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.
Adding 2nd stage to Solid State Marx Generator fails

Hi Friends

After my one stage Solid State Marx Generator started working, I have added a 2nd stage, but it does not work it should be a simple matter of adding stages

Screenshot from 2019-06-20 15-26-48.png

When adding the 2nd stage I expected that it would add ~1000V to the output, but instead it produces the same output as one stage ~2000V. I notice that the voltage across Q3 and Q5 is only 600V, half of the single stage Q3 voltage of 1200v, which is the reason for the low output. But I have no idea why it does that, 1200 is expected on both Q3 and Q5.

DUAL STAGE (not working as advertised)
dual-stage.png

SINGLE STAGE (working)
single-stage.png
 
Last edited by a moderator:

UF1007 breakdown voltage is a sufficient reason to get not more than 2000 V output from your circuit. You may evaluate this point by reducing the DC input voltage.

In addition it's not clear how the measurement loads the circuit. I'm stumbling upon the term "ground clip", there should be a fully differential high impedance probe.
 

Hi FvM

UF1007 breakdown voltage is a sufficient reason to get not more than 2000 V output from your circuit. You may evaluate this point by reducing the DC input voltage.

Yes that could be a reason, so first I changed D3 and D7 (post #21) to BY203 2000V and then I lowered input voltage to +/-300V, but the output is still the same ~1100V regardless of one or two stages. And if I remove one or the other stage, then remaining stage works fine.

SDS00014.png

In addition it's not clear how the measurement loads the circuit. I'm stumbling upon the term "ground clip", there should be a fully differential high impedance probe.

Sorry I forgot to mention that load on output is 100Kohm resistor. I don't have a differential probe yet.
 

you can fake a differential probe by measuring the voltage at each end and subtracting
if your 'scope has two probes, it may do the math for you
both probes must be referenced at the same point
 

you can fake a differential probe by measuring the voltage at each end and subtracting
if your 'scope has two probes, it may do the math for you
both probes must be referenced at the same point

Thank you for your suggestion, I thought I did fake a differential probe by floating the scope and measure as I did (see schematic in thread #21 where I indicated how the probe was connected)?

- - - Updated - - -

I made this sim perhaps it can give some clue to what is wrong, and yes maybe I'm interpreting it incorrectly.

I have draw a schematic that shows how the caps and diodes are configured during final output stage where full 3600V (6 caps x 600V) output is expected, I kept the component refs from the schematic thread #21.

sim33.png

On the left is how it is in the circuit. I'm wondering why it doesn't charge up to 3.6K. And as you can see the voltage C6 C9 and C10 is the same, it's like these caps doesn't exist.

On the right, it looks like if I remove either D1 D2 D6 or D8, then there is a voltage difference on C6 C9 and C10, but it still charges up to only 2.3KV. So I'm wondering if D1 D2 D6 D8 in fact are shorting out C6 C9 and C10?

- - - Updated - - -

I see the problem in the sim thread #25, the diodes are only 800V, so now i quintupled them and it charges to 3600V. As you can see on the right circuit in the sim below there is an even spread 600V across each cap, but that is only because I cut the connection from C5/6 to the diodes (marked in red). So I think now that a additional switch is required during output stage to cut this connection, would you agree?

Screenshot from 2019-06-21 16-42-30.png

- - - Updated - - -

Here is my problem: D8

dual-stage-crap.png

So this design will not work for additional stages. And user Easy Peasy warned me against believing everything on the internet, but I thought I was safe this time because it was a IEEE paper.

Well back to the drawing board :-(
 
  • Like
Reactions: FvM

    FvM

    Points: 2
    Helpful Answer Positive Rating
So this design will not work for additional stages. And user Easy Peasy warned me against believing everything on the internet, but I thought I was safe this time because it was a IEEE paper.
I agree it can't work. They claimed the circuit could be extended, apparently without ever checking it. Lesson learnt!
 


Why are there so many switching elements in your circuit ? It should be a straightforward RC network with a single switching power source.
 

Marx generator needs at least one switch per stage, more if active charging is intended. The suggested circuit looks reasonable.
 
  • Like
Reactions: Swend

    Swend

    Points: 2
    Helpful Answer Positive Rating
Why are there so many switching elements in your circuit ? It should be a straightforward RC network with a single switching power source.

Yes, you would be right if the final output was 800V, I choose 800V because that was the diode with the highest voltage rating in ltspice and that was just to check the topology. In reality I will have 8 stages of 1200V so the output will be ~10KV pulses at 20KHz.
 

OK, I got my PCB's and assembled them and problems start...

