160mm DRSSTC project

[Mathieu thm]

Hello everybody,

For a long time, I always wanted a DRSSTC but it was unimaginable for me to build one.
I started to make some small SGTC including a big RSGTC that I still have, and I started to do small SSTC.
A few months ago I built my first "little" DRSSTC to begin, which I never would have thought of when I was 12.
Now that I've built a little DRSSTC to understand how it works, and have fun, I am embarking on a project that I would love to do for a very long time, a fairly large DRSSTC.

Since it's a pretty big project, I started doing CAD design to see what it could look like and avoid as many mechanical problems as possible.
Here are the plans I was able to make with SolidWorks :





The box should be 46x25.5 cm (not counting the wheels and the primary support).
The entire DRSSTC should be about 1m40.

Some specs (it's not final, it can still change) :

• Toroid: ~150mm minor diameter, ~700mm major diameter
• Secondary : 160mm, 73cm winding with 0.315mm copper wire
• Secondary frequency : 74kHz
• I don't know yet exactly how the MMC will be (maybe Dawncap 3000v 0.47uf in 5s2p configuration)
• For lGBT, I don't know exactly, maybe cm300 or skm300 depending on price and availability, I do not know yet if I make it work in 325v or with a voltage doubler in 678v
• Driver :UD2.7 of profdc9
• Bus capacitance: maybe 2x6800uF (or 6800uf with the voltage doubler)

I have already started a little, starting with the secondary circuit,
Here is some photo:



I will try to start making the aluminum case, and of course I will post the progress !

[davekni]

I will try to start making the aluminum case, and of course I will post the progress !

Only concern that comes to mind is aluminum case top ring looks too close to primary coil.  See this post and many following ones related to the same issue with a large coil:
https://highvoltageforum.net/index.php?topic=1430.msg11047#msg11047
Top of aluminum frame makes shorted turn.  Issue in above link is that 8020 is anodized, so has a thin nonconductive coating on sides.  Initially frame top isn't quite a complete shorted turn due to this anodization.  Arc from secondary broke down this thin oxide layer.  Then the much higher current induced from primary coil melted screws.

[Mads Barnkob]

Great material and looks like you are well prepared for building a coil like this

I think you should add longer legs from frame to platform that supports the primary coil. A total height of 140 cm for a CM300 coil, that can easily do 2-2.5 meter sparks, will just give you a lot of ground strikes and not take advantage of such a powerful bridge.

Your wheels on the base is going to give you stability problems, the foot print is simply not large enough for something that is this tall. I have made the same mistake on both my DRSSTC 1 and 3. For my large coil I simply had to build a support frame of 4 pieces of 2 meter long wood frame, to give it a large enough foot print. Somehow there is never a even surface where you are going to run a Tesla coil, as its mostly outside

A 5s2p MMC of 0.47uF capacitors is only going to net you around 0.188uF, that is too small for such a coil, in my opinion. Did you take your design for a ride in the MMC calculator? https://kaizerpowerelectronics.dk/calculators/mmc-calculator/

In my guide, I recommend 0.3 to 0.6 uF for a 160mm coil https://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/mmc-tank-capacitor/

[Mathieu thm]

Only concern that comes to mind is aluminum case top ring looks too close to primary coil.  See this post and many following ones related to the same issue with a large coil:
https://highvoltageforum.net/index.php?topic=1430.msg11047#msg11047
Top of aluminum frame makes shorted turn.  Issue in above link is that 8020 is anodized, so has a thin nonconductive coating on sides.  Initially frame top isn't quite a complete shorted turn due to this anodization.  Arc from secondary broke down this thin oxide layer.  Then the much higher current induced from primary coil melted screws.

I think you should add longer legs from frame to platform that supports the primary coil. A total height of 140 cm for a CM300 coil, that can easily do 2-2.5 meter sparks, will just give you a lot of ground strikes and not take advantage of such a powerful bridge.

I originally wanted to insulate the upper part of the chassis like the brOdin coil, but that's a good idea, I would put longer legs from the frame to the platform that supports the primary coil to gain some height, and to avoid problems with the primary coil.

A 5s2p MMC of 0.47uF capacitors is only going to net you around 0.188uF, that is too small for such a coil, in my opinion.

I suspected that the MMC would have too small a capacity, I would try to make an MMC in this range of 0.3 to 0.6uf.

