Feedback system for induction heater / General Science and Electronics / Forums

Forums

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

« Previous topic | Next topic »

Feedback system for induction heater

1 2 3 next

Move Thread

LAN_403

nrhoades

Mon Oct 29 2007, 08:38PM

Registered Member #610 Joined: Wed Mar 28 2007, 09:44PM
Location: Middletown, RI
Posts: 110

I've come up with an idea for tracking the resonance of an LCLR induction heater load without the use of PLLs.

The idea is to watch when the work coil phase shift passes through -180 degrees. If a current sense is fed to a positive magnitude detector then a microcontroller can know which side of the peak the system is on. The microcontroller can then decide to either increase or decrease the frequency. The actual frequency coming out of the SPI Oscillator should be a multiple of the number of operating cycles desired within a period.

It is also desirable to let the binary searching algorithm rest when it is close to the resonant frequency. The phase shift of the voltage across L2 drastically departs from 0 degrees at resonance. If this voltage is connected to a fast comparator, then the microcontroller can also know when the system is in resonance. To detect this the controller would wait some small delay (a few operating cycles) and then read the input port. The delay would act as a phase shift threshold level thus providing a detection mechanism for the phase peak.

The Q of the load will increase if the work coil isn't well-coupled to the work piece. The rate of change of the phase shifts will increase. The resolution of LTC6903 is pretty good (close to 0.1kHz from 100kHz to 200kHz) so it should work well. Lower Q is better for induction heating; the work coil should be taylored to the work piece for best performance.

Marko

Mon Oct 29 2007, 09:15PM

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

It is not possible to build a feedback for LCLR with PLL basic circuit neither with any other direct means.. now you probably know that. Topic was debated here

to death.

If you lock to be in-phase with work coil you'l lose ZCS and if you tune phasing to the bridge output you'l lose resonant rise. Manually tuning with SG3525 seemed to give best results of all for now.

You need to sense bridge output current - not the voltage, since it's square and not useful for feedback in a voltage fed topology.

I'm not sure how would I do it with microcontroller either - it would need to be very fast in order not to generate it's own phase shifts. An even then it would probably just be trial-and error programming for phase shift. Good luck for it!

I myself really don't fully understand what's going on in there and this would need lots of simulations. It may be possible to create a working feedback with 2 CT's and phase comparators, don't know.

Rather a huge and complex project...

nrhoades

Tue Oct 30 2007, 02:02AM

Registered Member #610 Joined: Wed Mar 28 2007, 09:44PM
Location: Middletown, RI
Posts: 110

Topic was debated here to death.

Engineering never dies! :) I did read the thread many times though.

If you lock to be in-phase with work coil you'l lose ZCS

Mathematics show that the impedance of a parallel LCR circuit increases to log10(R) at resonance. I see this as gaining ZCS, i.e. lowest bridge currents at resonance.

if you tune phasing to the bridge output you'l lose resonant rise.

Sorry, I'm not sure what you mean.

You need to sense bridge output current - not the voltage, since it's square and not useful for feedback in a voltage fed topology.

The system I presented isn't using the bridge current as feedback. It uses a CT over the coil current, and the voltage across the matching inductor. The matching inductor will limit the bridge current anyway.

it would need to be very fast in order not to generate it's own phase shifts. An even then it would probably just be trial-and error programming for phase shift.

The speed of the microcontroller is always the bridge frequency. There isn't a need for that type of calibration.

Steve Conner

Tue Oct 30 2007, 09:13AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

I thought that adjusting for zero current switching on the inverter is all you need to do. It won't get you exactly on the resonant frequency of the tank circuit, but it will be close enough to get plenty of resonant rise.

Another way of saying that is that adding the matching inductor changes the resonant frequency of the tank circuit, and adjusting the inverter frequency for ZCS will put you on a resonant frequency of the resulting system. (Yes, "a" resonant frequency: as nrhoades' LTSpice simulation shows, the LCLR has two, one pole and one zero, but the zero is no practical use)

I know that adjusting for ZCS on the inverter gives maximum real power throughput in the matching transformer method I use, anyway, and I doubt it's any different in the LCLR.

nrhoades

Tue Oct 30 2007, 05:07PM

Registered Member #610 Joined: Wed Mar 28 2007, 09:44PM
Location: Middletown, RI
Posts: 110

I thought that adjusting for zero current switching on the inverter is all you need to do. It won't get you exactly on the resonant frequency of the tank circuit, but it will be close enough to get plenty of resonant rise.

