Linear Mode of IGBT FSOA/hotspots? - feasable?

On 24/04/2015 19:59, Robert Loos wrote: [...]

Are you driving the gate with an op-amp? If so, then you could try swapping the op-amp for an OTA like the first stage of a LT1228 (take the output from pin 1 not pin 6, and maybe add some extra capacitance on the gate), or try some op-amp that is intended for driving capacitive loads.

The open-loop output impedance of most op-amps combined with the gate capacitance of a big MOSFET (or even worse, the Cgd * Miller effect), will give you a nearly 90 degree phase lag at many frequencies, that was not allowed for by whoever designed the compensation of most op-amps. Together with another nearly 90 degree lag in the compensation of the op-amp, and a few degrees from other parasitic poles, and you have an oscillator.

Chris

Yes there are twp poles, but generally it's easy to add a zero in the loop to stabilize it.

That one is a bit extreme but for slow stuff it would work. Although I never use gate resistors you may have to place 100ohms or so if driving the FET with an opapmp. As Chris wrote many of those cannot take a capacitive load directly and a FET is always a capacitive load. You have to close the loop for higher frequencies before the bead/resistor, for example using 1000pF from opamp output to negative input.

BJT do not present much of a capacitive load. But you can't blame a FET for that to oscillate. That would be as if I blamed my mountain bike for the tree that was "in the way" last year :-)

How did you turn it on?

Can you post the LTSpice circuit file? It's ASCII so you can just copy that into a post here.

With BJT as well as FETs it's not great to burn that much power in a single device. Gets hot, unless you do that only for a short time. Best to spread into 2-3 devices on a large heat sink.

I'd have to rely on other people's smart phones because I don't have one and right now don't want one. Other than the occasional email I would have little use for it.

Thanks, maybe I do that. But first I have to reach semi-retirement :-)

I don't really need a web site, currently it's mainly used for people to find me back. Old clients after many years and so on. But later I want to start writing circuit design guidelines for the next generations because most universities do not teach analog properly anymore. That would mean proper schematic example linking, maybe PDFs, et cetera.

Can't find it. John, if you are reading this, can you post the link for your blown FET collection?

Am 24.04.2015 um 17:08 schrieb Joerg: ...

I thought so. I just picked the one with the highest impedance we have in our stock :-)

Glad that you are well up again.

Well, I just plugged the wall wart... see the spice file for details.

Here it comes:

Version 4 SHEET 1 2184 788 WIRE 1712 -320 912 -320 WIRE 1712 -208 1712 -320 WIRE 480 -160 480 -224 WIRE 912 -96 912 -320 WIRE 288 -32 240 -32 WIRE 336 -32 288 -32 WIRE 480 -16 480 -80 WIRE 480 -16 400 -16 WIRE 512 -16 480 -16 WIRE 544 -16 512 -16 WIRE 640 -16 624 -16 WIRE 704 -16 640 -16 WIRE 800 -16 704 -16 WIRE 864 -16 800 -16 WIRE -128 0 -432 0 WIRE 64 0 -48 0 WIRE 144 0 64 0 WIRE 288 0 144 0 WIRE 336 0 288 0 WIRE 144 32 144 0 WIRE 640 48 640 -16 WIRE 704 64 704 -16 WIRE 1712 64 1712 -128 WIRE 240 128 240 -32 WIRE 336 128 240 128 WIRE 480 128 480 -16 WIRE 480 128 400 128 WIRE 640 160 640 128 WIRE 704 160 704 128 WIRE 144 192 144 96 WIRE 240 288 240 128 WIRE 64 320 64 0 WIRE -720 336 -720 288 WIRE -432 336 -432 0 WIRE 240 416 240 368 WIRE 816 416 240 416 WIRE 912 416 912 0 WIRE 912 416 816 416 WIRE -720 448 -720 416 WIRE 912 448 912 416 WIRE -432 736 -432 416 WIRE -176 736 -176 576 WIRE -176 736 -432 736 WIRE 64 736 64 400 WIRE 64 736 -176 736 WIRE 912 736 912 528 WIRE 912 736 64 736 WIRE 1712 736 1712 144 WIRE 1712 736 912 736 WIRE 912 768 912 736 FLAG 912 768 0 FLAG -720 448 0 FLAG -720 288 P FLAG 704 160 0 FLAG 512 -16 QU1 FLAG 144 192 0 FLAG 368 -48 P FLAG 640 160 0 FLAG 288 -32 InM FLAG 288 0 InP FLAG 800 -16 Gate FLAG 368 16 0 FLAG 480 -224 0 FLAG 816 416 Sense SYMBOL nmos 864 -96 R0 SYMATTR InstName M1 SYMATTR Value Si4466DY SYMBOL res 896 432 R0 SYMATTR InstName R1 SYMATTR Value 3R4 SYMATTR SpiceLine tol=5 pwr=3 SYMBOL voltage 1712 48 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value PULSE(12 13 0 10u 10u 10m 20m) SYMBOL voltage -720 320 R0 SYMATTR InstName V2 SYMATTR Value 12 SYMBOL voltage -176 480 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V4 SYMATTR Value SINE(2.5 2.5 10) SYMBOL voltage -432 320 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V5 SYMATTR Value PULSE(0.01 2.5 0 1n 1n {Tperiod/2} {Tperiod}) SYMBOL res 640 -32 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R9 SYMATTR Value 56 SYMBOL cap 688 64 R0 SYMATTR InstName C1 SYMATTR Value 1f SYMBOL res -32 -16 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R10 SYMATTR Value 8k2 SYMBOL res 48 304 R0 SYMATTR InstName R2 SYMATTR Value 2k2 SYMBOL res 224 272 R0 SYMATTR InstName R3 SYMATTR Value 1k SYMBOL cap 400 112 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C2

