3x 18mF @ 350V -> What to do? / Electromagnetic Projectile Accelerators / Forums

Forums

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

« Previous topic | Next topic »

3x 18mF @ 350V -> What to do?

previous 1 2 3 4 next last

Move Thread

LAN_403

Pinky's Brain

Tue Jun 08 2010, 11:02AM

Registered Member #2901 Joined: Thu Jun 03 2010, 01:25PM
Location:
Posts: 837

Why IGBTs?

A thyristor will generally be pulse rated for a half sine pulse use with pulses lasting 10 msec .... the IGBT is only pulse (short circuit) rated for a couple us. For coil guns you need pulses in the range of Thyristor pulse rating, but you will need to use the normal rating for the IGBT. For the same bucks the Thyristor will do >10x the current as the IGBT (a small IGBT is still a good gate driver for the Thyristor though, they like a nice gate current pulse).

Of course since you have a big IGBT for free this might be a bit academic ...

DerAlbi

Tue Jun 08 2010, 11:11AM

Registered Member #2906 Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 727

Yeah thats the question what is more efficent. Railguns seems easyer.. but a coilgun is interesting too. With many stages the energy can be discharged too.
Problems with railgun would be to find a good solution current-switching and to reduce friction. The mecahnical lifetime of such a device seems verry limited
A coilgun instead is something that lasts a bit longer, if the IGBTs do not explode

So far a coilgun seems more interesting. (or is it really much much better?)

--
IGBTs: are you sure that IGBTs will not do the job if they are connected in parallel? And are you serious that the current maximal must not be calculated with the pulse-current? As far as i can see, IGBTs are short-circuit safe for some Âµs. The max pulsewitdh it mostly limited due to thermic problems. Lets say i overstress the IGBT with a current of (lets imagine:) 500A for 10ms.
The puls is not a rect.. so the normal current is exceeded only for ~8ms. And pulsecurrent is exceeded for a few ms only.
In this milliseconds the IGBTs voltage drop is about 10V with max gatevoltage (worstcase) this gives ~5kW of heat during conduction. For 10ms this is an energy of 50Joule. From 25Â°C to max 175Â°C the heat capacitance of the material should compensate this for a short time keepeing also thermic resistance in mind..
Or are there other physical effects that destroy the IGBT when it is overstressed?
Can you follow my thoughts? I would be pleased if someone can correct me in this matter.

To do some math for my example:

703J (for silicium)      per          1000g
--------------------        ==        -------------------- 
50J (deliverd in 10ms)                1g (die weight) * k

->gives me 70Kelvin. Acceptable. Issnt it?

So far i am looking at the APT100GN60B2 and i thought about 5 in parallel (within a steelcase

) for 1.2-1.5kA. Would Cost some dollars for 10 Coils.
Too Unsafe?

Pinky's Brain

Tue Jun 08 2010, 12:00PM

Registered Member #2901 Joined: Thu Jun 03 2010, 01:25PM
Location:
Posts: 837

Oops, I didn't mean pulse current for the IGBT ... yeah, you can pulse the IGBT a lot longer than an us. The pulse current is only moderately higher than the continuous current though, so that's not very impressive.

I was talking about the short circuit current, which is ~20x rated continuous at high gate voltages. Pulse current for thyristors is also ~20x rated continuous current.

That APT100GN60B2 is a lot more expensive than Thyristors with the same Itsm as it's pulse current.

DerAlbi

Tue Jun 08 2010, 12:11PM

Registered Member #2906 Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 727

And do yout think the IGBT would be enough? (some in parallel) Buying 50 pieces issnt that out of reach for me, as long as the fun will be propertional to the money. I found it with 5-6$ per IGBT.. and i am paying in euro.. so it issnt that expensive (maybe a months money spent from mama)

Due to my calculations in the last post, 3 in parallel would be ok to switch 1.5kA. so taking four for safety and a 9stage gun is a possible project.

Of course SCR would be cheaper.. but given to my big caps the possibilites are quite limited.
-> Are SCRs a possible replacement for the anitparallel diodes to shortcircuit CoilCurrent after open the IGBT? SCRs seem cheaper than HighCurrent diodes. am i wrong?

Pinky's Brain

Tue Jun 08 2010, 12:53PM

Registered Member #2901 Joined: Thu Jun 03 2010, 01:25PM
Location:
Posts: 837

One way or another IGBT's can switch any voltage at any amperage with sub us rise times, they are amazing devices which are relatively easy to drive and use (certainly easier than Thyristors, which you can't really turn off ... so you have to rely on resonance).

If money is not an issue IGBT will do just fine.

A P1000J will take 400A for a half sine at 50 Hz and costs 30 cents ... unfortunately no spec on reverse recovery (will certainly start conducting faster than a Thyristor though).

DerAlbi

Tue Jun 08 2010, 02:31PM

Registered Member #2906 Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 727

And what do you think about the usage of the 300A-types mentioned above? How much in parallel you would have if you want to switch 1.2-1.5kA. is 3-4 ok?
And by the way: is my math above correct? Just 70K temp. rise?

klugesmith

Tue Jun 08 2010, 02:33PM

Registered Member #2099 Joined: Wed Apr 29 2009, 12:22AM
Location: Los Altos, California
Posts: 1714

DerAlbi wrote ...
IGBTs: are you sure that IGBTs will not do the job if they are connected in parallel? And are you serious that the current maximal must not be calculated with the pulse-current? As far as i can see, IGBTs are short-circuit safe for some Âµs. The max pulsewitdh it mostly limited due to thermic problems. Lets say i overstress the IGBT with a current of (lets imagine:) 500A for 10ms....

