RDRSSTC - Project Build

[ZakW]

Hello,

While I am still wrapping up my RSSTC I'd like to start planning my next build. For that I want to step up to DRSSTC territory. I want this to be a ramped coil as well that will hopefully help prepare me for a QCWDRSSTC in the future.


High level plan:

• 120v AC input
• Full bridge using FGH75T65SHD https://www.mouser.com/ProductDetail/512-FGH75T65SHD_F155
• UD1.3 driver (modified)
• 360-450kHz operation

My aim is a compact design with the best secondary to arc length ratio.


Driver Questions:

• Considering the UD1.3b, how crucial is phase lead? I've noticed it in later UD versions. Should I integrate it into UD1.3? https://kaizerpowerelectronics.dk/files/stevehv.4hv.org_files/universal_driver_1.3b/DRSSTC_pndriver1_3b.pdf
• For a more compact RDRSSTC, would driving the IGBTs at around 18-20v vs. the UCC27423's 15v max make a significant difference, considering the improvement in current capabilities?
• Can I bypass the push/pull MOSFET output stage and simply use two UCC27423s, given that a single UCC27425 was adequate for my half bridge RSSTC?
• Are there notable differences between regular and ramped DRSSTCs in terms of component stress and operation, aside from the arc appearance?

Thank you!
 

[Lucasww]

Phase lead will be very helpful if you are planning to run long on-times, which you will be doing with a ramped coil.

With 15v input on your GDT, assuming 1:1 turns ratio, you will realistically be able to get 13-14v consistently on your igbt gates. The FGH75T65SHD datasheet shows that they should be fine at that voltage. However, what concerns me more is the possibility of ringing on your gates causing the voltage to drop potentially below 10v, which could easily cause desaturation. This isn't as much of an issue with the typical 18-24v gate drive most use. I'm using FGH75T65SHD in my QCWDRSSTC and have had no issues driving them at 19v.

Whether or not you can bypass the output mosfets is really up to what gate drive voltage you want. If you want the voltage to be above 15v, you will either need the mosfets, or a higher voltage gate drive chip. Most higher voltage chips with suitable rise times and propagation delays are SMD and quite small, so you may have trouble keeping them cool. I would recommend just using the mosfets.

As far as operation and component stress, You will generally want higher coupling, and a higher impedance primary circuit. The stress on your components ultimately depends on tuning, current and on-time. Properly tuned phase lead will greatly reduce switching stress, especially with long on-time.

[ZakW]

Hello Lucasww, I appreciate the insight!

Phase lead will be very helpful if you are planning to run long on-times, which you will be doing with a ramped coil.

Thank you for confirming that. I will see about adding it. I assume I can copy the phase lead portion from the UD2.7 into the UD1.3?

I would like to build the UD1.3 from scratch, is that typically not advised? It would be a good chunk of work but give me a lot of experience using PCB software, I have an okay foundation. Ideally, I would have the whole DRSSTC on one PCB or at most two. Similar to Loneoceans full bridge SSTC build on a single PCB.

With 15v input on your GDT, assuming 1:1 turns ratio, you will realistically be able to get 13-14v consistently on your igbt gates. The FGH75T65SHD datasheet shows that they should be fine at that voltage. However, what concerns me more is the possibility of ringing on your gates causing the voltage to drop potentially below 10v, which could easily cause desaturation. This isn't as much of an issue with the typical 18-24v gate drive most use. I'm using FGH75T65SHD in my QCWDRSSTC and have had no issues driving them at 19v.

Looking back on some scope captures of my gate drive signal from my SSTC using a UCC27425 at 15V Vcc I was getting 15.6-16V on the gate. I used a commercial 1:1 GDT.  Not quite as low as you mentioned but it sounds like a good idea to include the push/pull stage to achieve a higher gate drive voltage. I will aim for 18v or so.

As far as operation and component stress, You will generally want higher coupling, and a higher impedance primary circuit. The stress on your components ultimately depends on tuning, current and on-time. Properly tuned phase lead will greatly reduce switching stress, especially with long on-time.