SYMBOL cap 128 32 R0 SYMATTR InstName C6 SYMATTR Value 4n7 SYMBOL Opamps\\LT2078 368 -80 R0 SYMATTR InstName U1 SYMBOL res 624 32 R0 SYMATTR InstName R5 SYMATTR Value 1t SYMBOL res 464 -176 R0 SYMATTR InstName R11 SYMATTR Value 1t SYMBOL ind 1696 -224 R0 SYMATTR InstName L1

TEXT -752 648 Left 2 !.tran .3 TEXT 1016 488 Left 2 ;R-Sense TEXT 936 536 Left 2 ;3W TEXT -744 -104 Left 2 !.param Tperiod=.1

I hope there are no special parts in it. There are some teraohm resistors and femtofarad capacitors in it, I used them to play around. For example, if you reduce the pullup at the opamps output to 2k or less, oscillation stops.

Of course. But my design will only operate in the range of 10..20W so this is no problem.

Seems we have the same problem. I bought an android phone to keep me a little bit up to date. But of the three calls I get per year, two are from my wife when she get stuck somewhere with her car... But it is great to find a pizza nearby :-)

May come faster than you think. I know what I'm talking about.

I would highly appreciate if you started writing a little bit earlier :-) A web site is the cheapest form of advertisement.

Cheers

Robert

But the opamp won't like a direct capacitive load.

That happened the evening before a flight to Europe. I was on a narrow trail, it got late, so I shifted up and stepped on it. 15 miles an hour,

No, but the opamp is the problem.

You've used an expensive micropower opamp in there. Those are no good to drive any serious loads directly and this is a serious load. For example, if you replace it with the OP07 (I tried it because it happened to be in the LTSPice library) nothing oscillates anymore. But even then you have to simulate and measure the phase margin so your circuit won't be too marginal.

The graph on page 6 bottom right shows the problem. The output already starts to collapse at a mere 2mA. It doesn't have much oomph, wrong part for the job.

It's even more extreme here. A few years ago was the first time in our life that we ordered take-out pizza. Normally we always bake our own, outdoors over a wood fire.

I wish. Then I could bike more.

Time is the problem right now. I did reduce my business hours a little already but the free hours are often filled with volunteer work such as nursing home visits. We feel that it's the right thing to do.

I really don't need any advertizing but yes, in the past the simple web site I have has helped obtain interesting new assignments. It helped me get out of medical device design to a large extent which began to become a bit boring and that field was also largely destroyed by Obamacare.

Am 24.04.2015 um 15:40 schrieb Winfield Hill:

Winfield, it's easy for you! But I'm afraid I was too often away from important lessons... Can you give one or two sentences for dummies about how to do that?

Cheers

Robert

Instead of a feedback C from out to -in, use an R+C. That's the basic approach; it works quite well with 1/2 order systems (giving 2/3 order loops that can occasionally be tweaked to reasonably optimal response).

Tim

Right, see Robert, I said it was easy!

