QCWDRSSTC - Project Build

[ZakW]

Hello,

Creating this thread to chronicle my QCWDRSSTC project!

Currently, I am working on copying and recreating a lot of loneoceans' designs (credit goes to him!). Also, shout out to @Magneticitist on YouTube for documenting his builds so far; they have all been really helpful.

BrickDriver v1.1 schematic https://www.loneoceans.com/labs/sales/brickdriver/SchematicBrickDriver11.png
Loneoceans write up on the BrickDriver https://www.loneoceans.com/labs/sales/brickdriver/index.htm
HCPL-316J (2.5 Amp Gate Drive Optocoupler) Datasheet https://www.mouser.com/datasheet/2/678/av02-0717en_ds_hcpl-316j_2015-03-09-1827931.pdf


Like I said, I am creating my own schematics and PCBs. It would be great to get some feedback to ensure I understand everything correctly.

First up is the buck modulator schematic. I have opted to simplify the original design (BrickDriver v1.1) a bit and decided not to use the DESAT and FAULT capabilities of the HCPL-316J for this first build. I know they are valuable, but I am just trying to wrap my head around all of this at first. Referencing the datasheet, I think I have wired the HCPL-316J correctly to disable those two functions:



Schematic concerns:

• Isolated buck and logic power supplies: I will be using a separate power supply for the buck 24v input https://www.mouser.com/ProductDetail/709-RS15-24 as well as the UD driver https://www.mouser.com/ProductDetail/709-SCP35-24. For the buck logic, I will use 5v from the UD board since it has its own supply.
• DESAT (Pin 14): Connected it directly to VEE (Pin 9 or 10).
• FAULT (Pin 6): Left unconnected.
• RESET (Pin 5): Connect to a logic high to keep the fault output inactive.

If I understood the datasheet correctly, that should be all that is need to allow the IC to function without shutting down due to DESAT or FAULT conditions.

The initial PCB layout is almost finalized, just waiting to receive more parts to make sure everything fits nicely.

In the meantime I have also been working on a new secondary, topload, and MMC. I still need to coat the secondary and 3D print the primary coil form. I will likely copy a similar design to loneoceans 1.5 build since the coil is 2.5inches tall and needs some clearance from the topload.


-Zak

[Simranjit]

May I ask what is material for toroid core you used in buck modulator?

[ZakW]

May I ask what is material for toroid core you used in buck modulator?

I had my eye on the same one you mentioned in your post https://www.digikey.com/en/products/detail/magnetics-a-division-of-spang-co/0079908A7/18626744. I was hoping to see a reply to your post to help me figure out if it would work or not.

[davekni]

There's no way to know if this is a "good" core or not without knowing the rest of your buck design.  The most important factors are maximum ramp current (peak buck output current), buck switching frequency, and buck input voltage.  Inductor needs to handle maximum ramp current plus peak buck ripple current.  Ripple current depends on inductance, frequency, and voltage.

[Simranjit]

There's no way to know if this is a "good" core or not without knowing the rest of your buck design.  The most important factors are maximum ramp current (peak buck output current), buck switching frequency, and buck input voltage.  Inductor needs to handle maximum ramp current plus peak buck ripple current.  Ripple current depends on inductance, frequency, and voltage.

Thank you for clarification. It will get back to you in few weeks hopefully ☺️. Still trying to learn.

[ZakW]

Dave,

I am planning to copy Loneoceans QCW 1 pretty closely. For the buck modulator I am going to use two FGA40N65SMD in parallel https://www.mouser.com/datasheet/2/308/1/FGA40N65SMD_D-2313142.pdf. 650v 120A peak.

