Re: Supercapacitor coilgun
DerAlbi, Sun May 21 2017, 11:17AM

Hi and welcome to the forum i guess ^_^
Nice built - good to see some digital control.

I would be curious how you measured the 1.6kA.. this correspondents to 1.69mOhm which is indeed plausible.

Just a comment on:

I've seen in many places ideas that supercapacitors are not good for projectile launchers.
Here is a small device built to demonstrate that it can be done: [..]
The device is not optimized in any way..

See, the point is when you try to optimize it, the 2.7V is one of the first things you tackle. So all of your statement is still true (supercaps are not good; but it can be done; unoptimized). I mention this because got the feeling that you meant it in the way that because of your demonstration the general attitude towards supercaps for coilguns is wrong.. its not

The device is not optimized in any way
-pulse: fixed around 10ms, I tested various pulse widths to find an optimal value

Cheater hahaha.
I was about to ask you if you could do some PWM during the 10ms, then i noticed that this is maybe not a good idea. You do currently not have any diode in place that is keeping the inductive peak at bay. So you currently rely on the fact that your built has virtually no inductance (its more a wound resistor than not a wound coil) and the avalanche capability of your mosfets or slow switching. Just keep that in mind if you start to not optimize it further

Edit: i find it interesting since you have a small prototype where you can make quick adjustments: try to shoot aluminium and turn it into an induction launcher. for this you have to place a non magnetic slug behind / at the back-end of the coil... and i guess the pulse length can be way shorter. i would love to hear if this moves the projectile... your current may be high enough.

Re: Supercapacitor coilgun
hen918, Sun May 21 2017, 01:19PM

Yeah, interesting. I have a 2500F super cap here I might have a play with, it should be relatively easy to increase the number of stages using the same cap, as it gets nowhere near discharged. The only problem is the number of FETs needed to keep resistance down, and the long length of barrel.
I was intending to use my super cap for a battery spot welder, but the current is about a 1/3 too low with only one cap.

Re: Supercapacitor coilgun
Sulaiman, Sun May 21 2017, 01:27PM

I suspect that one way to significantly improve efficiency is to increase the load resistance,
by increasing the number of turns, e.g. x10
this will require a physically larger coil,
and a projectile of much increased mass.
... more like artillery than small-arms.

A reduced current for an extended period should reduce losses in the coil, switch, wiring and cell.
The recoil is equal and opposite to the impulse given to the projectile,
so for both your present setup and any other prototypes,
support the coil against recoil forces for efficiency.

e.g. tightly grip the coil of your launcher when you fire it,
... the coil should not waste energy by moving.

P.S. just in case you have not been there, Barry's coilgun site

Re: Supercapacitor coilgun
DerAlbi, Sun May 21 2017, 02:14PM

Hi hen918,
you might be on to something with a multistage design. The interesting part here is that the supercapacitor's energy density is 10x - 100x higher than aluminium caps (lets say 50x) according to wikipedia. This means you actually achieve a superior system by achieving just 0.4% efficiency which would give you the same energy density or weight-effectiveness as a 20% efficiency traditional coilgun*.
Not sure if 0.4% is doable however.. the low voltage makes the system sooo bad, that this could be quite a hard goal to reach.

Lets just do some calculations:
The video uses 2.7V * 1600A = 4.32kW * 10ms = 43.2J. (i assume no inductive component here, because the L/R-ratio in this build is negligible short)
At 0.4% this means 0.1728J would be within the projectile which seems about right or even underestimated, so the efficiency is likely to be higher than 0.4%

The problem will arise when you try to scale up the system due to the problems you mention. Since a low voltage system will likely to be restisively dominated.
The biggest flaw in the system above is the way-too-long coil.. it would be interesting whats possible actually.