Am 27.04.2015 um 22:56 schrieb Joerg:

...

hm. With the OP07 it does not oscillate but it also does not work since the OP07 has no rail to rail input. If I add a negative supply, it oscillates again, only at a higher frequency. Ferrite bead at the gate won't help either.

There is a video on the LT-site:

Here the Netlist:

Version 4 SHEET 1 2184 788 WIRE 1712 -320 912 -320 WIRE 1712 -208 1712 -320 WIRE 480 -160 480 -224 WIRE 912 -96 912 -320 WIRE 288 -32 240 -32 WIRE 336 -32 288 -32 WIRE 480 -16 480 -80 WIRE 480 -16 400 -16 WIRE 512 -16 480 -16 WIRE 544 -16 512 -16 WIRE 640 -16 624 -16 WIRE 704 -16 640 -16 WIRE 800 -16 704 -16 WIRE 864 -16 800 -16 WIRE -128 0 -432 0 WIRE 64 0 -48 0 WIRE 144 0 64 0 WIRE 288 0 144 0 WIRE 336 0 288 0 WIRE 144 32 144 0 WIRE 240 48 240 -32 WIRE 640 48 640 -16 WIRE 704 64 704 -16 WIRE 1712 64 1712 -128 WIRE 640 160 640 128 WIRE 704 160 704 128 WIRE 240 176 240 128 WIRE 288 176 240 176 WIRE 336 176 288 176 WIRE 480 176 480 -16 WIRE 480 176 400 176 WIRE 144 192 144 96 WIRE 240 288 240 176 WIRE 64 320 64 0 WIRE -752 336 -752 288 WIRE -608 336 -608 288 WIRE 240 416 240 368 WIRE 816 416 240 416 WIRE 912 416 912 0 WIRE 912 416 816 416 WIRE -752 448 -752 416 WIRE -608 448 -608 416 WIRE 912 448 912 416 WIRE -432 736 -432 0 WIRE 64 736 64 400 WIRE 64 736 -432 736 WIRE 912 736 912 528 WIRE 912 736 64 736 WIRE 1712 736 1712 144 WIRE 1712 736 912 736 WIRE 912 768 912 736 FLAG 912 768 0 FLAG -752 448 0 FLAG -752 288 P FLAG 704 160 0 FLAG 512 -16 QU1 FLAG 144 192 0 FLAG 368 -48 P FLAG 640 160 0 FLAG 288 -32 InM FLAG 288 0 InP FLAG 800 -16 Gate FLAG 480 -224 0 FLAG 816 416 Sense FLAG 288 176 FB FLAG -608 448 0 FLAG -608 288 M FLAG 368 16 M SYMBOL nmos 864 -96 R0 SYMATTR InstName M1 SYMATTR Value Si4466DY SYMBOL res 896 432 R0 SYMATTR InstName R1 SYMATTR Value 3R4 SYMATTR SpiceLine tol=5 pwr=3 SYMBOL voltage 1712 48 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V3 SYMATTR Value 12 SYMBOL voltage -752 320 R0 SYMATTR InstName V2 SYMATTR Value 12 SYMBOL res 640 -32 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R9 SYMATTR Value 56 SYMBOL cap 688 64 R0 SYMATTR InstName C1 SYMATTR Value 1f SYMBOL res -32 -16 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R10 SYMATTR Value 8k2 SYMBOL res 48 304 R0 SYMATTR InstName R2 SYMATTR Value 2k2 SYMBOL res 224 272 R0 SYMATTR InstName R3 SYMATTR Value 1k SYMBOL cap 400 160 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C2 SYMATTR Value 200p SYMBOL cap 128 32 R0 SYMATTR InstName C6 SYMATTR Value 4n7 SYMBOL res 624 32 R0 SYMATTR InstName R5 SYMATTR Value 1t SYMBOL res 464 -176 R0 SYMATTR InstName R11 SYMATTR Value 1t SYMBOL ind 1696 -224 R0 SYMATTR InstName L1

SYMBOL voltage 240 32 R0 WINDOW 3 41 61 Left 2 WINDOW 123 41 89 Left 2 WINDOW 39 0 0 Left 2 SYMATTR Value 0 SYMATTR InstName V1 SYMATTR Value2 AC 1 SYMBOL voltage -608 432 R180 WINDOW 0 24 96 Left 2 WINDOW 3 -65 10 Left 2 SYMATTR InstName V4 SYMATTR Value 0 SYMBOL Opamps\\LT2078 368 -80 R0 SYMATTR InstName U1 TEXT -752 648 Left 2 !.ac oct 12 100 10meg TEXT 1016 488 Left 2 ;R-Sense TEXT 936 536 Left 2 ;3W TEXT -744 -104 Left 2 !.param Tperiod=.1 TEXT -512 -240 Left 2 ;According to the LTspice video, plot the ratio of V(fb)/V(inm). The phase margin is\nthe phase at 0dB.

