Rule compliancy

[MarosMacko]

Hello Liv,

First of all, I would like to show my gratitude for open-sourcing a project like this! The documentation and calculations/simulations/measurements are done on a very high level. I've got multiple teams asking about this DC/DC, have some direct feedback from two teams who implemented it in their car (and had it explode multiple times), and I also see that the next revision is going to be designed.

This is why I would want to provide my feedback for the v9-3 revision, both for teams already having this revision or thinking about it for the upcoming season, and for the next revision development.

My feedback below is not about reliability or performance, but rather rule compliance, which is claimed in the readme. As a scrutineer in FSCZ and FSG, I would not let this DC/DC pass the technical inspection if it were built as stated in the manual.

Going off FS Rules 2025 v1.1:

EV4.5.13 All electric connections, including bolts, nuts, and other fasteners, in the high current path, see EV1.2.2, of the TS must be secured from unintentional loosening.
EV5.5.11 All fasteners used within or to mount the TSAC must comply with T10. Fasteners within the TSAC used for non-structural parts, e.g. PCBs, do not have to follow T10.1.2. Fasteners made of electrically non-conductive material within the TSAC used for non-structural parts do not have to follow T10.

EV4.5.15 Soldered connections in the TS high current path, see EV1.2.2, are only allowed if all of the
following are true:
• connections on PCBs
• the connected devices are not cells or wires
• the devices are additionally mechanically secured against loosening

This is not met on multiple occasions - glueing would be sufficient in most places, but it is only mentioned as an optional step

• Soldered input connector
• Soldered fuse holder
• Soldered wires for the transformer
• Other smaller components falling under "high current" requirement (EV1.2.2)

Other positive locking issues

• HV input connector wires are only screwed in - does not meet the rules
• The LV connector does not have any positive locking

Litz wire issues
EV5.4.7 Every wire used in a TSAC, regardless of whether it is part of the LVS or TS, must follow EV4.5.2, EV4.5.3, and EV4.5.5.
EV4.5.2 Insulation material that is rated for the maximum TS voltage must be used. Using only insulating tape or rubber-like paint for insulation is prohibited.

• Voltage rating is not sufficient
• Kapton tape is not sufficient
• Creepage from Prim to Sec unclear, possibly not sufficient? (as the Litz wire does not meet EV1.2.1 on itself - needs documentation)

Scruti document issue

• Plastik 70 datasheet is in German (scrutis don't like it), and not up to date - see www.kontaktchemie.com
• Plastik 70 is not considered sufficient (does not meet temperature rating {EV4.5.3} - see full-length datasheet, 100 °C rating is only for a short duration).
• One layer of coating is not usually recommended, as another coat after inspection is usually needed to get full coverage from "shadows". Plastik 70 also does not have an easy way for an inspection (e.g. UV light, as the Plastik 70 Super has)
• I see it fixed in the build document, but I already wrote this, so I will keep it here as a reference... :D

General feedback

• Adding cut-outs under SMD component(s) would improve the cleaning and inspection process - I see it partially done on TS-only HV SMDs
• Removing the soldermask will help as well. "Bonus points" when it is black, for ease of inspection
• The isolation barrier is nice and straight, 4.5mm, which is more than enough. But personally, I do not like seeing any copper under the line, which indicates the barrier itself. Currently, half of the line is underfilled with copper.

Thank you for considering my input. I will be glad to provide any feedback in the future, if requested.

Good luck!

[Rootthecause]

Hi Maros,

Thanks a lot! This is the kind of feedback I wish I had some years before (especially from the other side). So lets go trough it one by one - I hope I can address most of it in the next version and would highly value intermittent feedback in the upcoming version 10 design process. Sorry about the lenght - there is just a lot to discuss.