Your wheels on the base is going to give you stability problems, the foot print is simply not large enough for something that is this tall. I have made the same mistake on both my DRSSTC 1 and 3. For my large coil I simply had to build a support frame of 4 pieces of 2 meter long wood frame, to give it a large enough foot print. Somehow there is never a even surface where you are going to run a Tesla coil, as its mostly outside

Thanks for this information, this can indeed be quite an annoying problem.

[Mathieu thm]

I started to look at how much it could cost me to make the UD 2.9 of profdc9, but I noticed that it works in pulse skip, I also looked at the UD 2.7 of profdc9 and it would cost me almost the same to do the UD 2.7 (without pulse skip) or the UD2.9 (with pulse skip).

I wonder if it can be a good idea to make the pulse skip controller, or I keep it simple with the UD 2.7.

The pulse skip mode really brings a plus for this coil size ?

I imagine the pulse skip mode adds maybe more stress on the IGBTs,  so maybe it's better to stay with UD 2.7 ?
I don't know at all if it can be a good idea to use the pulse skip mode, what do you think?

I also started looking at the IGBT bricks, after looking at the table on the IGBT page (which is very useful !) : https://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/igbts/
I saw the CM300DY-24H bricks which I think might work well with my setup, I read several threads on where to buy IGBT bricks, so I looked on eBay for IGBTs and found several sales, one of which I spotted: https://www.ebay.com/itm/282631487695

I also looked at the SKM300GB125D: https://www.ebay.com/itm/322570650877 but I'm thinking of using the CM300s

Are there any special things to pay attention to when selling IGBT bricks ? Are there many counterfeits ?

[dbach]

Hey,

Im not sure if theres a better way to do this but i would be happy to sell you 2 cm300dy-24h bricks for nearly half that price, I can confirm they are genuine as well.

Most bricks on ebay will be genuine but make sure they are functional and also not in too poor condition.

Kindest regards,
Davis

[Mathieu thm]

Im not sure if theres a better way to do this but i would be happy to sell you 2 cm300dy-24h bricks for nearly half that price, I can confirm they are genuine as well.
Most bricks on ebay will be genuine but make sure they are functional and also not in too poor condition.

Thank you for the answer, it's an interesting offer, but the problem is that I live in France so I think the shipping can be very expensive

[futurist]

Nice work so far!

Are you planning CM300 half-bridge or full-bridge?
Check out my similar 160 mm coil https://www.highvoltageforum.net/index.php?topic=78.0
I'm using SKM400 half-bridge with 340nF MMC, originally I had CM200 half-bridge and both worked great
 

I would suggest you use CM300 half-bridge with voltage doubler for your coil, especially with 240 VAC mains here in Europe
Mine survived 7 years of occasional use and I'm sure the inverter won't fail anytime soon. I recently rebuilt it, changed DC power supply and rewired everything- now I'm waiting for some good weather to test everything out

[Mathieu thm]

Check out my similar 160 mm coil https://www.highvoltageforum.net/index.php?topic=78.0

Very nice DRSSTC ! Interesting way to cool IGBTs

Are you planning CM300 half-bridge or full-bridge?

I planned to make a full bridge with two CM300 which will maybe CM300DY-24H.

[Mathieu thm]

Good evening everyone,

Here is a little update, I finished making the aluminum frame and there are some changes compared to the planned plan.
I increased the height of the frame by a few centimeters (30cm instead 25cm originally planned),I will also extend the length of the bars between the frame and the primary coil (I think 15cm rather than 5cm originally planned).

Here are some pictures of the frame :

[Hydron]

Having the aluminium frame rails form a full loop without any insulating breaks may not be the best plan, see here for some fun times another coiler had: https://highvoltageforum.net/index.php?topic=1430.msg11047#msg11047
(I suspect at the power levels of a smaller coil it won't be quite as spectacular as the bolt-burning seen on that coil!)

If it were me building the coil I'd look at putting a few insulating breaks into it, maybe get some pieces of FR4 from a cheap PCB vendor to make insulating versions of the fixing plates.

[Mathieu thm]

Having the aluminium frame rails form a full loop without any insulating breaks may not be the best plan, see here for some fun times another coiler had: https://highvoltageforum.net/index.php?topic=1430.msg11047#msg11047
(I suspect at the power levels of a smaller coil it won't be quite as spectacular as the bolt-burning seen on that coil!)

If it were me building the coil I'd look at putting a few insulating breaks into it, maybe get some pieces of FR4 from a cheap PCB vendor to make insulating versions of the fixing plates.

