What sort of EM field is produced in an induction heater?

[grim3271]

What sort of EM field is produced in an induction heater? And how to measure it?

That's it, basically.

Assuming an induction heater is running on 200khz, and consuming about 3000 watts, how would I go about measuring the magnetic field produced in its coil? Now, I'm rather confused about all the B-field and H-field magnetic fields, and which of them (or both) exist in the coil, and how to measure them? What sensor would I require? My research so-far concludes that I'd probably need a hall effect sensor to measure this, is this correct?

 

[PietKuip]

The simplest thing to use is a loop of wire (and an oscilloscope).

 

[grim3271]

The simplest thing to use is a loop of wire (and an oscilloscope).

So basically currents would be induced into the wire, which I would connect to an analog pin, and measure. I realized I can do that, but how can I actually turn the voltage induced in said wire into units? Gauss? A/m? V/m?

 

[PietKuip]

"Analog pin" sounds like an Arduino or something. It would need to be fast enough to measure the AC waveform.

From the voltage, the frequency, and the area of the loop you can then calculate the amplitude of B in Vs/m^2 = tesla.

http://hyperphysics.phy-astr.gsu.edu/hbase/electric/farlaw.html

 

[grim3271]

"Analog pin" sounds like an Arduino or something. It would need to be fast enough to measure the AC waveform.

From the voltage, the frequency, and the area of the loop you can then calculate the amplitude of B in Vs/m^2 = tesla.

http://hyperphysics.phy-astr.gsu.edu/hbase/electric/farlaw.html

Actually, it's an STM32F7 Discovery board, which does have Arduino headers, which has a 12-bit ADC, which is higher than Arduino. I believe it would be fast enough to measure the AC waveform, considering the microprocessor has more than 400 MIPs of performance.

Now, your equation. Shouldn't turns be taken into account?

Also, if we bring the coil too close to the induction heater, won't it induce more than 3V3 (the logic voltage of the uC) into the pin and damage the board? I've seen videos where bulbs are illuminated wirelessly from an inductor which is pushed into the coil, this leads me to believe that our coil which will measure the magnetic field will have great voltages and current induced into it, which will damage the uC. Even if we place a resistor on the input, I don't think that'd be sufficient to protect it.

This leads to another question, is there a sensor which won't do that? Something that communicates via a protocol like i2c for example?

 

[PietKuip]

The maximum frequency will probably be limited by the conversion speed of the ADC. To see the waveform, you need to sample with at least twice the frequency of the magnetic field.

Of course, if you use a coil, number of turns must be taken into account. But if you are afraid of frying your inputs, you could start with a loop (one turn), small area, and large distance. Main question may be what negative voltages the input can tolerate. If the inputs are not diode protected, you can add them. And/or a dc bias.

Of course there are other sensors (Hall-effect, magnetoresistive), there are chips with built-in ADC's and serial protocols. Electronic compasses do not cost much, but those are slow.

Easier to borrow an oscilloscope. And to use the simplest of wire loops.

 

[jim hardy]

Connect a loop to an AC voltmeter and see what kind of numbers you get.
As Piet showed,

E = N dΦ/dt
so for sinusoids only,
Φ = E/(N X 2Πf) and you know area of your coil

i've done that for 60hz fields using ten turns on 1/10th square meterSince most true RMS voltmeters aren't very fast ( even good ones like Fluke 289 only go to ~100khz )
meaning it may well under-report flux at induction heater frequency
i'd make another test with a diode in series, 1N914 should do
switch meter to DC its input capacitance should filter
that'll give you a feel for what's in the air

to observe non-sinusoids you really need a fast hall sensor or electronic integrator, and oscilloscope .

 

[grim3271]

The maximum frequency will probably be limited by the conversion speed of the ADC. To see the waveform, you need to sample with at least twice the frequency of the magnetic field.

Of course, if you use a coil, number of turns must be taken into account. But if you are afraid of frying your inputs, you could start with a loop (one turn), small area, and large distance. Main question may be what negative voltages the input can tolerate. If the inputs are not diode protected, you can add them. And/or a dc bias.

Of course there are other sensors (Hall-effect, magnetoresistive), there are chips with built-in ADC's and serial protocols. Electronic compasses do not cost much, but those are slow.