*My above statement is tremendous false if you consider real world application btw.. in a traditional coilgun you use up all the energy in the caps while you store much more than needed in a supercap which wastes all the energy density in the end

Sulaiman:

I suspect that one way to significantly improve efficiency is to increase the load resistance,
by increasing the number of turns, e.g. x10 this will require a physically larger coil, and a projectile of much increased mass. [..]
A reduced current for an extended period should reduce losses in the coil, switch, wiring and cell.

Be careful with such proposals. First of all you mix up efficiency and overall power transfer. It is true that scaling up the thing will transfer more power, because simply said its just a bigger motor then. However this does not mean that its efficiency increases. Increasing the "load resistance" the way you describe it also wastes more energy while transferring more. It is however true that a change of the coil shape would increase the efficiency in this case, but only if you change it to a better shape ^_^.

Re: Supercapacitor coilgun
pisoiu, Sun May 21 2017, 06:25PM

Well, since it started some interest, I will elaborate a bit

Simplified DC calculation, I calculated the wire resistance, if I remember correctly, the total length is around 70cm..ish, there are 2 wires, 2.5mm diameter, connected in parralel. The voltage across the coil, during the pulse, is around 2V, measured with oscilloscope. From the total of 2.7 volts, some is wasted on the mosfet switch and some on the capacitor esr.

No pwm, just an off-on-off pulse, 10 ms wide. It is generated by a silabs 8051 development board. I do not have peak suppressing diode, as this is not a big problem so far. Looking with the scope, at mosfet on-off, there is a spike up to 40 volts or so (breakdown voltage of mosfet) but so far it looks capable to withstand the avalanche energy. To be honest I did not make calculations for this. However, the coil has an inductance in 10-100nH range, without projectile. The switching is not slow at all. Each mosfet is driven by a 9A capable mosfet driver with 12 volts levels.

That being said, my ideas toward optimizations are:

-first problem is mechanical in nature. The projectile is actually a beheaded screw. The plastic tube used to prevent contact between the projectile and the coil is very rough on the inside. This, plus the screw thread, creates a lot of friction during initial acceleration. Smoother surfaces are required.
-I would increase the projectile mass. More inertia keeps the projectile longer in the barrel, to accumulate energy.
-I would increase the current peak, maybe towards 10kA and would increase coil length. Consequently, the required readjustments on the coil resistance would seriously increase the required wire diameter, by my calculations, above 5mm
-there are some serious challenges when dealing with such low resistances. You just cannot connect two copper wires (in my case coil ends to bus bar) just by soldering them, as I did in the prototype. I am sure there is a serious loss in that place. Normally, they should be welded together, and this raises yet another issue, since for welding you have to use oxygen free copper and TIG welding procedure.
-soldering switching transistors to a 20x3mm bar, as in the prototype, was a real pain. Heating such copper mass is not easy, and when heated, it oxidize rapidly, the solder deposition is not even and there are a lot of residues which are standing in the way of a good connection.
-the coil should be inserted in some form of mechanical restriction. The intense current creates a mechanical shock in the coil windings and this propagates toward mosfets. Repeated mechanical shocks may damage them.
-during firing, indeed, little energy is spent from the full capacitor charge. There is a lot remaining for successive firing. But putting 1600A in 2x2.5mm wires is a lot. After 10 consecutive triggers with 1-2s in between, serious heat builds up in the coil and you cannot keep the hand on it.


LE:
You caught me :) Maybe 'optimal value' is not the correct term. With other values it barely worked. I had to try some values because this being my first such project, I had absolutely no idea where to start from. First pulse I tried was around 10 microseconds. I only knew that it is probably a very bad idea to keep mosfets on for 1 second or so.

Re: Supercapacitor coilgun
DerAlbi, Sun May 21 2017, 06:54PM

Looking with the scope, at mosfet on-off, there is a spike up to 40 volts or so (breakdown voltage of mosfet) but so far it looks capable to withstand the avalanche energy. To be honest I did not make calculations for this. However, the coil has an inductance in 10-100nH range, without projectile.