You're right, I've already planned to make an insulation so as not to have a closed loop, I haven't done it yet because I'm thinking about how I could do it properly

[Mathieu thm]

I received the pcbs of the ud2.7c, and I took the opportunity to take some Hydron "gate drive" pcb :



After soldering all the components on the ud2.7c, here is the result :



I plugged it in and tested it and everything seems to be working fine 

Now i have some question about Hydron gate drive pcbs, i started soldering a pcb and this is what it looks like :



Here's the schematic and the link to where I found it :



https://highvoltageforum.net/index.php?topic=493.msg3068#msg3068

What I don't understand is why not choosing a classic gate drive circuit (diode and resistor).
Why is there an extra diode and resistor ?
Also, putting two 5.6ohm resistors in parallel gives a value of 2.8ohm, it seems small to me but maybe I'm wrong.

If someone or Hydron himself could explain this to me, that would be great.

EDIT : I forgot to put the link to my Mouser cart, it took me so long to do it so if it helps anyone, here is the link :

https://www.mouser.fr/ProjectManager/ProjectDetail.aspx?AccessID=9bfd2b6a8c

[Mike]

The design gives the maximum flexibility. You have independent control of the gate charge resistor and discharge resistor allowing you to tune each to get the rising and falling edges as fast as possible with acceptable ringing.

With the classic single resistor and diode circuit you can't adjust the turnoff speed, it will always be as fast as possible which, depending on your layout and IGBT package inductance may cause excessive ringing during turn off. Under most circumstances this ringing doesn't do any harm but it could result in an EMI issue as it will likely be at quite a high frequency. You can still replicate this by setting the turn off resistors to 0R, which brings us to your other point about the low resistance.

The schematic notes that the 5R6 resistors are placeholders only. You are meant to set these correctly for your IGBTs and driver. In particular having the charge and discharge resistors equal negates the advantages of the diodes.

[Mathieu thm]

The design gives the maximum flexibility. You have independent control of the gate charge resistor and discharge resistor allowing you to tune each to get the rising and falling edges as fast as possible with acceptable ringing.

With the classic single resistor and diode circuit you can't adjust the turnoff speed, it will always be as fast as possible which, depending on your layout and IGBT package inductance may cause excessive ringing during turn off. Under most circumstances this ringing doesn't do any harm but it could result in an EMI issue as it will likely be at quite a high frequency. You can still replicate this by setting the turn off resistors to 0R, which brings us to your other point about the low resistance.

The schematic notes that the 5R6 resistors are placeholders only. You are meant to set these correctly for your IGBTs and driver. In particular having the charge and discharge resistors equal negates the advantages of the diodes.

Ok, I understand thank you



I made progress on the build, I finished putting the driver in an aluminum box and I put polycarbonate on the frame, here are some pictures and one that shows what it will look like :
I also added a signal led like the original ud 2.7c.





I don't know if I put too many pictures, but when I look at other people's builds, I like when there are lots of pictures, so I think other people must also like when there are lots pictures

[Manz]

I really appreciate all the details and pictures as I'm currently working on a similar sized coil
Did you buy the drivers aluminum box ?

[Mathieu thm]

I really appreciate all the details and pictures as I'm currently working on a similar sized coil

I'm glad it could help 

Did you buy the drivers aluminum box ?

I bought the aluminum box on amazon, here is the link :
https://www.amazon.fr/dp/B08QMGTLLZ/ref=sspa_dk_detail_6?psc=1&pd_rd_i=B08QMGTLLZ&pd_rd_w=KTm7I&content-id=amzn1.sym.a30983a5-b2e2-49b6-b674-0e622b88d06d&pf_rd_p=a30983a5-b2e2-49b6-b674-0e622b88d06d&pf_rd_r=XDV0Z03YX32QSGJRJ9CR&pd_rd_wg=yq5VN&pd_rd_r=657a4804-48c8-4164-873c-b7265ae7c832&s=industrial&sp_csd=d2lkZ2V0TmFtZT1zcF9kZXRhaWw&smid=A30I4SWEQ03TXF

[Manz]

Thanks.
Did you figure out the new mmc values as it was a bit too small for such a big coil?

[Mathieu thm]

Did you figure out the new mmc values as it was a bit too small for such a big coil?

Sure, at first I wanted to make a mmc with 10 Eurofarad SP2550 3.75uF 580Vrms in series but impossible to find and I think it would have been a bit expensive.
So I'm probably going to do a mmc with FKP1T031007E00, 4s15p or 5s20p

[Hydron]

Aha, glad to see someone is using some of the stuff I put on Github Really need to do some updates and add some new stuff there!

