Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Does Solid State Marx Generators actually work?

Status
Not open for further replies.
Hi kripa

1) Sure, please see my thread #1 the gate drive is identical for all igbt's (but now using 10/100ohm for off/on)......

..

Nope sorry. That circuit is a total mystery to me. I was just offering some 3rd person perspective.
 

Nope sorry. That circuit is a total mystery to me. I was just offering some 3rd person perspective.

Maybe it's because it's an isolated driver it looks more complicated than usual, please see this, it's just a simple driver however it has two power supplies.

Screenshot from 2019-07-01 17-33-16.png
 

Okay, that's easier to understand. So the Vcc1 is fine. And the dotted line implies galvanic/ optical isolation? Which is it ?

But then how is Vcc2H and L created? Looks very tricky to me. Since there is separate notations for the Vcc and the Vgnd for both. What's being done here ?

I'm not nit-picking, but sometimes problems are found in supposedly innocent places.
 

Okay, that's easier to understand. So the Vcc1 is fine. And the dotted line implies galvanic/ optical isolation? Which is it ?

Both, optical is also galvanic. In this driver it is a small coreless signal transformer though.


But then how is Vcc2H and L created? Looks very tricky to me.

H and L just infers that it is two seperate galvanically isolated power supplies, in my case I have 20 simple power suppliers each which consist of transformer+LM7815+caps, the LM7815+caps can be seen on the schematic in thread #1.

Since there is separate notations for the Vcc and the Vgnd for both. What's being done here ?

The left side is powered by the control board's power supply, the right side is powered by the power supplies mentioned above

I'm not nit-picking, but sometimes problems are found in supposedly innocent places.

I know that, don't worry - chill.
 

Okay... so its all floating supplies then.
Which leads me to question the GND point for each, and how that GND point voltage varies as you switch through your sequence.
I'm being lazy here and not actually working things out myself, just casting doubts on possible failure points.

cheers!
 

Which leads me to question the GND point for each, and how that GND point voltage varies as you switch through your sequence

I hope that is properly taken care of by nature, I don't know the exact tech jargon for it. But you can well imagine two power supplies in series, if you adjust one at the rate of 10 giga adjustments per second, the other one doesn't care and you will always be able to measure their combined voltage as whatever + a fixed value (offset).
 

I hope that is properly taken care of by nature, I don't know the exact tech jargon for it. But you can well imagine two power supplies in series, if you adjust one at the rate of 10 giga adjustments per second, the other one doesn't care and you will always be able to measure their combined voltage as whatever + a fixed value (offset).

I get the concept. Intuitively it should work fine.
What about the Vcc2 points ? Where do those come from ?
 

What about the Vcc2 points ? Where do those come from ?

VCC1 and GND1 is the control board's power supply, VCC2 and GND2 are the galvanically isolated power supply (in my case transformer+LM7815+caps).

So VCC1/GND1 is galvanically separated from VCC2,H/GND2,H and VCC2,L/GND2,L. And VCC2,H/GND2,H is galvanically separated from VCC2,L/GND2,L
 

That's an amazing number of galvanics. 8-O

So further, again without really knowing directly about your gate driver modules, I notice that these too have an upper value for isolation voltage. The part number whose specs I tried to read had values between 2500 - 3000. This lies right where your 4th stage voltage is.

Do crosscheck
 

Hi everyone

I assembled yet another stage which I will call "stage 5" and here is the interesting result when I add it to the four other stages:

The stages are like follows (I numbered them all now so I can distinguish them) 5+1+2+3+4 and you may remember from my thread #35 that 1+2+3+4 yielded 1800V where 2+3 showed the symptoms described in my thread #34 i.e. their caps were discharged instantly upon turning Qdischarge's ON.

Well now with five stages 5+1+2+3+4 it yields 2700V and stages 3+4 now suffers the symptoms... so to recap

With stage 1+2+3 = 2700V
With stage 1+2+3+4 = 1800V
With stage 2+3+4 = 2700V
With stage 5+1+2+3+4 = 2700V

The mystery continues... any exorcists among the honorable members perhaps?

I'm thinking the caps in question MUST somehow find a path to the 0V when the Qdischarge are turned on, but WHERE is it?

- - - Updated - - -

That's an amazing number of galvanics. 8-O

So further, again without really knowing directly about your gate driver modules, I notice that these too have an upper value for isolation voltage. The part number whose specs I tried to read had values between 2500 - 3000. This lies right where your 4th stage voltage is.

Do crosscheck

The entire idea of the Marx Generator/Modulator topology is that no component ever experiences any voltage difference in excess of the supply voltage which is 1200V, regardless of how many stages you have.
 
Last edited:

I believe I have a working theory now

With five stages cascaded I compare the CE voltage of Qcharge in Stage1 (working), with Qdischarge in Stage3 (not working). I see that Qcharge in Stage3 never actually experiences the full capacitor voltage when Qdischarge turns ON (well maybe for an infinity small moment).

