induction heating

[romy]

hi, first time poster here. i have a few questions on induction heaters for metals, as i decided i need such a device for my machine shop. for now im playing around, trying to get some understanding of these matters.

i do most of the electrical work myself and have some practical knowlegde of electronics (like cutting  an igtb out the welder or replacing a cap in a vfd) but consider myself totally clueless from the perspective of a professional. i have 3x380vac mains on 40aac breakers. so here goes:

1) power ratings

if i have this: https://www.ebay.com/itm/174540923486? and see 40adc at 37vdc its 1'480w.

however this:  https://www.ebay.com/itm/225251784975? is confusing. i found out (they are reluctand to give this out) it consumes 28aac (6'440w) and will run on a 32a breaker. the gage seems to be showing a max of 800a*. what are they trying to say? that 15'000w/800aac = 18.75vac is the voltage accross the work coil?

*i see this says 600a, i cant find the 800a-contraption right now. they all show 800a on youtube.

would this perhaps be real 15kw?

https://www.ebay.com/itm/324023764415?hash=item4b715245bf:g:4BUAAOSwyWZZXgqL

2) does anyone own this? how does it look to you? i assume it would run on two phases without the neutral?

https://www.aliexpress.com/item/1005002033519057.html?

3) i saw a thread on here where a member was building an induction heating circuit (must have a different title). it was 10+ pages but i cant find it. any idea what it might have been? (maybe im dreaming and it was somewhere else.)

im considering eventually getting a professional quality unit (support, spares etc.). 10kw is what i have in mind, as the prices seem to explode further up (>$20'000).

im reading all the other threads and have more technical questions also, but this is it for now.

[klugesmith]

The high frequency volts and amps in work coil are almost 90 degrees out of phase, so their product is measured in VAR, not in watts.   Typically the VAR is many times higher than the power input from wall, and the power output to workpiece.  It's because the coil and internal capacitor are operating at their resonant frequency. But VAR is not hard to measure, and can fairly be used to compare the possible heating ability of different induction heaters.

[romy]

thank you. i understand there is resonance in the coil. the question is, what would a power meter show on the input? i guess about 80% of that could be transferred into the work piece. yes?

i also believe the gain can be an order of magnitude in the coil, if the system is in resonnance and the curve is "sharp", so that the 15kw is way underrated in case is supposed to express whats going on in the coil. but maybe these contraptions are so lousy that its correct?

[klugesmith]

Here are some wild-ass guess numbers, from someone without much IH experience.

Suppose unloaded work coil and IH capacitor are working at 600 A, 100 V.
That's 60 kVAR, which might need 1 kW of power from wall to keep it going.
It would be a 1 kW heater for the stream of IH cooling water.

If the coil has 5 turns, then a 1-turn secondary in ideal transformer might theoretically approach 3000 A or 20 volts.

If a real workpiece had induced eddy current of 1200 A at 5 volts, it would receive 6 kW of heat.
Wallplug input power to IH would go up by a little more than that.  Coil current, voltage, and frequency would change by some percentage.

[davekni]

3) i saw a thread on here where a member was building an induction heating circuit (must have a different title). it was 10+ pages but i cant find it. any idea what it might have been? (maybe im dreaming and it was somewhere else.)

Perhaps one or more of these threads:
https://highvoltageforum.net/index.php?topic=530.msg11972#msg11972
https://highvoltageforum.net/index.php?topic=1561.msg12104#msg12104
https://highvoltageforum.net/index.php?topic=1976.msg14755#msg14755

The induction units you show appear to be almost all dishonest about power.  Claiming more power than is drawn from the line.  The one possible exception is the "2500W" 48V unit.  Under ideal conditions you might get close to 2500W.  However, you will need a 55A 48Vdc supply in addition to the unit, and adjust work coil and load to draw no more than 55A else unit fries.
At least the claims aren't as absurd as flashlights that could double as rapid camp fire starters:
https://www.ebay.com/itm/204177345517
900000 lumens is at least 2500 watts, all from a single 18650 cell.

[petespaco]

I'd be extremely suspicious of that particular Aliexpress offering.  I don't think the seller even knows what it is.  I sure don't.

