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.)