I have read that higher coupling and impedance is beneficial. Is a higher impedance achieved by more primary turns and a smaller resonant capacitor?

Good to hear that phase lead is so effective!


Thanks again!

[Lucasww]

I would like to build the UD1.3 from scratch, is that typically not advised? It would be a good chunk of work but give me a lot of experience using PCB software, I have an okay foundation. Ideally, I would have the whole DRSSTC on one PCB or at most two. Similar to Loneoceans full bridge SSTC build on a single PCB.

Building a board from scratch is fine, although it is quite a bit more work, as you mentioned. make sure you use large ground planes to help avoid interference, which is an issue when driving tesla coils.

Looking back on some scope captures of my gate drive signal from my SSTC using a UCC27425 at 15V Vcc I was getting 15.6-16V on the gate. I used a commercial 1:1 GDT.  Not quite as low as you mentioned but it sounds like a good idea to include the push/pull stage to achieve a higher gate drive voltage. I will aim for 18v or so.

Interesting that you got higher voltages on the gate than you have on the input. Still, the mosfet push/pull buffer is a good idea. 18v should be perfectly fine.

I have read that higher coupling and impedance is beneficial. Is a higher impedance achieved by more primary turns and a smaller resonant capacitor?

Yes. For QCW coils with >20ms on-time people usually use 7-15nF capacitance, and whatever inductance gives the right frequency. For a mains-ramped coil, your ontime will be probably under 5ms, so slightly more capacitance may be good to get a bit more current. Maybe somewhere around 20-30nF. Typical DRSSTCs use much higher capacitances like 100-500nF

[Mads Barnkob]

Looking back on some scope captures of my gate drive signal from my SSTC using a UCC27425 at 15V Vcc I was getting 15.6-16V on the gate. I used a commercial 1:1 GDT.  Not quite as low as you mentioned but it sounds like a good idea to include the push/pull stage to achieve a higher gate drive voltage. I will aim for 18v or so.

Interesting that you got higher voltages on the gate than you have on the input. Still, the mosfet push/pull buffer is a good idea. 18v should be perfectly fine.

A high voltage could indicate that you have ringing being rectified or your measurements are not fast enough to capture the ringing.

Why not drive the gates even harder, if you are worried about switching losses? Usually 24VDC is used in the UD1.x/UD2.x drivers. Limiting gate voltage in commercial applications is a way to limit the C-E maximum current, we are beyond trying to nurse our IGBTs like that.

[ZakW]

Building a board from scratch is fine, although it is quite a bit more work, as you mentioned. make sure you use large ground planes to help avoid interference, which is an issue when driving tesla coils.

I already started, much more complicated than my last SSTC. Going to be a lot of work.

Yes. For QCW coils with >20ms on-time people usually use 7-15nF capacitance, and whatever inductance gives the right frequency. For a mains-ramped coil, your ontime will be probably under 5ms, so slightly more capacitance may be good to get a bit more current. Maybe somewhere around 20-30nF. Typical DRSSTCs use much higher capacitances like 100-500nF

Good to know, I will keep this in mind!

A high voltage could indicate that you have ringing being rectified or your measurements are not fast enough to capture the ringing.

Interesting that you got higher voltages on the gate than you have on the input. Still, the mosfet push/pull buffer is a good idea. 18v should be perfectly fine.

Regarding the higher gate voltage. I attached a picture of my SSTC gate waveform. It is at 15.6V so probably a little bit of ringing, other than that I think it looks great.

Why not drive the gates even harder, if you are worried about switching losses? Usually 24VDC is used in the UD1.x/UD2.x drivers. Limiting gate voltage in commercial applications is a way to limit the C-E maximum current, we are beyond trying to nurse our IGBTs like that.

Haha great point, Mads! I was planning on driving them at a higher voltage but just never landed on a value. I agree though, no real point in going easy on them given how they are being used.