Looking at the circuit diagram the only way the cap can discharge instantly when Qdischarge turns ON, is through Qcharge.

I don't believe it's a timing issue, no matter how long a delay I put between charge and discharge, the result is the same. All Qdischarge's turns ON/OFF at the same time because they are connected to the same port, likewise all the Qcharge's.

I don't believe it's a defective component, because if I switch the stages around then Stage1 is not working and Stage3 is the one that is working, the total opposite.

So my question is now, under which circumstances can a IGBT (no body diode) possibly conduct from Collector to Emitter when it is OFF?


Stage1
Probe at Qcharge Collector, GND clip at Emitter
at 40uS Qcharge turns OFF, Qdischarge turns ON
S1Qc.png


Stage3
Probe at Qcharge Collector, GND clip at Emitter
at 40uS Qcharge turns OFF, Qdischarge turns ON
S3Qc.png
 

...The entire idea of the Marx Generator/Modulator topology is that no component ever experiences any voltage difference in excess of the supply voltage which is 1200V, regardless of how many stages you have...

Does not the output vs input of your 'galvanic' driver have an upper limit? Check output voltage levels Vs input levels during discharge cycle in 4th stage & beyond.

Seems to me this DOES get larger n larger.

What exact part number are you using?

- - - Updated - - -

If in fact problem is occuring because of failure modes occuring due to exceeding isolation voltage ratings, then the 'arbitrary' nature of results of your experiments also depends on precise timing differences in your discharge pulses to the different drivers.

- - - Updated - - -

Also not to forget that the charge-side drivers will also experience similar I/O stress during discharge cycle, and might be malfunctioning as well !
 

Does not the output vs input of your 'galvanic' driver have an upper limit?

Sure 1200V, and recognized under UL 1577 with an insulation test voltage of V(iso) = 3000 V for 1s

Check output voltage levels Vs input levels during discharge cycle in 4th stage & beyond.

Let me check.

Seems to me this DOES get larger n larger.

We'll see when I check, I personally doubt it because I saw clear Gate signals which is not indicative of a driver malfunction.

What exact part number are you using?

1EDC20I12AH
 

Seems to me this DOES get larger n larger.

I turned the input voltage down to 227V, so no component will ever ever experience more than 5*227=1135V if even possible.

Unfortunately the problem is the very same, Stage 3+4 does not work.

Please see below measurement, could the negative spike forward bias the diode on Stage3 and/or Stage4? And where does this negative spike come from? Stray inductance?

Stage3 100V/DIV
Probe at C1 positive side, GND clip at C1 0V side
at 90uS Qcharge turns OFF, Qdischarge turns ON

SDS00001.png
 

i have not found this on your current schematic, but several posts/days ago, the design
charged on the positive side and on the negative side and it didn't get to full voltage either.

in my simulations, (of that circuit) i found that +V of one cap was shorted to -v of another cap during discharge
making output voltage less than expected

is there a sneak circuit that ties two equally charged capacitors in anti parallel?
 

is there a sneak circuit that ties two equally charged capacitors in anti parallel?

That could definitely be a possibility but it won't be on purpose, but I can't see where it could happen in the circuit.

in my simulations, (of that circuit) i found that +V of one cap was shorted to -v of another cap during discharge
making output voltage less than expected

Could you please post your simulation so that I can compare with mine?

- - - Updated - - -

is there a sneak circuit that ties two equally charged capacitors in anti parallel?

I really don't see that happening, this is how the circuit looks now with stage 3+4 not working, the green line is how it should work in theory, and the blue line would be your scenario. Before your scenario would occur C3 would have discharged through Qcharge3, and C4 would have discharged through Qcharge4.

cap-short.png
 

Did you ever get around to checking whether Discharge CE drop was less than your diode drop?
 

Did you ever get around to checking whether Discharge CE drop was less than your diode drop?

I can't measure CE because there is a huge noise spike from the rapid cap discharge, so I really can't capture the low voltage in such resolution so that it is useful. Maybe you know how to get around that?

And I'm now abandoning the igbt CE short circuit theory, I replaced the igbt with a MOSFET and the result is unchanged.
 

No idea how to measure it, but thought that a look through respective specs would be enough. My theory was that if Vce is > Vdiode, then the diode path might be getting taken and hence cause problems.

Presumably with MOSFET the drop is << Vdiode? So need to look elsewhere.

- - - Updated - - -

Would be useful to know which mosfet. Especially the Rds-on?
 

Would be useful to know which mosfet. Especially the Rds-on?

A Cree Wolfspeed **broken link removed**

- - - Updated - - -

I made a little experiment, as you remember the setup is five stages in following order 5+1+2+3+4 where stage 3+4 malfunctions.

marx-25.png

IF I charge all five caps, and then

1. open Qdischarge3 ONLY, then C3 discharges instantly, C4 looks like it's slowly self-discharging and C5/1/2 looks like they discharge through a resistive path.

or

2. open Qdischarge4 ONLY, then C3 and C4 discharges instantly, and C5/1/2 looks like they discharge through a resistive path.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top