Seems legit. Well... 100nH/2*1.6kA^2 = 0.13J. So if you would do PWM the mosfets are dead within.... maybe 50 cycles.. good enough safety margin if you switch just once ^_^

Simplified DC calculation, I calculated the wire resistance

Awww dont! I mean yes, do! But, measure or higher confidence. Your soldering and contact resistances do matter at those levels. I recommend you use a mV-capable multimeter (if you have?) and use a 10A or 5A Power supply set it at high current and do a 4-wire measurement. Those experiments can tell you a lot in terms of where its correct to just do the math and where its important to be pessimistic. (its quite possible that doing the math is sufficient!)

-I would increase the current peak, maybe towards 10kA and would increase coil length.

A longer coil makes worse performance. The length of the coil should be as long as you projectile for your applications. Please go through Barrys website (theory and the prototypes) before you waste your time with bad ideas - i am just saying that because i am not sure if your setup withstands a lot of modifications .
Besides that you indicated that you want to use thicker wire meaning less turns.. less turns will pull less.. so you dont get any benefit from that - in general a constant coil geometry will perform equal, no matter which wire you use.

Re: Supercapacitor coilgun
pisoiu, Sun May 21 2017, 06:59PM

A new design would mean other capacitors and pretty much other everything (coil, mosfet, connections, etc). This one cannot give more than 1800A. There are some 3400F/2.85V, 4-5 of them could go there. They have 2kA per piece max I.
But this is just theory, currently I do not have time to put it to practice. However, I will study the theory, thanks for the link.

Re: Supercapacitor coilgun
DerAlbi, Sun May 21 2017, 08:41PM

Ok, i just looked up your capacitor and i see where your 1.8kA-limitation comes from. However this does not really apply to you since the 1.8kA is rated for a 1 second on-time. This thing can go higher for shorter time, no worries.
Your transistors are rated for 800A for 10ms, so 4 in parallel should give you quite some current headroom. However as i watch your video closely your mosfets are driven individually... which might be a bad idea. Due to driver mismatch the current sharing can be bad during switchting... not sure how much that matters ins your case, but on the other hand, paralleling all gates directly cant really hurt.

Re: Supercapacitor coilgun
Sulaiman, Sun May 21 2017, 11:34PM

directly paralleling fast transistors can cause parasitic oscillation at surprisingly high frequencies,
better to have individual gate drive resistors (with diodes, capacitors etc.) and sometimes ferrite beads on the gate leads too,
I have had this happen.

Re: Supercapacitor coilgun
pisoiu, Mon May 22 2017, 08:12AM

For the above design, I think this is it, I will not modify it further. Even if the capacitor can withstand higher current pulses, increasing current through coil would mean thickening the wire, but that comes with reduction of number of windings (to keep the same length), which is not desirable. Increasing the length creates other mechanical problems.

Re: Supercapacitor coilgun
klugesmith, Tue May 23 2017, 02:40AM

Nice work there, Pisoiu.

Before joining this party, I want to find a missing factor in an independent analysis of your coil. Mind checking to see where this goes off track?

Coil ID was given as 6 mm. If the OD is 11 mm, then a single winding of 34 turns would use 908 mm of wire, for R of about 3.24 mΩ and L of about 800 nH (by Wheeler's formula, the same as Barry's inductor sim). Putting the same mass of wire into 17 double turns, R and L come down to 0.81 mΩ and 200 nH. Resulting current with 2 V across the coil would be almost 2500 amps!

Pisoiu, somewhere you mentioned wire length of 70 cm. That's not enough to form 34 turns, so it must be the length of double wire to make 17 turns. Together, the two strands would use 140 cm of wire. Coil ID and OD would be about 10.6 and 15.6 mm. Now R = 1.25 mΩ. Current from 2.0 volts would be 1600 amperes, just like you said. 27,200 ampere-turns around the plastic tube. As long as you keep the wire at 20 °C.

