First DRSSTC: Problem with ZCS

[Phoenix]

Hello everybody 

I am sorry about my English, it is not the best. I am currently building my very first DRSSTC and i have a problem. The Fullbridge consists of 2 Semikron Skm200 Bricks and i am using the 1.3b Driver from Steve Wards. Everything is working fine and i conducted a static load test with a metal plate on the primary. The plate got very hot and water vaporized instantly when i dropped it onto the plate. My only problem is that the Fullbridge is not zero current switching like it is supposed to do. Here you can see a screenshot from my oscilloscope:  The red signal is the primary current measured via a 1:32:32 CT and a 100 Ohm Shunt resistor. The yellow signal is the output of the inverter. Here you can see a picture of my whole setup: I am quite surprised that the bricks don´t get hot even with 500A primary current. The voltage spikes are the big problem. If everything is working correctly, the spikes are getting suppressed by high power 440V TVS Diodes, but if they break the bricks are also going to break. For the Gate Resistors i am using 2.5 Ohms. Should i put higher valued resistors like 5 Ohms in front of the Gates? I also tried adding phase lead by adding a variable inductor and a resistor across the terminals for the feedback, but the Inverter did not want to start resonating with it.

I would be very grateful if someone could help me. I do not want to destroy my bricks.

Greetings
Phoenix

[Mads Barnkob]

Hi Phoenix

Welcome to HVF and you do not ever again have to excuse your English, you got no problems there!

You build a good looking inverter and you have implemented a lot of the best practises that is vital for a good performance, so there is not much to point at really.

I do have some questions before we look further into the switching voltage spikes.

What is your bus voltage during this test? The scope shot shows 400V with spikes, so I assume you are not testing at full voltage here. Low voltage testing will never show the real deal when we talk switching spikes, but more about that in another post if this was low voltage testing.

What size snubber capacitors are you using? Did you try different capacitances?

I do not think that you should worry about destroying your bricks,  you made a very clean build and being the first DRSSTC it is normal to nurse the coil, but they can really take some punishment when built so well as yours

The value of your gate resistor should be decided upon the waveform of the gate drive, excellent fast trouble shooting here: http://www.richieburnett.co.uk/temp/gdt/gdt2.html

[Phoenix]

Hi Mads 

I am glad to hear such nice words about my inverter from someone with a lot of knowledge about powerlelectronics

You are right, the DC-Bus voltage was around 160V when i took this image with the scope. I think it drew about 8A from my 20A variac. I have never gone over 220V bus voltage because i am scared something might blow up. Theoretically i could go up to 350V. I noticed that the voltage spikes never go over 440V even if i increase the voltage more. I think the TVS Diodes are clamping down the spikes. I am using this layout for my fullbridge: http://www.stevehv.4hv.org/drsstcdesign/ISSTC_sch2.JPG
My DC-Bus capacitor has 19000uF at 400V. I know that 19000uF is a lot for a DRSSTC this size. I also added a 100nF 1600VDC film capacitor from the negative rail of the bus to ground.

Right know i am using 2x Wima Snubber MKP Capacitors with a capacitance of 3,3uF each. So i have a total snubber-capacitance of 6,6uF.

The phase lead worked to some extent after i increased the value of the resistor in series with the inductor. The ringing and the spikes definitely decreased but the bad timing was still the same.

The page you linked me is very helpful. My signal looks good and there seems to be no difference between 5 Ohms and 2.5 Ohms gate resistance. The rise and falltimes are a little slow (400ns), but if i measure at the output of the inverter, the IGBT´s switch on/off in about 100ns which seems to be fine.

[Mads Barnkob]

I would also like to say that your MMC is a text book example of good current sharing, it is good to see that you have paid attention to all the small details.

Thanks for all the details, you are within all reasonable values for your components, that is not the issue.

Your only problem is a very common one and it has fooled everyone that has ever played with power electronics, I will say that even electronic engineers will at some point fall for these deceiving waveforms when doing tests.

The problem is the low voltage across the CE junction of your IGBT brick, it is the capacitances of the IGBT that plays us a trick.

Let of look at this cut-out from a IGBT datasheet and it is the capacitances vs. Vce graph that is interesting to us.



Coes, the output capacitance of the CE junction is what is causing the trouble, look at the capacitance value vs. CE voltage, so at low voltages you will have a high capacitance storing energy that gives you those violently looking spikes that might seem like 200% of your bus voltage, but that capacitance gets so minuscule at full voltage that the spikes are now down to 10-20% of your bus voltage with a clean layout like yours.

Another reason that you should not worry about these spikes is that the energy behind them is very low and a IGBT brick has a lot of mass to absorb these losses. If the voltage spikes was to go beyond the voltage rating of the IGBT, the anti-parallel diode will act as a zener diode and eat those spikes, again it can dissipate a lot of these small energies with its huge mass.

Turn on your video recording, full voltage and let it make some huge sparks! I am looking forward to see the test results
 

[Mads Barnkob]

Wow, excuse me for being blind. I completely missed the whole point of your first post, I was too focussed on the spikes being your problem and not that your waveforms shows you are hard switching at full primary current.

