Building my first DRSSTC

[drobotk]

Hello everyone!

I'd like to build my first DRSSTC. I have built a SSTC before but I'm not sure if I understand everything I need to build a DR one. Here is what I have at this point:

- Secondary coil: 110mm diameter pipe, 0.2mm wire, 400mm winding length, around 2000 turns (a bit less because real spacing)
- IGBTs: I have 4x DGTD65T60S2PT
- Primary coil: Not yet made but I have some pretty thick copper cable I could strip
- Top load: Will be made from aluminum air duct from the hardware store

I have made some tests in JAVATC and I think I can reach around 120kHz resonant frequency.

And I have a few questions for you experts here

- Are those IGBTs suitable for a DRSSTC this size? Could I make a full bridge or should I make a double half-bridge? Maybe with high L, low C and long on-times I could reach some safe peak currents for the full bridge (if I understand correctly )?
- Can I use doorknob capacitors for the MMC? (example: https://www.ebay.com/itm/2-pcs-470pF-15kV-40kVAR-K15Y-1-Ceramic-HV-Doorknob-Capacitors-USSR-NOS/202557133513?hash=item2f295892c9:g:uAQAAOSwjthcL-Op )
- What would be the simplest driver to make (not too complex, but preferably with an OCD )
- Do I have to use a toroid ferrite core for the GDT or can I use some EE cores I salvaged?

Thanks in advance!

David

[AstRii]

Hello David and welcome to the High Voltage Forum!
I see no problem with the IGBTs, for a frequeny of 120kHz they seem quick enough.
As you already said, high L and low C ratio will allow for longer ontimes at "safe" peak currents, consider about 7+ primary turns.
I'm not really familiar with specifications of doorknob capacitors, but i seen them many times in power electronics, therefore i assume they are pretty robust.
As a driver i would recommend you some version of Steve Ward's Universal Driver (UD2.7 / UD2.9 / UD3+...) as many people are using those and they are well documented.
For a GDT i would rather buy new ferrite toroidal cores instead of using some kind of salvaged cores. Making a GDT is an actual science and it's pretty much impossible to make a good GDT without knowing it's properties.

A few tips i wish i'd knew before i started my first DRSSTC:
- Low inductance is better than accesibility or design esthetics! Make sure you place all the components as close together as possible.
- It's not necessary to use all the large capacitors you can find at home for DC bus I use 450V 12000uF!! for a small DRSSTC (ended up using 50% of
  the whole DRSSTC space )
- If possible, do not use main's earth as a lower pole for your secondary, find some grounded objects around you.
- If you use any kind of protection as TVS diodes on your bridge, make sure they are rated to clamp let's say 500V. Even a quick low voltage spike can
  destroy your TVS diodes and make them go short circuit and thus destroying the bridge. Make sure they clamp only very high occasional voltage
  spikes.

Good luck and happy coiling!

[davekni]

Most DRSSTC designs use polypropylene (MKP) capacitors for primary resonance MMC.  I have no experience with door-knob capacitors.

EE cores are uncommon on this forum, but I use them more often than toroids for both GDT and CT.  EE cores are almost always ferrite.  One caution:  Make sure the mating surfaces are clean and that the two halves are well held together.  Even a 10um gap can cut effective permeability in half depending on core geometry and ferrite material.  Also make sure the EE halves meet on all three legs (center and two outer).  Occasionally you may find EE cores that are intentionally gapped (shorter center leg) for use in flyback/boost converters.  Don't use those.  Of course, buying a ferrite toroid of known parameters always a safe option.

I'd suggest a full-bridge, with high enough primary inductance to keep current reasonable.  Here's an example layout to keep inductance low:
https://highvoltageforum.net/index.php?topic=1324.msg9886#msg9886

Good luck with your build!

[drobotk]

Thanks for your replies! You really cleared things up for me

A few more questions that came to mind:

- My SSTC was based on the IR2153 ic which has a built-in dead-time, so things were simple. What I don't understand is how do Steve Ward's drivers achieve dead-time? I can't figure it out from the schematic. Also, when coilers show output of the inverter on a scope, there also seems to be no dead-time (?), but there has to be some, right? What am I missing?

