Design considerations for an induction coilgun / Electromagnetic Projectile Accelerators / Forums

Forums

4hv.org :: Forums :: Electromagnetic Projectile Accelerators

« Previous topic | Next topic »

Design considerations for an induction coilgun

1 2 next

Move Thread

LAN_403

Dr. Shark

Tue Jul 10 2007, 05:13PM

Registered Member #75 Joined: Thu Feb 09 2006, 09:30AM
Location: Montana, USA
Posts: 711

The thread about general coilgunnery and a lot of reading in the IEEE magnetics has left me wondering if a reasonably simple, reasonably efficient induction coilgun could be made. A pioneer in the field is certainly Thomas with his two guns discussed in this thread

and on his homepage

While achieving exceptional velocities, the efficiency of his apparatus seems quite low (about 2% in the four-stage version). On the other hand, the induction ring launcher described in great detail by FastMHZ here

has an efficiency reaching into the double digits even though it is very simple with only a single stage and a hefty 3kJ electrolytic capacitor bank. I attribute the good efficiency to the good magnetic coupling between the pancake coil and the aluminium disk. His data also shows a dramatic increase in efficiency when he switched to a projectile alomst an inch thick; aparently it takes a lot of metal to cancel the magnetic field of the coil and therefore create a large inductance gradient.

Some back-of-envelope calculations suggest that a single 350V 'lytics should be capable of supporting an induction launcher stage: With a 10-turn coil of 2uH and 10mOhm, mated to a capacitor with an ESR of another 10mOhm and ESL of 100nH (these figures come from a cornell dublier App Note), commuted by a perfect switch, should be able of discharging with a peak current of 10kA with a pulse as short as 50us. If this sounds overly optimistic, it is possibly the perfect switch. Reality will break my dreams - or my walltet - again.
Anyway this should create a field of about 10T in the coil, canceled by another 10T induced in the armature. Now, how to calculate the force I have no idea, but since the energy in an inductor is E=1/2 L I^2 and F=dE/dx, I assume it could conceivably be F=1/2 I^2 dL/dx.
In a perfect world the inductance would go from zero to 2uH more or less linearly from say x=0 to x=2cm, so the (peak) force would be 5kN.

This may not sound like a lot, but

tells us that using iron at the saturation point (1.7T) the force would be "one million newtons per square metre" or 100N on a 1cm^2 armature.

I have no idea if this makes sense, thats why I ask. If my equations don't make sense, please suggest better ones.

Regarding the electrolytic and the perfect switch, I wonder if it would at all be possible to use IGBTs, so all the stages of the gun could be powered by a single capacitor. This would save a lot of energy that would otherwise be wasted in a ringing discharge heating the coil.
I don't really know how to determine if an IGBT is up to this other than by the "turn it up till it breaks and then back a bit" method. I think it boils down to the action, i.e. the integral of squared current. There was a thread about this on 4hv, but google wont turn it up.
Even if affordable T247 IGBTs could survive this torture, I wonder if they would cound as a "perfect switch", or if they do have some resistance after all.

Otherwise I would have to use SCRs and one cap per stage with all the implied drawbacks, but even then if I can get the performance of FastMHZs launcher into a slug-shooting format I would be very happy.

...

Tue Jul 10 2007, 08:02PM

Registered Member #56 Joined: Thu Feb 09 2006, 05:02AM
Location: Southern Califorina, USA
Posts: 2445

regarding the switch, you are going to have an incrediby hard time finding an IGBT capable of 10kA. Unlike SCR's which usually have extreemly high Ipk (a 2" puck will take 10kA for a full ms), a 500A IGBT brick won't take more than 1kApk. Although for your induction launcher, you aren't worried about suckback, so you may as well just dump the whole cap in... Sure the eficiency will drop considerably, but that eficinecy drop doesn't hurt the velocity at all (in fact it should actually help it a bit) but rather just increases your recharge time a bit.