1) Clarification about the "Rules compliance"

What I wanted to say with "FSG-Rules 2025 conform" is, that you will be able to use the design at FS competitions. In total my DCDC's passed at least 7 official scrutineerings without mayor criticism. As the released version 9-3r was very similar in terms of rules/compliance to the previous versions, it felt right to state this point at the release. I'm also not aware of any other team not passing DCDC scruti this year either. But I'm aware, that some scrutineers are not going in-depth when scrutineering the DCDC and only check for the points on the inspection sheet.
In the end, because I could never be a 100% sure, I wrote "FSG-Rules 2025 conform" in italics. If you feel, stating it like this is totally misleading, I can remove it from the data-sheet for now.
But I like to address your critics in the following. Maybe an updated build guide can make it rules conform in your terms as well :) I'm also open on giving feedback on FSG rules regarding DC/DCs.

2) EV 4.5.13 and connected rules

EV 4.5.13 All electric connections, including bolts, nuts, and other fasteners, in the high current path, see EV1.2.2, of the TS must be secured from unintentional loosening. Fasteners must use positive locking mechanisms, see T10.2, that are suitable for high temperatures, see EV 4.5.3.

It raises the question, what would count as "secured from unintentional loosening".
In doubt I would apply T 10.2.2 but this is unlikely meet by any commercial module or team build converter for the components on the PCB. So I wonder, if there is any reference of glueing fulfilling 4.5.13 in the Rules?
Furthermore it raises the question, why would a solder joint not be enough (more on that later see EV 1.2.2).

EV4.5.15 Soldered connections in the TS high current path, see EV1.2.2, are only allowed if all of the
following are true:
• connections on PCBs
• the connected devices are not cells or wires
• the devices are additionally mechanically secured against loosening

So I guess, this rule specifically includes components on a PCB and "additionally mechanically secured against loosening" is allowing for glueing? But still, I wish there was a rule specifying it, because in doubt I would apply T 10.2.2 or is there a suiting rule which I'm not aware of?

For the DCDC I can make the input connector, fuse holder and wires for the transformer be required to be glued. However, I highly dislike the idea, if components like the MOSFETs in the "high current path" are failing. Removing them with the glue might damage the PCB and other components. But if sandwiching components between the PCB and a plate with insulating thermal pads is sufficient, then I can think of a possible design for the v10. But reading the rule like this is still odd to me, as many team built precharge circuits would disrespect this rule.

3) Other positive locking issues

• HV input connector wires are only screwed in - does not meet the rules

I see. How about this connector? It would fit on the current DCDCv9-3r without altering the PCB. If it is okay, I will change it in the next update for the v9-3r.
For the next Version 10 I intent to use crimped Molex Microfit 3.0 connectors. Please also let me know, if they are okay or if you can recommend something better :)

• The LV connector does not have any positive locking

The XT60 connector is LV and not TS, so EV 4.5.13 should not apply here? I'm also not aware of any other rules requiring it here. Have I overlooked something? My Team also uses them outside of the TSAC in the LV system without complaints.
If you can recommend a better connector, please let me know. I love XT60s because they hold up against high currents like nothing else despite their size and the clamping force is quite high. They're often found in drones and I've never heard of them getting loose. My second best Idea would be bolting on some cable lugs directly onto the PCB - but thats not a nice connector in terms of unplugging/plugging-in.

Litz wire issues
EV5.4.7 Every wire used in a TSAC, regardless of whether it is part of the LVS or TS, must follow EV4.5.2, EV4.5.3, and EV4.5.5.
EV4.5.2 Insulation material that is rated for the maximum TS voltage must be used. Using only insulating tape or rubber-like paint for insulation is prohibited.

• Voltage rating is not sufficient

Pheww. I wonder how this rules is meet by commercial converters. If interpreted like this, then the conductor of any wire wound coil and PCB trace needs to fulfil this as well. From research on the HVI I know, that it's quite hard to find voltage ratings for inductors, not to mention the enamelled wire used.
So what would you recommend in this case? If the converter/wire was housed inside a case which fulfils EV5.4.7 would that be ok? And is there another way to use it, like applying heat shrinking tube around the primary coil?
If I'm designing a PCB transformer for the next version, does the conductor on the PCB count as a wire too???

