# 30V 200A pulse circuit

1. Oct 16, 2012

### trini

Hey guys,

I'm working on finding a circuit which will allow me to pulse about 200A for 50-100 us. My external load is about 0.05 ohm and 3uH. The repetition rate of the pulse is 360 Hz. I require the pulses to be in alternating polarity.

Thus far, I have been thinking about using a push/pull setup using an nfet and pfet, but its hard to find pfets that have a 200A rating. My question is, given that my circuit has a net duty cycle of about 2%, can I use parts rated at 4A continuous?

If I can use lower amperages, I could possibly just buy a 5A rated H-bridge and use PWM to set my pulses up, is this correct?

Any alternative suggestions to achieving my goal would be appreciated.

PS:
the circuit will be used to generate magnetizing pulses for Alnico 5 magnets.

2. Oct 17, 2012

Hello Trini:

I think a MOSFET or IGBT H bridge will be the way to go. But you will need much more than 5A rated parts - the MOSFETs parameter to look for is Idm - Max pulsed drain current. Finding them > 200A is difficult.....I only find 2 on Newark and they are rated 30V. You need a Mosfet rated in the 100A Range, and 50 to 100V.

Also -this pulse looks like it would be higher than 200A peak ( 0.05 Ohm + 3uH and 50uS But NOT including the 2 X Rds ON - )

- Are you thinking of having higher frequency pulses in each power pulse - or just regulating the pulse width overall to control the peak current?

Also - is this a lab / research item - or for production equipment. I would steer you to a MOSFET or IGBT module esp for production for robustness and reliability - but they will cost more.

3. Oct 17, 2012

### trini

For the size of magnets I'll be using i need a 160A pulse to saturate, so I stated 200A for safety. my coil is 0.05 ohm and 3 uH, as you mention the voltage will be determined by the value of Rds on in the driver circuit. My plan was to use PWM to control the current output.

As I mentioned, each pulse is to last 50us, at a maximum of 360Hz rep rate (though it may be as low as 20Hz) leading to an max overall 2% duty cycle.
This is why I had hoped I could just find the average current and use this for rating (200A*2% = 4A) since my understanding was that to prevent device breakdown I have to avoid thermal overload of the junctions, which would be dependent on average values rather than instantaneous ones.

For the sake of example, lets say i use this driver, which is capable of 20kHz max, 28V, 12A continuous:
http://www.robotpower.com/products/MegaMoto_info.html

this device has an on resistance of 0.016 ohm meaning my total circuit resistance is 0.066 ohm. For a 200A pulse I need to run it at 13V, resulting in an average power of 200*13*0.02(duty @ 360Hz rep)= 52W, while the device listed can handle 800W.

My other option of course is to use low ESR caps and dump them using fets or igbts, but this requires more cost and work than buying pre-made H bridges.

EDIT:

Oh may I mention as well, that the Idm ratings are usually given for 350us pulses, which is 7 times greater than my pulse time.

Last edited: Oct 17, 2012
4. Oct 18, 2012

### Enthalpy

BEWARE it's an inductive load so you MUST protect your transistors against overvoltages when switching off. Switching not too quickly helps the protection circuit save the life of your transistors, even if conduction losses increase a bit then.

You don't need the full 200A rating but nearly that, certainly not 4A, because components limit the maximum current as well. Beware the peak current can be non-repetitive, which is not your situation. Anyway, the voltage drop will tell you soon that a small transistor can't conduct 200A.

Just take the proper components. 200A and 13V isn't that much, but it will have a cost and a size.

Do not take P-Mos. Take N-Mos and buy gate drivers with floatig drive voltage, they were invented for that purpose. Or build a voltage converter as suggested by Maxim for instance, to produce the gate drive voltage.

At 13V, Igbt and Darlingtons drop too much voltage, but you might consider bipolar transistors.