[Lucasww]

Waveforms look great. You might be able to get away with slightly less gate resistance for faster transitions, but it's probably perfectly good as is, especially if you decide to raise the voltage

[ZakW]

Waveforms look great. You might be able to get away with slightly less gate resistance for faster transitions, but it's probably perfectly good as is, especially if you decide to raise the voltage

Thank you. Just to clarify that scope capture was the gate drive from my Ramped SSTC that I was referencing when looking at potentially excluding the push/pull buffer stage from the UD1.3.

For a coil like this, providing a higher gate drive voltage seems to be the best way to go. I will be sure to include the push/pull buffer so I can increase the voltage.

[ZakW]

Hello,

Picked this project back up and have a few questions before I start working on the PCB.

1. Half wave vs full wave rectification for the bridge - I used a single diode in my RSSTC for halfwave rectification for the bridge. It was a SMD 600v 10A 100A peak standard diode, never had any failures with it. Since this is a Ramped DRSSTC currents should be much higher. My assumption is that a full bridge part/package would be better suited to handle the power and heat dissipation, could I also get away with using a beefier single diode instead? Although, I never did have a heating issue with using a single diode in my RSSTC. I am not sure which solution is better.
       - If I use a full bridge but my ZCD is only half wave rectification via a single 1N4148, would there be an issue with the bridge being powered from both cycles instead of just the half cycles like with a single diode? The ZCD wont trigger and power the
               bridge on for all cycles since it is only outputting a signal on the positive half cycle, which is fine. I only run the interrupter at low BPS, I don't care much for higher BPS.

2. I am aiming for as little bridge inductance as possible but I would still like to include TVS diodes across CE. Are two 1.5kW 220V in series better or can I get away with a single diode 1.5kW 400v type for each?
       - 220v https://www.mouser.com/ProductDetail/652-1.5SMC220CA
       - 400v https://www.mouser.com/ProductDetail/576-1.5SMC400CA

3.Since I am trying to simplify the design a bit I really just added the phase adjust to the UD 1.3b. However, the push/pull transistor stage parts after the UCC27423 are outdated. I have seen other use IRF530/IRF9530, 100v 17A parts but I don't think I need such highly rated MOSFETs. Plus they have much high Qg compared to lower voltage parts. While I have found other parts with better characteristics https://www.mouser.com/ProductDetail/Diodes-Incorporated/DMN6068LK3-13?qs=4YT44p7w2bsq47X783VrzA%3D%3D it is hard to find if they have a PNP counterpart, like the IRF530 vs the IRF9530. Does anyone know of a way to identify a close alternative for a complimentary NPN/PNP stage?
       - Most PNP parts I have found differ slightly, is this expected and still usable or should Vdss and Id match a close as possible?
       - Example of something that might match the PNP DMN6068LK3 above -  https://www.mouser.com/ProductDetail/Diodes-Incorporated/ZXMP6A16KTC?qs=7EMYER6HMn3fGPb2RALoHA%3D%3D

4.Finally, for my last few builds I have cascaded the logic power section instead of tying each linear regulator to DC input (~22v). I guess it comes down to causing more heating in the main 20v regulator and less in the 9v and 5v. With this method I haven't noticed excessive heating in the first regulator. I also have small SMD heatsinks I can also stick on if I need to help dissipate heat. Is there a 'right' way to do it?

Thanks,
Zak

[ZakW]

1. Half wave vs full wave rectification for the bridge - I used a single diode in my RSSTC for halfwave rectification for the bridge. It was a SMD 600v 10A 200A peak standard diode, never had any failures with it. Since this is a Ramped DRSSTC currents should be much higher. My assumption is that a full bridge part/package would be better suited to handle the power and heat dissipation, could I also get away with using a beefier single diode instead? Although, I never did have a heating issue with using a single diode in my RSSTC. I am not sure which solution is better.
       - If I use a full bridge but my ZCD is only half wave rectification via a single 1N4148, would their be an issue with the bridge being powered from both cycles instead of just the half cycles like with a single diode? The ZCD wont trigger and power the
               bridge on for all cycles since it is only outputting a signal on the positive half cycle, which is fine. I only run the interrupter at low BPS, I don't care much for higher BPS.

Going to just use the same single diode and I will see how it goes.

2. I am aiming for as little bridge inductance as possible but I would still like to include TVS diodes across CE. Are two 1.5kW 220V in series better or can I get away with a single diode 1.5kW 400v type for each?
       - 220v https://www.mouser.com/ProductDetail/652-1.5SMC220CA
       - 400v https://www.mouser.com/ProductDetail/576-1.5SMC400CA

Looks like others have used a single 400v TVS. I will do the same.

3.Since I am trying to simplify the design a bit I really just added the phase adjust to the UD 1.3b. However, the push/pull transistor stage parts after the UCC27423 are outdated. I have seen other use IRF530/IRF9530, 100v 17A parts but I don't think I need such highly rated MOSFETs. Plus they have much high Qg compared to lower voltage parts. While I have found other parts with better characteristics https://www.mouser.com/ProductDetail/Diodes-Incorporated/DMN6068LK3-13?qs=4YT44p7w2bsq47X783VrzA%3D%3D it is hard to find if they have a PNP counterpart, like the IRF530 vs the IRF9530. Does anyone know of a way to identify a close alternative for a complimentary NPN/PNP stage?
       - Most PNP parts I have found differ slightly, is this expected and still usable or should Vdss and Id match a close as possible?
       - Example of something that might match the PNP DMN6068LK3 above -  https://www.mouser.com/ProductDetail/Diodes-Incorporated/ZXMP6A16KTC?qs=7EMYER6HMn3fGPb2RALoHA%3D%3D

Sounds like the IRF530 vs the IRF9530 pair work well. I might try and find complimentary parts with similar but better characteristics if possible.

4.Finally, for my last few builds I have cascaded the logic power section instead of tying each linear regulator to DC input (~22v). I guess it comes down to causing more heating in the main 20v regulator and less in the 9v and 5v. With this method I haven't noticed excessive heating in the first regulator. I also have small SMD heatsinks I can also stick on if I need to help dissipate heat. Is their a 'right' way to do it?

I am just going to cascade the regulators and add a small SMD heatsink to the 20v regulator if needed.


Finished my first pass on the UD 1.3b with added phase lead. I incorporated the staccato interrupter onto the same PCB. Finally, I redesigned the full bridge layout to be similar to Loneoceans easy bridge setup. IGBTs will lay flat and mount to the top of the heatsink.

Here are my designs so far. Feedback would be appreciated. I still need to go through my schematic a few extra times to ensure I did not make a mistake.



















-Zak

[davekni]

Generally looks good.  You may want to add tape where IGBT leads cross other planes.  Solder mask by itself may be good enough, but not reliable at high voltages.

If you want to use higher current bridge rectifier, use it as a single diode.  Two bridge diodes can be paralleled to double current.  Wire bridge's two AC terminals together.  Use that and either + or - output (but not both) as the two diode leads.

[ZakW]

Generally looks good.  You may want to add tape where IGBT leads cross other planes.  Solder mask by itself may be good enough, but not reliable at high voltages.

Thanks, Dave. The IGBTs will be mounted either above or below the PCB (limitation of the 3d model) so the leads will be curved to allow some clearance from the PCB. 

If you want to use higher current bridge rectifier, use it as a single diode.  Two bridge diodes can be paralleled to double current.  Wire bridge's two AC terminals together.  Use that and either + or - output (but not both) as the two diode leads.

There isn't really enough room for a full bridge, worse case I could also stack two single diodes on one another to increase current capabilities. I did also find higher current diodes in the same package so I am not too worried.

[ZakW]

Alright, modified UD 1.3b is assembled and so is the new bridge. I have been working like a mad man the last couple of days on this.









Nailed the 3D printed clamp first try. Printed with 100% infill and seems plenty rigid, doesn't flex as far as I can tell. Had to redo the heatsink and ended up making the IGBT leads/spacing a little bit too long but I already have a couple things I want to change anyway.


Testing and first issue:

Powered up the driver with the GDT connected to the full bridge. Signal generator connected to + pin of the CT feedback, running at 400kHz. At 25v it draws around 0.11A, nothing gets hot. Onboard interrupter is working. The damn OCD LED is stuck on though...I have tried adjusting the variable 10k pot but I cant seem to get it to turn off.

Any ideas as to why it would be on as soon as it is powered up? I am not supplying any power to the full bridge yet. Retracing the 1.3b driver so far, I have matched the sections exactly in my schematic. I will continue probing around to see if I can figure anything out in the mean time. I am just not very familiar with this driver at all.

EDIT: Fixed the issue. Turns out I was missing a ground connection for the OCD feedback input section. Soldered a small jumper and the light is off now! I will start testing other parts of the driver now that everything isn't disabled.


Thanks!


Alright, not out of the woods yet. I think I made a mistake in my PCB. Here are the schematics I referenced, 1.3b and 2.1c



Links to both:
1.3b - https://kaizerpowerelectronics.dk/files/stevehv.4hv.org_files/universal_driver_1.3b/DRSSTC_pndriver1_3b.pdf
2.1b - https://kaizerpowerelectronics.dk/files/stevehv.4hv.org_files/universal_driver_2.1b/UD2_1revbschem.pdf

When I was looking at how I was going to add phase lead into the 1.3b schematic I excluded the 74hc14 inverting stage, circled in red in the 1.3b schematic. Looking at the 2.1b there is an extra inverting stage, also circled in red. I added all of the highlighted section in 2.1b nothing else. Scoping the output of my driver it looks like it is constantly running and stops only when the interrupter pulses, which is the opposite of what it should be doing. Seems like the missing inverting stage is the cause.

Here is my schematic for reference:



I would really appreciate if someone would be willing to take a look and confirm my error.

If I did miss a stage I might be able to cut some traces and use jumper wires to at least correct the issue and test the board. I can then work on fixing the issues and eventually ordering another one.

[davekni]

Nailed the 3D printed clamp first try. Printed with 100% infill and seems plenty rigid, doesn't flex as far as I can tell. Had to redo the heatsink and ended up making the IGBT leads/spacing a little bit too long but I already have a couple things I want to change anyway.

Clamps look great!
Yes, lead spacing is long and likely to cause problems.  I'd mount IGBT bodies up against ECB edges to get leads as short as possible.  Given ECB layout, leads can't be as short as would be ideal (ie. with pads at ECB edges).

Fixed the issue. Turns out I was missing a ground connection for the OCD feedback input section. Soldered a small jumper and the light is off now! I will start testing other parts of the driver now that everything isn't disabled.

Ground should be as close to a solid copper plane as possible.  If a section is floating, I'd connect it with several wires at different locations to be a closer approximation to a ground plane.

Overall looks like great progress.

[AstRii]

What is the purpose of the 3D printed clamp? You could mount the transistors directly to the heatsink. That extra metal pad between IGBT and thermal pad also seems unnecessary. It's all adding high thermal resistance and so the cooling will be worse.
I also believe a thermal pad on its own is not good enough, it should be used in combination with thermal paste (maybe I'm wrong on this?)

[ZakW]

Thanks, Dave! Looks like I missed your response while I was editing my last post with another issue.

Yes, lead spacing is long and likely to cause problems.  I'd mount IGBT bodies up against ECB edges to get leads as short as possible.  Given ECB layout, leads can't be as short as would be ideal (ie. with pads at ECB edges).

Ground should be as close to a solid copper plane as possible.  If a section is floating, I'd connect it with several wires at different locations to be a closer approximation to a ground plane.

Noted, I will make these changes to the next version.

What is the purpose of the 3D printed clamp? You could mount the transistors directly to the heatsink. That extra metal pad between IGBT and thermal pad also seems unnecessary. It's all adding high thermal resistance and so the cooling will be worse.
I also believe a thermal pad on its own is not good enough, it should be used in combination with thermal paste (maybe I'm wrong on this?)

