This is my first time playing with high voltages and I want to do it properly, so here is my question:
Say you have this RSFJL diode in attachement, good for up to 600V. For a PCB material the creepage distance for 600V should be more than 3.2mm, but the diode has only 1.9mm between pins.
Do you take the manufacturers word that that distance is enough for years of having 600V across it, or do you take the required creepage into account and use somthing in a much larger package? And what CTI do you use for the creepage distance tables, when the manufacturer doesnt specify one?

https://www.ptr.eu/fileadmin/template/ptr/media/images/informationen/Kriechstrom_Tab_4_eng.jpg table with creepage distances

 

This is my first time playing with high voltages and I want to do it properly, so here is my question:
Say you have this RSFJL diode in attachement, good for up to 600V. For a PCB material the creepage distance for 600V should be more than 3.2mm, but the diode has only 1.9mm between pins.
Do you take the manufacturers word that that distance is enough for years of having 600V across it, or do you take the required creepage into account and use somthing in a much larger package? And what CTI do you use for the creepage distance tables, when the manufacturer doesnt specify one?

https://www.ptr.eu/fileadmin/template/ptr/media/images/informationen/Kriechstrom_Tab_4_eng.jpg table with creepage distances

You need to take into account pollution degree, which is a function of the environment in which the circuit is intended to be used.

If you cannot meet the minimum creepage distances for specified pollution degree (regardless of component geometries), you must a) change the environment where the circuit is located so as to change the "local pollution degree" (i.e. sealed enclosures, air filters, etc.), or b) conformally coat your PCB.

 

How about a slot cut under the diode?

 

I am aware of that coating the whole pcb would be a solution, as it lowers the PD. The circuit is a part of a welder inverter, so I would rather go with PD3 if possible. Mind you this is my personal project and not some big production issue, but still I ´d like to do it properly.

But take for example the main IGBTs which are in TO247 (STGW80V60DF) where the worst case creepage is 2.4mm between pins. Would you also coat this part of the package with conformal coating to get lower required creepage?
Again the manufacturer doesn´t state the CTI, so I have no idea which number to choose for PD2.
I don´t really understand how such a part is supposed to be used, at 600V I would have to manage to get to PD1 if I wanted to fulfill the creepage distance with this package...

 

How about a slot cut under the diode?

A slot in the PCB is not a problem, but it doesn´t help with the fact that the creepage is insufficient over the body of the diode.

 

Helpful?

 

Thanks, it is a bit helpful as I didn´t consider the RMS, so my actual creepage will drop.
But still, theoretically if that FET has 600V across it 99% of the time, the RMS will be 600V and i would need to somehow pot or coat the bottom part of the package to get to PD1 where the creepage is enough?
I have never seen such thing done to these large packages, so either the manufacturers dont´t care or don´t know about this, or am I missing something?

 

I have never seen such thing done to these large packages, so either the manufacturers dont´t care or don´t know about this, or am I missing something?

It only counts if you are trying to get a UL or similar rating.

 

The 3.2 mm creepage you are referring to is most likely a specification for safety purposes, and far more than required otherwise. Half a millimetre is more than enough for 600 volts, if the board is clean.

I can't remember the manufacturer at the moment (I'll try to find it & post again later if I can), but a maker of high voltage surface mount resistors published some very useful info on work they had done to evaluate best practice for mounting high voltage devices (things like using rounded corners on pads, no solder mask under the parts, etc.)

Conformal coating of surface mount boards is far from easy if you require penetration of the coating under parts, especially given that there may be flux residue under the parts. No-clean flux is incredibly hard to remove even when it is fully exposed. Water soluble flux, which is really easy to remove where it is exposed, provided cleaning is done promptly, is hard to remove from under SMDs, and requires high pressure water impingment to do a good job. Once the flux is gone it is hard to get conformal coating under the parts without using a vacuum chamber.