But on the brigt side, that SCR (or IGBT) won't be dropping more than 5v--unless the active region is a ball of plasma at the time

Dr. Shark

Wed Jul 11 2007, 12:51PM

Registered Member #75 Joined: Thu Feb 09 2006, 09:30AM
Location: Montana, USA
Posts: 711

Thanks for your input, ..., I think this is a point that deserves further discussion. As I mentioned there was a thread (on the old 4hv?) where EVR explained the failure modes of overstressed IGBTs. IIRC, the real killer is overheating of the die, and this is determined by the time integral of the square of the current. Ignoring turn-on and turn-off, this means that there is no "peak current", only a maximum pulse time for any given current.

Sadly IGBT data sheets don't have any info whatsoever on this, they don't even give an example of peak current / pulse time which would allow to calculate the action integral.

However, some IGBTs are "short circuit rated" for e.g. 10us, I wonder what this means. The short circuit current is not given, but if it reaches 10kA, will the IGBT still survive for 10us? If so, that might be all that is needed.

I'd be most grateful if someone could have a look at my maths, since I am quite unsure if the 5kN I arrived at are reasonable. What worries me is the fact that the inductance _increases_ as the armature leaves the coil, and so does the energy that is stored in the magnetic field. Following this logic, the armature would supply energy to the coil, and not the other way round. Any thoughts on this?

Steve Conner

Wed Jul 11 2007, 01:18PM

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

OK, I'll take a stab at this one, since I've experimented quite a lot with overdriven IGBTs, and done a lot of background research.

First of all, the ratings published by the manufacturer are for reliable operation in an industrial context. They rate a 600A IGBT at 1200A peak because they know that the MTBF will be unacceptable above that peak current. For hobbyist applications, where they only need to last 5 minutes

they can be overdriven further.

Next, the peak current an IGBT will pass before dropping out of saturation is determined by the gate voltage. So, when overdriving IGBTs, you have to raise the gate voltage above what the datasheet recommends. This shortens life too, by stressing the gate dielectric. Also, an IGBT that was short-circuit proof will generally not be short-circuit proof any more at higher gate voltages. (They rely on the dropping out of saturation to limit the current to a safeish value.)

On this subject, short-circuit tests are performed with a real short circuit. The IGBT die drops out of saturation and sees several hundred volts across it while passing a couple of thousand amps. The dissipation and themal stresses are colossal. Manufacturers usually don't guarantee the device to withstand more than 100 such short-circuits over its whole life.

In a pulsed power application, you would expect that most of the voltage drop would be across your load, rather than across the die, which would stay in saturation because you used a higher gate voltage, and a load of a suitable impedance. Hence the dissipation would hopefully be much less.

Again, though, there are high current failure mechanisms that IGBT manufacturers don't like to talk about, presumably because it gives away commercially sensitive information.

I believe the highest current I ever dealt with was about 8-9kA through two 600A IGBT blocks in parallel. They survived and still seem to be in good working order.

Dr. Shark

Wed Jul 11 2007, 01:35PM

Registered Member #75 Joined: Thu Feb 09 2006, 09:30AM
Location: Montana, USA
Posts: 711

Thanks for filling me in on those details, Steve! This gives me hope that I might be able to pull this through with fairly standart IGBTs. Since I have a big bunch of small To220 devices I don't have an immediate use for, I imagine I should do some stress-testing myself, in particular to determine the maximum "save" gate voltage. I wonder if 50V would be a starting point?

I think the stress encoutered in DRSSTC operation is quite different from a pulsed power application since it is semi-repetitive. While an IGBT might blow out in a tesla coil after half a second at 60bps, 30 shots would be an "acceptable" lifetime in a gun, and lower operating temperatures might help even further.

I am amazed about the short circuit conditions you mention, it seems that if an IGBT survives THAT, it could take a beating dumping big caps into low inductance coils any day. But then, once out of saturation the high resistance across the IGBT automatically limits the current quite a lot, so for a TO247 devices we are probably not talking about thousands but "only" 100s of amps.

Eric

Wed Jul 11 2007, 11:55PM

Registered Member #69 Joined: Thu Feb 09 2006, 07:42AM
Location:
Posts: 116

joe wrote ...

I'd be most grateful if someone could have a look at my maths, since I am quite unsure if the 5kN I arrived at are reasonable. What worries me is the fact that the inductance _increases_ as the armature leaves the coil, and so does the energy that is stored in the magnetic field. Following this logic, the armature would supply energy to the coil, and not the other way round. Any thoughts on this?