Mike is correct - this was designed for maximum flexibility, I've dug up a photo (attached) and I used 2*2R2 in parallel for turn off, and 2*10R in parallel for turn on. Those values were somewhat guesswork though, but they seem to work OK with the CM300s I was driving.

BTW I'd avoid using any of the other boards on my Github other than the "Big_DRSSTC..." ones (of which this is one of them), and I'd check the compatibility with the terminals on your brick before using either of them. The others have known issues or could be improved, but I have a coil running using the "Big_DRSSTC..." ones (with the bridge ones using a stack of thin boards as noted in the README.md on github.

[Mathieu thm]

Aha, glad to see someone is using some of the stuff I put on Github Really need to do some updates and add some new stuff there!

I was looking for a pcb for the gates of  CM300dy-24h, and this pcb is exactly what I was looking for, Thanks for designing them !

Mike is correct - this was designed for maximum flexibility, I've dug up a photo (attached) and I used 2*2R2 in parallel for turn off, and 2*10R in parallel for turn on. Those values were somewhat guesswork though, but they seem to work OK with the CM300s I was driving.

BTW I'd avoid using any of the other boards on my Github other than the "Big_DRSSTC..." ones (of which this is one of them), and I'd check the compatibility with the terminals on your brick before using either of them. The others have known issues or could be improved, but I have a coil running using the "Big_DRSSTC..." ones (with the bridge ones using a stack of thin boards as noted in the README.md on github.

The resistance values ​​are much better than what I started to put, I will surely change for these values

[Hydron]

The 5R (equivalent of the 2*10R in parallel) I used might be a bit higher than is ideal for fast turn-on; the 1R1 (from 2*2R2) used for turn-off is close to the datasheet nominal value they used for the switching times, which is why I chose it.

If you already have 5R6 resistors then I'd use one to begin with, and look at the turn-on speed, if you think it could be a bit faster (and you're not seeing nasty spikes on turn-on) then maybe try and add another in parallel to speed things up. This is where the separate resistors for turn-on/turn-off are nice, as they allow the values to be tweaked independently. BTW I suspect that the actual values aren't too critical as long as they are in the right ballpark (and turn-off is faster than turn-on to avoid shoot-through).

[Mathieu thm]

The 5R (equivalent of the 2*10R in parallel) I used might be a bit higher than is ideal for fast turn-on; the 1R1 (from 2*2R2) used for turn-off is close to the datasheet nominal value they used for the switching times, which is why I chose it.

If you already have 5R6 resistors then I'd use one to begin with, and look at the turn-on speed, if you think it could be a bit faster (and you're not seeing nasty spikes on turn-on) then maybe try and add another in parallel to speed things up. This is where the separate resistors for turn-on/turn-off are nice, as they allow the values to be tweaked independently. BTW I suspect that the actual values aren't too critical as long as they are in the right ballpark (and turn-off is faster than turn-on to avoid shoot-through).

Thanks for the advice, i will do it when i finish the coil, i still have a lot works on it !

BTW, do you know the exact model of your TVS? because I could see that my TVS that I used (the same as in the diagram) are 24v TVS but have a blocking voltage of 38.9 V, isn't that a bit high?

[Hydron]

Haha, you've just discovered the dirty secret of TVS diodes.

Sadly you have to specify them based on the peak expected working voltage of your circuit (as you have done here), but they will clamp well above that number with any appreciable current. This means that it's hard to protect anything that has a narrow range between operating and maximum voltage by using a TVS.

Happily the IGBT won't pop even with a 40V spike on the gate (unless you're extremely unlucky and have a dud).

I don't know the exact model I used, but I can pretty much guarantee that it won't be any better than yours!

[Mathieu thm]

Haha, you've just discovered the dirty secret of TVS diodes.

Sadly you have to specify them based on the peak expected working voltage of your circuit (as you have done here), but they will clamp well above that number with any appreciable current. This means that it's hard to protect anything that has a narrow range between operating and maximum voltage by using a TVS.

Happily the IGBT won't pop even with a 40V spike on the gate (unless you're extremely unlucky and have a dud).

I don't know the exact model I used, but I can pretty much guarantee that it won't be any better than yours!

I had several 24v through hole TVS that triggered around 27-28v so I thought it was the same for these TVS !

Some update since last post,

I made the primary coil out of 1/4 inch copper tubing (starts at around 200mm in diameter, spaced approx 15mm center-center) I still have to make the strike rail :



I also put two power supplies (12v and 24v) for the two 120mm high speed fans and the ud 2.7 :

[Mathieu thm]

Some news here !