[romy]

kluge, i will have to brush up on reactive/apparent/active power (relating to motors i thought i had it worked out: cos phi). but did you actually say this: the 1kw resistive load is equivalent to 60kvar reactive power out of the transformer?

dave, unfortunately it seems even professional units are rated that way. my conclusion is that the dc units show real power, as you say, and the mains-units need degrading by 1.73 to be comparable. yes?

pete, are you referring to the 230vac board? does it seem to have a smoothing stage? it looks like a cooker to me (sure no expert here). what do you think? i wonder if i eventualy get any reply from them.

so nobody has  bought/used it?

4) oscillating frequency of apparatus

i understand how energy "vibrates" between the capacitor tank and the the work coil. it gets exited externally to compensate for the energy transfer to the work and internal losses. this input has to be in phase. frequency is defined by this hardware and is inversely proportional to the roots of C and L including (!) the work. at perfect resonnace the current is limited by the resistance of the coil and connections only.

how come then that:

a) apparently F can be adjusted externally? i see tuning from above to reach resonnance
b) i see external "self-tuning" cirquits being added
c) apparently the resonnant apparatus can be hooked up to the toroid of a welding inverter and F adjusted by a variable resitor in its cirquit
d) in the case of the "1800w-aliexpress" contraption F remains (seemingly?) unchanged when work is inserted into coil? im seeing a drop from about 51khz to about 48khz. no idea if its real, as im manually handling the workpiece.

i have come accross a circuit, where there is an inductance parallel to the work coil for the purpose of stabilizing F, without gaining deeper insight of how that works.

im i just confused? what are your thoughts.

lets see if i can get a schematic in here. this is what im looking at. what would be the correct name for this, btw? royer/jensen/baxdall/mazilli seem to be used loosely (sinus/square) and all have a center tapped coil, if im not mistaken.

[davekni]

dave, unfortunately it seems even professional units are rated that way. my conclusion is that the dc units show real power, as you say, and the mains-units need degrading by 1.73 to be comparable. yes?

Are you the professional units you reference fed 3-phase line power?  If so, they are probably accurately rated.  If AC line input voltage is listed as phase-to-phase, total power from all three wires is sqrt(3)*V*A = 1.73*V*A, presuming unity power factor and equal current from all three phases.  None of the links in your initial post appeared to be three-phase line input, though I couldn't tell for sure.  If one of them was 3-phase input, then perhaps it was accurately specified.

d) in the case of the "1800w-aliexpress" contraption F remains (seemingly?) unchanged when work is inserted into coil? im seeing a drop from about 51khz to about 48khz. no idea if its real, as im manually handling the workpiece.

Seems reasonable.  Presume your work piece was iron/steel to cause a frequency decrease.  Frequency will increase a little with non-magnetic work pieces, and with steel once above its Curie temperature.  How much frequency shifts depends on coil and work piece geometry, especially how tightly the coil fits around the work piece.

lets see if i can get a schematic in here. this is what im looking at. what would be the correct name for this, btw? royer/jensen/baxdall/mazilli seem to be used loosely (sinus/square) and all have a center tapped coil, if im not mistaken.

I don't know the correct name either.  I tend to call them ZVS oscillators for Zero Voltage Switching.  I gather "royer" is a square wave version without resonant caps and relying on inductor saturation to trigger switching.
The schematic you show does not have a center-tapped coil.  This is the most common version, presumably used in the 1800W unit you mention.  If L1 is center-tapped, then L2 and L3 can be combined into a single slightly-smaller inductor.

[Alberto]

I have one similar of the first link bit gets very hot and make noises.

It came without coil.  i made one but I'm not sure how many turns it should have. How can I calculate the number of turns? What effect does it have changing the number of turns?

[petespaco]

To: romy-

pete, are you referring to the 230vac board? does it seem to have a smoothing stage? it looks like a cooker to me (sure no expert here). what do you think? i wonder if i eventualy get any reply from them.

Yes, the 220 board.  Without a schematic or a better description, I, personally wouldn't touch it with a ten foot pole. That "Description" chart doesn't make any sense at all to me.