- What would be the max peak current that those IGBTs could handle? If I understand correctly, it mostly depends on the max power the device can dissipate, but how exactly do I calculate it based on that?
Datasheet (https://www.tme.eu/Document/dbf1ac2316f96cbe820db95a21100319/DGTD65T60S2PT.pdf) lists 428W@20*C and 214W@100*C

- Do I need 2 GDTs for a full-bridge or can I use 1 with 5 windings? Some Universal Driver schematics I've seen have outputs for 2 GDTs and I'm a bit confused.

- Is the design of the GDT critical? Could someone recommend me a good and easy to follow GDT design guide?

As for the doorknob caps: I was interested in using them because I've seen styropyro use them in his VTTCs and in his HFVTTC so I assumed that they would be suitable for a DRSSTC. But I will buy some regular PP caps if no one can confirm my assumption.

Sorry for all those questions, they're probably really trivial, but I want to avoid as many dumb mistakes as possible

Thanks
David

[davekni]

Two sources of dead-time.  First comes from the symmetric gate waveform, going as far negative as positive.  Ignoring parasitic inductances, all gate voltages cross zero at the same time.  So all switches are off as the gate-drive waveform transitions from +gate_threshold_voltage to -gate_threshold_voltage.  For FETs, that is typically sufficient.  My SSTC uses such GDT drive with nothing further for dead-time.
IGBTs have minority-carrier lifetime, so typically turn-off more slowly than turn-on, so require additional dead-time.  That's accomplished by adding a resistor in series with each gate to slow turn-on, with a parallel diode to keep turn-off fast.  There is often a second smaller-value resistor in series with the diode (or in series with the diode/resistor pair) to dampen gate-drive overshoot.  Damping is needed because of GDT parasitic (leakage) inductance.

One GDT is fine unless you are building a very-large DRSSTC or running fancier pulse-skipping drive.  Best to use 8 windings instead of 5.  Use four twisted pairs (ie. CAT5 cable).  One side of each pair for secondary.  Other side of each pair is primary, all four in parallel.  That reduces parasitic inductance.

GDT should minimize parasitic leakage inductance.  That means short twisted leads and relatively-few turns.  Core needs to be ferrite with reasonably high permeability.  Most GDTs have many more turns than needed.  That is because the cost of extra turns is just added leakage inductance, which can generally be handled with higher gate resistance and correspondingly slower switching.  The consequence of too few turns is core-saturation, more of a failure cliff.  If you know where the cliff is, there's no need to be too far away.  I'm sure there's some good GDT design guides around, but I don't know where off-hand.  Don't have time to write one in this reply.

Peak current is always a guess, and depends on gate voltage.  If driven with +-22V, I'd guess you could run up to 250A or perhaps even 300A peak.  As a rule-of-thumb, IGBT frying occurs around 4x rated average current or 2x rated peak current when driven at high Vge.

IGBT stress will be less when running with some phase-lead.  That way collector voltage transition occurs due to residual primary current after 2 IGBTs turn-off, before the other two turn-on.  If gate-drive phase lags primary current, IGBT diodes conduct before switching, and collector voltage changes only after the opposite IGBTs turn-on.  Diode reverse-recovery makes a current spike, which makes a voltage spike at the end.  That is the key reason low-parasitic-inductance is needed in H-Bridge layout.

[drobotk]

Thanks David, very useful info!

Now, I've finally decided to order some driver PCBs. I have found this thread by Daniel Marks (https://highvoltageforum.net/index.php?topic=353) where he released a pack of tesla coil related PCB files. And from here arise 2 more questions:

- Have anybody checked/tested those PCBs and could confirm whether they're good?
- In the pack there are files for UD2.7c and for UD2.9. Which one should I go for? I know that UD2.9 is the skip-pulse thing, but is it really worth it? What is the real difference between skip-pulse and regular OCD?

Again thank you all for your help! I can't wait to get this coil runnin'

David

[davekni]

I wouldn't use pulse-skip on your first coil.  It puts more load on the IGBT diodes, so you need to make sure your particular IGBTs can handle that much diode current.  It extends enable pulse width (rather than ending early) at maximum current, so requires lower IGBT transient thermal impedance.  Depending on how it is implemented, it can also make GDT imperfections more of a problem.

Concerning those specific boards, I have no experience.  However, I think I've seen those used in other forum posts.  Suggest searching here.

