zvs induction heating, water cooling my transistors, still friggin' hot / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

zvs induction heating, water cooling my transistors, still friggin' hot

1 2 next

Move Thread

LAN_403

Tiberius

Wed Jun 25 2008, 06:20PM

Registered Member #1484 Joined: Wed May 14 2008, 03:24PM
Location: Cary, NC, USA
Posts: 27

I've been playing with a induction heater based on the fairly standard "Mazzilli" driver for a few days now and my IGBTs (Fairchild 20N60A4) get *hot*. I would shut it off when the temperature got uncomfortable to touch, and previously it would take about 30 seconds to get unreasonably hot and I'd have to let it cool down for around 5 minutes. After failing for about the 10th time to maintain a reasonable temperature on the heatsinks while attempting to heat a test load to red heat, I decided to throw some money at the problem and rigged up a water cooling circuit straight out of the plumbing section of Lowes.

The water blocks aren't quite perfect, being soldered together from various fittings and pipe, but they still manage to move heat off of the transistors monstrously fast in comparison to the previous cooling setup. Now it'll run for a few minutes at a higher power level and seems to level off at some level of "uncomfortably hot", returning to luke-warm in tens of seconds. The water blocks are copper and I'm using a mercury alloy as the thermal interface material so heat conduction from the transistors should be (and indeed seems to be) excellent. That said, why are my transistors even still getting uncomfortably warm to the touch? This appears to be dissipating a significant fraction of my input power, even if thermal runaway is no longer an immediate issue.

The matching inductor, air cored 71 turns of 14AWG on 4" PVC calculated ~180uH, also gets obscenely hot and I believe may be dissipating a significant fraction of my input power as well. I'm planning to roll this again later tonight with 10AWG to reduce the heating. I've been wondering also what effect this inductance has on the circuit and if a higher or lower inductance will vary the operating characteristics of the circuit.

Dr. Dark Current

Wed Jun 25 2008, 06:32PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Hi Tiberius

Induction heater will run at significantly higher frequency than the flyback the driver was designed for.
At this frequency the pull-up resistors on transistors' gates cannot provide fast enough turn-on, and 2 things probably happen:

1) The transistors' switching is delayed to the LC tank so they disturb the oscillation (this means increased peak current), and they spend some time in linear region during the transition which increases dissipation even more.

2) Because one transistor is OFF and the other one still hasn't turned ON, the DC choke can create a flyback spike which can exceed the devices' voltage rating and they avalanche, burning the power away.

I would say the ZVS driver without modifications is useless for induction heating.

Tiberius

Wed Jun 25 2008, 06:56PM

Registered Member #1484 Joined: Wed May 14 2008, 03:24PM
Location: Cary, NC, USA
Posts: 27

Dr. Kilovolt,

Thanks for your thoughts.

I was somewhat aware of point 1 and currently have a 333 ohm gate resistance rather than the ~500ohm typically quoted. I attempted a run with a pair of 200ohm 5W wire-wound resistors but the circuit did not oscillate, I presume due to the inductance of the wire-wound resistors. Will further reduction of the gate resistance result in a large reduction in switching losses, or is there a phase lag issue as well?

I am only running up to around 60VAC input at this time with 600V devices, so I will have to assume that #2 is not significantly contributing to my losses unless otherwise convinced.

I definitely agree that the circuit is quite inefficient at induction heating in its unmodified form, and right now my goal is to soak off or eliminate my losses enough to see a test load through to red heat, then move on to a better inverter design for this.

Dr. Dark Current

Wed Jun 25 2008, 07:18PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Tiberius wrote ...

Will further reduction of the gate resistance result in a large reduction in switching losses, or is there a phase lag issue as well?

I had problems with reducing gate resistors, a "parasitic" oscillation of very high frequency kicked in and the circuit did not work (probably this can lead to transistor destruction as well).

I once simulated the circuit in my head, it was quite mind-twisting, but the unwanted very high frequency oscillation has something to do with the gate resistors, gate capacitance and threshold voltage and transconductance of the switches (probably saturation voltage too). I cannot tell you the mechanism but it was quite complex, note it did not involve any parasitic inductance (so shortening your leads cannot fix this problem).

Marko

Wed Jun 25 2008, 08:04PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Yes, for a large device like 20N60 with 10nF gate capacitance 330ohms is *way* too high, you might need like 10 times less, just few tens of ohms which implies large and hot resistors.

This way the IGBT's are probably spending large fraction of cycle in linear region, or they may not be turning on fully at all.

Still as Jan says this circuit has fundamental problems with parasitic oscillations once resistors get too small, and that is why it is avoided for higher power levels.

