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

Re: Terry's DRSSTC - Bench Test

Original poster: Terry Fritz <teslalist@xxxxxxxxxxxxxxxxxxxxxxx>

Hi Steve,

At 10:33 AM 1/22/2005, you wrote:
Hey Terry,

Just a comment about deadtimes and IGBT rise and fall times.  For the
BIG coil, i sorta fine tuned my gate drivers to give a 50nS dead band,
not much at all, but enough to see that there is no cross conduction.
Secondly, i went with NO gate resistance at all and was able to drive
the bricks *quite* fast, at about 100nS rise and fall times.

WOW!! Ok, Maybe I should not worry so much ;-) We are also driving a full resonant LC thing which can only help.

mention that the very fast transition times might might "tear the
IGBTs up", i assume this has to do with discharging the miller
capacitance?  I wasnt aware there was any negative effect to doing it
"faster", unless there is enough inductance for a dI/dt event to cause
some sort of transient within the IGBT?

Since we are driving a nice resonant thing, I bet the effects of super fast switching are far less of an issue. I would worry more with multi die IGBTs and all when things just might not all sing together. We also tend to use devices that are pretty oversized really. High speed switching tens to stress everything from wire bonds, die fingers, and all kinds of die level things. However, those things my not be such a big issue "now-a-days" and IGBTs may be better than say FETs too. Bottom line is "if it works, it works"!! Things certainly seem perfectly happy with super fast switching so far in my case. If you can run "for real" with it, maybe that is simply the way it "should" be!! Thanks for letting me know!!

I built a really cheesy SMPS to power all of my gate drivers... might
be something to think about later.  All it really is, is a half-bridge
running 40khz driving a transformer (old flyback core!).  To make all
4 outputs to the gate drivers "identical", i simply took a length of
cat-5 cable and wound about 8 turns on there, and seriesed each pair,
so that each winding was really 16 turns.  This method has been tested
up to 750VDC buss voltage, and no failures yet ;-).

I was going to do that too. But with 300VDC at tens of amps available right there, I thought I would just us a little of it and be done with it ;-)) Also, none gate drive transformer drive can do all kinds of fun things since each IGBT is totally independently controllable. And "I" hate GDTs anyway >:o) Really, we just need a TLP250 with some stupid power supply circuit and its all done...

Looks like its almost time to start driving coils ;-).

Yea, finally!!!

Dave Sharpe said...

Hi Terry

Outstanding work thus far! I ran into the thermal issue (at 30kHz CW) with
the TLP250's, but even without "juiced" drivers, I'm getting 150nS turn on and 350ns turn off on a high speed SSR for work.

The thermal thing also depends on how much drive the device needs. Usually no problem at all. But something to just be aware off especially when playing with them on the bench and driving them to say 2MHz just for fun ;-))

The floating supplies for the TLP250's, quick description of how they work? That is an elequent way to develop iso drive power for the optocouplers...

Referring to the schematic at:


Normally when off, the "+" to "-" voltage is 1/2 the buss voltage or 170VDC. For the lower IGBTs the "+" could also just go to the 300VDC opposite side buss...
.... Plenty of power around :-)) As the IGBTs switch on and off, those available voltages will fly all over the place but that is what big caps are for to store the voltage for the moments when the "+ -" voltage is low.

R1, D1, and C1 are just a 33 volt voltage source but with zero current drive. At steady state 300VDC (you always have to design for worst case) R1 will heat to just less than three watts even with just 9mA going through it. If the "+" drops out for a moment, C1 will easily sustain the voltage. One could use a blocking diode there too to help but the duty cycle in our case makes that a don't care and it really does not matter much anyway...

Of course, our little 33V power supply has no real current drive since R1 would burn up if we ask for any more current. So we just stick that voltage onto the gate of FET Q1. The source of Q1 runs about 3 volts less than the gate so big ol' 1000uF C2 is charged to about 30volts. Q1 is a big FET with heat sink and all. R2 limits any spike currents to two amps. Without it, a spike may reach like 10,000 amps =:-O You need the resistance just to limit fault current if something does not go right. Worst case, it is also the "fuse" ;-) So C2 stores a nice amount of power to run things and sustain the 30V when the "+" voltage drops low. The only other concern is the heat on the power current pass element Q1. The circuit would supply amps until Q1 explodes...

If it has say 270 volts across it and the TLP250 draws 8mA that is 2.16 watts (that goes way down if we hook the "+" to the 170V output instead... "Gate charging" power is only drawn when the gates are turned on since turning off just shorts them. According to the data sheet, it takes 160nC to charge the gate to 15V so will assume 600nC to get it to "our" 27V. If we charge the IGBT say 50*300 = 15000 times a second we get 90uC of total charge/second at 27V. That is like charging a (90e-6 / 27V = 3uF) 3uF cap to 27V every second. V=1/C x I x t so the current is 27 X 3E-6 = 81uA But probably close to a few mA with losses in the TLP250 so all that "thinking" is for not... One could think about it all a long time, but basically, Q1 can handle about 5W easily which at 170V is 29mA which is fine for our needs ;-)) Really, if R1 were say 10 watts at 10K ohms (add blocking diode now) then the FET booster would probably not be needed... But this is the way it all turned out. Probably many solutions... The TLP250 really does all the "work". Giving it a "little" 30V 20mA power is fairly trivial with thousands of available watts surrounding everything...