I started making a laminated BUS on solidworks :






(I forgot some dimensions on the pdf, ask me if necessary)

Some specs :

• 1.5mm aluminum plate
• insulation with 1.2mm plexiglass
• 150A 1600V rectifier
• SKM300GB128D (I also have SKM400GB128D,I think the SKM300 are more than enough)

I found two large 6000µF 400v capacitors on Ebay and a 150a rectifier on Amazon :




Here are some pictures of the bridge (I still have to find snubber capacitors, probably 1000v 2µF or something similar EDIT : probably this cap : https://www.mouser.fr/ProductDetail/80-C4BSNBX4200ZALJ) :







I also found a sort of trick to either make a simple rectifier or make a voltage doubler, for a 325v BUS voltage I just have to wire the rectifier normally, or if I want a voltage doubler, I connect the two rectifier inputs together on one mains wire, and the other mains wire, I plug it in between the two big capacitors.
I could test more simply either with 325v or 650v by changing a wire.

The next step will be to find a heatsink and fit everything in the case.


EDIT : to take some pictures I quickly assembled the bridge and I just noticed that the capacitors are mounted upside down !

[Mathieu thm]

Hi,

I found a perfect heatsink for my build, I drilled holes for the two IGBT and the rectifier (and three at the bottom to mount the heatsink to the frame) :





To mount the two large capacitors I 3d printed two supports :



I definitely glued the laminated BUS plates for more robustness, and this is the final result :




The two snubber capacitors will probably be two KEMET 1000V 2uF

For the precharge resistor I think to use a resistor of 180ohms 25w.
For the balancing resistors that go on both capacitors, I calculated that 15Kohm 7w resistors will be fine.
And I would put a resistor of 10K 20w on the output of the bridge.

[Mathieu thm]

Hi everyone

It's been a long time since I've done an update.

I received the last components to finish the drsstc (MMC, CTs, GDT, snubber....), I started placing components in the enclosure :



I've finished the CTs (N30 900 to 1 ratio ):



I have a question about the GDT (N30 10 turns), I did it this way with cat5 cable by removing the main sheath to make jointed spir :



Do you think it's a good idea or is it better to stay with a classic GDT with a classic cat 5 cable ?
I'm afraid of insulation defaults with what I've done but there can't be a problem at these voltages?

[davekni]

Do you think it's a good idea or is it better to stay with a classic GDT with a classic cat 5 cable ?
I'm afraid of insulation defaults but there can't be a problem at these voltages?

The high voltage (H-bridge voltage) is between one wire and the other within each pair.  The outer jacket provides no improvement.  GDT use voltage is far above specification for such wire.  Generally not a problem for the short term use DRSSTCs experience (compared to specifications that cover 24 hours per day for years).  I haven't heard of any GDT insulation failures reported even for designs using 560Vbus (peak of 400Vac).
I recommend one unrelated change:  Leakage inductance will be lower if leads of each primary winding are twisted with themselves, separated only at the very end for connection to driver.  In other words, four white twisted pairs, joined at the end.  The four twisted pairs can be twisted together as a group or not.  That makes little difference.  Individual pair twisting is better than twisting the two bundles of four primary leads as shown now.

[Mathieu thm]

Hi !

I recommend one unrelated change:  Leakage inductance will be lower if leads of each primary winding are twisted with themselves, separated only at the very end for connection to driver.  In other words, four white twisted pairs, joined at the end.  The four twisted pairs can be twisted together as a group or not.  That makes little difference.  Individual pair twisting is better than twisting the two bundles of four primary leads as shown now.

Sorry for the late reply, you mean like this ? :




I have finished (almost, the strike rail and the gnd of the secondary coil remain) the primary circuit, here are some pictures :









So I was able to do a first "test", I made a ct  for the current sensor and connected my oscilloscope.
I did the test with a Vbus of 60-70 v (with three 6s Lipo in series), and this is what it looks like without lead phase adjustment (7m3 223) :







I can't find in a thread that talked about what resistance value to put on the currant CT to measure on the oscilloscope, i had some problem of bad phase reading on the oscilloscope because of this ressitance, i put a 10ohm and it seems to be right. but what resistance value should we put on the measurement CT ?

The next step will be more oscilloscope capture (gate etc), connect the gnd of the secondary coil and the strike rail and try to adjust the phase lead to then do the first test with the secondary coil.

[davekni]

Sorry for the late reply, you mean like this ? :

Yes.  Looks great.