5. Oct 18, 2012

### trini

Hey guys,

the people at robotpower told me i can't use that driver, they have one I can use but its 500 bucks X_X

I snooped around a bit and found these fets:

P-channel: VISHAY SILICONIX - SUM110P04-05-E3 - MOSFET,P CH,40V,110A,D2PAK:
http://export.farnell.com/international-rectifier/irl1404zpbf/mosfet-n-40v-200a-to-220/dp/8659729
Data sheet: http://www.vishay.com/docs/73493/73493.pdf

N-Channel:INTERNATIONAL RECTIFIER - IRL1404ZPBF - MOSFET, N, 40V, 200A, TO-220
http://export.farnell.com/vishay-si...ch-40v-110a-d2pak/dp/1839011?Ntt=SUM110P04-05
Data sheet: http://www.irf.com/product-info/datasheets/data/irl1404zpbf.pdf

now the N channel is fine, and the P channel can hold pulses up to 240A. The only thing i am unsure about is that they list the PFETs voltage rating as -40V. what is the difference between 40V and -40V in terms of using it in an H bridge?

Also, I have a 7.5kW AC drive at home which I use to run a motor, can I use this to drive my FETs?
http://www.automationdirect.com/static/manuals/gs3m/gs3_quickreference.pdf

6. Oct 19, 2012

### Enthalpy

-40V because it's a P channel.

The N-channel 200A in TO220 might work IF you can cool it properly AND cable it properly... This package is meant for printed circuits, which conduct 200A uneasily. Look at a computer motherboard: the Cpu core receives ~100A at 0.9V and the tracks are wider than long, plus many contacts bring and scoop the current. Try to put some figures on the track resistance, including the current spread resistance.

I dislike the P-channel: the D2pack is even worse than the TO220. Where will you put the heat dissipator on a D2pack? At +150°C it has some 6mohm Rdson so you have about 7W to dissipate at 3% duty cycle.

Better take four N-channels, and use floating gate drivers. These exist commercially, and I believe to remember that Maxim has some chips for that. Or add a +30V rail with a standard smps circuit.

7. Oct 19, 2012

I do not know why you need the P channel - you can build an H bridge with all N Channel - and get a 2 channel driver with bootstrap ckt for the high side.

There are some H bridge drivers but all of the ones I am finding are self oscillating = not what you need. Search for 2 Channel MOSFET driver.

8. Oct 20, 2012

### Enthalpy

I'm more and more convinced that an output signal transformer would be better:

- You wouldn't need to supply power at 200A 15V, which is uncomfortable. For sure the 15V capacitors are to supply the peak current, but this means many capacitors in parallel, and they won't last for long. Cabling at 200A is uncomfortable as well.

- As opposed, you could supply the mean power directly from the PFC circuit, say at 500V, thus saving the down regulator. Capacitors for 500V are hugely better than for 15V; you could even use power ceramic parts, which last forever, or plastic film parts, instead of the bad electrolytics.

- With two primaries at the output transformer, you need only two switching transistors, both near the ground, thus avoiding the floating drive. You save (more than!) half the switching losses.

- The switch transistor would, for instance at 500V, need to withstand >2*500V, say 1500V, and only 4.8A: this is trivial for any technology, especially IGBT. The drive circuit is obvious, and individual conduction losses of 2V*5A are 20 times smaller than 5mohm*200A^2. Stray inductance eases a lot as well.