Parameters:
Buck smoothing capacitor: 500V, 25µF
Frequency: 30 kHz
Peak Current: 150 A (100A normal)
Peak Voltage: 340 V (300V normal)
Pulse Duration: 20ms MAX
Target Inductance: 112.4uH
Saturation Flux Density: 1 T (using 2 stacked Magnetics 0079908A7 cores)
https://www.mag-inc.com/Media/Magnetics/Datasheets/0079908A7.pdf

For these calculations I used online calculators as well as ChatGPT... Not trusting ChatGPT I checked its math with calculators and it seemed to be accurate. However, there was a large discrepancy with max flux density calculation. I included what ChatGPT said and how its calculation varied from what the calculator was doing.

Given that either calculation is below 1T, that means that I should be fine to use two stacked 0079908A7 cores, right?


Picture looks blurry, just have to right-click and open in a new tab.

[Anders Mikkelsen]

Try to figure out how to do the calculations yourself, this will give you a useful skill and is more likely to give a correct result.

For powder cores, talking about saturation flux density is a bit vague, since these materials saturate very gradually, over a range of currents. So the flux density you design with depends on how much inductance you can afford to lose at the operating current. This involves some design tradeoffs, since the ripple current and buck transistor turn-off losses will increase as the inductance drops, but there's otherwise nothing wrong with running the inductors well into the saturation region. I've done designs where there's less than 20 % of initial inductance at some operating points.

My recommended method is to look at the curves in the core datasheet, permeability vs ampere turns. Simply multiply your turns count by your operating current, and you can read out how much inductance you have left. For the core you proposed, 39 turns and 150 A  would give you around 14 nH/n^2 per core, giving a total inductance of 42 µH remaining at this current.

I'd recommend going with a material with better bias characteristics and higher initial permeability, maybe XFlux 40 or 60, but your proposed solution should work as it is. I would for example consider 25 turns on a 00X6527E040 core pair. This would give you 143 µH at zero bias down to 47 µH at 150 A. Overall cost would be lower, and you would also likely get a lower overall DC resistance.

The HCPL-316J is not ideal for the buck gate drive, since it has mediocre common-mode transient immunity (15 V/ns can be problematic with modern IGBTs) and is pretty expensive to boot.  If you don't need the desaturation detection, I would look at parts like UCC5390 or UCC25313.

[ZakW]

Hello Anders,

Thanks for the feedback and input!

Try to figure out how to do the calculations yourself, this will give you a useful skill and is more likely to give a correct result.

That's the goal

My recommended method is to look at the curves in the core datasheet, permeability vs ampere turns. Simply multiply your turns count by your operating current, and you can read out how much inductance you have left. For the core you proposed, 39 turns and 150 A  would give you around 14 nH/n^2 per core, giving a total inductance of 42 µH remaining at this current.

I'd recommend going with a material with better bias characteristics and higher initial permeability, maybe XFlux 40 or 60, but your proposed solution should work as it is. I would for example consider 25 turns on a 00X6527E040 core pair. This would give you 143 µH at zero bias down to 47 µH at 150 A. Overall cost would be lower, and you would also likely get a lower overall DC resistance.

Thank you for detailed explanation. There is a lot I still don't understand and inductors are top on that list. This helped a lot though. I like the idea of using the core you suggested since it is not only cheaper but will be easier to wind. Here is what I gather from looking at the datasheet:



Walking through what you said to make sure I understand and so that I have it for reference: Please correct me if I'm wrong.

This would give you 143 µH at zero bias

Zero bias meaning no DC current is flowing, so if I measured it with an LCR meter at 25 turns it would read ~143uH.

down to 47 µH at 150 A.

When a DC current is applied (bias), the core's magnetic flux increases, which causes its permeability to decrease, leading to lower overall inductance. Hence the graph, which can tell you approximately how much more inductance the core has before saturating.

lower overall DC resistance

Given the core high permeability I can afford to wind fewer turns using thicker wire to achieve less resistance losses that I would otherwise have due to increase turns.

The HCPL-316J is not ideal for the buck gate drive, since it has mediocre common-mode transient immunity (15 V/ns can be problematic with modern IGBTs) and is pretty expensive to boot.  If you don't need the desaturation detection, I would look at parts like UCC5390 or UCC25313.