Easier to borrow an oscilloscope. And to use the simplest of wire loops.

Well, the ADC on the uC is said on the datasheet to be 2.4 million-samples-per-second each, with 3 in-total on the controller, which means a total of 7.2 million samples per second if we interleave them all. That means I can measure 1.2 Mhz with 1, and 3.6 Mhz with all 3. Considering the highest frequency the induction heater will run at is 200 khz, I'd say the ADC is plenty enough.

I realize it's simpler to just use an oscilloscope and math, but the person who I'm making the heater for said he doesn't want to bother with that. Considering he probably can't even use an oscilloscope, and that he asked me to make using the device as "comfy" as possible, that means maximum idiot-proofing, which is also the I'm using a capacitive touch screen to allow him to adjust the values. (Power level, an option to disable automatic resonance tracking and enable him to set a frequency of his choosing, the magnetic field's strength, plus whatever other gimmicks he asks for)

So, as I understand, the magnetic field produced by the work-coil of the induction heater diminishes with distance, and since I know he'd probably want to measure the intensity directly INSIDE the coil, which he can't do because doing so would empty hundreds of watts into the probe/coil/whatever, so another question is, can you compensate for distance? I think you can by taking samples at increments of distances (10, 20, 30, 40 cm, etc.) and then finding the average percentage of decrease in power, and then using that to deduce the strength of the field inside the coil, is this reasonable?

Also, in the light of all the above, can you recommend me any chip/sensor that I can use to measure it? Something that preferably poses no risk to the controller. (preferably specific part ids, or general parameters I should search by.)

 

[PietKuip]

Magnetoresistive sensors are probably not suitable. I would expect that they might get very hot in the ac magnetic field.

Hall-effect sensors would have a different problem. Trying to measure small hall voltages is not easy when wires pick up induced voltages from the radio frequency field.

 

[grim3271]

Connect a loop to an AC voltmeter and see what kind of numbers you get.
As Piet showed,

E = N dΦ/dt
so for sinusoids only,
Φ = E/(N X 2Πf) and you know area of your coil

i've done that for 60hz fields using ten turns on 1/10th square meterSince most true RMS voltmeters aren't very fast ( even good ones like Fluke 289 only go to ~100khz )
meaning it may well under-report flux at induction heater frequency
i'd make another test with a diode in series, 1N914 should do
switch meter to DC its input capacitance should filter
that'll give you a feel for what's in the air

to observe non-sinusoids you really need a fast hall sensor or electronic integrator, and oscilloscope .

So.
Tesla = Volts / (Turns * 2(3,14*frequency))
If so, what does area even have to do with that? You just need to know the turns.
Also, please refer to my previous post.

Magnetoresistive sensors are probably not suitable. I would expect that they might get very hot in the ac magnetic field.

Hall-effect sensors would have a different problem. Trying to measure small hall voltages is not easy when wires pick up induced voltages from the radio frequency field.

I think I'd agree with you about the magnetoresistive sensors. Hall-effect sensors were what I was thinking of using so far. The problem you suggest seems logical, since I read that they produce tiny signals, I even read that you shouldn't use long wires because they'd drown out the signal, does that have any basis in reality?

Also, I think that simply using coax wire or just any shielded wire for the signal and power should work?

Also, according to the reading I did, Hall effect sensors have 3 pins, voltage input, ground, and signal, the signal goes into the ADC and its voltage, which is derived from the logic voltage, changes according to the surrounding magnetic field, and the accuracy by which I can measure depends on the ADC.

But my 2 questions are-
1- How can I calculate Tesla or A/m from such a reading from the sensor?
2- Provided it gets too close to the device, can it get overwhelmed by the magnetic field and just return the full logic voltage back to the input? If so, I would assume different devices are made for different ranges?

Also, can you reply to this?

I know he'd probably want to measure the intensity directly INSIDE the coil, which he can't do because doing so would empty hundreds of watts into the probe/coil/whatever, so another question is, can you compensate for distance? I think you can by taking samples at increments of distances (10, 20, 30, 40 cm, etc.) and then finding the average percentage of decrease in power, and then using that to deduce the strength of the field inside the coil, is this reasonable?