To: Alberto-
That first link offers several different models.  If you go there and select the one you bought, several thumbnails for it show up to the left of the main picture.  One of them shows the coil and I can count the turns and estimate the diameter, so I am sure that you can, too.

[romy]

(off topic: looking for a simple, clean and fast way to heat up a crucible? make a thin cylinder out of silicon (silicium) carbide. stick your carbon crucible in it, put insulation around, below and on top of it and insert into you microwave. my "1.5kw"-labeled (whatever it means) microwave will heat up a 80 mm crucible to bright red in a few minutes.)

so there is no real answer to (4)? that probably means im confused. am i mixing up different topologies?

ad 4b): oscillating F is inversely proportional to sqrt of L. L is proportional to relative magnetic permeability. if i insert a piece of steel into the coil i fill half of the area with material that can have a permeability up to 2 orders of magnitude higer than air/free space. (did i get it right?) so why do i see a negligible (if at all) drop in F on the "1800-contraption"? it cant be that much out of phase, right?


5) penetration deph and appropiate coil frequency

standard pd is defined as the distance where magnetic flux reaches 37% (a little less, where does such a number come from?*) of max. the recommendation is to use 1/3 of diameter max. this makes sense, after that the flux tends to cancel out.

so taking steel (e.g 1045) 30mm diameter and 60 mm long (360g) we would aim for 10mm pd. this would be achieved at way below 500hz at room temp. pd would rise to about 25mm at currie and 30mm at 1'200°c. to get it down to 10mm F has to increase to around 6khz. this is even more pronounced with thiner work. flat stock of 6mm needs 2mm pd. this is the case at 500hz at room temp and way above 100khz at 1'200°c.

the above are very rough empirical numbers, they greatly vary with the type of alloy, the exact setup and probably all other (interdependent) factors you might think of. there are formulas, but my understanding is they are not especially usefull. however, fact is we need a frequency controll of at least an order of magnitude for efficien operation.

how do we manage to do that?

i see small (5-45kw) commercial units operate on 35-80khz, 20-150khz. so they clearly are not efficient for thinner stock or small parts. on top of that, those i looked at will "tolerate" those F depending on the working coils used, but will not adjust F during operation in any way. is this the reason you need a 45kw (45kva?) unit to heat up 1"x1" stock i short time? im getting the impression, much less power is neccessary if the system is tunable/autotuning.

thoughts?


*oh, 1/e.

[klugesmith]

>> *oh, 1/e.

In EE trade, what you call pd (penetration depth) is more commonly called skin depth.

For induction heating, it's very common to use frequencies where skin depth is a tiny fraction of the workpiece thickness.   Whether we are heating a frying pan or a 25-mm-thick bar for forging. 
On the plus side, that makes the sheet resistance relatively high, so we get more power for a given amplitude of eddy current.  High frequency also means more power density from magnetic hysteresis loss in ferrous material below Curie temperature.
On the minus side, sometimes, that thick bar could get incandescent at its surface while still cold at its core.

[Alberto]

To: Alberto-
That first link offers several different models.  If you go there and select the one you bought, several thumbnails for it show up to the left of the main picture.  One of them shows the coil and I can count the turns and estimate the diameter, so I am sure that you can, too.

Thank you!

Yes, I didn´t think about that

[romy]

kluge, thanks, you helped me realize that "skin effect" and "penetration depht" are the same. while s.e. in conductors is often given as 86% of power density it corresponds to the 37% current density used for p.d. in induction hardening and apparently is derived from the same formula. question: what do we do about the positive dependency of p.d. on magnetic field intensity? how does that enter the calculation?

ad (5)

so in a nut shell: a much higher frequency is needed to get steel to forging temp than offered by most commercial units (at least as i see it).

in industrial applications this apparently is compensated by large power densities and short heating times. this strategy is wastfull for the obvious reason and additionaly for two others, that mittigate the effect:

- p.d. increases with magnetic fiel density
- "magic wave" phenomenon: second (or even first) max. of power density under the surface (→max. in permeability/critical field intensity/currie)

question: why are there no devices that automatically increase working F from say 10 to 500 khz (that i understand is about the limit for current igbts)? is it technically impossible or is it because maybe they were designed before it was possible?

