First DRSSTC: Problem with slow rise/fall times

[Phoenix]

Hello everybody 

The following post is long and contains a lot of scope shots but i would be very thankful if you could help me

As you may know I have rebuilt my DRSSTC with 2 SKM200 Bricks in parallel. I always had issues with slow rise/fall times even with only one IGBT Brick. The SKM200 have a very low gate capacitance (13nF) so i do not understand why my rise/fall times are so slow. I am using a modified 1.3b Driver with added phaselead and a second GDT output.

Here i have a list of things i have tried to improve my rise/fall times:

I tried to make all my connections as short as possible
I tried 2 GDT´s and 2 driver outputs but the rise/fall times did not really improve and i got a lot of glitching as soon as i applied power
I bought a new ferrite toroid to make absolutely sure i have N30 material
I changed the gate protection diodes to zener instead of TVS, because i heard they have a lot of capacitance but it changed absolutely nothing
I bridged the coupling capacitors at the GDT Output and nothing improved in a positive way
I tried different amounts of turns on my GDT but nothing significant improved (only +- 50nS)
I lowered my gate resistors to 2,35 Ohm
If i connect the 2 Outputs in parallel my gate signal starts to look ugly but maybe this has to do with the metal plate of the case underneath the driver

I am starting to get tired because nothing helps. How can someone get such fast switching times with CM600 Bricks that have a gate capacitance in order of magnitudes higher?

Here i have some scope shots for you:

This is the gate signal. As you see it is extremely slow (over 1us rise/fall times)

Here i marked the points from 4 to 15V because this is the voltage range were i would have the highest losses (the IGBT´s turn on in this range)


Here are some screenshots of the voltage and current during operation. I got rid of nearly all the switching spikes with phase-lead. For the remaining ones i have 440V TVS-Diodes and snubber capacitors. I know it looks like the bridge is switching to early but this has to do with compensation of my probes. Voltage spikes are a better indicator of bad ZCS. As you can see here it takes the Bricks 340ns to turn on and 100ns to turn off. I have heard i should simply let the bridge run to see if anything gets hot. But it is difficult to tell the real temperature of the junction.





For the last image i have a picture of my GDT and the wiring:


I really hope that someone more experienced can help me. I set the OCD to 1000A and i would really like to see some sparks again. I am just worried that everything will blow up with such slow rise/fall times.

Greetings
Phoenix

[profdc9]

Could you create a "dummy gate load" to connect to your GDT which consists of a 10 nF capacitor in series with a 2.5 ohm resistor to try to see if it can charge and discharge a capacitor with a low rise/fall time?  You could also attach the dummy load directly to your driver as well and see if it works differently to see if its the GDT.  This won't model the Miller capacitance but it might help you troubleshoot the driver/GDT setup.

[Phoenix]

Thats a good idea thanks for your response. I noticed that the signal at the gates of the driver mosfets has only 200nS which is not much. Tomorrow i will trace the signal back with my scope and see at which point the signals gets slow. I will try to replace the foil capacitors with ceramic ones maybe that helps.

[Mads Barnkob]

Huge thumbs up for the thorough explanation of your problem with all the data and measurements available to help you!

It looks like you are doing some static dummy load tests with low voltage across CE/DCbus?

This is from page 23 in ST's "Datasheet Explanation" AN4544, application note (I also attached the PDF to this reply)
 

Reverse transfer capacitance (Cres)

This is the reverse transfer capacitance measured between the collector and gate terminals
with the emitter connected to ground. The reverse transfer capacitance is equal to the gate
to collector capacitance:

The reverse transfer capacitance, often referred to as the Miller capacitance, is one of the
major parameters affecting voltage rise and fall times during switching.

Figure 19. STGW40V60DF capacitance variation
Figure 19 shows a graph of typical capacitance values versus collector-to-emitter voltage.
These capacitances decrease over a range of increasing collector-to-emitter voltage,
especially the output and reverse transfer capacitances. This variation is linked to the gate
charge data. These parameters are not tested in production.

[Phoenix]

Thank you very much for your response, Mads

You are right, this test was at low voltage and with static load. After trying out different things i finally got it to work without glitching using 2 driver outputs 
The solution was a small 1nF capacitor between the output of the driver. Now everything is working fine and my rise/fall times are arround 100ns at the inverter output which is fine.

Here i have some technical data if someone is interested:
I am using 4 SKM200 Bricks from Semikron for the bridge (2 in parallel)
The secondary coil has 1850 turns of 0.4mm wire and is 77cm tall
The MMC consists of 60 WIMA Fkp 1 and i am getting 375nF at 6400V
I succesfully added phaselead and a second GDT output to the 1.3b driver
The interrupter schematic is also from Steve Wards
The electrolytic bus capacitor has 19000uF at 400V: Probably way to large for this coil but you can never have to much 
The coil consumes arround 5-6kW from my Variac at 350V bus voltage.
Without spark loading the resonance frequency is arround 83kHz, with spark loading it is arround 75kHz
For my topload i am using 20cm aluflex tubing. The topload has a diameter of 80cm.

I did a lot of high power testing during this weekend and i am more than happy with the results. I have got a lot of images of sparks and the coil for you:

Here is the whole setup. The DRSSTC is arround 155cm tall.

Here is an image of the nice power electronics

This is my modified driver

This is my interrupter

Here are some images of the sparks. The OCD was set to 1100A during these tests. The longest straight spark was arround 220cm, the one image of the two discharges at once is arround 320cm in total! I tuned my coil to a point where the OCD is only limiting at full power and during heavy ground strikes.






Again, thank you very much for your help during my build. I am happy that i managed to finish my project and i am happy with the results.

Greetings,
Phoenix

[Hydron]

Wow, looks very nicely made, and great spark pics
220cm is about the max length I got on a similar sized coil, so it sounds like it's running well too.

[Phoenix]

Thank you very much for your nice comment, Hydron 

I am going to leave the OCD at 1100A because I do not want to risk anything. 220cm is more than enough for me. You can never know how good the current sharing between the bricks is in reality. But i have read somewhere that the SKM200GB128D have a special technology which makes paralleling easier.

I really like your methode of measuring the currents from and to the topload with your scope. It is very interesting to see the current waveforms and how everything changes durring a ground strike.

Greetings,
Phoenix