My spreadsheet gives that coil an inductance of 470 nH, which is much higher value than other values seen in this thread. Brings L/R to almost 0.4 milliseconds, if we don't count the ESR of the supercapacitor. How much is that, by the way?

Re: Supercapacitor coilgun
pisoiu, Tue May 23 2017, 07:03AM

Thanks for helping with the calculations, I don't mind at all. I'm not very good at those, mainly because I have some gaps in the theory, here and there. I usually replace those gaps with intuition. Sometimes it works, sometimes not, but I like smoke shows and fireworks anyway :)
Coil ID is most certainly 6mm because I used a 6mm aluminium rod to build the coil, then I extracted it out and replaced with that plastic tube which has also 6mm diameter. Also, the copper wire is 2.5mm OD.
But the wire length is very approximate, because I initially prepared 2 wires, aprox. 1m length, then winded those on the aluminium rod, then cut the excess off. But I was not wise enough to exactly measure the length before winding it, so I cannot say for sure how much wire is inside. What I can say is that the coil length is very close to the capacitor length, which is 108.5mm according to datasheet.
The DC ESR for this capacitor is 270 microohms and the combined rdson of mosfets is 137.5 microohms typical. But rdson is measured at 100A Id, so, I cannot say for sure what is the rdson at 1000 amps.

Re: Supercapacitor coilgun
klugesmith, Wed May 24 2017, 08:38AM

Thanks for clarifying the dimensions. From geometric calculations I'm still getting about 810 microohms for the coil (two 454-mm wires in parallel). That doesn't match your posted voltage and current values, even if we accounted for the change in cross-section and resistivity from tightly coiling the wire.

You've told us the resistance of your supercapacitor and MOSFETs with high precision. Why not do the same for the coil resistance?

You could weigh the coil and look up ohms per kilogram (or per pound) in a standard wire table.
That depends on being sure of the wire gauge. 2.5 mm OD is too small to be 10 AWG, especially if measured over the insulation. And it's not close to any standard square-mm conductor size.

The best way to get the coil resistance is to measure it. Can you run a steady current through the coil, perhaps about 10 amperes AC or DC, and measure the voltage drop? You can get measurements 4 times larger by temporarily separating your bifilar winding at one end, then driving the strands in series instead of parallel. Don't forget that copper resistance goes up 10% for each 25 degrees C.

Der Albi wrote: >> supercapacitor's energy density is 10x - 100x higher than aluminium caps (lets say 50x) according to wikipedia.

Yes, but supercapacitors have lower power density. You need seconds to get all that energy out, compared to milliseconds with the aluminum electrolytics. And microseconds with high voltage (multi-kV) caps.

Re: Supercapacitor coilgun
pisoiu, Wed May 24 2017, 08:56AM

I told you parameters for capacitor and for transistors with high precision because it's in the datasheet. The coil was built on the rush and unfortunately I was not paying enough attention to its details. But I will try to measure it as soon as I have some time. I have means to run high currents through it, even if with not great precision (it's a 10A benchtop power supply) and I can also measure voltage with precision of 4 decimals, 2 volts full scale. Will get back with data when it's available.

Re: Supercapacitor coilgun
klugesmith, Wed May 24 2017, 09:11AM

Sounds good. I forgot to mention that it's easy to protect your MOSFETs from overvoltage when they turn off. Just put a suitable pulse-rated diode in parallel with the coil, as you would with a relay coil. It makes the current decay more slowly, but that's insignificant in this case, because L/R is so short.

Re: Supercapacitor coilgun
pisoiu, Wed May 24 2017, 09:36AM

Thanks, I'll keep that in mind. However I have some doubts, mainly because the physical connection is long considering the coil dimensions and I'll have to put some wires(read inductance) on the diode leads. And if I remember correctly, that pulse looked extremely steep on the scope, under 20ns. It may be even steeper, as my scope is not a very high speed one, and those 20ns may not be the real signal, rather the limitation of the scope...