[klugesmith]

>> why are there no devices that automatically increase working F from say 10 to 500 khz (that i understand is about the limit for current igbts)? is it technically impossible or is it because maybe they were designed before it was possible?

It might work if you abandon the use of resonant circuit in induction heater.  To get a 50:1 change in resonant frequency, without changing the coil geometry, you would need to change C by a factor of 2500.

[Anders Mikkelsen]

The way I like to think of it is as follows:

Start with a coil, just a winding of copper tubing. This coil will have some inductance and some effective series resistance. Inductance is more or less constant with frequency, while the resistance increases roughly with the root of frequency. The ratio of reactance (2*pi*f*L) to resistance is called the Q factor, and this is also the ratio of stored energy to lost energy per cycle.

Let's say we wind a 1 uH coil of 5 mm brake tubing, this will be some five turns of 50 mm diameter with a 8 mm pitch. At 160 kHz, this coil will have a reactance of 1 ohm, and a resistance of 10 mohm, giving a Q factor of 100, or a power factor of 0.01. If we feed the coil with 100 V, it will draw 100 A, and dissipate 100 W, while circulating 10 kVAR.

Now we put a piece of steel pipe inside the coil, and the effective resistance has gone up. Some of the field is coupled into the workpiece, and the workpiece has higher resistivity and higher permeability, giving much higher eddy current losses than in our coil. Now the series resistance has gone up to 100 mohm, of which 90 mohm is from the workpiece. If we feed the coil with the same 100 V, the current will still be around 100 A, but now we have 100 W of losses in the coil and 900 W in the workpiece. The Q factor is now 10, power factor is 0.1, and coil efficiency is 90 %.

I like to use Q factor and coil efficiency, because the former will illustrate how much reactive power we need to transfer a certain amount of power, and the latter will tell how much power will end up in the coil and how much will end up in the workpiece. Both can be measured without involving high voltages, high currents or expensive equipment, as long as you have a coil and a capacitor to resonate it with.

If we change the frequency, the ratio of workpiece and coil resistance will stay the same, so it has no impact on how efficient the system is, but once we go low enough then the eddy currents in the workpiece go deep enough to start cancelling out. Below this critical frequency, coil efficiency goes down. This frequency depends on the size, conductivity and shape of the workpiece



So why would we want to operate much above this critical frequency? Firstly efficiency will still continue increasing up to around the workpiece being 10 - 15 skin depths. Secondly a given coil will have a higher impedance at higher frequencies, meaning you need less current and more voltage to drive it. It's generally easier to make a driver that provides 500 amps at 500 volts, compared to one that provides 5000 amps at 50 volts. One could also wind the work coil with thinner wire to trade current for voltage, but there are limits to how far you get with this.



So now we have a coil, where hopefully most of the losses come from our workpiece. The inductance is still much larger in magnitude than the resistance is, so we need to provide 10 - 100 times the reactive power to the coil, for a given amount of power delivered. The cheapest way to do this is often with a capacitor, and if the frequency is chosen so that f = 1/(2*pi*sqrt(LC)) then we are left with pure resistance. The higher the Q factor is, the narrower the frequency window is, and inductance changes with the workpiece, which is the reason for the active feedback used in a lot of circuits. There are two concerns here, one is that the phase (and therefore power factor) changes rapidly with frequency with high-Q tanks, which firstly impacts the power factor as seen by the inverter, but more critically also affects the conditions under which switching happens. Voltage fed transistor bridges are a lot more efficient when a controlled amount of inductive power factor is present in the load, due to zero voltage switching, and one of the main tasks of the control circuit is to maintain this condition.

ad 4b): oscillating F is inversely proportional to sqrt of L. L is proportional to relative magnetic permeability. if i insert a piece of steel into the coil i fill half of the area with material that can have a permeability up to 2 orders of magnitude higer than air/free space. (did i get it right?) so why do i see a negligible (if at all) drop in F on the "1800-contraption"? it cant be that much out of phase, right?