But as you say that the bricks and heat sink stays cold, something is not adding up.

Could you please verify that your oscilloscope is set up correct without time difference between channels.

Check your probes/current transformers according to page 10 in this guide, if extreme skew is your problem: https://highvoltageforum.net/index.php?topic=111.0

[Uspring]

Perhaps your 100 Ohm shunt resistor is too large. That may cause a phase shift between the primary current and the voltage across the shunt. Try 10 ohms or less. Also check the phasing between the feedback CT and bridge and between feedback CT and your measuring CT.

[Phoenix]

@ Uspring yes you were right, thank you very much
The 100 Ohm shunt was causing the phaseshift. I replaced it with a 10 Ohm shunt and now it looks way more like ZCS. Of course i will never be able to get perfect soft switching with a driver that is not able to do phase lead.

Thank you very much for your explanation with the voltage spikes Mads, you were absolutely right. Today I went up to 300V bus voltage and my scope was only showing 330V in total instead of 408V. I set the OCD to 520A today and let the bridge run at that current. In the static load test i made the bridge draw 22A. The baking tray got up to 300°C. I used my infrared thermometer to measure the temperatures of the components. The IGBT Bricks only increased 10°C in temparature (20°C, 10°C ambient temperature in my workshop. The only thing i worry about right now is the temperature of my MMC. It got about 45°C hot. I may need to buy more of these FKP 1 Wima capacitors and increase the RMS current capability of it. Luckily, i know a seller that sells them very cheap.

In the evening i am going to add a second layer of epoxy to the secondary. The coil turned out to be absolutely beautiful. I payed attention to the termination. There is no metal inside the coil. The secondary has a diameter of 160mm and is wound with a 0,4mm wire. In total it has 1800 Windings.

The last thing i need to construct is the topload. I bought some aluflex tubing for that. Then i will finally be able to produce sparks after tuning   

[Hydron]

Looks great!

Now that the spike/ZCS questions are cleared up, the only other suggestion I'd have would be to twist the gate drive wires going from the transformer to the IGBTs. This will reduce their inductance and make the switching a bit faster/cleaner.

Can I ask which IGBTs you're using? Mads will be interested (he has a table of IGBTs tested in different DRSSTCs).

[Phoenix]

Thanks for your reply Hydron 
I have already twisted the gate drive wires to the IGBT Bricks. I am using the SKM200GB128D Bricks from SEMIKRON. Tomorrow i am going to do another test to make sure that the bridge is soft switching. I can input a signal into the feedback input of the driver with my function generator and then i can check the delay between the input and output. There needs to be no bridge connected durring this test. But i think that everything is working correctly. 520A at 300V would produce 156kW of switching losses. The temperature is going to raise way beyond 20°C if that is the case.

[Mads Barnkob]

I will certainly add the newer generation SKM200 brick to the table when you have some "maxed out" results.

It is actually Hydron that has the older SKM200GB125D sitting at 750A@80kHz: http://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/igbts/

[Phoenix]

I have good news: Today i replaced the CT for the Feedback and now the bridge is switching perfectly at zero current. Here you can see a picture from the scope: The red signal is the current, the yellow one the voltage.

The only problem i have right now is my MMC: Today i lowered the resonance frequency to 70kHz. Suddenly i heard a banging and hissing noise. I took the whole bridge apart and everything was fine when i checked it with my multimeter, so it must be the MMC. This happend because i increased the primary inductance to lower the frequency and therefor the voltage across the MMC increased and exceeded it´s rating. The good thing is that the Wima FKP1 capacitors i am using are self healing, so i can reuse them. I checked them with my LCR Meter and only 6 of them lost 3nF in capacity. I have already ordered my new capacitors for the new MMC. It is going to consist of 60 Wima FKP1 100nF 1600VDC capacitors. I will put 5 of these in series and 12 of these packs in parallel. Therefor, the finished MMC is going to have 240nF at 8000VDC. The old one had 233nF at 4800V.

[Phoenix]

I have another question about the MMC: Is it better if i connect 4 capacitors in series and 15 of these packets parallel? Then i would get 375nF instead of 240nF. Maybe i can only run a
 shorter pulse-width with the DRSSTC after that, but i think my discharge length is going to increase. The RMS current capability would also increase which would lead to less heating of the MMC.

[Uspring]

The issue of choosing the right size MMC is somewhat complicated. A larger MMC implies less primary turns for a given frequency and thus a faster rampup of primary current. Initially all the energy goes into the primary. The power transferred to the secondary is proportional to the product of primary and secondary current. Since secondary current is initially zero, the energy in the secondary will somewhat lag behind the primary energy. A short burst might get you not enough energy into the secondary, perhaps a lot less than in the primary.

The fast primary current rampup of a too large MMC might cause the overcurrent protection to trigger. From my experience it is better to design a coil in such a way, that arc loading will limit the current in the primary so that the OCD wont trigger. Much longer bursts are then possible. Ideas on the right choice of MMC capacitance were discussed here: https://highvoltageforum.net/index.php?topic=90.msg462#msg462 . They are based on letting the arc loading limit the primary current.