You worry about a 500$driver: then you should have a look at the capacitors that supply 200A with <<10µs rise time... About cabling: if 200A shall rise in 1µs (the narrowest output pulse being 10µs) and lose 5V (over 12V) in the wires and parts, you allow only 25nH inductance in the complete circuit... That's seriously difficult! Already one TO220 has 15nH stray inductance on each pin. An electrolytic capacitor is much worse. 25nH is 40mm of wire. So at 12V you'd end up with 4* 10 MOS in parallel, many dozens of fragile electrolytic capacitors to achieve I, L and C, all cabling with busbar... Due to commutation noise in the source you'd have floating drivers anyway, even on low-side N MOS... Stray inductance is the killer argument. 9. Oct 20, 2012 ### trini hi enthalpy, that is indeed a very interesting suggestion. I am by no means an electrical expert so please tell me if I understand this correctly: -charge HV caps to desired voltage with the help of a voltage regulator (primary side circuit) -hook the caps up to the collector terminal of two IGBTs -hook the IGBT emitter terminals up to the primary side coils of a dual primary:single secondary transformer so that the polarity of the current in the secondary can be changed by choosing which IGBT is in the on state -control the IGBT using a pwm source (such as an arduino) If this is correct, what factors should I look for when choosing my caps? Does it matter whether or not they are low ESR? 10. Oct 21, 2012 ### Enthalpy Wait! Re-read your post #3 telling that the load is a coil... Instead of winding a transformer to adapt the generator to the coil, you may better wind directly a coil with more turns of thinner wire, so it accepts a reasonable current (and voltage) without a transformer! Could you tell the use? If you want to make magnets for instance, the proper setup is a single pulse controlled by a thyristor. Yes, two IGBT driving one primary each, for different polarity, and one secondary making the output. If no transformer is used, the coil should have two windings, which may be possible or not. I had wrongly read 13V... You want 30V output pulses, so the current at 500V supply is bigger than my figures. The trend stands. For 30V output pulse, transistors that guarantee only 40V aren't enough, because of voltage spikes during commutation, which protective circuitry can't completely cut. Especially at small voltage and big current. PWM: usually means that you're interested in the mean voltage at the output, averaged by a low-pass filter, and adjusted by the duty cycle. In case you're interested by the pulse itself at the output, something like "pulse width control" would be clearer. 11. Oct 23, 2012 ### trini Hey H, let me give you some background. My equipment requires that I remagnetize Alnico magnets with alternating polarity on a repeated basis, which isn't too hard given its low saturation intensity of about 4*Hci = 240 kA/m. This is the H field my coil must produce to saturate the alnico. From FEM simulations which factor in geometry, and considering the two reels of AWG18 and AWG32 i have lying around, I have have found that I can use: a 24 turn coil using AWG #18 (req'd: 15V 80A); L=13 uH; R = 0.1357 ohm a 514 turn coil using AWG #32 (req'd: 190V 6A); L=153 uH; R = 31.94 Ohms I do however require a relatively fast pulse time and the inductance of the 512 coil kills me. I've been doing my homework and I went with your advice by looking around at suitable IGBT's to make H bridges with. Among the cheap fast switching IGBT's I found these: FGA180N33ATD, 330V 180A PDP Trench IGBT: http://www.fairchildsemi.com/pf/FG/FGA180N33ATD.html these have a continuous rating of 180A so they should handle my 2% duty fine along with any spikes. I have 6 of these coils to power in all, and all are to be fired simultaneously. With this in mind I got the spice models for the IGBT's from Fairchild and simulated it in LTSpiceIV (see attached diagram). The source voltage is 20V and has a resistance of 0.012 ohm, and comes from these: http://www.ebay.com/itm/Maxwell-350...049?pt=LH_DefaultDomain_0&hash=item564c54db59 The peak draw on the 20V source is around 650A which occurs as a spike according to my sim, though I suspect this may be because I did not include diodes to reduce computing time. Even so these spike values are within the tolerances of my devices (these ultracaps have an SC current of 1200A ;D). I plan to control the IGBT gate at 24V by using an arduino to drive a FET which in turn drives the IGBT gate. I hope to be able to use a single arduino to drive all the bridges. If you can think of something to reduce the amount of bridges I use that would be awesome (24 IGBTs ~$100 bucks :S)

#### Attached Files:

File size:
4 KB
Views:
148
File size:
9.3 KB
Views:
111
• ###### 20V80A.jpg
File size:
36.9 KB
Views:
124
12. Oct 23, 2012

### Enthalpy

That's clear.

You decide how to build the coil, so design one with more turns, for higher voltage and smaler current. Its response time is exactly the same (any simulation telling the opposite would be inaccurate) since energy and power don't depend on the number of turns, provided you fill well with copper the available volume.

Even better: coils with thinner wire are of better quality, because they reduce the eddy currents within the wires. The resistance computed by FEM is probably very wrong at 10µs, especially if computed as a DC resistance.

One possible sensible supply voltage - among others - is the one resulting directly from rectified mains. Only a reduced current will make your circuit feasible. 30V and 200A are just too difficult.