So far i am looking at the APT100GN60B2 and i thought about 5 in parallel (within a steelcase

) for 1.2-1.5kA.

1. A problem with running transistors in parallel is uneven current sharing, due to normal mismatches between devices. Details depend on the transistor type and operating region, but it could help to put a balancing resistor (to drop a few V at your max I) in series with each collector terminal.
2. Your thermal analysis of the pulse rating is an excellent start!. The physics of heat diffusion leads to cascaded models where shorter pulses have smaller effective thermal mass. Your datasheet
]go-apt100gn60b2.pdf[/file] has a fancy model in Fig. 19a and 19b -- I think 19b has a decimal point error.
3. How much gate voltage will your IGBTs need to be "saturated" at 500 A? (see Fig. 2, typical VCE curves). Note that in the current-source operating region (flat part of curves), the currents will be substantially lower at room temperature, as well as different from device to device.

Sulaiman

Tue Jun 08 2010, 05:54PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

Since you have a very high voltage IGBT try to make use of it well,
e.g. use the three capacitors in series (with balancing resistors) for a ht supply of about 1 kV, use a string of diodes and TVSs of > 2kV total from collector to gate, to clamp the inductive flyback at a high voltage allowing rapid ramp down of the coil current for no 'suck-back'.
If the string is 2 kV then current ramp down time is about equal to ramp up time,
optimising available energy.

Initially I would use one mains transformer to give a low voltage ac, e.g. 24 Vac.
Feed this 24 Vac into THREE 240:24 transformers in reverse so that you get three 240 Vac 'floating' supplies which can be rectified for one capacitor each.
Put a mains voltage rated filament lamp in series with the mains input and you have a good 1 kV supply.

Since you only have one IGBT I recommend putting a HRC semiconductor protection fuse in series with the collector, e.g.

The fuses are expensive, but not as expensive as your IGBT !
Check that the I2T rating of the fuse is less than the IGBT.

DerAlbi

Tue Jun 08 2010, 08:55PM

Registered Member #2906 Joined: Sun Jun 06 2010, 02:20AM
Location: Dresden, Germany
Posts: 727

Thanks for the idea for how to charge the caps. I already thought about this. Its a easy but heavy solution. So i decided the weight is too high.. maybe if i would have some light transformers for switching-powersupplys.. but where to get a good and efficient transformer for e.g. 12V->230? PC?

@Klugesmith:
1) The proble of uneven currents does not matter for these IGBTs. They have a postive temp. coefficient, so the automatically conduct bad if they heat up. Can be read at the first sheet

2) Interesting. after your hint i was cuious to understand the drawing. It seems that for 1ms single pulse, the thermal impedance shinks to 0.08K/W. So 2kW would already make 160K that brings us to the maximum limit of 175Â°C. :-/ is that correct? This completly is against what i wrote above. I do not understand currently how Watts without respect to any time-constants can heat up something :-/ Arent Jouls the real matter? Meybe i need some help here. Or are the 1ms already the timebase :-/

3) This was a simple extrapolation

i see that every Volt at the gate increasse saturation ~50A

30Volts are alowed, and so i looked waht curve could hit the 10V..

But for now, i have to understand the heating-problem at first.

klugesmith

Wed Jun 09 2010, 07:36AM

Registered Member #2099 Joined: Wed Apr 29 2009, 12:22AM
Location: Los Altos, California
Posts: 1714

DerAlbi wrote ...
It seems that for 1ms single pulse, the thermal impedance shinks to 0.08K/W. So 2kW would already make 160K that brings us to the maximum limit of 175Â°C. :-/ is that correct? This completly is against what i wrote above. I do not understand currently how Watts without respect to any time-constants can heat up something :-/ Arent Jouls the real matter? Meybe i need some help here. Or are the 1ms already the timebase :-/

You got the concept, just had a wrong coefficient.

Let's study Fig 19b, "transient thermal impedance model", a rough two-lump approximation of the real thermal behavior. It agrees with fig 19a if we change the upper "resistor" from 0.949 to to .0949.

The total thermal resistance (theta_jc) is .09 + .12 = .21 degress/watt, a value also found elsewhere. 45% of total resistance comes from the upper "R" and "C", with time constant of 0.7 ms -- it reaches its steady state thermal rise after about 5 time constants (3.4 ms). The other 55% of thermal resistance has a time constant of 28 ms and takes about 140 ms to settle.

If you set up Fig19b in SPICE, and hit it with various width pulses of 100 "amps" (watts), you can see how far the "voltage" (temperature) rises toward DC value of 21 "volts" (degrees); I found it to match the roller-coaster curve in fig 19a.
For example, 100W for 1 ms increases junction temp by about 8 degrees (0.08 deg/W). 1000W for 10 us increases Tj by 1.45 degrees. In such a short time, cooling by conduction is negligible, and the "short pulse" heat capacity (upper "C" in the figure) is 0.00708 joules per degree. What volume of silicon would that correspond to? What volume of Si or Cu would the "long pulse" heat capacity correspond to?

Is this model consistent with the short-circuit pulse rating SCSOA = 6 us?
That's with 600V across the device and 15V on the gate. From Fig. 2, the current will be about 400A. Package resistance and inductance are not very significant here. 240 kW for 6 us is 1.44 joules, which from the heat-capacity model gives a rise of 200 degrees. (comparable to component bulk temperatures during reflow soldering.) To me it looks like reasonable agreement. Such a large temp. change would substantially change the short-circuit current during the pulse.

I hope that helps.

-Rich

previous 1 2 3 4 next last

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