Long Pulse IGBT GDT Design Discussions
Edited/Updated: October 23, 2004
First Post
Date : Sat, 09 Oct 2004 11:39:06 -0600. Subject : Long Pulse IGBT GDT design
Original poster: sparktron01@comcast.net
All: Need some design guidance for a GDT for an IGBT power switch. Need
to turn on for ~300usecs, off time of 1.5millisecs (PRF = 400Hz). If drive
voltage is 30Vpk to pk, results in a V*t product of 9000V-usecs (typical GDT
product is in 200-500V-usecs).
Transformer may saturate, but will DC restore during off time so not
particularly a big deal. Any design guidance out there for DC restoration using
standard toroids for long pulse IGBT gate drive?
Thanks!
Best Regards,
Dave S.
Date : Sun, 10 Oct 2004 12:58:22 -0600. Subject : Re: Long Pulse IGBT GDT design
Original poster: Jim M. Perhaps some optocouplers?
Date : Sun, 10 Oct 2004 12:58:49 -0600. Subject : Re: Long Pulse IGBT GDT design
Original poster: Jan W.
On Sat, 9 Oct 2004, Tesla list wrote:
>Original poster: sparktron01@comcast.net
<snip>
>Transformer may saturate, but will DC restore during off time so not particularly a big deal. Any design guidance out
>there for DC restoration using standard toroids for long pulse IGBT gate drive?
Something like
http://www.irf.com/technical-info/appnotes/an-950.pdf ?
cheers, Jan.
Date : Sun, 10 Oct 2004 17:31:25 -0600. Subject : Re: Long Pulse IGBT GDT design
Original poster: sparktron01@comcast.net
Hi Jan. Yes, I downloaded that AN today; will try to "throw together"
Figure 5 with IGBT's and IEEE dynamic voltage equilization circuit, and see what
happens (like "run for cover!!! ;^))))
Speed looks right (~100nS turn on and off), FET is rated to nearly 20A
pulse. Might even be able to use a Radio Shack type IRF510 / IRF511 variant for
driving to keep cost down.
Thanks again. Regards, Dave S.
Date : Mon, 11 Oct 2004 08:10:40 -0600. Subject : Re: Long Pulse IGBT GDT design / HV-OLTC
Original poster: sparktron01@comcast.net
Hi Jim:
Optocouplers are available to 3.75kV isolation (Toshiba TLP250 / TLP351 family).
However to make work on a series switch requires an isolated current loop drive
for each level of the switch for isolated +/- DC power for the opto-coupler
driver. Then you still need a fiber-optic link between your main pulse
generator and each level of the switch. I've seriously looked at the Toshiba OC
approach, and the part count, isolated DC power supply availability (and cost)
and circuit synchronization issues make that approach quite unreasonable, at
least for "us guys"... :^D
The AN950 International Rectifier approach as mentioned by Jan appears to be, at
least on the surface, a much more reasonable and maybe reliable alternative.
I'm feasibility testing at this point, but at least I see a vision where I want
to go... the HV-OLTC based on a series stack IGBT switch. 2.4kV PT powered by
an IGBT AC chopper (pulse by pulse current limiting) into a resonant charging
stage as previously used by other experimenters (notably Steve Conner among
others). 8 series switch sections with Vbalancing circuit and TVS protection
using 1.2kV IGBT's powered at a peak charging voltage of 6.8kV; shooting for a
HONEST 1.0MW ring wave power (4.0MW peak [6.8kV * Ipeak ring wave] ). He.., I'm
using samples at this point, and have 20 more coming, might as well enjoy the
sparks (and IGBT detonations)... :^)
But first steps will involve building a prototype pulse generator controller and
SSVariac, and 1.2kV level switch running at 1.0kV, and take it to detonation to
determine maximum Ipk my devices can tolerate, what drive levels are needed for
optimal performance, and GDT / gate drive design. I have some new IXYS gate
drivers that can sink or source up to 30A pulse (8A DC continuous) that will be
used for switch triggering loop. The point is the hardware and published
technology (circuits and validation of dynamic balancing) has recently become
available; so there is
at least a glimmer of hope of making this concept work. Regards, Dave S.