• Kapton tape is not sufficient

There is no Kapton tape used inside the transformer. Its just an additional layer of protection below the transformer, if the primary insulator would break. See page 23 in the build guide. The 3D-Printed and insulation tested coil formers and the bottom insulator are used as primary insulator. I could see, that this might be not enough on the fan and temperature sensor described on page 26. Afaik the rule (no tape as insulator) was implemented because teams used it in their TSAC to cover the walls, but it got scratched easily when removing the segments - this is not a problem here. It's a bummer, that it was disallowed in general. Imho Insulating tape should be allowed for non structural parts or something like this.

• Creepage from Prim to Sec unclear, possibly not sufficient? (as the Litz wire does not meet EV1.2.1 on itself - needs documentation)

The required creepage distance for 600V is 4 mm. The whole converter including the transformer is conformal coated, so 4 mm is required here as well. At the initial release the creepage of the litz wire from the primary to the secondary was just around 4 mm, but I changed it to at least 6 mm with the commit on Feb 12, 2025 .
Indeed, the transformer was tested for EV 1.2.1 but not the wire. But as the creepage is meet. Is this okay now?

4) Scruti document issue / Plastik 70

I got informed about this a while ago and changed it to Plastik 70 Super. I will change the link to an english data sheet :)
However, I could not find the data sheet on their website. The most recent english document I could find is from 2012. But there is a german one from 2022. Help?!

About the two times coating: I heard about it somewhere before, but I was not sure, as the scrutineering document asks about "applied according to data sheet" and the data sheet does not mention any second coating. To make sure, that it is applied thoroughly I wrote on page 26 "apply the conformal coating thoroughly from all sides so that it drips off". Also considering that you can check Plastik 70 Super with a UV torch for weak spots after first coating. I've done this and could not find an issue. But I can add the two times coating to the scrutineering document and build guide :)

5) General feedback

Adding cut-outs under SMD component(s) would improve the cleaning and inspection process - I see it partially done on TS-only HV SMDs

The reason I've done it below the gate driver was because it was recommended in the data sheet. But below the resonant capacitors has a whole different reason than cleaning or inspection: I would not have believed this if I hadn't seen it with my own eyes - C0G capacitors can burn if operated outside of the safe operating area. So the cutouts prevent burning the PCB and resulting in an isolation issue due to the carbonisation of the PCB. For other components it's not desirable to make cutouts in order to not waste space on the other side of the PCB. I might consider some additional cutouts in the next version, as the PCB will be a bit larger. But still, there are barely components which would benefit from it.

Removing the solder mask will help as well. "Bonus points" when it is black, for ease of inspection

OH! THIS IS BRILLIANT! I will totally keep that in mind for the next version! Loving it!

The isolation barrier is nice and straight, 4.5mm, which is more than enough. But personally, I do not like seeing any copper under the line, which indicates the barrier itself. Currently, half of the line is underfilled with copper.

I see. Thats something with my own design rules and the definition of reference points.
It's imho not "clean" to have the copper stopping before the line, as everyone uses different line widths and measures the distance differently. In many cases, I use graphics/lines "centred", because if you change the line with, the center line is still consistent with the end of the copper below. My line is usually 0.3 mm so in doubt it's still 4.2 mm clearance.
I will have a thought about this, for the next version, but in doubt the copper is what's relevant, and there may be also manufacturing tolerances on the silk screen, having it end up on the copper anyway.
FYI: Maximising the copper surfaces is very crucial to my design, as they function as heat sink. Thus their priority is higher than silkscreen lines.

6)

Regarding the rules, there are some points I've stumbled across and would appreciate to see some improvements:

EV 1.2.1

"The isolation test voltage RMS, AC for 1 min"
There is no test frequency defined. This would allow for a loop hole using 0.01 Hz "DC" and then swap the polarity for AC testing ;D So maybe something usual like 50/60 Hz should be added if the test insist on AC.
Personal opinion: Sometimes I found it hard to find components with an AC isolation rating, despite being rated multiple times the DC Voltage of what is needed according to the rules. It's also inexpensive of getting a 5 kV DC insulation meter (e.g. RQ-6688B ~50€). For testing the self built transformer, even a inexpensive AC HiPot like the Tonghui TH9310B (~600€) would be 2 to 3 times the costs of a DCDCv9-3.
Aside from this, I would try to design the PCB transformer in a way, that the PCBs doing double duty as insulation. So rather than testing, more relying on the specs of the PCB. I will still test it to be sure, but I like to keep this option open for teams without the test equipment. What is your opinion about this from a rules but also engineering standpoint?