I have finished (almost, the strike rail and the gnd of the secondary coil remain) the primary circuit, here are some pictures :

Looks like plenty of MMC caps, so hopefully they don't generate much heat.  Many designs include some gap between cap rows for cooling air flow.

I can't find in a thread that talked about what resistance value to put on the currant CT to measure on the oscilloscope, i had some problem of bad phase reading on the oscilloscope because of this ressitance, i put a 10ohm and it seems to be right. but what resistance value should we put on the measurement CT ?

10ohms is a common value for CT burden resistance.  Presume CT cores are appropriate ferrite, not powdered iron.  Ratio somewhere around 1000:1?
Too-high resistance has two possible issues.  One is excess power dissipation.  Other is short L/R time constant for CT output inductance L.  In other words, resistance becomes a significant fraction of CT output impedance at operating frequency.  For accuracy, R should be <1% of CT output impedance.
Main issue with too-low resistance is small signal amplitude.  Usually scope is also measuring bridge output at comparatively high voltage.  Can introduce measurement errors to current with any channel-to-channel coupling.  Other possible issue is resistor inductance.  Low value resistors are often wire-wound.  If impedance of resistor series inductance becomes more than ~1% of resistance, adds measurement error.

BTW, even with battery power, scoping is often more accurate measuring one bridge output at a time with scope probe ground to Vbus-.  Avoids rapid voltage swing between scope ground and H-bridge components (heat sink etc.).  Reduces inductive drop along scope probe ground leads etc.  May be fine as is, but can obscure subtle details of measurements.

[Mathieu thm]

Yes.  Looks great.

perfect

Looks like plenty of MMC caps, so hopefully they don't generate much heat.  Many designs include some gap between cap rows for cooling air flow.

I did some tests at 500a OCD and they are cold after a few minutes. From what I've calculated, they'll heat up very little.
BTW the drsstc draws about 20a max at about 130-140Vac from the variac and with the OCD set to 500a (which doesn't trip very often), the arcs are about 1m20.

Ratio somewhere around 1000:1?

the ratio is 900:1 for all three CTs.

Too-high resistance has two possible issues.  One is excess power dissipation.  Other is short L/R time constant for CT output inductance L.  In other words, resistance becomes a significant fraction of CT output impedance at operating frequency.  For accuracy, R should be <1% of CT output impedance.
Main issue with too-low resistance is small signal amplitude.  Usually scope is also measuring bridge output at comparatively high voltage.  Can introduce measurement errors to current with any channel-to-channel coupling.  Other possible issue is resistor inductance.  Low value resistors are often wire-wound.  If impedance of resistor series inductance becomes more than ~1% of resistance, adds measurement error.

Thanks for the explanation, it's always important to understand problems.

I've finally decided to insulate the upper part of the frame, i've printed 3d parts and it seems to do the trick, but i know it's not ideal :
There are four spacers for the screws, which also go through the aluminum, kapton under the aluminum square and two black insulating pieces between the 20x20 profiles, all 3d printed.


I made several oscilloscope captures (Vbus at 90-100v with battery) :

• bridge output
• bridge output with gnd at Vbus-
• Vge
• primary current

OVERVIEW (bridge output = yellow and primary current = purple) :



Start of pulses :



End of pulses :


 
At the end of the pulses with the transition when the freewheel diodes of IGBTs take over :



Both sides of the bridge with the oscilloscope's gnd to Vbus- :



Vge of one IGBT :



And primary current :



There's something bothering me, i've set the phase lead to have the least amount of ringing and spiking on the bridge output, but I get the impression that there's a lot of phase lead.
What do you think? should we set less phase lead? is setting for less ringing not a good solution?

Sorry for the dumb question, but I'd like to know more about how to get a perfect phase lead (or at least the best possible setting).
I'd like to be sure of my phase lead before testing at full power.

[Mads Barnkob]

You should adjust your phase lead for somewhere between zero-current-switching and least-possible-spikes. If you only adjust for not spikes, you risk having high dissipation losses from turn-on or turn-off losses being higher than it could be.

Regarding your MMC, I am afraid that you have to do something about it. You need active cooling on it, with a fan, forced air cooling. You need to make room between acrylic side and MMC for air to flow on both sides.

Capacitor dissipate 2/3 of their losses through the terminals and 1/3 through the body. This makes your heat dissipation very uneven in the very tight configuration. This will lead to hot spots and that is bad for current sharing. I predict that your most inner/middle capacitors will eventually fail.

"Microsoft Paint Simulation" of heat dissipation in your MMC.