Date : Mon, 11 Oct 2004 12:00:03 -0600. Subject : Re: Long Pulse IGBT GDT design
Original poster: Jan W,agner
On Sun, 10 Oct 2004, Tesla list wrote:
>Original poster: sparktron01@comcast.net Hi Jan
>
>Yes, I downloaded that AN today; will try to "throw together" Figure 5 with IGBT's and IEEE dynamic voltage equilization circuit, and see what happens (like "run for cover!!! ;^))))
If you use a resistor across IGBT gate and emitter, watch out it's not too small
resistance, otherwise you might really have to run for cover! :) Same situation
if there are to be expected some heavy voltage transients that might couple back
to the gate signal... ;)
>Speed looks right (~100nS turn on and off), FET is rated to nearly 20A
pulse. Might even be able to use a Radio >Shack type IRF510 / IRF511 variant
for driving to keep cost down.
Btw there's also a slighlty simpler circuit at
http://dsms.ajusd.org/~fritz/AN1.pdf that runs very nicely, but has some
"startup" time, and you can't change to duty cycle too fast either. Other
circuits like Dan's ultrafast gatedrive
are at the end of: http://www.hut.fi/~jwagner/tesla/SSTC/general-sstc-notes-gatedrv.htm
But if you need a really rigid drive that can also handle arbitrary duty cycles,
then the only way is to set up auxiliary floating supplies for each IGBT. E.g.,
1MHz 50% duty oscillator and some small ferrite cored 1:1:1:... xfmr (with good
RF/HV insolation...) to generate the supplies, some minor voltage regulation,
and a standard gate drive chip for each IGBT, plus fast optos (cheap
HCPL-3020's?) to steer these.
At higher cost you could use dedicated opto/driver combo chips that have all
sorts of extra features, like fault and igbt saturation detection, and whatnot.
For example chips from Agilent:
http://www.agilent.co.jp/spg/products/pdf/optocoupler/HCPL316J.PDF, cheers, Jan.
Date : Tue, 12 Oct 2004 09:51:12 -0600. Subject : Re: Long Pulse IGBT GDT design
Original poster: "B2"
<bensonbd@boo.net>
e-mail-19263
Hi Dave, I used a Radio Shack 120V to 12V transformer for the same purpose
(choice arrived at by trial and error). I tried a few toroids. They were all
too fast and the L/R was too short for my need. A floating pulse stretcher
circuit was considered, but the design time was not available. In the end, I
decided on a long L/R with a clipper. I had to build a simulator that
duplicated a bomb dropping off a jet and being armed while it was simultaneously
being struck with lightning (fun test). I needed to simulate a 0.1 second pulse
using power mosfets (IRF450). The gate (series) resistor was around 100 Ohms
(adjust for your L/R). A 15 Volt zener diode and 50 ohm resistor in parallel
with the gate clipped the extra voltage. That is all that I can remember, as I
do not have access to the circuit. Hope that is helpful. Barry.
>Need some design guidance for a GDT for an IGBT power switch. Need to turn on
for ~300usecs, off time of 1.5millisecs (PRF = 400Hz). If drive voltage is
30Vpk to pk, results in a V*t product of 9000V-usecs (typical GDT product is in
200-500V-usecs).
>
>Transformer may saturate, but will DC restore during off time so not
particularly a big deal. Any design guidance out there for DC restoration using
standard toroids for long pulse IGBT gate drive? Thanks!