Re: Supercapacitor coilgun
pisoiu, Thu May 25 2017, 08:31AM

I did the measurement, with a current of 4.67A, the voltage at coil terminals is 5.22mV, results R=1.117 miliohms. With this value, the current through coil, at 2V voltage, is around 1700A.


LE: I'm very sorry, I was mistaking about diameter of the coil. The ID of the coil is 8mm, not 6. The ID of that plastic tube is 6mm actually. There was a time when I was younger and my memory was better....

Re: Supercapacitor coilgun
klugesmith, Fri May 26 2017, 01:27AM

Oh good. With your new coil diameter, my resistance model is only 10% different
from your new, measured resistance value (using method suggested by Der Albi).
Close enough to say there's no mystery.

Where did you get your coil inductance numbers, given as 10 to 100 nH? Wheeler's formula is now giving me around 250 nH.

Those voltage spikes at MOSFET turning-off time indicate a circuit dominated by inductance, at the time scale of switch transitions. Just as it's dominated by resistance, at the timescale of the projectile-launching pulse.

You could get the inductance value by measuring the height and width of a braking-voltage spike.
For example, suppose FET drain to source voltage is 40 V for 5 microseconds, as the coil current ramps from its final on-state value to zero, and the other end of coil and supercapacitor remain at +2 volts.
That would be (40-2)*5 = 190 volt-microseconds on the coil.
Same as the product of coil L and the on-state current, in ampere-microhenrys. e.g. 1000 A * 0.19 uH, or 1700 A * 0.11 uH.

We don't need to ignore the resistive aspect of the circuit. The I*R drop from the coil's own wire resistance helps to stop the currrent. It's about 2 V during most of the 0.01-second ON pulse. Let's say 1 V, on average, during the on-to-off ramp. So it contributed 5 volt-microseconds of stopping power, bringing the total to 195. This "refinement" would increase the bottom-line inductance result by about 2.5% in this example.

Suppose you deliberately made the MOSFET gate voltage slew more slowly, enough that the "braking" voltage amplitude did not exceed the "breaking" V_DS specification on the data sheet. I think that would be less abusive to the transistor, even though the total energy is the same. Maybe a power MOSFET expert will speak up here. Isn't this a place for a snubber network?

I stand by my earlier recommendation of a plain clamp diode. It would be nice to find one with an impulse current rating (I^2*t, given in A^2*s). Not counting inductance of the diode circuuit, the braking voltage (between coil terminals) is then set by Vf(diode), on the order of 1 volt. When helped by the coil's IR drop, which we said averages about 1 volt, it might take about 100 us to bring the current to zero.

Re: Supercapacitor coilgun
pisoiu, Fri May 26 2017, 07:21AM

Coil inductance numbers were mostly to indicate range, rather than an exact value. They were taken from some calculator from net, don't remember which one. If your calculation is 250nH, then it must be close to reality.
I don't know if I can measure true ramp of that spike. I have several scopes, but the fastest have 100mhz bandwidth. In this case I'm not sure if the scope shows the true ramp of the spike.
Thanks for analysis, it will sure help if I decide to proceed to next design. But this is uncertain due to lack of time.

Re: Supercapacitor coilgun
pisoiu, Tue Jun 26 2018, 07:56PM

I plan to move further with the project and make some tests with higher numbers. But I need a bit of help. In the next design I plan to use some maxwell supercapacitors, a number of them, around 5-6. The current they will be able to supply will be around 10kA. In order to properly manage such currents, the main connection will be made with copper busbars, for now I have 20x3mm available. It is easier this way, as I can solder switching mosfets directly on them using some pcbs as support and connection for driving circuits. However, the problem is related to connecting copper busbars to capacitor ends, which are threaded with M10 and made from aluminium. I do not have experience in aluminium-copper joints which must withstand such currents (around 1500A per connection). Does anyone have any advice? Thank you.

