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

Re: SISG IGBT Timing Resistors

Original poster: Vardan <vardan01@xxxxxxxxxxxxxxxxxxxxxxx>

Hi Mark,

At 07:54 AM 7/13/2006, you wrote:

Last nite I decreased the IGBT Timing Resistors to 680 ohm in my SISG
array(4 boards - 12 SISG circuits - 10.8KV threshold) on my 4.5" dia
coil.  This gives an IGBT "on" time of 107 uS.  For my coil 1st notch is
40 uS and full decay around 180 uS.

I should mention that if the IGBT turns on at 3 volts and the charge voltage is 24 volts. Then the turn off time is:

3 = 24 x e^(-t/RC)

Since C is 100nF, the time function can be reduced to:

t = 208e-9 x R

Where t is the dwell tie and R is the R4 gate drain timing resistor. In the notes I always mention 4.7k ohms but in practice that value is usually less. I have been keeping the IGBTs on until the system just drains completely with no quenching.

Most recently, I had been operating with a 1.1 Kohm resistor(173 uS).


The improvement was significant.


MOT Primary Current dropped from ~15
amps to ~12.2 amps
and voltage increased from 98-99 VAC to 105-106 VAC.  Spark length
increased a few inches.  Have not measured exact length, but roundabout
4 foot.

So to review the Resistors tested so far:

Ohm   "on" time
4.7K    738 uS
2.2K  342 uS
1.1K  173 uS
680   107 uS

I have seen a steady improvement as I progressed down the list.  Note
that the 680 ohm is the first case that turned off the IGBT before decay
was complete.

Wow cool!!! I have been working on my SISG coil too but mostly cutting wood and making parts rather than anything "technical".



I have never done anything yet with quenching or trying to "sync" the gap.

For those of you experimenting with these boards, I suggest you
temporarily solder the IGBT Timing resistor(R5) to the bottom of the
board to make it easy to change until you are satisfied with the choice
for your application and then permanently mount it to the top of the
board.  Makes it easier to change.

One could also use little square 10 turn trim pots if you have them. In the future, I will put them in place of the fixed resistor. BTW - Mark's boards or super nice!!!

IGBT cooling does not appear to be much of an issue at this time, but we
are only pushing 1300 VA into the system.

Normally they should stay cool and happy with maybe a little air flow.

Note that when this coil was a DC Resonant charging type, it took
upwards of 1800-2000 VA to achieve the same performance(spark length).
Thus, it appears that SISG has improved efficiency by nearly 30%.  Of
course, these numbers are VA not watts so we are not comparing real
power only apparent.

The big point is that it "works" :-))) MOT DC charging coils have sort of thrown me too. I never played with them before and they are sort of complex!!

Also note that I am not using a Variac

Do you want a variac?

so it is not a matter of simply
increasing voltage(in which case current would rise as well).  Current
in my system is limited by an inductive ballast, which remained
unchanged during the test.  Thus, the relatively large change is
current/voltage documented above was purely due to the change in IGBT
"on" time.  I find this very interesting.  I had always wondered how
much effect varying the pin size and disc diameter would have with RSG,
but such an experiment was too labor intensive for me.

Glad to know all is going well!!

Gerry wrote:

Hi Terry and all,

A little thinking out loud..... (and hopefully, a little collective brainstorming). It seems like the SISG is a great replacement for static sparkgaps cause of the reduction in losses (and noise). Currently, the SISG circuit is unipolarized requires a DC supply for charging. Im wondering how this will compare to the SRSG (an AC device) since firing after peak can allow full charging of almost twice the capacitance that an AC power source can charge with a static gap.

I am just rectifying the AC with a high voltage bridge rectifier so not much changed there. A "true DC" source would be cool, but rectifying the AC off a MOT is "easy".

Taking the SISG one step further, Im thinking that it may be possible to use the SISG concept in an AC application (no HV rectifying required) to trigger a circuit that determines a delay so firing will occur after peak.


One possible implementation would be to use a common circuit that detects peak voltage of the charging sine wave and initiates a delayed trigger for a series string of IGBTs. This circuit would need to detect both positive and negative peaks so a trigger control signal can be generated for both directions. The control signal would then need to control a series of IGBT's, each at a different ON/OFF potential within the string. An optical control signal comes to mind to allow for the needed isolation. A three terminal (two power terminals and an optical control terminal) module could be built, each containing two IGBTs (one for positive switching and one for negative switching), resister equalizer across the IGBT pair, and a converter to allow the optical control input to determine the gate voltage.

But is it "easier" just to rectify the high voltage?? My rectifier here is like $10 of 1N4007 diodes, a pipe cap and a lot of epoxy:


The IGBT would need to NOT have the "reverse" diode so the switch can be turned off in both directions (hopefully this part exist).

Oh!! I am not sure IGBTs can standoff significant reverse voltage... They are basically standard transistors so maybe...

If a suitable part exists similar to a FET where current can flow in both directions, then maybe the two IGBTs can be replaced with a single part. It would be perfect if there was such a FET like device that directly had an optically controlled gate (I'm just not up on high power analog devices).

There are optical FETs and IGBTs used for insane power applications the high voltage power transmission. But we can't afford them and ebay does not have them.

If not, it would be desirable to have the gate control circuit be low power and use the optical energy to generate the gate voltage. If the converter needed some electrical power, one could extract it from the voltage difference between the two power terminals when the switch is OFF.

I think it is easier just to rectify the high voltage and stay with "one" element instead of two back to back. Two element also have different common emitter voltages which is a real pain too.

I have not done anything with quenching or sync stuff, so all that is wide open...

Any comments or suggestions welcomed.

Gerry R.