I've added a negative supply that I set to 0 for the LT2078. Something very interesting happens if you raise it. Up to 9.99V the curve stays nearly the same. But at 10V it looks completely different! Can this be real? A problem in LTspice? What I find also curious is that the oscillation is nearly independent of the feedback C. You cannot stop is even with arbitrary big C. Not even the frequency changes. Also the inductance I added in the load path to simulate line inductance does not significantly change it. What is the timing factor in this circuit? A different opamp changes it. But how? Such a simple circuit but this babe drives me crazy.

Cheers

Robert

I'm ashamed. Thank you both!

Cheers

Robert

On 2015-04-29 4:30 AM, Robert Loos wrote:

.AC is a small signal analysis method and does not work well for power stuff, it has to be measured.

The OP07 was just one example and if I go above 120ohms it no longer oscillates even with negative supply. For this kind of application you need an opamp with a more powerful output or have a complementary pair follower. I only used the OP07 because it was in the LTSpice directory. Bottomline this is never going to work well with the LT2078 unless you make it very slow. Why do you want a micropower opamp in there?

Not likely. Opamps and other models are at least to some extent behavioral models and there can be parameters under which they fall off the cliff. It shouldn't with this opamp though because I believe it can go to +/-18V.

Strange. If I use, for example, 120ohms it is stable until about 0.02uF. At 0.01uF it oscillates. Kind of what I'd expect. Here it is with 0.1uF:

Version 4 SHEET 1 2184 788 WIRE 1712 -320 912 -320 WIRE 1712 -208 1712 -320 WIRE 480 -160 480 -224 WIRE 912 -96 912 -320 WIRE 288 -32 240 -32 WIRE 336 -32 288 -32 WIRE 480 -16 480 -80 WIRE 480 -16 400 -16 WIRE 512 -16 480 -16 WIRE 544 -16 512 -16 WIRE 640 -16 624 -16 WIRE 704 -16 640 -16 WIRE 800 -16 704 -16 WIRE 864 -16 800 -16 WIRE -128 0 -432 0 WIRE 64 0 -48 0 WIRE 144 0 64 0 WIRE 288 0 144 0 WIRE 336 0 288 0 WIRE 144 32 144 0 WIRE 368 48 368 16 WIRE 640 48 640 -16 WIRE 704 64 704 -16 WIRE 1712 64 1712 -128 WIRE 368 144 368 128 WIRE 640 160 640 128 WIRE 704 160 704 128 WIRE 144 192 144 96 WIRE 240 224 240 -32 WIRE 336 224 240 224 WIRE 480 224 480 -16 WIRE 480 224 400 224 WIRE 240 288 240 224 WIRE 64 320 64 0 WIRE -720 336 -720 288 WIRE -432 336 -432 0 WIRE 240 416 240 368 WIRE 816 416 240 416 WIRE 912 416 912 0 WIRE 912 416 816 416 WIRE -720 448 -720 416 WIRE 912 448 912 416 WIRE -432 736 -432 416 WIRE -176 736 -176 576 WIRE -176 736 -432 736 WIRE 64 736 64 400 WIRE 64 736 -176 736 WIRE 912 736 912 528 WIRE 912 736 64 736 WIRE 1712 736 1712 144 WIRE 1712 736 912 736 WIRE 912 768 912 736 FLAG 912 768 0 FLAG -720 448 0 FLAG -720 288 P FLAG 704 160 0 FLAG 512 -16 QU1 FLAG 144 192 0 FLAG 368 -48 P FLAG 640 160 0 FLAG 288 -32 InM FLAG 288 0 InP FLAG 800 -16 Gate FLAG 480 -224 0 FLAG 816 416 Sense FLAG 368 144 0 SYMBOL nmos 864 -96 R0 SYMATTR InstName M1 SYMATTR Value Si4466DY SYMBOL res 896 432 R0 SYMATTR InstName R1 SYMATTR Value 3R4 SYMATTR SpiceLine tol=5 pwr=3 SYMBOL voltage 1712 48 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value PULSE(12 13 0 10u 10u 10m 20m) SYMBOL voltage -720 320 R0 SYMATTR InstName V2 SYMATTR Value 12 SYMBOL voltage -176 480 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V4 SYMATTR Value SINE(2.5 2.5 10) SYMBOL voltage -432 320 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V5 SYMATTR Value PULSE(0.01 2.5 0 1n 1n {Tperiod/2} {Tperiod}) SYMBOL res 640 -32 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R9 SYMATTR Value 120 SYMBOL cap 688 64 R0 SYMATTR InstName C1 SYMATTR Value 1f SYMBOL res -32 -16 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R10 SYMATTR Value 8k2 SYMBOL res 48 304 R0 SYMATTR InstName R2 SYMATTR Value 2k2 SYMBOL res 224 272 R0 SYMATTR InstName R3 SYMATTR Value 1k SYMBOL cap 400 208 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C2

