HV Resistor layout

[jcconverters]

Hello!

I was wondering why in a Isolation meter (like a High voltage multimeter that checks isolation between HV battery lines a chassis in EV) this strange layout is used in the voltage divider for HV+-, is because of optmizing the space, clearance, creepage... would not be another layout be more appropiated.

Thanks!

[exmadscientist]

Creepage and clearance are the concerns. Also fitting it into the available space as the requirements can call for a lot of distance. Clever designers will make use of the voltage divider's actual division to relax the spacing closer to the end of the division.

This one... does not look clever. Or even sane. I guess they needed that conformal coat to get it working?

[T3sl4co1l]

Looks fine to me.  Seems to be adequate clearance between components depending on position in the string, and it more or less fills the available area.  A more orderly arrangement might take up less space, but if they weren't limited on space, that doesn't matter.  The stray capacitances are surely not going to be very straightforward, but I'm guessing this is low frequency where a flat frequency response is more or less trivial (no compensation needed).  That leaves a large number of available states, that is to say, positions and orientations of the available components, and without any other constraints, the layout shown is about as likely as any other.

Personally, I always apply the constraint that, sure it doesn't matter, but if I can, why not?  Or, if it doesn't affect anything else (dev time included!), make those as optional constraints.  So, a more orderly and symmetrical layout would give a somewhat flatter frequency response, or have better CMRR near cutoff, or other metrics like that; or the constraint to make it look pleasing, symmetrical, that sort of thing (a generally low-entropy goal, except when I do shit like this, https://www.seventransistorlabs.com/Images/AltiumStepRoute.png ); or even to abuse the available entropy and, say, spell out something cheeky with the components/traces.

Tim

[floobydust]

The concern is creepage and clearance between/across resistor strings, the connectors, most of the odd looking layout is to keep generous spacings to the corner mounting screws.
I'm not sure what the operating voltage and surge voltage specs are for an EV here.

I would say it's not a really great design, looks more like something from a worried, inexperienced designer and PCB layout person. The exposed metal (for mounting) on the TE/AMP connectors is a concern, what is that about. There's no conformal coating there, you can see the mask. Oops shoulda had a slot and them further apart. Common noob mistake is thinking there's conformal coating somewhere it isn't, and then using the incorrect spacings.

Vishay MELF MMA 0204 operating voltage is 200V each, 1.5mm pad spacing and 19 of 10k resistors to the divider node... is a bit crazy to make it good for 3.8kV operating? Yes they need to be AEC-Q200 and high stability.

This kind of circuit (voltage monitoring) was a problem with smart energy meters. Without really impulse testing it, if the spacings are adequate is just a theory especially with moisture present.
Smart meters being in OVC Cat. IV would breakdown and arc inside, and because there are no fuses, it started bad house fires. Then UL/CSA realized they should probably have some standard (UL 2375) and get smart meter manufacturers to comply and actually test their products.

It's a good question - to what standard are the EV makers testing to for HV spacings? Safety standards always lag behind so it's likely this is loosy-goosy.

[T3sl4co1l]

Those are standard connectors for CCFL application, I *think*.  Unless they're just bog standard Micro-Fit or whatever.  Not sure.

More to the point, the pads for those mounting nails are much closer than the thru-air distance indicated, and the creepage even closer still.  That's likely the biggest compromise in this design.  Possible there's board routs, invisible under the connectors, but clearly they don't cover the width of the connector so will be of limited value if present.

Tim

[jcconverters]

Thanks to all for the responses!

That is more or less what I was thinking, although I did not realise the connector itself, that is in fact a major flaw. This should work up to 1000V DC. More pictures attached if you are interested.

Thanks again.

[exmadscientist]

What was bothering me looking at this earlier is that its design doesn't really match the component values. (Though those make slightly more sense now that I'm not reading those resistors as 20k when they're really 220k.) Max inputs around 1000VDC makes sense given the values in use. But the physical design looks like it's done (psychotically, but still) for 10kV. You just... don't need to do that for 1kV.

At least the creepage and clearance is probably fine given that. I did one of these for 8kV max inputs and it was a real headache to get the creepage distances OK, get the frequency response decent to 1kV1kHz, and keep the power dissipation under control. It... did not look like this when it was done.