 

[jim hardy]

the person who I'm making the heater for

can you link to any information on how you're building it?
I just pulled the inverter board from a junk microwave oven for such a future project ... just a thought
you have so much energy available in your magnetic field
assuming your ADC board has decent input impedance at that frequency,,,,,,,
i'd try a small search coil
with a home-made "poor man's approximate integrator", just a RC with time constant at least 10X period of frequency of interest,
200Khz = 5usec period you'd want 10X that or 50 usec
so using 1Kohm and 0.05uf , voltage across capacitor will reproduce the flux waveform
attenuated by transfer function of that RC circuit, so twist your leads from search coil to computer

i've done that at 60 hz never higher

keep us posted?

 

[jim hardy]

So.
Tesla = Volts / (Turns * 2(3,14*frequency))

No, that's Webers
Tesla is Webers per square meter

 

[grim3271]

can you link to any information on how you're building it?
I just pulled the inverter board from a junk microwave oven for such a future project ...just a thought
you have so much energy available in your magnetic field
assuming your ADC board has decent input impedance at that frequency,,,,,,,
i'd try a small search coil
with a home-made "poor man's approximate integrator", just a RC with time constant at least 10X period of frequency of interest,
200Khz = 5usec period you'd want 10X that or 50 usec
so using 1Kohm and 0.05uf , voltage across capacitor will reproduce the flux waveform
attenuated by transfer function of that RC circuit, so twist your leads from search coil to computer

i've done that at 60 hz never higher

keep us posted?

Well.
Your idea is nice, but 200khz is only an approximation of the real frequency. The switches are a full-bridge of FGH60N60SMD IGBTs, connected in this configuration (with some added protection for the bridge)

You should also ignore the values on the capacitor and the inductor, as I just pulled that image from a website. The inverter will actually run at 170 KHz, but considering that since the resonant frequency of the LCLR tank changes as its surroundings change(stuff is inserted and taken out of the coil) therefore I consider the frequency to be variable within a range, the 200 KHz I was using earlier is just a rounded number.

The controller circuitry is a STM32F7 Discovery board, which runs at 216 MHz, which will be used to track the resonant frequency of the tank by measuring the output changes through the measuring sensor we're discussing. As I understand, going below or above resonance decreases output, so resonance can be found and tracked by simply increasing/decreasing the frequency and checking against the values returned from the sensor. The initial frequency which will be increased/decreased from will be the calculated resonant frequency of the LC tank, which I am unsure of, because this project is still in the form of ideas and blueprints.

Further, the powerful microprocessor will display an interface on the capacitive-touch screen that comes with the board that allows changing the power (amplitude) by adjusting the phase-shift between the 2 "legs" of the full bridge inverter, turning off frequency tracking and setting the frequency manually, the interface will show also the power of the magnetic field, which is also derived from the sensor we're discussing, in-addition to any other gimmicks that I'm asked for.

So, from the switching devices, it will indeed be a pretty powerful device, considering it should be able to take 120 amps in if I could keep the bridge cool enough, but I'm not planning to do that, because I simply can't source 300V (rectified EU mains) at 120 amps, so it'd be run at 20 amps, which is about 4kW, at max power. I'm planning to cool the bridge using 4x (1 per IGBT) tiny (when compared with the monstrous copper-piped heatsinks you find today) processor heatsinks (think the stock ones that come with intel processors, but slightly smaller and without fans), which will have a single bigger fan blowing air on all the four of them.

The work-coil will be cooled via water, since it's basically copper pipe, and since the capacitor will be a bunch of CDE 940C20P1K-F, which cannot be cooled via water, I suppose I'll simply just use another fan on them.

So, all that above is simply the current ideas about the project, with probably a lot of un-related stuff, but it helps to have background, I suppose.

So, I really don't want to use a coil, since I'm pretty sure that if it gets too close to the IH, it'd induce currents into said coil as if we had a transformer, and then I'd be kissing my 50$ board good-bye.

No, that's Webers
Tesla is Webers per square meter

Aren't they both equal?
T = Weber/meter squared, so assuming just 1 meter squared, it'd be T = Weber. I still don't get what the area of the coil has to do with this. Sorry, but this is kind of confusing me.

 

[jim hardy]

T = Weber/meter squared, so assuming just 1 meter squared, it'd be T = Weber. I still don't get what the area of the coil has to do with this. Sorry, but this is kind of confusing me.