[AstRii]

What is the real difference between skip-pulse and regular OCD?

The difference is that when regular OCD trips the enable pulse will be pulled down to 0 and this will turn the H-bridge off until the next enable pulse comes in.
Thus when reaching maximum primary current, your bridge will turn off.
Pulse-skip is more sophisticated. When pulse-skip OCD trips, your bridge will turn off but it will turn on again when the primary current rings down below your OCD current.
This way you can run longer on-times without primary current growing beyond a point of failure.

Even though usually you set OCD lower with pulse-skipping than you would set it to without pulse-skipping, it heats up your IGBTs way more, because of the longer on-times at high current.
Therefore i would not try to drive TO-247 package IGBTs with it. I think pulse-skipping is more of a brick IGBT feature:)

Although if you're planning to make a pulse-skipping coil in the future, it may be beneficial to order let's say 2 UD2.9 boards instead of making UD2.7 and then a UD2.9.
As many of components usually gets cheaper when buying in larger quantities.
You can run UD2.9 without pulse-skipping by not connecting the J16 jumper on the board and therefore run it in UD2.7 style:)
This way you could even sometime in the future (when you gain more experience) run your first coil in pulse-skipping mode simply by shorting the jumper and setting lower OCD threshold.

[drobotk]

Ah, I see. Thanks for the clarification!

So I decided to go with the 2.7c as I highly doubt I'll be pulse-skipping any time soon.

Now it's time for the MMC. I have been thinking of buying WIMA MKP1T031006B00MSSD capacitors (100nF; 1.6kVDC; 1,4kV/μs). I have found only one datasheet and unfortunately it's in German, but here it is: https://www.tme.eu/Document/c9c3612d9384576529f461ace7f3add8/WIMA_MKP_10.pdf
I think they should be absolutely fine in a 300A max. MMC. But before I buy anything, I'd like to hear from you guys if these are a good choice.

So as always, thanks in advance!

David

[Mads Barnkob]

Ah, I see. Thanks for the clarification!

So I decided to go with the 2.7c as I highly doubt I'll be pulse-skipping any time soon.

Now it's time for the MMC. I have been thinking of buying WIMA MKP1T031006B00MSSD capacitors (100nF; 1.6kVDC; 1,4kV/μs). I have found only one datasheet and unfortunately it's in German, but here it is: https://www.tme.eu/Document/c9c3612d9384576529f461ace7f3add8/WIMA_MKP_10.pdf
I think they should be absolutely fine in a 300A max. MMC. But before I buy anything, I'd like to hear from you guys if these are a good choice.

So as always, thanks in advance!

David

Hi David

I did not have time yet to chime on your build thread here, looks really good and you put some effort into doing it the best possible way to start with! Cudos for spending the time on research!

From page 6 of the datasheet, you can see the difference from WIMA MKP10 to FKP1 on derating of the voltage specification vs. operating frequency. MKP is approximately 10-fold worse, so keep that in mind when designing the overhead voltage rating.

I got the data for the WIMA FKP1 capacitors listed in my good MMC capacitors for Tesla coils list and you should be able to use the same data for the MKP with above considerations. The same WIMA FKP1 capacitors are also available in the tesla coil MMC calculator tool

[drobotk]

Well, I think it's about time I finally revive this thread

School got the best of me and I sort of lost motivation to work on this project, also because I would rarely have enough energy/free time to go into the shop. Now that I've graduated I set out to finally build this thing and I aim to have it finished (or at least make a few sparks before blowing up) by the end of summer.

A few specs/plans have changed since I last posted here, so I'll probably edit the first post to reflect that.

But here's where I'm at right now:

Interrupter, THT UD2.7c, IGBT full bridge, GDT and CTs are all built. Currently got it all setup with a function generator at 150kHz working as feedback and nothing connected to the bridge yet. I've done some tests with it and for the most part, everything seems to be working great.

Two IGBT gates:



Crossing seems OK I think, but let me know if something should be faster/slower.


My only concern right now is some ringing I'm seeing on the gates. After the drive disappears, there seems to be a consistent delay of about 30us and then some damped oscillations occur, which according to my scope can reach well above the IGBT's Uge threshold. It seems to be about 40kHz.