[T3sl4co1l]

Weird, are those not pad rings on the top?  But there's none on the bottom, nor hole plating (which is as I'd expect those nails to be used, well, they can accept solder so a plated hole is more solid, but they're forked so do alright in NPTH still).

I wonder if the designer either had a transient voltage spec they took too literally (resistors are still rated for some transient, the 200V or so figure is nominal not peak), or took the IEC 60950-1 or other calculation based on nominal voltage only (perhaps not realizing this includes some CAT x overvoltage rating) and chose resistor count based on that peak voltage, or perhaps by dividing the creepage figure by the pad spacing to get count.

The connector spacing and metal might not be a fault as line-to-line spacing could be different from line-to-ground (or whatever else the circuit is considered to be at); or, the rules could be easily misinterpreted, once again; who knows.

Tim

[floobydust]

OP, there's no conformal coating on the PC board top-side at the connector's pins, which is a mistake with exposed pins and pads where you expect dirt and moisture. Why bother coating the bottom-side then? Your calipers need to go between metal (not the pins) include pad's OD. It ain't 7.62mm lol.
Why are you handling the board with HVDC connectors plugged in? That's kind of cringe, they must connect to something HV...

We don't know what kind of enclosure the board in housed in, or where it is located i.e under the hood, operating altitude.
Pollution Degree II "Normally, only non-conductive pollution occurs. Occasionally, however, a temporary conductivity caused by condensation may be expected."
1,000V PCB clearances 5.5mm minimum, hi-pot 3.31kVrms as a Cat. II example but you always overdesign for the safety margin and lack of fuses here.

[exmadscientist]

The connector spacing and metal might not be a fault as line-to-line spacing could be different from line-to-ground (or whatever else the circuit is considered to be at); or, the rules could be easily misinterpreted, once again; who knows.

UL (interpreting 61010 at least) doesn't care in the least about line-line, only hot to chassis or hot to accessible parts (which should usually be the same as chassis). This surprised me a bit at first but made more sense the more I thought about it.

(Also, someone else was responsible for interpreting regulatory and talking to our UL guy, so I might have that wrong. Might.)

It is worth mentioning that the creepage/clearance stuff gets kind of weird when a single part fails the rules. Like certain Murata HV resistors do. You can either argue for an exemption because obviously the damn manufacturer didn't make a useless part or just add another board slot and call it done. Or pick a 10x more expensive part instead to save a series of meetings with UL. I forget which option we took, but it wasn't the arguing one, and it was a good choice.

[floobydust]

UL does care about insulation line-line in polyphase and a certifier can and will bridge any two points at their discretion inducing a fault.
Because it's not only the shock hazard they look for with insulation failure, it's the fire hazard as well from lack of a fuse or too big a branch feed.
Are the battery connections to the board fused? To what current? If an arc happens, you'd like it to clear voltage-sense fuse(s) and not burn up the board, car and garage/house.

EV HV DC bus is not chassis grounded, that I have seen. I believe the isolation was to give an extra level of protection against shock and confuse all mechanics until the end of time.
Although OP's resistors likely ruin that, there is a low value path. It might be a requirement we don't know. I'd expect an artificial 1/2 battery voltage to chassis ground.

There should not be the same transient overvoltages as seen from mains on the DC bus but I see no harm designing using the tried and true. It's a few extra mm.
Fluke multimeter rated 1,000V Cat. III design uses 8.5mm spacings and clever 4-layer traces. But applying 61010 to a vehicle? That's the loophole- I've seen a lot of automotive skirt safety standards because they are not a building/permanent installation apparently.

[T3sl4co1l]

EV standards are probably a very new and active area of development, anyway... not that they're actually very new things, they've been around literally forever, just not to commercially significant levels -- what's new is the market shift, simultaneously towards lower production costs (better profits from quantity production) and to higher regulatory thresholds (partly for worker and user safety, or ostensibly so, but also as barriers to entry against new manufacturers!).

But that's just my assumption, that along with the increase in available onboard electrical power, also comes standards for working with them.

In any case, clearly there's no more than the usual automotive transients on the battery -- very different from nominal mains voltage plus category transients environment seen in traditional standards like 60950, 61010, etc.

And I wouldn't be surprised if the usual automotive transients indeed have to be modified, as for example the impulse from a battery contactor opening/closing under load (or into supply bypass caps, etc.) will be different levels from traditional relay or starter transients.

Tim