Then, I still don't understand why you have PWM and short pulses. Magnets are made by one single pulse (and erased by LC alternative ringing when needed). This is far simpler: charge a (...big) capacitor to an adjustable voltage, disconnect from mains, discharge the capacitor through a thyristor into the coil.

I did it for Sm-Co magnets in 42mm*42mm*100mm, with 1m3 of 350V capacitors. The current was 20kA (single pulse through a 2kA thyristor), the duration 10ms. Is there a reason to go to 10µs or 100µs pulses?

13. Oct 23, 2012

### Enthalpy

The L and R values for both coil hypothesis aren't consistent.

On the same shape, L must vary as N2. (514/24)2 doesn't fit 153µH/13µH.

The L/R ratio, or time constant of the coil, should depend essentially on the shape and size of the coils, not primarily on the wire size if it fills the available space.

The L/R ratio given by FEM is probably wrong. Unless the coil is tiny, its time constant is much longer than that.

Winding 500 turns is nothing special. It's done quickly by hand and even faster on a rotating mandrel. 2000 turns would be boring by hand. Enamelled wire is available from components distributors. D=0.2mm is easy to win by hand, D=0.1mm gets difficult.

-----

In any normal magnetizing machine, the current would be limited by the inductance, not the resistance. In other words, coupled with the capacitor, you'd get an RLC which is under-damped, and add a flywheel diode to avoid reverse voltage in the capacitor and later reverse current in the coil.

Would you tell the size of the magnets you make, and the field you need? I can estimate the coils by hand then. Do you need to erase the magnets as well?

Once the coils use comfortable V and I, the drive circuit is much easier.

14. Oct 23, 2012

### Enthalpy

One more worry: AlNiCo is an electrical conductor, so it will need time until the induction reaches its core. And because AlNiCo has a big permeability, the coils won't behave as when they're empty.

About the coil of 514 turns: if it has D=36mm the vacuum inductance should be around 15mH.

15. Oct 23, 2012

### trini

I am using this for electropermanent magnets, and I have to saturate in alternating directions frequently. In my application, the faster the switching time the better as I want to use minimal switching energy.

The magnets are 1"x 1" x 1/4" blocks. My sim was done on COMSOL using their built in coil option, which tells me the strength of the field, but does not give me the inductance of the coil or its resistance since I solve it in stationary. Solving in stationary is fine because i just want to know what the peak current is that will achieve saturation. To calculate the resistance and inductance I use a matlab file I wrote up. I know it calculates the right resistance(wire length*ohm per length), but the inductance I am not so sure about.

Unfortunately, it's difficult to find consistent equations online for a multi-layer rectangular solenoid. I originally used a formula I found on wiki for a single turn rectangular loop and multiplied by N, though i guess N^2 is what i should have used:

I took that formula and multiplied by N.u0.uR for total inductance.

If you could help me find a proper formula to calculate rectangular solenoid inductance that would be a great help.

Some I have found are (air core for all, so have to multiply the answer by 3.67 for the uR of alnico):

http://www.technick.net/public/code/cp_dpage.php?aiocp_dp=util_inductance_rectangle (This is for a multi-layer flat rectangular loop though, not a rectangular solenoid since it doesn't ask for coil length input)

http://electronbunker.ca/InductanceCalcRc.html (rectangular solenoid, but it says the calculation is for a single layer)

So basically I need to find a proper inductance formula to properly set up my circuit.

Last edited: Oct 23, 2012
16. Oct 23, 2012

### trini

I just realized i said 24 turns for the AWG#18 but it should be 48 (6 turns x 8 layers).

Do you think I should use the rectangular solenoid calc and, for example with the AWG18, multiply the single layer result by 8^2 = 64?

17. Oct 24, 2012

### Enthalpy

The AlNiCo magnets are surprising...

- I thought AlNiCo had disappeared, because ferrite magnets have only advantages.
- An AlNiCo magnet wide and short is likely to demagnetise
- I imagine the repetition frequency is related with a mechanical movement, like going from one magnet to an other, and then 360 Hz is a lot.

Is this something different, like Transcranial Magnetic Stimulation for instance?