EV 1.2.2

"High Current Path – any path of a circuitry that, during normal operation, carries more than 1A."

I don't understand the intent of this rule. Why 1 A ? What should happen above it? Why is it acceptable up to 1 A? Why would it be ok to solder the components without additional mechanical restraints against loosening if the current is 1 A? I could vanish my efforts of fulfilling this rule by placing a 1 A Fuse on the DCDC. Thats still 600 W at 600 V. Or even better: "during normal operation, carries more than 1 A." - I know teams not even requiring more than 200 W on the LV side. So they are always below 1 A. But also due to the LLC topology of the converter, the input current might be 100 mA and there are still cycling 2 A inside the resonant circuit, which is somewhat in the high current path. Where to draw the line there? I feel like, we need a "FSG Rule's intent" document so everyone is on the same boat.

There is also another aspect I like to address here: If I would use a connector with 3 Pins and only 2 of them are carrying the "high current", would the 3rd pin without any current make this connector "secured from unintentional loosening"?

EV 4.1.1

Speaking strictly, DC/DCs and many voltage indicators are not allowed due to
"The maximum allowed voltage that may occur between any two electric connections is 600 V DC and for motor controller and Accumulator Management System (AMS) internal low power control signals 630 V DC."
This would forbid any high side switching component (except AMS and motor controller) in the TS, as it uses the TS Voltage + Gate drive voltage for switching. Voltage indicators as well as DC/DCs are not listed for the low power control signals exception. Fix: Add DC/DC and Voltage indicator, or remove AMS and motor control altogether so "internal low power control signals" is the main separator here.
Note 1: Teams using 48 V as LV Voltage would have a TS component if 13 V for the gate drive is used ( > 60V).
Fix: Remove "[CV Only]" from T 11.1.5.
Note 2: If commercial modules are used, the internal voltage is likely unknown, prohibiting commercial modules in doubt.

7) Reliability issue

I've got multiple teams asking about this DC/DC, have some direct feedback from two teams who implemented it in their car (and had it explode multiple times).

There was a team approaching me on Reddit with this issue and I think it was same as I had, where the half bridge MOSFETs failed, but nothing exploded there - you couldn’t even see it. See issue #8
The team seemd not have identified all affected components for a replacement, resulting in additional failures. But after they figured it out, they finally got it working successfully.
The issue was addressed in the commit on Jul 18, 2025.
The team is from a city with 4 letters. If it was a different team and you can provide more information on this, please do so (new issue please, thanks!).
Also any other feedback is highly valued! If you or the teams with their feedback not already found my survey, please forward it to them :) https://forms.gle/GCBGRqUn9ecCr8ko6

[Rootthecause]

Edit: Never mind - I just saw that the FSG feedback was open till last sunday... welp 😅

Original Message:

@MarosMacko
FIY: I'm considering writing official FSG rules feedback, but I like to consider your input first, as it might resolve some points.
I would wait until end of this week. Let me know, in case you need more time :)

Best regards!
Liv

[MarosMacko]

Hi Liv, thank you for the extensive reply with interesting discussion topics. I am compiling my answer, but I run out of time today, unfortunately :) I will post the full answer as soon as I can.

[MarosMacko]

Hi Liv, quite a long message, so I hope I won't get lost in my thoughts... :D

@ FSG Rules feedback
Yap, it is closed.... But rule questions will be open I think. But that is no place to validate design, just to clarify the rules, if something is unclear.

@ My position
I myself do not agree with some rules, or want some change to be made. For some, I have submitted a feedback ticket, or after discussion with relevant people I have decided not to submit a feedback at all (when I feel I don't have a strong case). I try my best to have comments about the rules as much "objective" as I am able to. But please note, that I am in no position to provide official rule explanations, and none of my explanations should be used as an argument during scrutineering (maybe in FSCZ yes, where you can find me and make me accept it 😅 But no guarantees ).

@ This project vs. rules
I agree that the rules can be little bit short for a DC/DC like this, especially where you (or the team) is in charge of designing and building parts that are usually bought (and have relevant certifications and/or testing - hopefully....). During competitions I was consulting with a team member their DC/DC, where they also wanted to hear feedback regarding legality. I must say that the design was on a high level, but still there were some issues with the rules. It is hard to apply them here, but it is also a deep rabbit hole, if we would make some "global" exceptions... That is why I was kind of no-compromise in the feedback I have provided - and it is up to you (or teams) to defend the design, that it is "as safe as rules intend, but not compliant". This is sometimes solved at competitions with an exception from scrutineering team, or some other penalty let's say. But it differs from scrutineer to scrutineer (especially wheter they notice at all), and from competition to competition, and it also creates a discussion which delays the team in the scruti process. So I would not advise to do anything against the rules (even if the intent is right) - or to have a rule question / communication about legality before competitions.

1) Clarification about the "Rules compliance"

• I think the v9-3 can be modified to be fully rule compliant. And as you have pointed out, it passed multiple inspections for some reason, so it either is rule compliant, or very well hidden :) But from the scrutineering side, it is little bit of a meme to hear teams say phrases like it passed scrutineering last 3 years or we never had a problem with this in other events.

2) EV 4.5.13 and connected rules

• Regarding fasteners and TS, it's quite simple. Is it TS? Is it a fastener? It must have positive locking (or alternative*). Biggest culprit here is screw terminals in the connector. There is nothing stopping the screws from loosening. One of the alternatives to literall positive locking (T 10) to comply with "stop from loosening" is for example maintenance plugs, which teams often secure against loosening by the TSAC lid itself (but it must be done properly and they must prove that it functions as intended). You can ask yourself, is there something which stops the screws from loosening? Friction does not count :) And I think the answer is clear.
• Regarding other teams: yes, there are some modules which do not fulfill this criteria. I have seen some "industrial" (not automotive) modules with screw terminals used in DC/DCs in the TSAC. Long story short, we did not accept it. But usually there are ways to fix it...
• Regarding glue, it is more for the EV4.5.15. My personal explanation / reasoning for this one: consider fail scenario of a current passing through the solder. It is either poor connection (bad soldering, oxidized cables, ...), or high (possibly short circuit) current. What will happen? The wire / component will literally desolder itself. Now you have a loose component flying somewhere where it is not supposed to. Sounds unreal? Trust me, I have experienced it multiple times. Once I had a large electrolyte cap "desolder" itself, and after few minutes, I still measured ~60V on it. If you have a TS shunt resistor, for example, properly glueing it can help mitigate this.

For the DCDC I can make the input connector, fuse holder and wires for the transformer be required to be glued.

• Careful here. First thing, wires themselves are explicitly forbidden to solder in the high current path. Regarding connectors, my personal thought, easiest way is ust to glue them down. Food for thought: If you would argue with commercial transformer... It is a little bit different story. There, the wire is wrapped around the pins on the base, then soldered to them. Wire wrapping is technique on itself :) And I would still require you to secure the transformer itself.

However, I highly dislike the idea, if components like the MOSFETs in the "high current path" are failing. Removing them with the glue might damage the PCB and other components. ...

• Yes, I do not like that solution as well. You have already proposed some solutions. See if they do not contradict any other rules, or if they will not weaken parameters of the DC/DC (isolation, mechanical strength, ease of repair..)

But reading the rule like this is still odd to me, as many team built precharge circuits would disrespect this rule.

• Yes, well... But it is enough to have 600 Ohms resistor, and you are not high current path anymore. And also, there are usually ways to fix it.

HV input connector wires are only screwed in - does not meet the rules

• Crimped connections - yes yes yes. IMO best way about it - assuming the team is capable of doing a decent quality crimp. Another story.
• For the wurth connector, I don't see any rule that would be violated by that, assuming that it holds the wire properly, and it is also strain-reliefed somewhere (e.g. tied with a ziptie).

The LV connector does not have any positive locking

EV 4.5.13 should not apply here?

• There is no explicit need for positive locking for LV connectors as far as I know. But it is something that I highly encourage, and what most of the teams usually do themselves. EV 4.5.13 and EV 5.5.11 communicate the intent well. XT60 would be on a safe side from my point of view, it holds very well (if not melted or something). If you would switch this to other connector with 3+ pins, keep in mind EV 3.2.2.

Litz wire issues

So what would you recommend in this case? If the converter/wire was housed inside a case which fulfils EV5.4.7 would that be ok? And is there another way to use it, like applying heat shrinking tube around the primary coil?

• This is a harder one... The wire itself, especially TS one, is problem on itself with the beforementioned soldering. That would be the problem one, really... Case covering the transformer would be a speculation right now, which I do not want to go into.
• Using heat shrink for the cable is definitely possible, but not sure how would you then wrap it around the transformer.
• First I would re-evaluate the way wire is used.
• For the PCB, it should not be hard to prove isolation properties of used FR-4 and also use standards like IPC-2221B / IPC-9592 - i.e. for isolation in internal layers.

Kapton tape is not sufficient

Afaik the rule (no tape as insulator) was implemented because teams used it in their TSAC to cover the walls, but it got scratched easily when removing the segments - this is not a problem here. It's a bummer, that it was disallowed in general. Imho Insulating tape should be allowed for non structural parts or something like this.

• AFAIK it was implemented because of that too. But it is best to design without it, as it has other problems too, like tear, weak sticking properties, sensitivity to surface type / preparation, execution etc.

Creepage from Prim to Sec

• Yes, if you fulfil EV 1.2.1 and wire rating is okay, 6 mm is the target creepage.
• The core of the transformer is not "conductive", but is treated as such for galvanic isolation evaluation, as far as I know.

Scruti document issue / Plastik 70

• That is why I have linked the manufacturer's website :) Click on "TDS" tab and find Plastik 70 Super there for most up-to-date datasheet.
• Regarding two coats, how do you ensure that the coating done has enough thickness? I am not saying that just by doing two you ensure you did it correctly... There is still issues with shadows, process errors, etc.
• But let's do a model situation, for fun, which could be done during scrutineering :) According to the datasheet, one coat thickness should be (not is) 20 - 40 um. If you do the math with 85 kV/mm dielectric strength, you will not reach 3*TS withstand voltage. Yes, it is not a perfect scenario, disregarding some parameters which would help. But at the same time, even those 20um are not guaranteed. Why not be on a safer side? I would much rather do 3 coats to cover everything properly, than to rely on just one, believing my calculations and cleaning / application process.

C0G capacitors can burn if operated outside of the safe operating area. So the cutouts prevent burning the PCB and resulting in an isolation issue due to the carbonisation of the PCB.

• Why not. But I guess it could still burn a bit, but I guess easier to clean? :D

For other components it's not desirable to make cutouts in order to not waste space on the other side of the PCB.

• I don't think you will lose any space? Main part I meant was U9 (optocoupler from LV to TS). What you do completely on TS or LV side is not relevant to scrutineering, other than trying to help you out, and possibly T 2.1.1 (but that would be a faaaaaar stretch, I reserve that rule to much more critical things - like some dodgy connectors in TSAC, even though they are LV for example).