Back to Faraday, Piet's hyperphysics link

flux encircled by your search coil is Φ and is Webers. Flux used to be called Lines in cgs days when i started, later Maxwells, when we went SI it became Webers

Flux density is B, Webers per square meter or Tesla, used to be called Gauss in cgs days and was one line(Maxwell) per square centimeter, ~same order of magnitude as Earth's magnetic field. a far weaker field than a Tesla ,
10^4 gauss is a Tesla
that'd be 10^4 lnes per square cm which is 10^8 lines per sq meter
and 10⁸ Maxwells make a Weber(that's my own internal cross check on my thinking - i can't remember that stuff so have to reason it out every time and check against Wiki and I'm mistake prone so you should check me too)

so -
a search coil of diameter 2/√pi cm would have area pi X (2/√pi)²/4 = 1 square cm = 10^-4m^2
and when placed in in a sinusoidal AC field of 1 tesla peak would encircle flux B of 1_Tesla X A of 10^-4_m^2 sin(wt) Webers
so it'd produce 10^-4 wcos(wt) volts
200khz is w of 1.257E6
so your search coil should produce ~126 volts per turn per tesla at that frequency

2/√pi = 1.128 , so i described a search coil of diameter 0.444 inch

If one connects 126 volts 200khz to series combination of 10K and 0.05 uf
what current will flow? What's voltage across the capacitor? How much power is dissipated in the resistor?
Is that a practical experiment to make with a DMM in order to figure out whether you dare connect an expensive ADC ?

I notice your capacitors are rated 8 amps RMS so it'd be nice to know whether your machine is producing sinusoids or something with higher crest factor. Polypropylene has low melting point and we had some SCR snubbers made of it melt, replaced them with oil filled metal can capacitors made for commutating service . Feel of them early in your testing.

Thanks for the diagrams of your project
i'm not close to my project yet
but your resonance idea is intriguing it should encourage sinusoids...
hmmm... current gain of a parallel resonant circuit... watch those caps

old jim

 

[meBigGuy]

Can you just measure the coil current? After a one time calibration of the system, wouldn't coil current and magnetic field be related?

I'm a bit confused about whether we are talking real time control without intervention, a single calibration, or manual display and adjustment.

There are lots of gauss sensors with digital spi interfaces. Can you place one in proximity and get repeatable results after a 1 time calibration?

 

[jim hardy]

I'm a bit confused about whether we are talking real time control without intervention, a single calibration, or manual display and adjustment.

I was musing about a simple and cheap exploratory instrument to make measurements while developing this thing.
If it works he could build it into the delivered device.
Seems to me measurements of flux would be mighty handy while optimizing his coils and tuning the controls.
Simultaneous readouts of coil current and flux give an idea of what's going on inside the metal core he's trying to heat
flux is in proportion to current, but that's total current in both his coil and in the workpiece.

Difference between coil current waveshape and flux is due to 'other current '
Φ = μμ₀NIA/L_ength where I is sum of coil current and 'other current', mostly eddy current in core?
There's no derivatives in that formula so current and flux signals ought to be in phase and resemble one another unless 'other current' is significantly out of phase,
skin effect at his frequencies i expect constrains eddy current to outer mm or so

i've only worked at power line frequencies up to 400 hz so take with a grain of salt
looking forward to learning from this thread

old jim

 

[grim3271]

Back to Faraday, Piet's hyperphysics link

View attachment 94133

flux encircled by your search coil is Φ and is Webers. Flux used to be called Lines in cgs days when i started, later Maxwells, when we went SI it became Webers

Flux density is B, Webers per square meter or Tesla, used to be called Gauss in cgs days and was one line(Maxwell) per square centimeter, ~same order of magnitude as Earth's magnetic field. a far weaker field than a Tesla ,
10^4 gauss is a Tesla
that'd be 10^4 lnes per square cm which is 10^8 lines per sq meter
and 10⁸ Maxwells make a Weber(that's my own internal cross check on my thinking - i can't remember that stuff so have to reason it out every time and check against Wiki and I'm mistake prone so you should check me too)

so -
a search coil of diameter 2/√pi cm would have area pi X (2/√pi)²/4 = 1 square cm = 10^-4m^2
and when placed in in a sinusoidal AC field of 1 tesla peak would encircle flux B of 1_Tesla X A of 10^-4_m^2 sin(wt) Webers
so it'd produce 10^-4 wcos(wt) volts
200khz is w of 1.257E6
so your search coil should produce ~126 volts per turn per tesla at that frequency

2/√pi = 1.128 , so i described a search coil of diameter 0.444 inch

If one connects 126 volts 200khz to series combination of 10K and 0.05 uf
what current will flow? What's voltage across the capacitor? How much power is dissipated in the resistor?
Is that a practical experiment to make with a DMM in order to figure out whether you dare connect an expensive ADC ?