Re: Supercapacitor coilgun
DerAlbi, Wed Jun 27 2018, 12:28PM

The way you calculate you current hints that you plan to use the capacitors in parallel. Thats a big problem since the low voltage is your biggest issue.
Put them in series: it will give you at least 15V to work with which leads to much more manageable currents.
I feel like you have the misconception that coilguns are magic beast that operate at ridicules high currents the higher the better. Thats not true. See it as a motor. Depending on how you wind your coils you can build equivalent motors with the same power output with 1V and 10A or 10V and 1A.

Re: Supercapacitor coilgun
pisoiu, Wed Jun 27 2018, 02:26PM

Yes, I do want to use them in parallel, just forgot to say that :)
I usually avoid misconceptions, just want to make tests and see the results. Putting them in series is also a good idea but for now I just wanted to avoid cell balancing issues. Series connection also does help from a mechanical point of view.

Re: Supercapacitor coilgun
DerAlbi, Wed Jun 27 2018, 07:16PM

Wrong focus. Solve the balancing issue, its the way easier problem compared to all the effects that the low voltage does.

At 2.5V @ 10kA, your setup only allows for 250µOhm total resistance. That is the sum of Cap-ESR, Mosfet-Rds_on and Coil-ESR and every solder joint between the components. Add the condition that the Coil-ESR should at least pose half (better 3/4) the overall resistance your setup is simply impossible to manage. All that because you dont want to balance the capacitors? Bad trade.

Re: Supercapacitor coilgun
klugesmith, Wed Jun 27 2018, 07:57PM

Yes, all that Der Albi said.
Let's focus on the part about rewinding coils (as in motors) to match different supply voltages and currents, with no difference in coil size, weight, or efficiency.

Suppose you are trying to set a record for lowest voltage, in a coilgun with respectable output energy.

You could start with a coil size, shape, and weight that performs well in somebody's 40 V or 400 V system. To adapt it for 2 V, the winding area needs to be partitioned much more coarsely into series-connected turns. Current density (A/mm^2) and volts per turn are invariant.

If the reference design had more than 2 volts per turn, you are out of luck.
Suppose the reference design had 1/2 volt per turn. Then the 2 V version needs 4 turns, each filling 1/4 of the winding area. (Let's ignore the interconnection and switch requirement.)

It might be practical to take a heavy wire with many strands, replace the factory-applied insulation with something much thinner, and wind it around a form. Or you could saw a spiral slot in a thick metal tube, like a flexible shaft coupling:



If you configure six capacitors in series instead of parallel, the optimized coil will have six times more turns and 36x as much resistance and inductance. For the same energy, and same stress on each capacitor. The losses due to switching and interconnecting wires can be much smaller.

Re: Supercapacitor coilgun
pisoiu, Wed Jun 27 2018, 08:54PM

Thanks for the tips. I will give more thought to series connection, I have to make some calculations.
At an earlier stage I made some calculations in which the coil resistance was around 1/2 of total resistance and indeed the losses looked quite high. The coil I considered was made with a 6mm copper rod.

Re: Supercapacitor coilgun
hen918, Thu Jun 28 2018, 01:50PM

The issue with very low voltage isn't the different coil specifications (yes the coil will still obviously need fewer, larger turns), but the parasitics, most importantly the parasitic inductance and resistance. The time constant of your system is approximately L/R where R is the total resistance and L is the total inductance. When you have to wind your coils with so little inductance to get the greatest current and an appropriately short rise time, the inductance of the capacitor connecting wires, and the MOSFETs start becoming an issue. (I would expect the ESL of the capacitor is similar to the inductance of your coil, but the datasheets don't mention this unfortunately). A multistage coilgun is going to need a very long barrel to accommodate the long pulse width.

Re: Supercapacitor coilgun
pisoiu, Fri Jun 29 2018, 07:09AM

Well, the barrel will be long. All 6 capacitors stacked in series have around 1 meter.