EV 1.2.1

• The frequency is not explicitly defined in the rules. But I think that if you would dive into standards, which the rule is going off, you would see that using 0.01Hz would not be adequate.
• Regarding the rules, you can give a feedback on the wording, or later ask a rule question on a testing procedure which would be deemed acceptable.
• For the PCBs acting as a TS to LV separation, I have asked about this, but either the link does not work anymore, or my question was left unanswered... But as far as I know this is highly discouraged, mainly for the inspection process I think. If you want to do something like this, I would definitely ask for clarification regarding legality.
• From engineering standpoint, high voltage PCB transformers exist and work. But there are other pitfalls, like the engineering itself (how to calculate it properly, what materials to choose, what failures might happen) and the certification (if any safety is required). I remembered this nice article: https://www.magazines007.com/pdf/High-Voltage-PCDesign.pdf and this site I use as a quick reference sometimes: https://www.smps.us/pcbtracespacing.html

EV 1.2.2

• Good points. Here I will prefer not to give any specific point of view on this matter, as I don't have a definitive answer for you.

EV 4.1.1

• Definitely agree here. Me personally, I see the intent, so for the gate drive purposes I see no point enforcing 600 V limit instead of 630 V.

Reliability

• I have relayed them your form, hope they will give you proper feedback.

[Rootthecause]

Hi Maros,
Thanks for the many answers!

@ My position

:D I see - totally understandable.

@ This project vs. rules

Yep, agree. In the end, the goal is a functional and rules conform DC/DC to make it easy for teams to get through the comps.

I will focus in the following on the things where I would like to her further feedback from you.

Regarding glue, it is more for the EV4.5.15.

I think the idea of fixing components in place must be categorised:
A) loosening due to vibration
B) loosening due to self desolder

For A) glue is an easy choice here. No issue with that.
For B) If it can desolder it self - what kind of glue/temperature rating should be used? It must be withstand at least 200°C then, but where is the upper limit?! Especially if it can't desolder it self (like a thermal fuse :D), it might rise way above that temperature.
Also for replacing components, the glue should be not too strong. Loctite 272 can be removed at 250°C - but there is no insulation resistance available. Silicone might be interesting too. Something that loosens at 350°C would be perfect for soldering.
What would you recommend?

I can make the [...] wires for the transformer be required to be glued

wires themselves are explicitly forbidden to solder in the high current path

Yeah, you’re right - It is still odd to me to define the HF-Litz as a wire here. Wire wrapping is out of reach as a simple solution I guess. I could try finding a solderable clamping connector (like for the HV-Input) and then crimp the HF litz wire into a cable-end sleeve. However, the HF-Litz still needs to be soldered before, as every litz wire is enamelled. So not a good solution at all.

Maybe I need to elaborate the initial idea a bit: In the first designs the transformer was mounted with the litz wires being under tension. Also depending on the litz, it was rather sturdy. There was much less movement possible then with other large components (like the output capacitors) - therefore I didn't feel the need to glue down the transformer.
When I think about the rule, that no wire in the high current path is allowed to be soldered, then I believe the issue might be the wire being flexible and not having strain relief, making the connection brittle at some point.
Now for the transformer "wires": If the transformer was glued and the wires are fixed in place (it's all on the same PCB), then there is not a lot room for mechanical movement. The transformer is somewhat a strain relief for the cables.
A good connector is a strain relief for a cable as well, but technically just a short wire in a housing, soldered to a pcb. If I can meet the idea behind a connector with my soldered „wire“ whats the difference then?

Litz wire issues

So what would you recommend in this case? If the converter/wire was housed inside a case which fulfils EV5.4.7 would that be ok? And is there another way to use it, like applying heat shrinking tube around the primary coil?

This is a harder one... The wire itself, especially TS one, is problem on itself with the beforementioned soldering. That would be the problem one, really... Case covering the transformer would be a speculation right now, which I do not want to go into.
Using heat shrink for the cable is definitely possible, but not sure how would you then wrap it around the transformer.
First I would re-evaluate the way wire is used.
For the PCB, it should not be hard to prove isolation properties of used FR-4 and also use standards like IPC-2221B / IPC-9592 - i.e. for isolation in internal layers.

I see.
Heat shrinking the complete cable would be not possible from the required space as far as I can tell without testing it.
The idea was rather to heat shrink the whole transformer with one large heat shrinking tube, and not the cable. So it's isolated against the outside - otherwise as said before, every Inductor would need to have rated enamelled wire.