 

Thank you ebp, all very valid points that I didn´t realize till now. I know that for my toy-project design this is a non-issue and I can do it however I please and it will still work if I just observe the clearance (yes for longer term reliabillity I want to get much better than that).

But for my actual job there most likely will be times where I come to such border cases where the package is not good enough, and this really helps me realize what possibilites and obstacles there are and how to get over this.

 

When things like this are spec'd as RMS, it is almost invariably assumed that the waveform is a "pure" AC sine. It is the peak voltage that is of concern, not the ability to dissipate heat in a resistor.

There are tonnes of things like variable frequency drives made that have small creepage distances. They get safety agency approval.

Among the things I hate in electronics are safety standards. Many are abominably badly written and difficult to interpret. That should NEVER be the case, but I think a lot of this stuff was put together by committees of aged electricians.

 

When things like this are spec'd as RMS, it is almost invariably assumed that the waveform is a "pure" AC sine. It is the peak voltage that is of concern, not the ability to dissipate heat in a resistor.

I have to disagree here, peak voltage or impulse voltage is relevant for air breakdown and resulting arcing and flashover, therefore it is used to define for clearance.
Creepage and tracking on the other hand really do depend on the RMS, because the more voltage for longer time you apply, the more the material degrades and little parts of it become charred and conductive, making the voltage gradient across the remaining isolator higher.
And it is not average but RMS (i´m guessing) because the closer you get to the limit the faster the degradation occurs, therefore the squared relationship seems appropriate.

 

A slot in the PCB is not a problem, but it doesn´t help with the fact that the creepage is insufficient over the body of the diode.

For the creepage distance over the body of the diode you can rely on the PIV rating of the diode.

 

Here's the paper I was thinking (wrongly) of - it's from Johanson Dielectrics (I scared this up when looking for capacitors for telephone line coupling - they need to have a high voltage rating and the circuit I was working on was to be surface mount):
https://www.johansondielectrics.com/downloads/jdi_arc-design-observations_2006-03.pdf
Have a look at johnason's web site for other similar works.

I don't think we disagree, I think you haven't understood what I'm trying to say.
My point regarding RMS voltages is that if a creepage distance is specified based on RMS, the implication is almost always that a pure AC sine wave is assumed by the specification and hence the peak voltage, which does indeed matter, is well-defined with respect to the RMS voltage specified. A 10 kV pulse with a width of 10 ms (equal to a half-cycle at 50 Hz) and a duty cycle of 0.1% has an RMS value of about 316 volts, but clearly is of much greater concern with regard to creepage and clearance than 316 volts DC or a pure AC sine wave of 316 VRMS. Specifying a limit as an RMS value makes no sense UNLESS a pure sine wave is assumed or additional information is supplied. The peak value matters, but assuming a pure sine wave gets you to the peak value by implication.

RMS will matter once a conductive track has been formed, but by that time it's way too late and a voltage a great deal lower than the initial failure-causing event can cause cascading damage. I've seen lots of PCBs with holes with the epoxy completely burned out by excessive heating by components and possibly by current through the resulting carbon. I've never seen one where it looked like that was initiated by arcing. I have seen several boards that were the victims of lightning strike - they tend to have copper tracks completely blown off with little damage to the substrate and have parts blown to bits. I've seen a few boards with tracks evaporated by excessive current, but that's a different matter.

Some safety regs allow smaller creepage and clearance distance for the same voltage if the available power is limited.
Some safety regs will allow a high-pot test as an alternative to demonstration of adequate creepage and clearance distance. I certainly would not allow that. I know of a board that received safety agency approval that had provision to use 0603 zero ohm jumpers to power the board from a low voltage DC supply instead of AC mains - instead of installing the transformer the jumpers were used. When AC mains was used, the creepage and clearance between the line and what should be the safe low-voltage part of the circuit was the gap between the pads for the jumpers - about half a millimetre. It must have passed based on a high-pot test. I would have denied approval and sent the fools back to learn a thing or two. I've seen lots of stuff from China that wouldn't meet safety regulations anywhere.