Thanks for sharing these! Not sure how I missed the UCC5390, cheaper and provides 10A gate drive.

mediocre common-mode transient immunity (15 V/ns can be problematic with modern IGBTs)

This is new to me as well. "15 V/ns means the driver can handle voltage changes of 15 volts per nanosecond without malfunctioning.. it can lead to erroneous switching or damage.".
Looking at the datasheet for the UCC5390ECDR https://www.mouser.com/ProductDetail/Texas-Instruments/UCC5390ECDR?qs=BZBei1rCqCCYvNb%2F7TMRZA%3D%3D it looks like it has transient immunity of 100-120 V/ns.

Since it can source 10A, I don't think I will be needing the buffer stage, and I don't plan on driving huge bricks. This is also just the first version and I can always add it later if I need more drive current.

Here is my updated schematic:




-Zak

[Anders Mikkelsen]

Hello Anders,

Thanks for the feedback and input!

Try to figure out how to do the calculations yourself, this will give you a useful skill and is more likely to give a correct result.

That's the goal

It might have come across as a bit harsh from my side, but seeing chatbots get things wrong and people trusting them just rubs me the wrong way. Glad to hear that you have the right motivations.

My recommended method is to look at the curves in the core datasheet, permeability vs ampere turns. Simply multiply your turns count by your operating current, and you can read out how much inductance you have left. For the core you proposed, 39 turns and 150 A  would give you around 14 nH/n^2 per core, giving a total inductance of 42 µH remaining at this current.

I'd recommend going with a material with better bias characteristics and higher initial permeability, maybe XFlux 40 or 60, but your proposed solution should work as it is. I would for example consider 25 turns on a 00X6527E040 core pair. This would give you 143 µH at zero bias down to 47 µH at 150 A. Overall cost would be lower, and you would also likely get a lower overall DC resistance.

Thank you for detailed explanation. There is a lot I still don't understand and inductors are top on that list. This helped a lot though. I like the idea of using the core you suggested since it is not only cheaper but will be easier to wind. Here is what I gather from looking at the datasheet:



Walking through what you said to make sure I understand and so that I have it for reference: Please correct me if I'm wrong.

This would give you 143 µH at zero bias

Zero bias meaning no DC current is flowing, so if I measured it with an LCR meter at 25 turns it would read ~143uH.

down to 47 µH at 150 A.

When a DC current is applied (bias), the core's magnetic flux increases, which causes its permeability to decrease, leading to lower overall inductance. Hence the graph, which can tell you approximately how much more inductance the core has before saturating.

lower overall DC resistance

Given the core high permeability I can afford to wind fewer turns using thicker wire to achieve less resistance losses that I would otherwise have due to increase turns.

Spot on.

The HCPL-316J is not ideal for the buck gate drive, since it has mediocre common-mode transient immunity (15 V/ns can be problematic with modern IGBTs) and is pretty expensive to boot.  If you don't need the desaturation detection, I would look at parts like UCC5390 or UCC25313.

Thanks for sharing these! Not sure how I missed the UCC5390, cheaper and provides 10A gate drive.

These chips are a game changer for gate drive. Full galvanic isolation, and better output drive capability and transient immunity than most gate drive transformer based options. I use them for both hard-switched converters, tesla coils and induction heater, without issues. There are also half bridge versions with overlap protection, automatic dead-time insertion and other fancy features.

mediocre common-mode transient immunity (15 V/ns can be problematic with modern IGBTs)

This is new to me as well. "15 V/ns means the driver can handle voltage changes of 15 volts per nanosecond without malfunctioning.. it can lead to erroneous switching or damage.".
Looking at the datasheet for the UCC5390ECDR https://www.mouser.com/ProductDetail/Texas-Instruments/UCC5390ECDR?qs=BZBei1rCqCCYvNb%2F7TMRZA%3D%3D it looks like it has transient immunity of 100-120 V/ns.