After input voltage reaches one level the circuit will drop into a potentially damaging oscillation, without warning, and blow the devices or the fuse.

I explored the possibility of using ferrite beads for prevention of parasitic oscillations but they just don't seem to work. Small beads have no effect, while larger ones prevent the circuit from oscillating completely!

It's good you actually made to ~85V DC input, I had parasitic oscillations start at like 40V!

If I wanted to get most from this circuit, I would use a lower voltage source (rewound MOT?), and appropriate mosfet's instead of IGBT's.

IGBT's have high collector voltage drop which leaves relatively high voltage on gate of other device while it's *OFF*, allowing instabilities, and are in overall slower than mosfets, and mosfet's Rds On gets very small as voltage rating drops, winning over IGBT's.

Still, this circuit has limited capabilities and it may not be works trying to fix it instead of going to another topology.

I've still thought about what could be done...

Most logical way of preventing parasitic oscillation is removing the another switch. The circuit would work with just one class E amplifier as well, but the original feedback would not be possible anymore.
Actually, feedback is very difficult due to odd shape of drain waveform, and I'm still unsure how to do it properly.

Ferrite voltage transformer could be used, but it's output waveform would probably not be 50% duty cycle;
Current sense may be used with integrator to perform 90 degree phase shift - I'm unsure if that would work.

voltage transformer feedback with gate drivers can be easily used for push-pull version of the circuit though, and some people reported success eliminating the parasitic oscillations just by altering the type of feedback.

Marko

Dr. Dark Current

Wed Jun 25 2008, 08:22PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

I have succesfuly built an "electronic feedback" for this circuit. Works well no parasitic oscillation and fast switching. Voltage transformer is used for feedback and start-up oscillator for initiating the oscillation.

I'm now designing a half-bridge version of the circuit, it is near-complete I just need to sort out little things. And for testing I would like to get some igbt samples which st refused to send me. With commonly available semiconductors (30A IGBTs) it should be good for several kilowatts.

However I'm not sure how well it would work for induction heating; but for sure a magnitude better than the original circuit.

Tiberius

Wed Jun 25 2008, 09:25PM

Registered Member #1484 Joined: Wed May 14 2008, 03:24PM
Location: Cary, NC, USA
Posts: 27

Incidentally before I switched to the 20N60s I was using salvaged ST MOSFETs that ran noticeably cooler while transfer of energy to the work piece seemed to be greater for similar input voltage. Ironically, I switched due to the greater power handling capability of the 20N60s, but I think this evening I'll bump the gate resistance down a little more, rewind my matching inductor, and switch back to the MOSFETs to see if I can't accomplish my symbolic goal of red heat.

Marko

Wed Jun 25 2008, 10:17PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

The inductor doesn't really perform impedance matching, so it's just called DC link choke, not a matching inductor.

You can wind it on several material 26 cores and put all in series, it would save a lot of copper loss (but core loss must be watched for if there's not enough inductance).

To increase power throughput, just reduce characteristic impedance of the tank circuit. Simply increasing the capacitance will be OK.

I'm curious, what capacitors are you using, work coil, resonant frequency?

The DC link inductor also doesn't need to go into center of the work coil if it's impractical. Instead, you can use two inductors between supply voltage and drains of both switches.

Marko

Tiberius

Wed Jun 25 2008, 11:44PM

Registered Member #1484 Joined: Wed May 14 2008, 03:24PM
Location: Cary, NC, USA
Posts: 27

Marko,

The capacitors are CDE942 series 1600V .56uF, right now I have 6 in the tank circuit but will add the remaining 4 I have handy now that you mention it. The work coil is 3.5/3.5 center-tapped 1/4" copper tube, inductor is on the center tap. Unfortunately I have absolutely no idea what the resonant frequency is at the moment and nothing convenient to measure it with.

general update:

I swapped the MOSFETs (ST W34NB20) back in and managed to throw the 15A breaker at around 60VAC input without things getting too drastically hot, will try with more capacitors next.

Marko

Thu Jun 26 2008, 12:43AM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Be careful with voltage as the peak voltage devices see is 3.14*supply voltage, I'm not sure how your mosfets survived 85V for a longer time...

You should better get somewhat larger MOSFET's if you intend serious induction heating, like IRFP260 or higher current.

What are you using for power supply? You should better have some large iron transformers around for power supply.

Also, how are you mounting your devices? A single screw is usually poor since it lifts the die from the surface, a clamp from top should be used to provide proper pressure.

You said you are using mercury for thermal interface, isn't that a bit worrying? I bet even cheap thermal grease would work up to full dissipation for the package as long as it is properly applied, surfaces are clean and mounting is proper. Grease probably plays just a minor role after that I think.

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob