QCWDRSSTC - Project Build

[ZakW]

Hello Dave,

If I'm understanding your design correctly, this supply is for gate drive power.  Using typical values for your IGBTs, gate charge is ~200nC each at 24V, so 400nC total.  30kHz * 400nC = 12mA.  Gate drive supply average current should be 12mA plus current consumed by driver chip itself, which should be available in its data sheet.

Thanks for that, I will double check my IGBTs for their specific values and run through the calculations again so I can better familiarize myself with them. I will also double check the UCC5390E datasheet.

Does your buck converter switch high side (positive supply rail) or low side (negative supply rail)?

I am planning to switch high side like in this schematic.

If switching positive rail (as is more common), then emitter voltage includes fast switching edges.  Standard line-powered isolated supplies work only if there is enough stray inductance in wiring to filter high edge slew rates.  Adding a common-mode choke to a standard supply would likely work, at least much better than relying on wiring inductance alone

That makes sense, thanks for breaking it down more.

Or as suggested, use a small supply designed for gate drive (designed to handle high slew rate edges of switching waveforms)

I agree, that is sounding like the right thing to do. I appreciate everyone's input regarding the power supply, suggesting alternatives, and being patient as I work through the advice. Glad I asked! I will take a look at the recommend alternatives and work on a new design.

If designing a continuous-use buck converter, E-cores have a disadvantage.  Wire experiences much more proximity-effect loss than for toroid cores (more magnetic field attempting to cross through wire, causing eddy current losses).  Toroid coils, especially ones with single layer windings, have much less proximity-effect loss.  Stranded wire also has a disadvantage of increased skin-depth loss.  (Unless strands are individually insulated.  In that case, called litz wire, losses are lower.)

That is good to know, I will keep that in mind. Given the increased skin-depth loss of stranded wire, would simply using 12awg solid perform better? I will likely see how the stranded wire works, but just curious if I do decide to rewind it later on.

-Zak

[Anders Mikkelsen]

One more question regarding these power supplies - the power output is so low at only 2.4W.

If I'm understanding your design correctly, this supply is for gate drive power.  Using typical values for your IGBTs, gate charge is ~200nC each at 24V, so 400nC total.  30kHz * 400nC = 12mA.  Gate drive supply average current should be 12mA plus current consumed by driver chip itself, which should be available in its data sheet.

Does your buck converter switch high side (positive supply rail) or low side (negative supply rail)?  Less common to switch negative rail, but that's what I did for my QCW buck converter.  That way IGBT emitter is tied to rectified line voltage with no high frequency switching voltage.  Standard isolated supplies work fine.  If switching positive rail (as is more common), then emitter voltage includes fast switching edges.  Standard line-powered isolated supplies work only if there is enough stray inductance in wiring to filter high edge slew rates.  Adding a common-mode choke to a standard supply would likely work, at least much better than relying on wiring inductance alone.  Or as suggested, use a small supply designed for gate drive (designed to handle high slew rate edges of switching waveforms).

That's an option in this case, given the fiber optic isolation of the buck switch driver.

Adding series inductance to limit the common mode currents through the supply Y capacitor adds an underdamped series resonance into the mix, which might be worse or better than not having it, but surely not good. Ferrites also have low enough loss at the resulting series resonant frequency (assuming a few turns through a 10 nH/n^2 core and a 4.7 nF Y capacitor) to give an underdamped response. There is also the issue of saturation of the choke due to applied volt-seconds. Residential EMI current limits into protective earth are in the tens of microamps in these frequency ranges. Putting peaks of tens of amps into earth just to save a five dollar isolated PSU seems a bit risky to me, and having switching power supplies go out of regulation from common mode transients is a problem I've had in the past, leading to death of power electronics from overvoltage.

Here is my inductor core:
    26 turns came out to 139uH (25T calculated to be 143uH)
    12awg stranded wire

I expect this will work well.  QCW use is typically low duty cycle (longer gaps between ramps than ramp time itself).  Inductor power loss is not critical.

If designing a continuous-use buck converter, E-cores have a disadvantage.  Wire experiences much more proximity-effect loss than for toroid cores (more magnetic field attempting to cross through wire, causing eddy current losses).  Toroid coils, especially ones with single layer windings, have much less proximity-effect loss.  Stranded wire also has a disadvantage of increased skin-depth loss.  (Unless strands are individually insulated.  In that case, called litz wire, losses are lower.)

E cores are worse than toroids in some respects (fringe field losses in windings, stray field EMI issues) and better in others (practical window fill factor, ease of winding), neither of which will make or break this application. Here I recommended E-cores due to the availability and low cost of a XFlux in 40 permeability. For CW type converters, toroids or edge-wound block core assemblies tend to win out, but by far the biggest factor to making dense and efficient magnetics are the core materials, and the recent advances in this field are staggering, and core shape is a secondary concern.

Stranded wire is better than solid, but much worse than litz, when it comes to eddy current losses. There's some data in this paper: http://inductor.thayerschool.org/papers/stranded.pdf

Keep in mind that the AC resistance of the wire only matters for the ripple fraction of the current, not the low-frequency ramp component. I don't think eddy current losses in the wire here will be a major issue, but litz wire is something to look into in case you have overheating of the winding.

[ZakW]

I think I have landed on some overall specs:

• Driver: Custom 1.3UD with phase lead
• Buck modulator: Asynchronous buck | dual FGH75T65SHD
• 8200uF bus cap (for now) 120vAC + doubler
• 139uH inductor + 25uF buck output cap
• Full bridge FGH75T65SHD | 200A target max current
• MMC: 10nF @ 10kV | 10S 3P | B32642B0333J
• Secondary 2.55"(64.88mm)x3.25"(82.55mm)Diameter
• Ramp generator: Fin Hammers ramp code + Arduino nano | fiber Tx

Completed:

• Secondary + topload (may have to change topload later)
• Buck inductor

Work in progress (WIP):

• Custom 1.3UD with phase lead
• Asynchronous buck driver (high side)
• Ramp generator + enclosure

Not started:

• Full bridge PCB
• MMC construction
• Buck modulator
• Primary construction | Planning to use solid 12AWG

Based on this thread (https://highvoltageforum.net/index.php?topic=2621.0) I have decided to use B32642B0333J for MMC caps as they seem to work great without excessive heating as well as the a dual FGH75T65SHD buck setup with a full bridge of the same parts. I have A LOT of them on hand so that should give me acceptable overhead.

Why do they have their MMC configured as such? Is there a load balancing benefit or does it not matter? I can't seem to find an answer online.



Why do I see some projects where a dedicated CT is used to measure primary current (also in the above picture)? If I already have CT feedback on the UD why not just probe the PCB to monitor primary current? It seemed to work just fine in my RDRSSTC.

-Zak

[davekni]

Why do they have their MMC configured as such? Is there a load balancing benefit or does it not matter? I can't seem to find an answer online.

Some MMC designs include bleed resistors across caps to keep voltage divided equally among them.  With first configuration shown only 10 resistors are required.  Second configuration needs 30.  If not using bleed resistors, only difference would be what happens if a cap fails.

Why do I see some projects where a dedicated CT is used to measure primary current (also in the above picture)? If I already have CT feedback on the UD why not just probe the PCB to monitor primary current? It seemed to work just fine in my RDRSSTC.

Measuring on UD works fine.  Measure across just the burden resistor.  Do not include phase lead inductor (ie. not directly across UD CT input if using normal UD2.7 style phase lead).

As to why others use a separate CT, hopefully someone who does that will answer.

[ZakW]

Hello Dave,

Some MMC designs include bleed resistors across caps to keep voltage divided equally among them.  With first configuration shown only 10 resistors are required.  Second configuration needs 30.  If not using bleed resistors, only difference would be what happens if a cap fails.

Thanks, I think I like the first configuration better. I will omit the bleed resistors.

Measuring on UD works fine.  Measure across just the burden resistor.  Do not include phase lead inductor (ie. not directly across UD CT input if using normal UD2.7 style phase lead).

Good to know, that is how I measured it before so I will plan to do that this time as well.

[ZakW]

Hello,

Here is an update on my custom UD 1.3 PCB.

Feedback on my PCB layout would be much appreciated! It is my first 4 layer board.

layer 1 - signals
layer 2 - solid ground
layer 3 - power plane. Mainly 5v but also 24v
layer 4 - signals

Main changes I have made so far:

• I chose to use a set of IXDI614YI & IXDN614YI. https://www.mouser.com/ProductDetail/849-IXDI614YI. I don't plan on driving more than a full bridge at most, I think 14A should be plenty. Has anyone used these before?.
• Found some tunable inductors on Amazon. Going to try winding various coils to get the right amount of phase lead. To make things easy I am going to socket them using 2.5mm female socket pin headers so I can just pull it out and rewind it. *I hate desoldering*
• Added a pin header for an interrupter input as an alternative to fiber.
• Added self oscillation mod plus jumper to turn it on/off.
• Modified some component values per previous feedback. Notes in schematic.
• Added phase selection jumpers
• Added interrupter selection jumpers

Schematic will be cleaned up later, just adding my notes for context for myself.

Is it okay that I split the 5v power plane for the 24v plane under the driver IC's?

Edit: There was an error with my 24v bulk cap footprints on the PCB (C2&C3). I have since replaced them with a single though hole 1,300uF cap to help prevent the 24v rail from dropping too much.




















-Zak

[flyingperson23]

why not use the RC phase lead described in a recent post? that seems a lot easier than winding your own inductor

[ZakW]

why not use the RC phase lead described in a recent post? that seems a lot easier than winding your own inductor

I considered it, but the LR phase lead worked fine on my previous board so I am going to stick with it for now. Plus, with the pin headers, it will be easy to swap out inductors. After all, this is still just a prototype—I’m focused on creating and testing rather than making it 100% perfect. I'm also learning a lot of other skills along the way.

[Mads Barnkob]

Great to see that the UD1.3 lives on and gets an update with "modern" and available components!

Once you get it tested, finished and if you want to share the PCB files, I'll for sure add it to the DRSSTC Valuable Thread list of UD designs for download.

[ZakW]

Great to see that the UD1.3 lives on and gets an update with "modern" and available components!]Great to see that the UD1.3 lives on and gets an update with "modern" and available components!

Thanks, Mads! I am not really sure if it is more 1.3 or 2.7 at this point, but I started with the 1.3 and borrowed a lot from the 2.7.

Once you get it tested, finished and if you want to share the PCB files, I'll for sure add it to the DRSSTC Valuable Thread list of UD designs for download.]Once you get it tested, finished and if you want to share the PCB files, I'll for sure add it to the DRSSTC Valuable Thread list of UD designs for download.

If all goes well, and it seems to work as expected I would be more than happy to make everything available! Thank you for the opportunity to contribute to this awesome community

[davekni]

Looks workable, other than a schematic error around U4D and U7B (two outputs shorted together).

BTW, have you looked at this driver variation:
    https://highvoltageforum.net/index.php?topic=2054.msg15611#msg15611
If I were designing a new project, that has a nice input stage.  Avoids a second CT for OCD.  Of course, the recent RC phase lead option Anders proposed looks great, but of course doesn't have much history of actual use.  Makes sense to avoid changing too far away from your comfort zone of UD1.3 and UD2.7.

[ZakW]

Hello Dave,

Looks workable, other than a schematic error around U4D and U7B (two outputs shorted together).

Thank you for pointing out the error, must have occurred when I was moving everything around.

BTW, have you looked at this driver variation:
    https://highvoltageforum.net/index.php?topic=2054.msg15611#msg15611
If I were designing a new project, that has a nice input stage.  Avoids a second CT for OCD.  Of course, the recent RC phase lead option Anders proposed looks great, but of course doesn't have much history of actual use.  Makes sense to avoid changing too far away from your comfort zone of UD1.3 and UD2.7.

I did see that thread and forgot about it, I appreciate the reminder. The comfort zone is real, but I do like the idea of using the fixed inductor value and a pot to adjust instead. I know you mentioned it also removes the need for an additional CT which is good, among other benefits. I will have to go back and take a closer look through the thread again.

Since I had to adjust the 24V rail caps, I'm now concerned about the 24v rail sagging while the coil is running. I'm planning to use a 1300µF cap but am also considering switching to a regulator that can provide more current. I don’t have much experience with switching regulators, so this would be new for me as well. I know they can be designed to supply more current than a typical 1A linear regulator, but would it be enough to prevent voltage drop in the instance the coil is running (assuming that's a concern)?

Edit: I’ve looked into it a bit more, and it seems that increasing the power supply output capacity might not be as effective or as simple as just adding some additional caps near the gate driver ICs. I have a decent cap on the 24V rail now, along with two SMD caps mounted near each gate driver.

[davekni]

Edit: I’ve looked into it a bit more, and it seems that increasing the power supply output capacity might not be as effective or as simple as just adding some additional caps near the gate driver ICs. I have a decent cap on the 24V rail now, along with two SMD caps mounted near each gate driver.

Is the 1300uF cap you mentioned on +24V?  Capacitance is my thought on how to handle peak current.  Typical for normal DRSSTC.  QCW's long enable times may be an issue for just capacitance.  Depends on total IGBT gate charge, switching frequency, and enable on time.  QCW coils have long enable times, so higher requirement for 24V capacitance.  Calculation is as follows:
    FGH75T65SHD gate charge extrapolated to +-24V looks to be about 400nC.  1.6uC for all four devices.  This charge must be supplied twice per H-Bridge cycle (every switching event).  Presume 400kHz coil frequency (800kHz H-bridge edge rate) and 25ms enable (ramp) time:
    Total charge = 1.6uC * 400kHz * 2 * 25ms = 32mC.
Would require 16mF (16,000uF) for 2V sag if regulator supplied no current during enable time.
Or, with little capacitance, regulator needs to supply 32mC/25ms = 1.28A.  So regulator may be the better choice here.

Above does not include any margin.  Regulator of at least 1.5A, better 2A, would be advised.

Switching regulators may have an issue with transient load (step load), so have significant voltage drop and recovery/overshoot depending on switching frequency and regulation loop dynamics.

[ZakW]

Is the 1300uF cap you mentioned on +24V?

Yes, I will have a bulk cap on the output of the regulator as well as two more local bulk caps by the gate drivers themselves. I did have to scale them back a bit to fit everything so I downgraded them all to 470uF for now. I want to keep the PCB as compact as I can without hurting performance too much. If I need to add more capacitance in the end I will have to make room. I am also working on a new version from the freewheeling thread you shared. That is cutting down on some parts so I will likely have some additional board space to use.

QCW's long enable times may be an issue for just capacitance.  Depends on total IGBT gate charge, switching frequency, and enable on time.  QCW coils have long enable times, so higher requirement for 24V capacitance.

The long on times were my concern. Plus if I understand correctly, current is rising throughout the time the coil is on. If my gate voltage sags, that will lower my IGBT current handling capability at its max point, right?

Calculation is as follows:
    FGH75T65SHD gate charge extrapolated to +-24V looks to be about 400nC.  1.6uC for all four devices.  This charge must be supplied twice per H-Bridge cycle (every switching event).  Presume 400kHz coil frequency (800kHz H-bridge edge rate) and 25ms enable (ramp) time:
    Total charge = 1.6uC * 400kHz * 2 * 25ms = 32mC.
Would require 16mF (16,000uF) for 2V sag if regulator supplied no current during enable time.

Thanks for this! Glad to see I wasn't too far off. I did not extrapolate the gate charge from the chart but rather used the Qg in the datasheet which I have since learned is not as accurate as the chart if I am using a different gate voltage.

Or, with little capacitance, regulator needs to supply 32mC/25ms = 1.28A.  So regulator may be the better choice here.

Above does not include any margin.  Regulator of at least 1.5A, better 2A, would be advised.

I can upgrade my 1A 24v regulator to this 1.5A one https://www.mouser.com/datasheet/2/389/l78-1849632.pdf. I can't seem to find any 24v 2A regulators on Mouser or Digikey.

Looks like in the datasheet I can further increase output current by following their example.


Other than increasing current like that, I spent some time on Google and it sounds like putting two regulators in parallel is a bad idea so I won't go that route.

Switching regulators may have an issue with transient load (step load), so have significant voltage drop and recovery/overshoot depending on switching frequency and regulation loop dynamics.

I thought there might be a catch, thanks for the heads up.

[davekni]

I can't seem to find any 24v 2A regulators on Mouser or Digikey.

Quick search shows this adjustable 5A regulator:
LM338T/NOPB
Not quite a drop-in, but might be reasonable by lifting pin 1 (adjust pin) and scabbing in resistor divider.  I've made similar scabs for adjustable regulators when I don't happen to have the correct voltage part around.

[ZakW]

Quick search shows this adjustable 5A regulator:
LM338T/NOPB
Not quite a drop-in, but might be reasonable by lifting pin 1 (adjust pin) and scabbing in resistor divider.  I've made similar scabs for adjustable regulators when I don't happen to have the correct voltage part around.