The clamps are better for distributing the force of the mounting screw to the body of the IGBT or MOSFET for increased thermal transfer. The screw, if too tight, can cause the IGBT to even lift a bit too. The small aluminum pieces are there for thermal mass to absorb rapid spikes in temperature due to the ramped nature of the coil, the heat is then more slowly distributed to the larger heatsink. I do have thermal paste on the IGBT to the aluminum pieces.

[davekni]

The small aluminum pieces are their for thermal mass to absorb rapid spikes in temperature due to the ramped nature of the coil, the heat is then more slowly distributed to the larger heatsink. I do have thermal paste on the IGBT to the aluminum pieces.

Thought about suggesting this on my previous reply:  The "small" aluminum pieces would be better if larger, and even better yet if copper.  If necessary to shorten leads, they could be thicker.  Thickness is a trade-off.  Increases thermal mass and reduces thermal spreading resistance, but does also increase vertical thermal resistance a bit.  Overall I'd guess somewhat thicker would be net benefit for thermals besides allowing shorter leads.  Larger area would make the most improvement.
Are your thermal pads compliant?  If rigid such as mica or ceramic, then thermal paste is needed on each side of pads too.

Looking at the 2.1b their is an extra inverting stage, also circled in red. I added all of the highlighted section in 2.1b nothing else. Scoping the output of my driver it looks like it is constantly running and stops only when the interrupter pulses, which is the opposite of what it should be doing. Seems like the missing inverting stage is the cause.

That "also circled in red" inverter appears to be a limit on maximum on-time and duty cycle.  Would cause trouble for a coil running 5ms on-time.  Sounds like you have another inversion mistake somewhere, perhaps a FF Q instead of QN output or such.  I have a hard time following schematics drawn without normal FF and gate symbols, so not sure where your inversion error is.  (BTW, I moved to KiCad from DesignSpark largely because KiCad has better symbols available.)

[ZakW]

Thought about suggesting this on my previous reply:  The "small" aluminum pieces would be better if larger, and even better yet if copper.  If necessary to shorten leads, they could be thicker.  Thickness is a trade-off.  Increases thermal mass and reduces thermal spreading resistance, but does also increase vertical thermal resistance a bit.  Overall I'd guess somewhat thicker would be net benefit for thermals besides allowing shorter leads.  Larger area would make the most improvement.

Tried using what I had on hand. If it gets too hot I can always add thicker aluminum.

Are your thermal pads compliant?  If rigid such as mica or ceramic, then thermal paste is needed on each side of pads too.

I am using soft silicone pad material.

That "also circled in red" inverter appears to be a limit on maximum on-time and duty cycle.  Would cause trouble for a coil running 5ms on-time.  Sounds like you have another inversion mistake somewhere, perhaps a FF Q instead of QN output or such.  I have a hard time following schematics drawn without normal FF and gate symbols, so not sure where your inversion error is.  (BTW, I moved to KiCad from DesignSpark largely because KiCad has better symbols available.)

I use KiCad as well. I have a plug in that I use to download footprints and 3D models directly from Mouser, it is really useful and streamlined. Download the file and it auto adds it to the library.

I will look through my schematic some more tomorrow. Been working on this for about 9 hours straight today.

Update:

Using my function generator I input a 200kHz signal into the CT input and measured the interrupter output as well as the driver output. Confirmed that the coil is running until the interrupter signal starts which is backwards.

Yellow = GDT output
Purple = Interrupter output

[flyingperson23]

The small aluminum pieces look very close to the heatsink and are electrically connected to the igbt. Maybe make the thermal pad a little bigger than the aluminum on all sides so it doesn't arc over like in https://highvoltageforum.net/index.php?topic=2643.0

[ZakW]

The small aluminum pieces look very close to the heatsink and are electrically connected to the igbt. Maybe make the thermal pad a little bigger than the aluminum on all sides so it doesn't arc over like in https://highvoltageforum.net/index.php?topic=2643.0

Good call, thanks for pointing that out. Better safe than sorry. The pack of sheets I have come in varying thicknesses I will get a thicker sheet with more clearance.