The F=1/2 I^2 dL/dx part is right. I know it seems counter intuitive but the system does seek to increase inductance / minimize field strengths. Works the same way with a rail gun if you think about it. I'm not sure if your 5KN is reasonable, I usually model these things in FEMM and make a small perturbation in x to get an accurate dL/dx for that calculation. I'm guessing that the dL/dx gradient is not at all linear in your example. It should become very high where the two coils come closest.

Dr. Shark

Thu Jul 12 2007, 03:31PM

Registered Member #75 Joined: Thu Feb 09 2006, 09:30AM
Location: Montana, USA
Posts: 711

I have typed up a little maths on

I hope it is a little clearer with proper equations and all. I will try and verify the stuff by experiment this weekend, but it would be nice to hear some comments on the manuscript (warning: work in progress!).

I apologize for the pseudo-professional tone, while I don't want to sound like Sam B., I am hosting this on my work server so I thought I'd better give it a scientific appearance

So far the main challenges I have identified are:
1) Coupling between armature and coil: Ideally this should be as high as possible, firstly to get the current risetime faster, and secondly because this is the transfer mechanism for energy from the coil to the projectile.

2) Resistive losses are a bitch, what a surprise! In order to get a reasonalbe field of 10T or so even in a small volumeof space requires one to burn in excess of 1MW in as little as 50mOhm of stray resistance. Ideally one would like to recover the energy that goes to the coil with e.g. a "diagonal half bridge" topology, but in practice it is hard to get even critical damping.

In this regard it is probably a good excerise to use a low voltage like 350V, since using a higher voltage and less capacitance will in practice only make the i^2R losses worse.

Dr. Shark

Fri Jul 13 2007, 11:46AM

Registered Member #75 Joined: Thu Feb 09 2006, 09:30AM
Location: Montana, USA
Posts: 711

I have performed a few measurements: To my absolute amazement, the inducutance of a coil does indeed decrease when conductive material is brought close to it. Of course this is quite logical if you think about it, since the mutual inductance with the "conductive material" is shorted out and only the leakage inductance remains. Still, I did not intuitively expect this to happen!
I've put the data on a little table in

(top of p.3)

This makes for a nice link between the reluctance and the induction coilgun: both try to maximize their inductance, one does it by sucking in an iron armature, the other does it by pushing out a copper armature. The change in inductance (from my measurements) is comparable in magnitude in both cases. Nice, huh?

Dr. Shark

Mon Jul 16 2007, 08:52AM

Registered Member #75 Joined: Thu Feb 09 2006, 09:30AM
Location: Montana, USA
Posts: 711

This thread is turning into a blog, but hey, _if_ this thing turns out, that might be a good thing. As promised I made a few experiments and tried to capture waveforms, which was only marginally sucessful.

This is the lashup, messy but rather low inductance I dare say, with a slighly oversized SCR, but that's one less thing to worry about. The coil to the far left is the work coil, just right of it partially hidden the current transformer I used to capture the following waveform:

This is 100us/ division, my scope doesn't do any better in single shot mode. Top trace is the trigger, bottom the current in the sense transformer. A rather skewed waveform, maybe one of the GDT-gurus can comment on that? I suppose it could be saturating, but then I am only giving it 100mV / turn or so.
I guess the only thing this shot really tells me is the period of the RLC-circuit, so from this I can work backwards to determine the true ESL and ESR of my cap. The I will go ahead and compare this with discharge waveforms with an actual projectile in place, but first I need to fix that current transformer

Edit: Hang on, I am trying to pass DC through that transformer since my circuit should be more or less critically damped. No wonder than that it is doing something funny. So maybe the little negative dip after 400us is not an actual voltage reversal (which it should't be since the SCR only conducts in one direction) but just the core resetting? Mysterious....

Steve Conner

Mon Jul 16 2007, 12:17PM

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

If it's an air-cored current transformer (Rogowski coil) then it senses di/dt, not i(t). Otherwise, it probably is saturating. Do you know the volt-second product of your core?

1 2 next

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