I did not think to use a higher current adjustable regulator. I haven't used them before but it seems like a great solution!

I hope this isn't a stupid question but the freewheeling driver is just a different approach to dealing with an over current condition? Instead of just rapidly shutting the drive signal down, the 'freewheeling' aspect allows the drive signal to continue to allow the current to decay over time?

Edit:Is pulse-skipping different than freewheeling? I am not finding much information on builds using a freewheeling driver like this.

-Zak

[davekni]

I hope this isn't a stupid question but the freewheeling driver is just a different approach to dealing with an over current condition?

Yes, that's how I see the term being used.

Instead of just rapidly shutting the drive signal down, the 'freewheeling' aspect allows the drive signal to continue to allow the current to decay over time?

Freewheeling causes current to remain roughly constant, varying from slightly under to slightly over OCD limit, for remainder of enable pulse.  Decay is same as normal, just delayed until after full enable pulse width.

Edit:Is pulse-skipping different than freewheeling? I am not finding much information on builds using a freewheeling driver like this.

I think those terms are often used interchangeably.  There are two variations.  Freewheeling likely refers (at least more often) to the version that continues to enable one half-bridge during skipped pulse.  Pulse-skipping may refer more often to the version that turns off both half-bridges for the skipped pulse.  Turning off both half-bridges is a bit simpler, but causes more current ripple.

[ZakW]

Thank you for explaining that. Glad to know I'm using the terms correctly.

Is freewheeling generally considered better or more safe for the bridge instead of disabling the drive signal during an overcurrent event like previous UD versions?

[ZakW]

I am taking a look at the LTspice sim from this post here: https://highvoltageforum.net/index.php?topic=2054.msg15605#msg15605

Green = the output of B1 where I marked 'Green_Trace'
Blue = After D2 'Blue_Trace'

I'd like to know if the 4.7uH inductor will provide sufficient lead at my target Fres of around 400kHz. I assume LTspice can help me figure that out? Problem is this is the first time I have used it and am not sure how to tell what frequency this is even running at. Do I need to tell it to run at 400kHz, can I sweep a range?

Any advice would be great, I will keep on messing with it to get more familiar with the program.

Edit: Dave, you mention here (https://highvoltageforum.net/index.php?topic=3097.msg22164#msg22164)

Fixed inductor can be a toroid or gapped E-core or pot-core, all less sensitive to stray magnetic fields than typical adjustable inductors are.

Would a shielded inductor be ideal? I am finding small RF inductors on Mouser, like this https://www.mouser.com/ProductDetail/Pulse-Electronics/PE-1008CLH4R7STS?qs=vHuUswq2%252Bsw5OlhdL5g%252BEA%3D%3D or this https://www.mouser.com/ProductDetail/Wurth-Elektronik/744766904?qs=HXx4m3XcTe1xF7KnJqgGSg%3D%3D.

Am I on the right track, or should I be looking at a different kind? Looks like Mike used a wire wound radial type in his build.

[ZakW]

I am taking a look at the LTspice sim from this post here: https://highvoltageforum.net/index.php?topic=2054.msg15605#msg15605

Green = the output of B1 where I marked 'Green_Trace'
Blue = After D2 'Blue_Trace'

I'd like to know if the 4.7uH inductor will provide sufficient lead at my target Fres of around 400kHz. I assume LTspice can help me figure that out? Problem is this is the first time I have used it and am not sure how to tell what frequency this is even running at. Do I need to tell it to run at 400kHz, can I sweep a range?

Any advice would be great, I will keep on messing with it to get more familiar with the program.

Edit: Dave, you mention here (https://highvoltageforum.net/index.php?topic=3097.msg22164#msg22164)

Fixed inductor can be a toroid or gapped E-core or pot-core, all less sensitive to stray magnetic fields than typical adjustable inductors are.

Would a shielded inductor be ideal? I am finding small RF inductors on Mouser, like this https://www.mouser.com/ProductDetail/Pulse-Electronics/PE-1008CLH4R7STS?qs=vHuUswq2%252Bsw5OlhdL5g%252BEA%3D%3D or this https://www.mouser.com/ProductDetail/Wurth-Elektronik/744766904?qs=HXx4m3XcTe1xF7KnJqgGSg%3D%3D.

Am I on the right track, or should I be looking at a different kind? Looks like Mike used a wirewound axial type in his build.