transistors are alternating between cutoff and saturation. During this time, current is allowed to increase relatively unimpeded through the two chokes. When one transistor fully shuts off and the other turns on for the first time, the current in the choke on the shutoff side can't change rapidly, so it gets forced through the tank. This eventually rings back and causes the transistors to switch the other way, and the oscillation grows. But the problem is, the current built up through both chokes probably does not match the current needed to drive the tank steadily. If the power supply was switched on too fast, it will be too low, if it was switched on too slow, it will be too high. This can cause all kinds of nasty voltage spikes on the tank, heating in the transistors, etc.

Practically, it seems like the limited current output of the power supply seems to dampen some of these effects. But this definitely doesn't make the design very resilient. @TommyDee had a really good simple idea in the half-pint thread, of switching the power supply to the gates separately from the main power, in the Halfpint guide, where the trace was cut. In this case, the osillation always has to start somewhat hard, with zero current in the chokes, but that is better than the starting design.

The gates are not driven that strongly, so the losses in the transistors are high-ish
The gates are driven low by the opposite side of the tank, through D1/D2. No problem there, they get shut off very quickly. But they get driven high only by R1 and R2 pulling them up. These are big power FETs with substantial gate capacitance, so there is a delay in increasing the gate voltage. They spend some time in the linear region, which causes them to dissipate a lot of power. It gets better if you make R1 and R2 smaller, but you can't really push that too far, since each is conducting power to ground about half the time.


I think that these issues could be solved if the gates were driven by a proper gate driver, controlled by a phase-locked-loop based on the tank voltage. The PLL could be digital, run on microcontroller, or more analog using a PLL chip. A microcontroller may allow for some power control and safety features as well. But it could get pretty complicated. I think it would be worthwhile to elevate the open-source/DIY induction heaters, but I'm not sure how much use it would be, practically, when you can buy a cheap ZVS for $10. If anyone has interest in some more of the numbers I used, LTspice model, simulation results, etc, let me know. This is a pretty fascinating circuit, with a lot of room for improvement.

 

[TommyDee]

@jackrod Is that 470 ohm and 480 ohm intentional?

Edit: Note what the specs are for the FETs 24N06. If I'm not reading it wrong, they are good to 60 watts. The ZVS is noted to be 120 watts. In the real world, true, a 50% duty cycle but each FET takes on the whole load. This means many of us are running each FET at 70-80 watts consistently. You might also see that we are already working at the 60V threshold. Upgrading the standard issue FET is a worthwhile consideration in any reinvention of the heaters.

You might also consider clipping those little spikes I see in the scope reading with voltage snubbers.

 

[jackrod]

The 470 and 480 is to help out the simulation a bit. In reality there is always some small tolerance which will make one FET get ahead of the other, but if the simulation is perfectly balanced it can get stuck in a marginally stable state.

The transistors could definitely use an upgrade, they are considered obsolete. I’ve been looking at this datasheet: https://www.onsemi.com/download/data-sheet/pdf/ntd24n06l-d.pdf
But I think 62.5W is for power which is actually dissipated by the transistor, not for what flows through. See the notes 1 and 2 on the first page, where the dissipated power rating is much less (1.88 or 1.36 W) when mounted to a PCB normally. The dissipation happens mostly due to switching losses, and is much smaller when the transistors are fully on and fully off. Current design seems to dissipate about 1-2W in the transistors on average.

The voltage spikes can get pretty nasty, especially during startup. They are likely getting big enough to cause drain-source breakdown, according to the sim. I would have to see how snubbers affect the oscillation. I think that they come from shutting off the FET quickly and getting an inductive kick from the choke. Part of that energy gets pushed into the tank, which is good, but clamping the maximum voltage would be a good idea too. Maybe some TVS diodes across the transistors could do it. This may also be easier to deal with in a full-bridge inverter design.

 

[TommyDee]

Interesting trick on the sim.

That 60 watts comes in when you use the FET along with a PWM... they get really hot when you lower the pulse width. That's one reason we don't regulate onboard.

If you look carefully at the startup pulse, it seems to be 1-1/2 cycles long... that's where things can get crossed up. One failure mode I noted in the past was a switch resonance. Using the CapAsSwitch, a loose connection would oscillate with the circuit. You could hear it as a sonic vibration between the cap and the contacts. Within moments, one of the FETs will be dead.

 

[jackrod]

Yeah, not sure if it's the best trick to use, but it seems reasonable. I think you sort of have to do something, otherwise you may be waiting for numerical errors to kick off the oscillation in the sim? But I'm not entirely sure how the implicit solver in LTspice affects that marginally stable case.

You mean PWM on the power supply to the whole module? That would make sense, the startup period is definitely the hardest part on the FETs. Starting up many times per second would make a lot of heat. If the oscillation never starts properly, all the power supplied could get dumped into the FET, too, so 60W like you say. What frequency was the PWM on?

Not sure what you mean by the startup pulse length. Talking about the time for the gates to start reaching full voltage? 1.5 cycles sounds about right there. What frequency was the work coil running at then? My ZVS ran about 190kHz, which seems a bit high but in the right range for the coil and capacitors.

Would have to do some more testing to validate the sim model, though it seems to be pretty good so far. Modeling the power supply behavior is a bit tricky. I ran mine off a bench supply with the current limit set at 5A. But it seems like current limiting/voltage sagging of the supply just makes the startup process take longer. The current spike from the power supply at startup is longer, about 10 cycles. That's with R_work at 12 ohms for a cap in the coil (gives about 60W input power), and the input voltage rising in 1us.

The audio-frequency resonance with the cap is very interesting. Think the cap was getting pushed away magnetically and breaking the circuit?