I notice your capacitors are rated 8 amps RMS so it'd be nice to know whether your machine is producing sinusoids or something with higher crest factor. Polypropylene has low melting point and we had some SCR snubbers made of it melt, replaced them with oil filled metal can capacitors made for commutating service . Feel of them early in your testing.

Thanks for the diagrams of your project
i'm not close to my project yet
but your resonance idea is intriguing it should encourage sinusoids...
hmmm... current gain of a parallel resonant circuit... watch those caps

old jim

So.
Field strength in Tesla * Area of coil * 2 * 3.14 * Frequency = Volts per turn per Tesla
therefore
Peak field strength in tesla = V/turn / (Area of coil * 2 * 3.14 * frequency)

But there's one thing, we don't exactly know the field strength that would be inside the induction heater. Also, we may add resistors and capacitors, but once the voltage across our coil goes past just 3.3 volts, it'll fry the controller, I'm saying this again because, I have nothing of which I'm talking about yet, I'm still planning/designing/deciding on the finer details of controlling, the gatedriver, actual values of the matching network, etc, I'm merely just trying to create a total "parts list" to buy at once, and assemble.

I don't mind experimenting and even sacrificing a cheap multi-meter, but I'm fairly certain that it'd be more trouble than it's worth to use it safely, considering that I'd need a step-down transformer, isolation, then probably even voltage dividers, just to give me the variable DC voltage between 0 and 3.3 that I'd need to then multiply inside the controller to get the actual V/turn.

Also, using the inverter board from a microwave seems like a nice idea, you can even use the ferrite from that transformer that comes with it for the matching inductor, then you'd only need to supply a capable capacitor, a coil, and cooling for the coil. But if it uses PWM for power control, then that'd be a problem for higher powers, because flyback currents, also I think you'd have trouble changing the frequency, unless it's a self-tuning inverter, which I doubt it is since manufacturers like to skimp as much as they can on these things.

Well, I found those capacitors to be recommended around the web, 20 of them should give 160A Rms, I think they should be fine with a fan. Those were the most economical (Specs per dollar) capacitors I could find, if you have any suggestions for more economical ones, please give them to me!

Here's an excellent resource on the LCLR topology. ==> http://www.richieburnett.co.uk/indheat.html

Can you just measure the coil current? After a one time calibration of the system, wouldn't coil current and magnetic field be related?

I'm a bit confused about whether we are talking real time control without intervention, a single calibration, or manual display and adjustment.

There are lots of gauss sensors with digital spi interfaces. Can you place one in proximity and get repeatable results after a 1 time calibration?

The way I imagine it is, a mix of both. There is a button on the interface to disable automatic frequency tracking, which locks the bridge at the last determined resonant frequency before the pressing of the button, this should be pressed before taking the sensor out of its stationary position, which it will be taken out of to take measurements from different "perspectives." Since the user wants to know the value of the magnetic field directly inside the coil, and he can't measure that directly because it would induce thousands of volts and amps into the coil and just heat it up as if it was a piece of metal.

So I suggested placing the sensor at increments of a meter (10, 20, 30, 40 cm, etc.) to find the average percentage that the magnetic field's intensity decreases by every X cm, then we can simply deduce the intensity inside the coil.

Then once that's done, the sensor is just put back into its stationary position, and auto frequency control is turned on back again.

Can you link me to any sensors that won't die near the Induction heater?

skin effect at his frequencies i expect constrains eddy current to outer mm or so

True, that's also the reason why the matching inductor should be made out of litz wire, as only 0.146mm of the wire will be used. (assuming it's copper)

 

[jim hardy]

But there's one thing, we don't exactly know the field strength that would be inside the induction heater.