[davekni]

There's something bothering me, i've set the phase lead to have the least amount of ringing and spiking on the bridge output, but I get the impression that there's a lot of phase lead.
What do you think? should we set less phase lead? is setting for less ringing not a good solution?

Sorry for the dumb question, but I'd like to know more about how to get a perfect phase lead (or at least the best possible setting).
I'd like to be sure of my phase lead before testing at full power.

The "ideal" situation is to have IGBTs turn off with just enough remaining current to cause bridge output voltage to transition due to just that remaining current.  Ideally, opposite IGBTs turn just as output voltage transition completes, which would be just as current crosses zero.
Of course, no real build can be "ideal", especially since IGBT delays increase with temperature.
You may have more dead-time than necessary (time between IGBT turn-off and opposite IGBT turn-on).  That may be why phase lead needs to be high.  (Doesn't look bad to me as-is.)  What does Vce (output) look like with slightly less phase lead?  I'm guessing what I refer to here as triple-transition.  Vce rises, then starts to fall after current crosses zero, then Vce rises again as opposite IGBT turns on.  As Mads pointed out, a little bit of that is fine.  To much triple-transition (or past that with even less phase lead) and you may add as much power dissipation there as you save by lower turn-off current.

I did some tests at 500a OCD and they are cold after a few minutes. From what I've calculated, they'll heat up very little.
BTW the drsstc draws about 20a max at about 130-140Vac from the variac and with the OCD set to 500a (which doesn't trip very often), the arcs are about 1m20.

How do those conditions compare to your intended operating conditions?  Cap heating will be proportional to duty cycle and to current squared.  If you end up needing just a bit more cooling, say 1.5x, thick copper foil (0.1-0.2mm) strips could be soldered across cap middle wires to provide crude heat-sink fins.  That and some fan air flow will help noticeably.

[Mathieu thm]

You should adjust your phase lead for somewhere between zero-current-switching and least-possible-spikes. If you only adjust for not spikes, you risk having high dissipation losses from turn-on or turn-off losses being higher than it could be.

The "ideal" situation is to have IGBTs turn off with just enough remaining current to cause bridge output voltage to transition due to just that remaining current.  Ideally, opposite IGBTs turn just as output voltage transition completes, which would be just as current crosses zero.
Of course, no real build can be "ideal", especially since IGBT delays increase with temperature.
You may have more dead-time than necessary (time between IGBT turn-off and opposite IGBT turn-on).  That may be why phase lead needs to be high.  (Doesn't look bad to me as-is.)  What does Vce (output) look like with slightly less phase lead?  I'm guessing what I refer to here as triple-transition.  Vce rises, then starts to fall after current crosses zero, then Vce rises again as opposite IGBT turns on.  As Mads pointed out, a little bit of that is fine.  To much triple-transition (or past that with even less phase lead) and you may add as much power dissipation there as you save by lower turn-off current.

I set the phase lead like this to have less dissipation losses, I think this setting is better, what do you think ? :



pulse start :



end of pulses :



on both sides of the output with gnd to Vbus- :




The "interference" on the primary current signal when current crosses 0 is due to what?
parasitic inductance of the oscilloscope probe or something like that?

Regarding your MMC, I am afraid that you have to do something about it. You need active cooling on it, with a fan, forced air cooling. You need to make room between acrylic side and MMC for air to flow on both sides.

Capacitor dissipate 2/3 of their losses through the terminals and 1/3 through the body. This makes your heat dissipation very uneven in the very tight configuration. This will lead to hot spots and that is bad for current sharing. I predict that your most inner/middle capacitors will eventually fail.

"Microsoft Paint Simulation" of heat dissipation in your MMC.

I'm well aware that the MMC is not designed in the best way but good idea I'll leave some space between the acrylic and the MMC for cooling.

How do those conditions compare to your intended operating conditions?  Cap heating will be proportional to duty cycle and to current squared.  If you end up needing just a bit more cooling, say 1.5x, thick copper foil (0.1-0.2mm) strips could be soldered across cap middle wires to provide crude heat-sink fins.  That and some fan air flow will help noticeably.

I think I'll set the OCD in the 700-750a range.