Thanks for bringing up IPC9592B, I was not aware of this one. It seems better than the IPC2221 which changes in large steps.
Also IEC 60664-1 is implemented as a calculator in KiCAD. Maybe I'll adopt IEC 60664-4 for high frequencies if I can get my hands on that.

Kapton tape is not sufficient

[...] But it is best to design without it, as it has other problems too, like tear, weak sticking properties, sensitivity to surface type / preparation, execution etc.

I'm currently working a bit on the new transformer design and came up with using flex PCBs as isolation between windings (< 50 V between windings). The flex PCBs are held in place by the core, but also by the PCBs which press down on it using glue and/or cable ties (is there an issue with cable ties? ofc having proper temperature rating). But the thing is: The flex PCBs are rather expensive just as an insulator (no copper on them, the windings are on FR4 PCBs). In this case, using Kapton would be much more inexpensive. The sticking aspect of the Kapton tape is not really relevant here too. So why not? FYI: The main insulator will be a 1.2 mm PCB. The Primary/Secondary and ferrite cores are separated by it. This should pass multiple kV insulation with ease.

Creepage from Prim to Sec

• The core of the transformer is not "conductive", but is treated as such for galvanic isolation evaluation, as far as I know.

Just FYI: The transformer core is actually conductive - something in the realm of tens of kΩ. This is the reason, why I designed the 3D-printed Insulator for in between the core and the PCB.

Scruti document issue / Plastik 70

That is why I have linked the manufacturer's website :) Click on "TDS" tab and find Plastik 70 Super there for most up-to-date datasheet.

Urg, I see. TDS/TDB (Ger) is not an abbreviation I was looking for in the first place. Found it, fixed it, will push it in the future. Thanks!

Regarding two coats, how do you ensure that the coating done has enough thickness? I am not saying that just by doing two you ensure you did it correctly... There is still issues with shadows, process errors, etc.
But let's do a model situation, for fun, which could be done during scrutineering :) According to the datasheet, one coat thickness should be (not is) 20 - 40 um. If you do the math with 85 kV/mm dielectric strength, you will not reach 3*TS withstand voltage. Yes, it is not a perfect scenario, disregarding some parameters which would help. But at the same time, even those 20um are not guaranteed. Why not be on a safer side? I would much rather do 3 coats to cover everything properly, than to rely on just one, believing my calculations and cleaning / application process.

Jup, agree! It seems that I had the isolation voltage for one coat memorised falsely (~3kV). Three coats seems a bit overkill to me, but I can life with two. I clean and rinse my PCBs with isopropanol before coating. Maybe I will do three, if I can confirm that the thermals are not affected too much. Thanks for pointing out!

For other components it's not desirable to make cutouts in order to not waste space on the other side of the PCB.

I don't think you will lose any space? Main part I meant was U9 (optocoupler from LV to TS). What you do completely on TS or LV side is not relevant to scrutineering, other than trying to help you out, and possibly T 2.1.1 (but that would be a faaaaaar stretch, I reserve that rule to much more critical things - like some dodgy connectors in TSAC, even though they are LV for example).

Ah, yeah, for the opto it is true. Will do that for the next version.

EV 1.2.1

For the PCBs acting as a TS to LV separation, I have asked about this, but either the link does not work anymore, or my question was left unanswered... But as far as I know this is highly discouraged, mainly for the inspection process I think. If you want to do something like this, I would definitely ask for clarification regarding legality.

Okay, will do a rules question at some point.
In my case, it's pretty much a plain PCB (at least where the transformer is located). It would be possible to show the PCB without the coils/cored and also remove the solder mask, so you can shine through. Not too hard to inspect, or am I overlooking something?

From engineering standpoint, high voltage PCB transformers exist and work. But there are other pitfalls, like the engineering itself (how to calculate it properly, what materials to choose, what failures might happen) and the certification (if any safety is required). I remembered this nice article: https://www.magazines007.com/pdf/High-Voltage-PCDesign.pdf and this site I use as a quick reference sometimes: https://www.smps.us/pcbtracespacing.html

Thanks! Very nice references! Will go through that more in depth after my thesis ;)