Agreed, and was that not your original question ?
Its should be approximable from
http://hyperphysics.phy-astr.gsu.edu/hbase/magnetic/solenoid.html
and
http://hyperphysics.phy-astr.gsu.edu/hbase/magnetic/curloo.html#c1

i was suggesting an approach to get an actual measurement

the proposed integrator makes a voltage divider that at 200khz would give across its capacitor ~200 millivolts per Tesla in the 1cm² search coil.
just keep the concept in back of your mind in case you need it.

Well, I found those capacitors to be recommended around the web, 20 of them should give 160A Rms, I think they should be fine with a fan. Those were the most economical (Specs per dollar) capacitors I could find, if you have any suggestions for more economical ones, please give them to me!

So long as they stay below 180F they'll be fine
if you have trouble
this series should be overkill but certainly not less expensive
http://www.cde.com/resources/catalogs/SCR.pdf

Thanks for sharing those DIY links !

old jim

 

[grim3271]

Agreed, and was that not your original question ?

Actually, I think there might have been a misunderstanding. My original question was how I could go about measuring the coil's magnetic field, but that's not an actual constant, since it'd constantly change as different objects are brought into the coil, and change as they heat up, too. (For example, iron rod is inserted, it becomes an electromagnet of sorts, until iron reaches 700C and then it's just eddy currents and the skin effect from there.)

In your first link, assuming 5 turns of 1 cm copper pipe, and a relative permeability of 1, assuming air has no magnetic qualities, and RMS current of 120A, we get 151~ gauss at peak. However, just put in an iron core to be heated, you get 3~ Tesla at peak, until you reach 700C, of course. And surely, the peak intensity produced will vary a lot depending on the size of the iron core, and will also vary if something with a relative permeability other than 200 (the one stated by the site for iron) is inserted. And considering that the user informed me that he'll be playing with iron oxide nano-particles, and also intends to melt metals.

So, our sensor will need to measure accurately 150 gauss to 3 tesla at max.

With my 12-bit ADC, I'd be only able to measure in crude steps of 60 gauss, which I don't think is ideal.

this series should be overkill but certainly not less expensive
http://www.cde.com/resources/catalogs/SCR.pdf

Sweet mother of Joseph, these are much worse value than the other caps, and are much heavier, too. I don't think build for that level of duty are necessary for our 4000W IH

Thanks for sharing those DIY links !

Honestly, I should be the one to thank you for giving me all the help you've given me.

 

[jim hardy]

However, just put in an iron core to be heated, you get 3~ Tesla at peak, until you reach 700C, of course. And surely, the peak intensity produced will vary a lot depending on the size of the iron core, and will also vary if something with a relative permeability other than 200 (the one stated by the site for iron) is inserted.

Now you're getting down to why some measurement of flux would be desirable .

Iron in an AC field magnetizes from its outside inward
so only the outer skin is 'magnetically there' at significant frequency
meaning when you insert the iron your flux will increase not nearly so much as one expects
because the whole middle of the workpiece is magnetically not there

the experiments i did were at power line frequency
a solid stainless steel bar about two inch diameter was our core
it had μ_relative of a couple hundred we thought
but we hadn't realized that'd be frequency sensitive
inside a ~8 inch coil
whole assembly twelve feet long
at 60 hz its μ_relative was more like 15
above 400 hz the coil was oblivious to whether the core was inserted, ie μ_relative of 1
at 3hz it was a way better core with μ_relative approaching 100

at 200 khz I've no idea what to expect especially with your iron oxide nanoparticles
and it'll depend on your workpiece's resistivity anyway

seems to me a small search coil somewhere inside that copper tube inductor
for use as a flux detector
might come in handy for the first one of these gizmos that a fellow builds.
I'll sure put one in mine.

I've already learned a lot from your link - still trying to digest that business about commutating ...

thanks again

 

[jim hardy]

Sweet mother of Joseph, these are much worse value than the other caps, and are much heavier, too.

Yes they're industrial grade and very expensive.
Our failures were spectacular so took no chance of recurrence

I see what looks like same caps you chose in those photos of actual working device,,, and you can't argue with success.
Feel your caps and if they're running cool you're okay
if not look for motor run capacitors they should be less costly than those i linked