[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [TCML] OLTC theory

Hi Carlos,

On Wed, Nov 28, 2012 at 12:53 AM,  <lightningfor@xxxxxxxxxxxxxxxxxxxxxx> wrote:
> couple of questions for the solid state guru's...
> OLTC's have a VERY high capacity in the tank circuit in order to get
> enough pulsed energy into the secondary per burst.
> As a result, the primary has a very SMALL inductance...one turn most the
> time...
> I have read that coils like this have a VERY sharp tuning point.
> Would it be fair to say that the sharpness of tune would be a result of
> the high capacity to inductance ratio in the tank.

Its all relative, but yes, they are certainly more sensitive to small
*inductance* changes because they have so little primary inductance to
begin with.  I would expect the Q of the coils to be similar to any
other type of TC, so really they are just as sensitive to tuning in
terms of frequency.  For this reason, it becomes more practical to
fine tune these machines by changing the tank capacitance by
adding/removing little capacitors from a big bank of them.  Also, the
low impedance nature of an OLTC could make all the connections more
sensitive to contact resistance, so its best to have very solid and
permanent connections within the primary circuit.  The mechanical
layout of the circuit is also far more critical as any inductance that
is not part of the primary coil is essentially parasitic and serves to
lower the coupling coefficient between the coils.

> The other question I have relates to IGBT's.
> I have 2 very large bricks rated 6000v 200A cont, or 400A pulsed.
> I would like to use them in a big OLTC set up, but the ones I have are the
> slightly older models and are only rated to 5khz as a switching speed.

Can you supply a datasheet?  The most critical aspect will be the turn
ON time, and whether or not you can over-volt the gate input to 30V or
so.  Some IGBTs will show an internal zener diode from gate to
emitter, which means you cant drive the gate extra hard.  Driving the
gates to 30V allows the IGBT to pass more current.  You want the turn
on time to be a small fraction of the resonant frequency period.  Id
probably just be really sure the turn on time is fast enough that the
IGBT is fully ON before the resonant current exceeds the IGBTs nominal
current rating.  This would ensure that you are within the SSOA (safe
switching operating area) of the device.

> What I am wondering is how they might fare in an OLTC set up, where they
> are being switched at low pulse rates, but the oscillations across the igbt
> will be much higher than its rating...?

The difference in inductance between the IGBT dies within the brick
will mean that they dont share the current evenly.  The really big
modules might have something like 200-300nH of circuit inductance.
You can view this inductance as a series impedance at the resonant
frequency (Z = 2*pi*f*L).  Once you know this impedance, you can work
out the voltage drop at some current level and compare it to the IGBTs
VCE ON voltage.  IF the inductance voltage drop is similar in
magnitude to the VCE drop (from the datasheet), then you would
probably experience balancing issues within the module, which would
mean some dies are gonna get much more current than others.  The
higher current forces the VCE of the IGBT to go up, which will restore
balancing to some degree, but the IGBTs with more VCE drop are being
stressed harder, and will be the first ones to fail if you push too

Lets do a hypothetical example.  Lets say the IGBT has some dice with
only 100nH of stray inductance, and others with 200nH.  Lets assume
the coil rings at 100khz (probably far higher than is practical).  The
impedance of 100nH is .063 ohms at 100khz, and 200nH is .126 ohms.
Say you want to get 1000A pulses through this IGBT,  Ignoring VCE
drop, the current would split 2/3 through the 100nH and 1/3 into the
200nH to get the same voltage drop across those inductances.  So what
is the drop?  666A*.063ohm =  42V, and as expected, 333A*.126ohm =
42V.  This 42V is significantly larger than the VCE voltage of some
10V i would guess for a module of that voltage rating.  The VCE drop
will help restore the current sharing to some degree, but its obvious
that the inductance causes more imbalance.

My example is really just to explain the issue as to *why* high
frequency currents through a IGBT module could compromise it.  The
numbers are wild guesses, the reality is probably that the mismatch in
inductance probably isnt so bad, i really cant say.  But, it does say
one thing, if you want to run higher peak currents there is likely
something to be gained by lowering the operating frequency so that the
IGBT inductance is less important.

> Would I just de-rate the current rating or are they likely to fail
> miserably?

The OLTC guys found that most IGBTs can tolerate rather big pulses of
current, sometimes up to 10X the nominal current rating.  The limit is
essentially thermal in nature.  IGBTs can only pass so much current
before the junction "de-saturates" and the voltage drop goes up
*dramatically* causing a large instantaneous power dissipation in the
junction which can burn it out pretty much instantly.

If i had a number of these modules and was willing to sacrifice one, i
would build a OLTC tank circuit, and simply gate the IGBT on with 30V
and let the system ring.  Id use an oscilloscope to observe the
voltage drop across the IGBT at ever increasing pulse currents until
you observe some drop thats maybe 4X the datasheet specified value for
VCE on.   One really important detail is that you do not turn the IGBT
off while current is flowing through the IGBT, as opening up that
circuit current will produce a devastating voltage transient across
the IGBT switch which WILL destroy it more likely than not (at least
at high currents).  You could add a big MOV across the IGBT to protect
it possibly.  For testing purposes, where you dont have a high current
power supply charging the tank cap, you can simply make the IGBT gate
pulse *really long* so that the tank energy will be fully dissipated
by the circuit resistance and there will be negligible current flow by
the time the gate pulse is terminated.

Luckily for you, the energy stored in an OLTC is small compared to
what DRSSTCs use, so IGBT failures shouldnt be as explosive.  I once
tested some 300A modules up to 5200A, as i wanted to know what the
maximum current that the IGBT could pass before desaturation (turns
out its 3600A at 24VGE, and 5200A at 30VGE).  This required about 800J
of energy storage in big capacitors to source this 5200A,  The IGBT
failure is loud and sends bits of module plastic across the room!  Its
just part of the fun of playing with pulsed power silicon :-).


> Thanks,
> Carlos
> _______________________________________________
> Tesla mailing list
> Tesla@xxxxxxxxxx
> http://www.pupman.com/mailman/listinfo/tesla
Tesla mailing list