I disagree. All power that goes into the work piece must be supplied by the inverter, so there shouldn't be a zero-current situation. We want the load to be very lossy, and thus we are not going to get much resonance anyway. I would expect a few amps out of the inverter when running properly.

Richie Burnett used a ridiculous coil which produced a very high Q. Because of this only a small fraction of power leaked into the work piece. The inverter only needed to supply that small bit of power.

Steve Conner

Tue Oct 30 2007, 05:22PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Zero current switching means that the instants where the MOSFETs switch are the same instants that the current waveform passes through zero. It doesn't mean that the current is always zero.

Ridiculously high Q's are the very nature of induction heating. The loaded Q can be maybe 50. The higher a Q your heater can run at, the more leeway you have in work coil design, and the better you can heat low-loss workpieces like aluminium and steel above its Curie point.

That's why commercial heaters use these water-cooled banks of what are in effect power factor correction capacitors. A 2kW heater will be handling maybe 100kVA of reactive RF power. This is delivered by the tank capacitor, so the inverter only needs to supply the 2kW of real RF power.

nrhoades

Tue Oct 30 2007, 11:25PM

Registered Member #610 Joined: Wed Mar 28 2007, 09:44PM
Location: Middletown, RI
Posts: 110

Steve Conner:

Zero current switching means that the instants where the MOSFETs switch are the same instants that the current waveform passes through zero. It doesn't mean that the current is always zero.

Whoops, you're right.

But... why wouldn't we want to try to drive the tank EXACTLY at resonance, even at the cost of making the bridge supply some reactive current?

Steve Conner

Wed Oct 31 2007, 09:09PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

I argued above that you are in fact driving the LCLR system at its resonant frequency. It's just that this isn't quite the same as the frequency you calculate from the tank inductance and capacitance, because of the influence of the matching inductor.

Making an inverter supply reactive current tends to be costly, especially if it's current with a leading power factor. Richie Burnett has written at length about the need to isolate MOSFET body-drain diodes with extra components if you intend to run at a leading power factor. The situation is somewhat better with fast IGBTs, but I don't think these are quite up to tabletop induction heater frequencies yet (my last one ran at around 250kHz)

BTW, isn't your system just a PLL implemented in firmware? I once used a PIC to make a 50Hz signal generator that locked itself to the AC line, and when I was done, I realised that I had just made a digital PLL.

nrhoades

Wed Oct 31 2007, 11:07PM

Registered Member #610 Joined: Wed Mar 28 2007, 09:44PM
Location: Middletown, RI
Posts: 110

Making an inverter supply reactive current tends to be costly, especially if it's current with a leading power factor. Richie Burnett has written at length about the need to isolate MOSFET body-drain diodes with extra components if you intend to run at a leading power factor.

Thanks, this is what I needed to know.

I still have yet to build anything more complicated then a flyback driver, though definately learning a lot from this forum.

nrhoades

Fri Mar 28 2008, 01:23PM

Registered Member #610 Joined: Wed Mar 28 2007, 09:44PM
Location: Middletown, RI
Posts: 110

OK, so I got something working. Nothing too great but capable of heating a quarter inch bolt to glowing red in 20 seconds. Yay. No feedback, ZCS, or anything. If I sneeze at it, it breaks.

The lesson learned is that I shouldn't expect reliability out of a dumb system. I'm going to make a smarter one now. The next easiest thing to do is use a simple state machine with a current monitor and let it run at whatever frequency it needs to. No resonance, just hard switching for now.

I attached a .ppt of my idea. I'm sure this is peanuts to you all but I like to travel along the learning curve nice and steady.

]hbridge-states.zip[/file]

1 2 3 next

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