-----

Some alternatives:

A permanent magnet of Nd-Fe-B is enough to make an AlNiCo magnet perfectly. Or two Nd-Fe-B taking the AlNiCo in sandwich. No electronics nor electric power.

An electromagnet with an iron core can make an AlNiCo magnet. No electronics, very little power, just a strong mechanical unloader.

Maybe you could make two magnets at once and avoid the polarity change in the drive electronics. Or even, make at once all magnets of a rotor, provided there is something like a rotor.

18. Oct 24, 2012

### Enthalpy

Let's go on in the hypothesis of making 1" x 1" x 0.25" magnets using an air coil...

The simple computation of the coil is without the magnet...
After the pulse, it will probably be round (OK, I had 7T and you need maybe 0.7T), so it better is round from the beginning. Take ID=40mm OD=60mm h=20mm. Hold the coil externally, maybe with fibreglass-epoxy; I had plywood. And hold the power wires together, and mind the contact quality.

Inductance is N2 * 0.6µH/m. As accurate as complicated formulas, simpler. That's 94nH/turn2.
Copper foil, 20mm wide, can take 80% of the volume, or 18µohm/turn2 when cold, or about 27 when warm.
You can insulate the turns with thin fibreglass-epoxy, polymerize, and then the coil is sturdy.

To make strong AlNiCo (!) you need 560kA/m at the centre of the coil, or 28kA*turn2.
The (empty) coil stores 37J and dissipates 21kW (in pulse BUT from DC resistance!) when warm. Its time constant is 3.5ms when warm. Forget the 10µs, 100µs and 360Hz.

You might consider accelerating and decelerating the current in less than 3.5ms BUT the AlNiCo itself, being a conductor, will have a similar time constant, so B wouldn't reach its core.

The influence of the permanent magnet is, at 160kA/m, some 0.7A*m2 as compared with >60A*m2 for the empty coil, hence is neglected.

19. Oct 24, 2012

### Enthalpy

Let's suppose you charge the capacitor by rectified full-phase 117V, adjusted with a thyristor. You get 160V DC for full energy.

Energy transfer can be better than exp(-2) so 21,000µF 250V (160V?) is enough - please check and optimize, I didn't. Take high quality capacitors that deliver the peak current and survive it many times! Cable many in parallel with good conductors. Put the capacitors in a casing, as the beasts sometimes explode.

Say we want 5ms half-period, then the inductance is 121µH or 36 turns of 0.25mm foil. The peak current is 780A.

Take a moderate thyristor to switch: it should be around 200A and 200V, check the semi-repetitive pulse. This one is exaggerated:
http://www.vishay.com/docs/93685/93685irk.pdf
Or an IGBT, but it costs more. Add a ~200A flywheel diode antiparallel to the capacitors to avoid reverse charge.

For reverse polarity, either add the aforementioned transformer, or use an H-bridge of thyristors or IGBT.

Maybe your magnets don't require 560kA/m (some accept half that) but we're far from 30V 200A, and the capacitors alone cost ~500USD. More than 160V would be more comfortable for the electronics. Please check the exp(-2), I didn't and am no more sure.

20. Oct 24, 2012

Hello All:

One question - I am not a magnetic guy - but to concentrate the field into the magnet to be magnetized - would it be be better to make a broken ferrite core and then put the AlNiCo magnet in between them like a sandwich - centered in the middle of the coil? Or will the increase in inductance cause the current rise time to be too slow.

On the power electronics side - again is this for "research" like grad material studies, or for production. For some type of research where the understanding of the power electronics is to be reported - I guess the way you are proceeding seems OK, but for a production system - and reliability I would consider using IGBT / MOSFET modules not discretes and possibly a solutions provider to make the core elements ( Select and match the IGBT/MOSFETs, Drivers and the DC Link Capacitors).

The selection of the caps and design of the DC bus will affect the operation of the IGBTs/MOSFETS.

I do not think a thyristor will work here as you cant turn it off. MOSFET or IGBT --

Note - I looked at the IGBT you had selected and it would drop 1.6 V at 180A and 25C - so in an H bridge you will have 600W of power in the IGBT per Puse - granted the duty cycle is low - but the device life time is defined by the Delta T of the junction - so again back to the production question.