SMD Pad Layout for HV (~1000VDC)

[SparkFly]

Hi all,

I'm doing some PCB layout for a circuit and require a capacitor to withstand around 350VDC. I came across a VJ0805A221JXGAT5Z (datasheet http://www.vishay.com/docs/45197/vjhvarcguard.pdf) which is rated at 1000VDC.  However, checking the dimensions on page 3 for 0805 packages, I note that worst case, there will only be 0.38mm between termination pads. When drawn onto a PCB this results in pads which are far closer than what is recommended by various specs, for example IPC 2221B Table 6-1 suggests that 0.38mm isn't suitable for even 50Vpk.

Is it really worth laying out the capacitor pads for the worst case? If not, how should I go about deriving suitable dimensions? Is it worth emailing Vishay asking for their recommended footprint?

Many thanks for your time,

SF

EDIT: Or is this a case of the manufacturer's 1000V+ rating being in idealised, non PCB placement? Ie, the part itself is rated for xV, but because of physical constraints when placed on a PCB you'd never run it at that because of the gap between pads?

[tszaboo]

I have 1mm between 0805 pads in my library. I think you confused the imperial and metric values.
But you are right, IPC2221A (dont have B) states larger distance for 1000V, even with conformal coating. I guess you always need to mill out the PCB below the 0805 capacitor. 1KV clearance is 0.8mm worst case, creepege is 3.2mm.

[texaspyro]

Mount the capacitor vertically with one pad soldered to the board.  Run a wire from the top of the cap to another pad far from the PCB pad.   Also, look for a cap in a larger package... I'd use a through hole device. 

[SparkFly]

Thank you both.

NANDBlog, just ensure I am not making silly mistakes, on page 3 it states minimum length would be 1.8mm, with maximum termination pad lengths of 0.71mm each. So up to 1.42mm of the device could be pads, leaving 0.38mm clearance, correct?

Texaspyro, I'm loathed to mount the caps in such a way; doesn't seem very secure. I guess the only way forward would be to use a larger package, or one with tighter physical tolerances.

[floobydust]

You would not operate with 1kV across an 0805, even with conformal coating IMHO.

I show 1.2mm spacing for 0805 pads but that assumes the part is centered and no overlap. You can get a bit more with custom pads.
SMT Pad spacing: 0805 1.2mm, 1206 2mm, 1812 3.5mm

Very good technical paper on it:
Johanson - Impact of Pad Design and Spacing on AC Breakdown Performance

[tszaboo]

Thank you both.

NANDBlog, just ensure I am not making silly mistakes, on page 3 it states minimum length would be 1.8mm, with maximum termination pad lengths of 0.71mm each. So up to 1.42mm of the device could be pads, leaving 0.38mm clearance, correct?

Actually, that is correct. Though, I think it is only a datasheet issue. Probably a copy from the standard size chart. There is an email to contact.

[SparkFly]

Thanks for the new replies.

Floobydust, thanks for sharing your dimensions and that paper, it was an interesting read. I'm trying to create IPC compliant footprints so will need to look over their recommendations and see how I might be able to incorporate those into the design. Agreed 1kV over 0805 is pushing things, but the tolerances for this part seem so large that designing for worst case seems to be overly limiting. 50V IPC2221B limitation over a 1kV rated device hmm.

NANDBlog, I've sent an email asking for clarification as well as if they may point to any appropriate guidelines etc. I will post here with any new information should it arrive.

Regards,

SF

[texaspyro]

Texaspyro, I'm loathed to mount the caps in such a way; doesn't seem very secure.

Sometimes you do what you gotta do in order to do what you want to do...  and unless you are in a harsh vibration environment it should be fine.   You can always pot the cap / wire in silicone.  Anyway, an 0805 device is NOT the proper device for these voltages.  I did use some 1206 100V caps in an EL backlight inverter circuit and thought that was pushing things.

[floobydust]

Is this 1kV DC all the time (operating) or is it only during transients?
Careful some of those unicorn capacitor ratings are short-term (dielectric test), not continuous operating rating, and they do not include the realities possible on a PCB.

I would design considering safety standards based on IEC 60664-1, which feeds into other standards like 61010, 60730, 60950 etc.
why? It's just a pile of research and experimental data has been done on spacings that are reliable in practice, with aging, to avoid tracking.

This IEC 60664-1 graph which is interesting showing creepage/clearances for high voltages on a PCB, considering pollution degree and PCB laminate (material group).

For pollution degree 2 (non-conductive, dry, no condensation/only temporary) and usual FR-4 laminate CTI <175 (Mat. group IIIb), altitudes under 2,000m, are suggesting minimum creepage of 3mm for 707VRMS/1kVDC, as I read. Just as a number to bounce around, unless you are dealing only with short-term overvoltages, or a safety aspect.
An 0805 (1.2mm pads and lands) roughly 350VDC off the graph.

In a safety function, HV capacitors like X and Y-rated are also subject to mains overvoltages so the spacings are generous with an extra safety-factor applied. They are mainly through-hole parts to achieve the large spacings, as an arc would be disastrous.

I hope this brings to mind some issues to consider for your design.

[SparkFly]

Thanks floobydust, food for thought indeed. The expected 1kV would only be short transients, typically around 120Vexpected.

I have just received a reply from Vishay. They say that whilst the termination pads are unlikely to be so large, the tests were performed in idealised lab conditions. They point out that these caps have a "special  internal construction" to resist arcover. The mention that if I have specific creepage/surge criteria to meet, then this device may not be suitable for the aforementioned reasons. They'd still recommend a conformal coating and inhouse testing of my own.

Seems fair enough.

I think for me, the physical tolerances are just too large for me to guarantee reliable construction and meet safety criteria on this specific device.

[T3sl4co1l]

It's worth noting that ceramic capacitors are ceramic, thus have the best possible CTI.  So they can stand off much more voltage along the surface -- at least, assuming no contamination, metal migration, etching or other chemical effects on the ceramic, and so on.  (Because, while it is ceramic, it's not exactly diamond.  It's fairly soft, as technical ceramics go, and could be more susceptible to chemical etching.  Not that anything normally used in manufacture should have appreciable effect, but contaminants might -- who knows.)

Since the PCB does not have a good CTI, you'd have to route a slot under the cap to use it at ratings, while meeting whichever applicable standard.  That really hurts mechanical stability though -- a slot puts maximum bending force on the terminations, making failure more likely.  (Flex termination type caps would be a good idea!)  It also prevents you from running ground plane immediately beneath the capacitor, if that's a problem (in a switching or RF circuit).

You can also thoroughly wash the assembly (to ensure no flux and voids beneath the cap), and pot it in resin (using vacuum to ensure no voids).  Then you can use 0603s for capacitors and resistors at whatever voltage you like.  (I've seen HV resistors of just such a description, that carefully note in their datasheets that they're only valid when potted.. ).

Tim