SYMBOL cap 128 32 R0 SYMATTR InstName C6 SYMATTR Value 4n7 SYMBOL res 624 32 R0 SYMATTR InstName R5 SYMATTR Value 1t SYMBOL res 464 -176 R0 SYMATTR InstName R11 SYMATTR Value 1t SYMBOL ind 1696 -224 R0 SYMATTR InstName L1

SYMBOL Opamps\\OP07 368 -80 R0 SYMATTR InstName U1 SYMBOL voltage 368 144 R180 WINDOW 0 24 96 Left 2 WINDOW 3 24 16 Left 2 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V3 SYMATTR Value 6 TEXT -752 648 Left 2 !.tran .3 TEXT 1016 488 Left 2 ;R-Sense TEXT 936 536 Left 2 ;3W TEXT -744 -104 Left 2 !.param Tperiod=.1

The inductance is 10uH, very small. The opamp can change it because they differ in drive capability. That makes a huge difference when driving capacitive or other hard loads. Small signal analysis doesn't help that much with it. For example, the slew rate matters, big time. This is why at university I never really bought the small signal theory they taught us. Full confession: It resulted in me skipping class and just writing in the exam what they wanted to hear.

Wood-fire baked pizza tastes beyond compare. We've never looked back, haven't baked one in the oven since years because it simply isn't as good.

Am 29.04.2015 um 22:40 schrieb Joerg:

I found a way to simulate it: the good phase margin is just because the input is grounded. If you insert a 2V voltage source the plot shows a terrible phase and it is obvious that it must oscillate.

Historic reasons. In my first board I used a TLC272 (without thinking much about it, just because it was on stock) but the high offset made it impossible to set the current to 0. Then I was searching for a pin compatible type with rail to rail input and low offset that is easy to get and ended up with the LT2078. Since I'll have to make a new board anyway, I will now search for a type with rail to rail input, low offset and high drive capability that is easy to get :-)

... The gate resistor is the clue. If you plot the .ac analysis (with the 2V source) you can see how the phase improves. 220R would be even better.

That I believe unseen.

Unfortunately I'm now away from this thread. I'm going on a cruise tomorrow :-) Don't know if it will still be on our news server when I come back.

Joerg, I'm deeply grateful for your help. I feel I have learned a lot in the past days, not only from you but also from the simulations.

Thanks to everybody who contributed!

Cheers

Robert

As I said, it makes no sense to drive this with a micropower amplifier. It can't do it. Put an OP07 in there and all is well even at 2V.

Trying to drive a big FET with an LT2078 is like trying to pull a 5-ton trailer up a steep hill with a Fiat 500.

Coincidentally there is a very similar discussion going on in de.sci.electronics, subject "Schneller Strompuls / LED-Blitz". Since you understand German, take a look. Timm is now using the THS4631 for controlling a big old FET in a fast manner.

Yeah but at some point you start losing drive. Depends on how fast you have to swing that gate.

But the occasional *PHUT* can be fun as well. When our Skyworks rep was at my office he said "If I ride my mountain bike and there wasn't a spectacular crash then it probably was a boring ride".

Well, if the cruise lasts more than three months, maybe not :-)

Have fun. And easy on that big buffet.

As a SW engineer once put it (and the bible does), we are here to serve :-)