Permeability of steel is high, but for higher frequency, the field doesn't penetrate deep into the material (skin effect), so most of the core does not contribute to the path reluctance. At higher frequencies, a tiny fraction of the outer skin is contributing high permeability, but the inside of the core is contributing zero permeability, so putting an iron core in a coil can actually reduce inductance. Putting conductive non-magnetic materials in the core always reduces inductance.

question: why are there no devices that automatically increase working F from say 10 to 500 khz (that i understand is about the limit for current igbts)? is it technically impossible or is it because maybe they were designed before it was possible?

It's possible, but due to the fact that the coil will have a very different impedance at the two frequencies, then completely different levels of current and voltage are needed at the two operating points, making it hard to do a single inverter that can do it. There are heaters that run at two frequencies at the same time, but that's mainly to tailor the heating pattern for things like case hardening of gears.

[romy]

The high frequency volts and amps in work coil are almost 90 degrees out of phase, so their product is measured in VAR, not in watts.   Typically the VAR is many times higher than the power input from wall, and the power output to workpiece.  It's because the coil and internal capacitor are operating at their resonant frequency. But VAR is not hard to measure, and can fairly be used to compare the possible heating ability of different induction heaters.

"The high frequency volts and amps in work coil are almost 90 degrees out of phase":

i thought at resonance in the work coil the effect of inductance and capacitance cancel out and the coil behaves as a resistive load. i can see how some shift occurs because in case we are not at perfect resonnance (low quality factor?), but 90°? please help me there.


ad (4a/c): im realizing this is a misunderstanding on my part. they are not changing the inherent resonant frequency of the oscillator. they are adjusting the frequency of the exiting input to match the former. it applies to bridge-circuits like below. the "zvs" circuits are self tuning. yes?

[klugesmith]

>>i thought at resonance in the work coil the effect of inductance and capacitance cancel out and the coil behaves as a resistive load. i can see how some shift occurs because in case we are not at perfect resonnance (low quality factor?), but 90°? please help me there.

The combination of coil and capacitor behaves as a resistive load at resonance.   Current in coil always lags the tank voltage; current in capacitor always leads the voltage. At resonance the reactive currents in L and C cancel.
Anders's excellent description puts some numbers on an example coil.   I-V phase is very close to 90° with no workpiece, and slightly less close to 90° with workpiece.

[romy]

anders, thank you so much for taking the time to provide such a concise and yet comprehensive explanation. i think i can put away about 20 pages of (handwritten) notes and refer to this. im still digesting some parts and will surely have question later on.

6) for now im not clear about the following: if over time we see constant current and sharply dropping induced power after curie, under what conditions/ceteris paribus is it the case? what variables of the system have been kept constant? if you have a constant voltage supply, does the supplied current simply drop along with the dropping voltage in the work coil? maybe im totally off the track.

[klugesmith]

Applause for Anders' detailed explanation of critical frequencies etc, drawing from real industrial references.   (IH technology proliferated in the 1930's, using the same physics and work materials as today.)

One minor correction is called for.
>> The ratio of reactance (2*pi*f*L) to resistance is called the Q factor, and this is also the ratio of stored energy to lost energy per cycle.
Nope, it's the ratio of stored energy to lost energy per radian time.  I was reminded of that when brushing up on Q a couple months ago. My example L and f were identical to those chosen by Anders!  https://highvoltageforum.net/index.php?topic=2299.msg16805#msg16805

Consider the example Anders gave a few posts back.  L = 1 microhenry, f = 160 kHz (radian frequency = 1 million/s), reactance Z magnitude is 1 ohm, resistance = 10 mohm.
So Q = 100.  The energy loss per cycle is not 1%, it's 2 pi %.

Let's check that.  Example current is 100 amps, RMS, so the real power loss is I^2R = 100 watts.
Peak energy in coil is 0.01 joule.  If we had a 1 uF capacitor for resonance at the designated frequency, peak energy in capacitor (with 100 V RMS) would also be 0.01 joule, and at all intermediate phases the L + C energy is 0.01 J.
Our 100 watt loss is 0.0001 joule per microsecond, or 0.000628 joule per cycle.