>
>Best Regards, Dave Sharpe, TCBOR/HEAS, Chesterfield, VA
Date : Tue, 12 Oct 2004 09:52:09 -0600. Subject : Re: Long Pulse IGBT GDT design / HV-OLTC
Original poster: sparktron01@comcast.net
Hi Jan
-------------- Original message --------------
> Original poster: "Jan Wagner"
>
>
> On Sun, 10 Oct 2004, Tesla list wrote: Original poster: sparktron01@comcast.net
Hi Jan
> >
> >Yes, I downloaded that AN today; will try to "throw together" Figure 5 with
IGBT's and IEEE dynamic voltage equilization circuit, and see what happens (like
"run for cover!!! ;^))))
>
> If you use a resistor across IGBT gate and emitter, watch out it's not too
small resistance, otherwise you might really have to run for cover! :)
> Same situation if there are to be expected some heavy voltage transients that
might couple back to the gate signal... ;)
> ------------------------------------------------------------
SNIPPO
The dynamic equilization network (DEN) is going to put a + equilization
(compensation) bias on the IGBT gate unless it is hard driven OFF with -Vee. So
a floating drive system won't work... :^CCCC
The latest alternative would be use a pulse transformer + rectifier driving a
TLP250 (up to 1.5A pulse) so with no drive pulse train (1-3 Mhz, 50%DC), Vge is
biased -Vee. When triggering pulse train arrives, OC is rapidly biased ON,
and
with appropriate speed up circuits will have a propogation delay of ~ 300 nS. If
I can hold triggering within +/-250nS, the DEN will take care of equilization
during turn on transistion (most critical). Below is link to equilization
circuit...
<http://palgong.knu.ac.kr/~kimhg/High-voltage%20switch%20using%20series-connected%20IGBTs%20with%20simple%20auxiliary%20circuit.pdf>http://palgong.knu.ac.kr/~kimhg/High-voltage%20switch%20using%20series-connected%20IGBTs%20with%20simple%20auxiliary%20circuit.pdf
Will also have to scale Requilization from network, the 30K, 3K are scaled based
on leakage currents of a 1.2kV 400A brick, my candidate device is a Fairchild
HGTG27N120BN (1.2kV, Ic = 54A, ICM pulse 216A), with separate RHRG75120
Hyperfast Recovery Diodes, both in a TO247AC package. If you "guess" that
leakage current scales approximately linearly to Ic, I should be able to scale
resistors by 5X (larger) and not impact circuit performance (more efficient
though, 5W dissipation versus ~27W per switch level). Rg may have to be
adjusted as well. The caveat of "don't change the DC bus voltage too fast" in
article was also caught. My comment is I need DEN critically when I'm turning
the series stack on to full conduction, once the GE junctions are enhanced and
saturated, following decrements will be significantly less critical (Voltage and
current stress). Only way to find out definitely will be to build a two level
switch, operate at 2kV, and see what smokes (or arcs over...)
Thanks much for the links Jan, I've downloaded and put in my "literature search"
tab of my Excel design folder... :^D
> cheers, Jan. Helsinki University of Technology, > Dept. of Electrical
and Communications Engineering,
http://www.hut.fi/~jwagner/ -
jwagner@cc.hut.fi Best Regard, Dave S.
Date : Tue, 12 Oct 2004 09:53:55 -0600. Subject : RE: Long Pulse IGBT GDT design
Original poster: "Steve Conner"
<steve.conner@optosci.com>
>If drive voltage is 30Vpk to pk, results in a V*t product of 9000V-usecs (typical GDT product is in 200-500V-usecs).
This can be handled by simply using more turns on the transformer.
Unfortunately, then the leakage inductance tends to get excessive, leading to
sluggish rise and fall times and/or excessive ringing.
It might be worthwhile investigating a transmission line transformer. This is
wound with coax cable- the screen is the primary and the core is the secondary.
This gives very low leakage inductance at the expense of high inter-winding
capacitance.
A variant with even lower leakage inductance is the ribbon cable transformer
where all the even numbered cores are connected in parallel to form the primary,
and all the odd numbered ones likewise form the secondary. But this
has even higher interwinding capacitance, and considerably worse isolation than
the coax version (the insulation is thinner and only PVC) I don't imagine for a
minute it could take 2.4kV. Steve C.