[davekni]

I set the phase lead like this to have less dissipation losses, I think this setting is better, what do you think

Looks great!  Definitely not excess dead time as I'd speculated previously.  Third scope capture, final falling edge before enable ends, shows a very brief pause center-transition.  Looks like turn-off is early enough to cause half of the output voltage swing, then opposite IGBT turn-on pulls the output the remainder of the way down.  If there was excess dead time, that brief pause would rise briefly before falling (triple transition).
The one possible concern:  When coil is driven hard and IGBTs get hot, IGBT delay increases.  I personally prefer to start with a bit too much phase lead, enough to compensate for increased delay due to temperature.  Before making any change, best to scope at higher bus voltage and higher current.  IGBT capacitance is lower at high voltage, and higher current will cause faster bridge output transition at turn-off.  So you may have some margin (some "excess" phase lead) at operating conditions.
However, I have relatively little experience with IGBT bricks, mostly smaller TO247 devices.  So if Mads or others with more experience with bricks says otherwise, follow their advice.

The "interference" on the primary current signal when current crosses 0 is due to what?
parasitic inductance of the oscilloscope probe or something like that?

Such interference is difficult to avoid.  CT signals are relatively low voltage compared to bridge output swings.  Capacitance between CT winding and bridge output wire through center will inject some signal.  Can be mitigated, but usually not worth the work.  Grounded shield around bridge output wire inside CT core can help, though inductance in that ground lead can limit effectiveness.  Common-mode choke in CT output helps.  CT can be wound as a center-tapped coil, with center grounded and one half used for scoping, other half "unused".  If coil is symmetric, capacitance to both winding halves cancels.

[Mathieu thm]

Looks great!  Definitely not excess dead time as I'd speculated previously.  Third scope capture, final falling edge before enable ends, shows a very brief pause center-transition.  Looks like turn-off is early enough to cause half of the output voltage swing, then opposite IGBT turn-on pulls the output the remainder of the way down.  If there was excess dead time, that brief pause would rise briefly before falling (triple transition).
The one possible concern:  When coil is driven hard and IGBTs get hot, IGBT delay increases.  I personally prefer to start with a bit too much phase lead, enough to compensate for increased delay due to temperature.  Before making any change, best to scope at higher bus voltage and higher current.  IGBT capacitance is lower at high voltage, and higher current will cause faster bridge output transition at turn-off.  So you may have some margin (some "excess" phase lead) at operating conditions.
However, I have relatively little experience with IGBT bricks, mostly smaller TO247 devices.  So if Mads or others with more experience with bricks says otherwise, follow their advice.

Ok I see thanks.
Yes, of course I have to try with higher bus voltages and higher primary current, but I don't have a differential probe.
I've tried this technique, but I can't get a clear oscilloscope capture : https://highvoltageforum.net/index.php?topic=1373.msg10212#msg10212

Such interference is difficult to avoid.  CT signals are relatively low voltage compared to bridge output swings.  Capacitance between CT winding and bridge output wire through center will inject some signal.  Can be mitigated, but usually not worth the work.  Grounded shield around bridge output wire inside CT core can help, though inductance in that ground lead can limit effectiveness.  Common-mode choke in CT output helps.  CT can be wound as a center-tapped coil, with center grounded and one half used for scoping, other half "unused".  If coil is symmetric, capacitance to both winding halves cancels.

Okay, I understand, so this interference isn't a problem.

[davekni]

Ok I see thanks.
Yes, of course I have to try with higher bus voltages and higher primary current, but I don't have a differential probe.
I've tried this technique, but I can't get a clear oscilloscope capture : https://highvoltageforum.net/index.php?topic=1373.msg10212#msg10212

Some digital scopes capture multiple waveforms and display the composite on screen (simulates behavior of an analog scope).  That may be making a confusing display result.  Either turn that mode off if possible, or go into single-capture mode.  Single-capture mode requires pushing run/stop button for every new capture, but should give you a stable view of each capture.

Okay, I understand, so this interference isn't a problem.

For scope measurements, just ignore the interference.  Drivers such as UD2.7 have a small capacitor across CT to filter out such high-frequency noise.  Could be a problem if using a driver that doesn't filter out high frequency noise.

[Mathieu thm]

Hello everyone !

This thread needs a little updating.
I made some aesthetic additions and made a support for the two CTs to secure them so they don't move on the wire :



I've also made a "wattmeter", which is easier to see the consumption :



Now for the most interesting part, the tests. I'd already done a few tests, but they weren't very interesting to show because they were just tuning tests.
Here's a video of the DRSSTC operating at 170Vac, 200µs on-time and 600A OCD if I remember correctly :

/>
and a MIDI test at 170Vac, 200µs, and 600A OCD :

/>
During these tests, everything remained cold, only the wires connecting the primary coil to the MMC and IGBTs were barely warm.

The next step, I hope, will be to test it outdoors in two weeks.