I've read in the UD2.7c documentation that the RC network across the DC blocking capacitor of the output stage "damps ringing with the leakage inductance of the GDT and circuit capacitance, which might accidently turn on the IGBTs on the GDT secondary". This makes me think that maybe I need to adjust their values to make it work better?

I also have a few concerns about the startup of the driver, but I think that's enough questions for one post after such a long time

Glad to be back

David

[flyingperson23]

How are your GDTs constructed?

[drobotk]

Ah yes, should have shared.

I have a single GDT (since the THT version of UD2.7c only has one output) made of 9 turns of 4 twisted pairs from network cable. This is the core I got: https://www.tme.eu/en/details/b64290l0618x830/ferrite-cores/epcos/



The 4 primaries are all connected in parallel. I'm 99% sure I didn't mess up the phasing of them

Cheers

David

[davekni]

I have a single GDT (since the THT version of UD2.7c only has one output) made of 9 turns of 4 twisted pairs from network cable. This is the core I got: https://www.tme.eu/en/details/b64290l0618x830/ferrite-cores/epcos/

GDT looks great.

My only concern right now is some ringing I'm seeing on the gates. After the drive disappears, there seems to be a consistent delay of about 30us and then some damped oscillations occur, which according to my scope can reach well above the IGBT's Uge threshold. It seems to be about 40kHz.

Normal damped oscillations are immediately after last gate pulse.  There is some issue with UD2.7c board, probably with driver chip or FET buffer between driver chip and GDT.  Are you using an inverting driver chip as is needed for UD2.7c?  A non-inverting driver chip could explain delayed ring.  But I'd also expect an issue with start of pulses, which look fine in your captures.  Scoping driver chip outputs would help, should both be high when disabled.  If those look good, then scope FET gates.

[drobotk]

Huh, well this is awkward. The BOM for the driver lists TC4427 as a cheap UCC27423 substitute. Now I looked up its datasheet and it's indeed not an inverting driver chip, and I was probably meant to use TC4426. Is this an error in the BOM or is there a way to use a non-inverting chip? I'm confused


David

[flyingperson23]

I think it might be an error in the BOM. I ran into the same issue a few years ago with my first drsstc.

[Admiral Aaron Ravensdale]

Yes there was also an error in the order list from Mouser.
There was listed a LM8365BALMFX45 instead of LM8365BALMF27, which caused that the OCD stays on!
I got an wrong delivery from Mouser and get one xx27 and xx45 and one one board worked.
Take me some time to figure out what was wrong :-)

[davekni]

Is this an error in the BOM or is there a way to use a non-inverting chip? I'm confused

Must be a BOM error.  There are a couple patch (modification) options for using non-inverting driver, but easier and safer to change driver chips.  Simplest patch is to add zener diodes across capacitors coupling driver chip outputs to PFETs.  That works only if 24V supply is stable and accurate and you happen to have ~16V zener diodes available.

Good work scoping signals carefully before full power runs!  Much cheaper to fix now than after IGBTs are fried.

[drobotk]

Good news.

Replacing the driver chip with a TC4426A seems to have solved all my previous issues. The ringing is no more and the driver correctly starts up with no feedback, which it previously could not. I also noticed that previosly the first gate pulse wasn't synchronized with the interrupter edge, but now it always is, which seems good.

Here is the same scope setup as last time - Yellow is IGBT gate and green is interrupter signal.


With that I think I'm set. Is there anything I might wanna still test in the current setup before I try powering things up?

Thanks

David

[davekni]

With that I think I'm set. Is there anything I might wanna still test in the current setup before I try powering things up?

Have you already "powered things up" with a bench supply or other lower voltage supply for VBus (for H-bridge power)?  I'd suggest doing that before running at full line voltage.  Though not everyone does that - just the cautious sequence.

Either way (at low or full Vbus to H-bridge), initially test at low duty cycle (low repetition rate and short enable pulse width).  Minimizes chance of frying IGBTs if there is some issue.

Replacing the driver chip with a TC4426A seems to have solved all my previous issues. The ringing is no more and the driver correctly starts up with no feedback, which it previously could not. I also noticed that previosly the first gate pulse wasn't synchronized with the interrupter edge, but now it always is, which seems good.

Yes, all those previous issues are explained by wrong (non-inverting) driver chip.