I have two stages, when I test each one independently it delivers 1100V on the output, however when I connect both together (see below) the output remains 1100V when it should have been the double. I checked both caps charge up to 1100V and measuring the Qdischarge IGBT's across Source and Drain they both show ~0V when ON, which is expected. So why only 1100V on the output (measured across R1) with 2 stages?

marx-22.png
 

If I understand right, the circuit worked in your simulation, thus there's obviously a problem with the implementation, either actual circuit or control pattern.
 

If I understand right, the circuit worked in your simulation, thus there's obviously a problem with the implementation, either actual circuit or control pattern.

Yes, it's an implementation problem for sure.

Sequence of events: I assembled four stages, tested all four individually and they worked. Joined stage 1+2 they worked (output 2200V), felt confident that everything was a-ok and added stage 3+4 so had 1+2+3+4 connected together but output was 1100V. Messed around stage 1+2 for four hours couldn't get them to work together so I wrote thread #31 for help. Subsequently tried with only stage 3+4 (output 2200V) they worked. So now I'm trying to compare whats different between 1+2 vs. 3+4 and so far can't figure out what's wrong with 1+2.
 

I found the culprit, it was Qcharge1 that was defunkt. So now it looks like this and the output with one stage 1100V, two stages is 2000V, three stages 2700V, four stages 1800V - disappointing to say the least.

marx-23.png

So I measured across the caps (200V/DIV), and for C1 and C4 it looks like it should:

SDS00015.png

But for C2 and C3 the caps are completely discharged instantly, that's not good, but why?

SDS00016.png

Here is the commutation
Screenshot from 2019-06-30 21-45-49.png
 
Last edited:

I did some more investigation.

With stage 1+2+3 = 2700V

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

With stage 2+3+4 = 2700V

So the problem seems to occur only when the fourth stage is added, but it doesn't make any sense to me, here is the analysis.

marx-23-path.png
 

i guess i missed it
please show how the capacitors get charged, as the only path through the caps is the discharge path
 

i guess i missed it
please show how the capacitors get charged, as the only path through the caps is the discharge path

Charging is Qsupply1, Qcharge(1-4) = ON and Qdischarge(1-4) = OFF so follow the red and blue lines to each capacitor, the charging is all caps in parallel.

Discharge is Qsupply1, Qcharge(1-4) = OFF and Qdischarge(1-4) = ON so follow the green line through eachcapacitor, the discharging is all caps in series.
 

Red + blue is charge path, as indicated in the drawing. Also explained in the post #27 link.
 

My few bits... I could be totally wrong here, but thought to mention anyhow. Maybe it'll trigger some ideas somewhere.

1) Would it be feasible to share your gate drive circuitry ? It seems to me that in your discharge cycle, in the later stages the IGBT's emitter/ collector will be raised to ever higher voltages. How do you drive the gate in such a scenario ? Some error there vis-a-vis higher voltage operation might explain why each stage works independently but fails when cascaded.

2) What is the 'delay' time in your commutation ? Maybe that's too small - not enough time for some of the charge IGBT's to switch off fully & hence shorting out one complete stage.

3) The voltage drop across Discharge IGBT should surely be lesser than the diode drop in the following stage ? If not, then again this would short out an entire stage during discharge. Maybe put in 2x diodes in series to ensure ?

Do let me know if these are obvious/ irrelevant points, and I'll go back to spectator status.

:popcorn:

- - - Updated - - -

And oh !! What about that free-wheeling diode in the IGBT's ? Doesn't that come into the picture somewhere ?
 

My few bits... I could be totally wrong here, but thought to mention anyhow. Maybe it'll trigger some ideas somewhere.

1) Would it be feasible to share your gate drive circuitry ? It seems to me that in your discharge cycle, in the later stages the IGBT's emitter/ collector will be raised to ever higher voltages. How do you drive the gate in such a scenario ? Some error there vis-a-vis higher voltage operation might explain why each stage works independently but fails when cascaded.

2) What is the 'delay' time in your commutation ? Maybe that's too small - not enough time for some of the charge IGBT's to switch off fully & hence shorting out one complete stage.

3) The voltage drop across Discharge IGBT should surely be lesser than the diode drop in the following stage ? If not, then again this would short out an entire stage during discharge. Maybe put in 2x diodes in series to ensure ?

Do let me know if these are obvious/ irrelevant points, and I'll go back to spectator status.

:popcorn:

- - - Updated - - -

And oh !! What about that free-wheeling diode in the IGBT's ? Doesn't that come into the picture somewhere ?

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). But three cascaded stages do work, it's when you add stage 4 then the stages 2 and 3 does not work. (see thread #35). Then when you remove stage 1 or 4, the three remaining stages work.

2) It's 40uS as can be seen in the scope traces in thread #34. But yes timing could be an issue so I experimented with various delays at various points, even tried to turn the Qd's in sequence forward/reverse with various delays in between, but none of it made any difference unfortunately.

3) I will check it.

4) These igbt's don't have any fw diode, but if they did it still works in the sim.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top