Transient immunity is the rate of change of voltage across the device that it can tolerate without giving invalid output. For a high side driver, the gate driver sees large common mode transients during switching as the output is connected to the switching node while the driver input is referenced to power or signal ground. To a first approximation, the common mode dV/dt is simply the bus voltage divided by the turn-off and turn-off time. The IGBTs you mentioned switch in around 15 ns, so with a bus voltage of 400 V you get close to 30 V/ns. With superjunction MOSFETs and wide bandgap transistors like SiC, it's easy to exceed 50 V/ns.

In an asynchronous buck converter, unintended turn-on is not necessarily a disaster, but it can lead to some ugly glitches and additional losses. It's a lot worse if it happens in a half bridge where the opposite transistor might be on. I now always try to use parts with a minimum of 100 V/ns CMTI, and never had any problems since doing that, even with fast wide bandgap transistors.

Since it can source 10A, I don't think I will be needing the buffer stage, and I don't plan on driving huge bricks. This is also just the first version and I can always add it later if I need more drive current.

Here is my updated schematic:

[attachment=2,msg21708]


-Zak

It should be good enough here without an additional driver stage

[ZakW]

It might have come across as a bit hars from my side, but seeing chatbots get things wrong and people trusting them just rubs me the wrong way. Glad to hear that you have the right motivations.

No offense taken, you're correct though. I use ChatGPT as an enhanced search engine, never taking what it says as correct. It has been really helpful in providing more context into certain topics and has help me to understand a lot more. I was trying to do use its calculations and fact check them with online tools to see if the answers were close, as well as check with knowledgeable people here.

Transient immunity is the rate of change of voltage across the device that it can tolerate without giving invalid output. For a high side driver, the gate driver sees large common mode transients during switching as the output is connected to the switching node while the driver input is referenced to power or signal ground. To a first approximation, the common mode dV/dt is simply the bus voltage divided by the turn-off and turn-off time. The IGBTs you mentioned switch in around 15 ns, so with a bus voltage of 400 V you get close to 30 V/ns. With superjunction MOSFETs and wide bandgap transistors like SiC, it's easy to exceed 50 V/ns.

In an asynchronous buck converter, unintended turn-on is not necessarily a disaster, but it can lead to some ugly glitches and additional losses. It's a lot worse if it happens in a half bridge where the opposite transistor might be on. I now always try to use parts with a minimum of 100 V/ns CMTI, and never had any problems since doing that, even with fast wide bandgap transistors.

Awesome! That is good to know, I think I will start using these more often. That makes sense then why I see some coils that have unstable ramps, I wonder if it could be due to transient noise causing unintended turn-ons and glitches in the ramp output.

It should be good enough here without an additional driver stage

Excellent!

I ordered some fiber transmitters/recievers, fiber, and that core. Will have more updates to come.

[ZakW]

Here is the my final draft of the buck modulator driver schematic and PCB. Only 48mmx48mm! Layout and design feedback would be much appreciated, I am always trying to improve my designs.

Switching to the UCC5390E simplified the design a lot. I referenced this schematic for the overall designhttps://www.loneoceans.com/labs/sales/brickdriver/SchematicBrickDriver11.png

Couple of design questions:

• On the PCB I have two ground pours on the bottom layer. The default clearance between these layers is 0.5mm. Is that going to be a problem?
• Gao placed the TVS diode on the PCB, I assume he did this because his PCB was deigned to be mounted on the Brick itself placing it really close to the package pins. But in practice shouldn't the TVS diode be mounted as close to the body of the IGHTs as possible? My plan was to use a short twisted wire connection from my PCB to my IGBTs, if that is the case I should probably remove the TVS diode from the PCB and solder one on the IGBT leads then. Is that correct or will this be sufficient?

Here is my inductor core:

• 26 turns came out to 139uH (25T calculated to be 143uH)
• 12awg stranded wire

Working on my updating my UD 1.3 board next.

Thanks!
-Zak

 

[flyingperson23]

I'm not sure about your isolated 24 and 5v inputs. Why not input non isolated 5v and use something like https://www.mouser.com/ProductDetail/XP-Power/IHL0205D1509?qs=w%2Fv1CP2dgqrMGeGh%2FuoQBA%3D%3D to get the isolated gate drive voltages

[ZakW]

I'm not sure about your isolated 24 and 5v inputs. Why not input non isolated 5v and use something like https://www.mouser.com/ProductDetail/XP-Power/IHL0205D1509?qs=w%2Fv1CP2dgqrMGeGh%2FuoQBA%3D%3D to get the isolated gate drive voltages

Definitely an option, are you saying use a single 24v supply to power two of them, one for 5v and another for +15 & -9v output? The 15v output current seems a bit low at 66mA.

I was already planning on buying two separate enclosed supplies for the buck modulator as well as the UD. I was just planning on using the UD 5v rail since the board would be physically near by.

Thanks for the link, interesting little power supplies.

[flyingperson23]

I'm saying you can use non isolated 5v for the data stuff. I don't see any reason why it needs to be isolated/floating. Then use that as the input to the linked converter, which takes in 5v and outputs isolated 15/-9v.

Idk exactly what buck igbt/switching frequency you're using, but if that one can't give you enough output current there's probably a different converter that can.

I'm not sure I'd trust the isolation on standalone power supplies. The small converters will definitely give you proper isolation.

[ZakW]

Taking a look at the datasheets for both of these supplies that I plan on using:
https://www.mouser.com/ProductDetail/709-RS15-24
https://www.mouser.com/ProductDetail/709-SCP35-24

Here is what is says:



Withstand Voltage:
I/P-O/P: 3KVAC
I/P-FG: 2KVAC
O/P-FG: 0.5KVAC

Isolation Resistance:
I/P-O/P: Isolation resistance between Input and Output.
I/P-FG: Isolation resistance between Input and Frame Ground.
O/P-FG: Isolation resistance between Output and Frame Ground.
100M Ohms: The resistance value is 100 Mega Ohms.

Isolation shouldn't be a concern.

[Anders Mikkelsen]

Layout looks decent, but I'd aim for a bit more separation between the sides. I'd go for a minimum of 2 mm, but more is always better.

Using a mains-powered PSU for the buck IGBT gate drive directly is dodgy. Most mains supplies have a capacitor between the output negative and protective earth, or across the transformer in case of units without an explicit ground connection. This capacitor is shown in the RS15 datasheet for example. Since the power supply negative is connected to the IGBT emitter, and this moves around by hundreds of volts at a fast rate during switching, this cap will couple a lot of current into your mains earth and cause havoc.

Y capacitors like this are often in the range of 2.2 to 10 nF. Let's say it's 4.7 nF for example. If your bus voltage is 320 V, this gives a current through this cap of I = C dV/dt = 4.7 nF * 320 V / 22 ns (IGBT turn-on time) = 68 A. This will induce a lot of noise and possibly blow up the capacitor or something else. Supplies like this also tend to really dislike high common mode voltage swing on their output, due to limited transient immunity of the feedback optocoupler.

For gate drive you want the lowest possible isolation capacitance you can get, and ideally characterized common mode transient immunity. Anything below 10 pF should be good, and again 50 V/ns is a good target. The supply mentioned by flyingperson would be an option, but there are also cheaper options in stock at Digi-Key https://www.digikey.com/en/products/detail/gaptec-electronic/2-4S7SIC-151505D6UP/16985744?s=N4IgTCBcDa4HQBYDKB2JBJAwgfQIwFYCAGfAEQDYBVABRAF0BfIA . Mornsun used to be the best option for converters like this, but since the trade sanctions were put into place by the US government in May, they have become impossible to get.

[ZakW]

Layout looks decent, but I'd aim for a bit more separation between the sides. I'd go for a minimum of 2 mm, but more is always better.

That should be easy enough, I figured it was a bit close together. Besides isolation and general clearance is there a design technique or principle behind adding more space? It would be great to know the why behind it.

Using a mains-powered PSU for the buck IGBT gate drive directly is dodgy. Most mains supplies have a capacitor between the output negative and protective earth, or across the transformer in case of units without an explicit ground connection. This capacitor is shown in the RS15 datasheet for example. Since the power supply negative is connected to the IGBT emitter, and this moves around by hundreds of volts at a fast rate during switching, this cap will couple a lot of current into your mains earth and cause havoc.

Y capacitors like this are often in the range of 2.2 to 10 nF. Let's say it's 4.7 nF for example. If your bus voltage is 320 V, this gives a current through this cap of I = C dV/dt = 4.7 nF * 320 V / 22 ns (IGBT turn-on time) = 68 A. This will induce a lot of noise and possibly blow up the capacitor or something else. Supplies like this also tend to really dislike high common mode voltage swing on their output, due to limited transient immunity of the feedback optocoupler.

For gate drive you want the lowest possible isolation capacitance you can get, and ideally characterized common mode transient immunity. Anything below 10 pF should be good, and again 50 V/ns is a good target. The supply mentioned by flyingperson would be an option, but there are also cheaper options in stock at Digi-Key https://www.digikey.com/en/products/detail/gaptec-electronic/2-4S7SIC-151505D6UP/16985744?s=N4IgTCBcDa4HQBYDKB2JBJAwgfQIwFYCAGfAEQDYBVABRAF0BfIA . Mornsun used to be the best option for converters like this, but since the trade sanctions were put into place by the US government in May, they have become impossible to get.

I appreciate the explanation and I think I am following where you are going, I guess at this point I am confused for a couple reasons. Mainly, that supply (Meanwell RS-15-24) is what Loneoceans used and recommends. From this page https://www.loneoceans.com/labs/sales/brickdriver/index.htm they say:

• "Note that the driver can be powered either with a ~20VAC transformer, or via an isolated 24VDC supply..."
• "Isolated 24VDC Power Supply - Recommended over the transformer, e.g. Meanwell RS-15-24 x 1"
• "First, plug in 24VDC or 20VAC into the power input (this needs to be isolated, i.e. from a regular iron transformer or isolated SMPS, up to the bridge bus voltage especially if it's the driver for the high side switch in a half of H Bridge for example)."

I am willing to use a different supply and modify my design, but I would like to better understand why the same supply worked for them versus the potential issues that I could run into. I am not in a rush and want to do things right to avoid as many pitfalls as I can. I am willing to listen and learn as much as everyone is willing to share.

but there are also cheaper options in stock at Digi-Key https://www.digikey.com/en/products/detail/gaptec-electronic/2-4S7SIC-151505D6UP/16985744?s=N4IgTCBcDa4HQBYDKB2JBJAwgfQIwFYCAGfAEQDYBVABRAF0BfIA .

One more question regarding these power supplies - the power output is so low at only 2.4W. Is this okay since I have a lot of bulk capacitance (560uFx2), so when the buck driver is switching it is drawing all of the current from them and not the actual supply itself. Am I understanding that correctly?

[flyingperson23]

I appreciate the explanation and I think I am following where you are going, I guess at this point I am confused for a couple reasons. Mainly, that supply (Meanwell RS-15-24) is what Loneoceans used and recommends.

Just because loneoceans did something doesn't mean it's the best way to do it. He may have just been using what he had on hand. A full isolated smps would probably work yes, but it's much more sketchy than the small well isolated converters actually meant for this purpose.

One more question regarding these power supplies - the power output is so low at only 2.4W. Is this okay since I have a lot of bulk capacitance (560uFx2), so when the buck driver is switching it is drawing all of the current from them and not the actual supply itself. Am I understanding that correctly?

Yes, you just have to make sure that the overall power draw is less than what the supply can provide. Instantaneous power while switching will be much higher.

[ZakW]

Just because loneoceans did something doesn't mean it's the best way to do it. He may have just been using what he had on hand. A full isolated smps would probably work yes, but it's much more sketchy than the small well isolated converters actually meant for this purpose.

Thanks, flyingperson. That was not my intent, I was simply pointing out the discrepancy and asking for clarification. They have an extensive background and I can only imagine they would have been aware of the drawbacks to using such a SMPS. That is why I am confused. There is not a lot of documentation or posts pertaining to QCW coils that I can find which go into detail on how their isolated supply was implemented.

Yes, you just have to make sure that the overall power draw is less than what the supply can provide. Instantaneous power while switching will be much higher.

I will have to look more into this.

[davekni]

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).

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.)

[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.

[davekni]

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?

Improves performance (usually), increases average power draw from line, and increases power dissipation in IGBTs (longer time at max current) and especially of anti-parallel diodes within IGBT package.  Not safer for bridge.

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?

LTSpice can do a good job of predicting how much phase lead you will get.  Harder to determine how much is needed as accurate IGBT models are hard to find (and often have difficulty simulating in my experience).

Would a shielded inductor be ideal?

Yes from an interference viewpoint, presuming other parameters are appropriate.  The Yageo shield part is a typo in parameters and data sheet.  Appears to be actually 4.7nH, not 4.7uH.  Wurth part (unshielded) should be fine.  Small enough that it likely picks up little stray field from TC primary etc.

[ZakW]

If I were designing a new project, that has a nice input stage.  Avoids a second CT for OCD. 

I might have misunderstood what you meant here. Were you suggesting I copy the input stage/phase lead from the freewheeling design and splice that into my current design (UD1.3/2.7) or were you suggesting I copy the entire freewheeling driver altogether?

I do like that the input is simplified and updated, I am just not entirely sure about the freewheeling portion at this point.

Improves performance (usually), increases average power draw from line, and increases power dissipation in IGBTs (longer time at max current) and especially of anti-parallel diodes within IGBT package.  Not safer for bridge.

That makes sense, I think I will stick with the traditional UD1.3/2.7 OCD then.

Regarding the input stage of the freewheeling driver, I was not sure how to adapt that to the 2.7 since it uses a different comparator and I have very limited understanding them as well as the subsequent flip flop logic. I will have to look more at the datasheet of the TL3116 versus the TLV3501 to figure that out.

LTSpice can do a good job of predicting how much phase lead you will get.  Harder to determine how much is needed as accurate IGBT models are hard to find (and often have difficulty simulating in my experience).

Yes from an interference viewpoint, presuming other parameters are appropriate.  The Yageo shield part is a typo in parameters and data sheet.  Appears to be actually 4.7nH, not 4.7uH.  Wurth part (unshielded) should be fine.  Small enough that it likely picks up little stray field from TC primary etc.

Thanks for taking a look at the parts. I did not notice the error in the reported value.

I am going to try to find an inductor series that comes in the same package but buy a few different ones just incase I need more or less. I saw your recommendation about adding a resistor in parallel to reduce the lead if the minimum was not enough.


As always, thanks for the advice Dave!

[davekni]

Regarding the input stage of the freewheeling driver, I was not sure how to adapt that to the 2.7 since it uses a different comparator and I have very limited understanding them as well as the subsequent flip flop logic. I will have to look more at the datasheet of the TL3116 versus the TLV3501 to figure that out.

Several options on how much to copy.  Minimal change is to copy just the phase lead part:

Connect output of that to R27/C5 (100k/100nF) of UD2.7.

The comparitor part of the freewheeling schematic is self-oscillating, a feature I'm fond of.  It is not necessary.  I also like rail-to-rail input and output comparitors like TLV3501.  However, TL3116 will work fine with the self-oscillating schematic.  To match lower output voltage of TL3116, change R6 from 1k to 470 ohms to center voltage divider closer to half of TL3116 high level output voltage.  This will add self-oscillation without freewheeling.  (Freewheeling is a function of logic after comparitors.)

One more option is to also use OCD input of this freewheeling schematic without freewheeling.  Still avoids an extra CT for OCD.  Keeps same OCD functionality as UD2.7.

[ZakW]

Hello Dave,

Thank you for breaking that down for me.

Does this look about right? Sorry, about the reference designators being all changed again, I don't like having C36 next to C2.

At this point I am not sure what to call this UD. It was based off the 1.3, I added stuff from the 2.7 and the freewheeling driver, and modded the output section to use IXD driver ICs. Any suggestions?

I wanted to keep the UVLO from the freewheeling schematic since it sounded like it worked well and I excluded it from the previous version. That shouldn't be a problem, right?

Since I no longer need the 9V output, I am considering removing it; however, I am concerned that the 5V regulator may generate excessive heat when reducing 24V to 5V. After reviewing the datasheets of the ICs drawing 5V, their current consumption appears to be quite low. I tested the heat generated with a 100Ω resistor simulating a 50mA load at 24V, and it seems manageable.

[davekni]

Does this look about right?

Input circuitry looks fine.  Still two logic outputs connected together.
BTW, any reason for R8 value of 5k rather than UD2.7's 1k?

I wanted to keep the UVLO from the freewheeling schematic since it sounded like it worked well and I excluded it from the previous version. That shouldn't be a problem, right?

Looks good to me.

Since I no longer need the 9V output, I am considering removing it; however, I am concerned that the 5V regulator may generate excessive heat when reducing 24V to 5V. After reviewing the datasheets of the ICs drawing 5V, their current consumption appears to be quite low. I tested the heat generated with a 100Ω resistor simulating a 50mA load at 24V, and it seems manageable.

Sounds good.

[ZakW]

Thank you for taking a look!

Still two logic outputs connected together.

Dang, I thought I corrected the error, I will check again.

Edit: Found it, not sure how the pin out flipped. Thanks for spotting that, I never would have.

BTW, any reason for R8 value of 5k rather than UD2.7's 1k?

This was your advice to make self-oscillation more stable. As a result of increasing R8 you also suggested changing C33 (C8 in mine) from 220pF to 1nF. I liked the relevant posts below.

https://highvoltageforum.net/index.php?topic=2651.msg20276#msg20276

Thank you.  Makes discussion much clearer.  BTW, self oscillation may work better if R6 is increased from 1k to ~5k as mentioned near the end of this post:
    https://highvoltageforum.net/index.php?topic=1914.msg14854#msg14854
Above post recommends changing R2, the label from common UD2.7 schematics.  This is R6 for your schematic.  Increasing R6 value should help make self-oscillation frequency more stable and easier to adjust.

https://highvoltageforum.net/index.php?topic=2651.msg20288#msg20288

Just looked back at your updated schematic from a few posts ago.  C15 will need to be larger than 220pF, ~1nF, to go along with 5K for R6.  I'd recommend changing C15 rather than going so low for R6 value.

Sounds good.

Excellent, then I will remove the 9V regulator and clear up some board space.

[davekni]

This was your advice to make self-oscillation more stable. As a result of increasing R8 you also suggested changing C33 (C8 in mine) from 220pF to 1nF. I liked the relevant posts below.

https://highvoltageforum.net/index.php?topic=2651.msg20276#msg20276

Thank you for the prompt to look back at my post and more at the version you have here.  That post was for a self-oscillation modification for existing UD2.7 boards designed to minimize required changes/patches.  That does fit best with what you have here including 5k value for your R8.  So keep that value.

Another option is to tweak this schematic to use TL3116 and your existing input phase-lead and OCD circuitry:
    https://highvoltageforum.net/index.php?topic=1336.msg9894#msg9894
Change R3 to 470 ohms to match reduced output swing of TL3116.  Add a variable resistor in series with R1 for adjustment.  Replace L4 and R5 with existing phase lead circuitry (with your L1, C18, R7, R11, and RV1).  This version uses fewer components and avoids having two 1k/470 resistor dividers that need to match well.