Exceeding a resistors rated voltage

[coppice]

Multiple resistors in series to match the expected voltage spike?

That's a very common and effective technique, but its prohibited by regulations in a few applications, so take care.

Huh, got any cites?

Tim

I haven't hit this directly, but I have several instances of customers looking for alternatives to a chain of resistors, because they had problems with approvals. Two examples are German smart meters (energy meters in most countries use a chain of several SMD resistors), and a power quality meter that had issues with UL approvals when it used a chain of resistors.

[Simon]

need a what ?

[coppice]

I would have thought that if you are pushing resistors in series to increase their voltage rating would need to make sure they are placed in a straight line end to end increasing the distance between 2 points where the potential is applied.

They don't need to be all in a straight line, but you do see some very dumb designs, like a zig-zag layout which brings every second connection close together.

[Simon]

I would have thought that if you are pushing resistors in series to increase their voltage rating would need to make sure they are placed in a straight line end to end increasing the distance between 2 points where the potential is applied.

They don't need to be all in a straight line, but you do see some very dumb designs, like a zig-zag layout which brings every second connection close together.

yes that is what I meant by head to tail I high voltages would just jump across the board from top to the bottom of the ladder or across sections. After all a a higher rated resistor is nothing more than the same device in a bigger package which increases the distance between the entry points so I can't see a lot against replicating that with multiple parts particularly where you are just adding a little safety margin and peace of mind for another 2p.

[tautech]

Not everything you see in manufactured electronics equipment, would be a good idea to copy. There are many cases of poor design, unfortunately.

That is also true I have seen some pretty horrific things myself some of them designed for safety critical systems. Yes as I said the standard resistor voltage ratings are pretty much what is required but I'd like to be a little bit more resilient. As the maximum voltage I expect is 250 V a 250 V part on paper is fine but as we all know what on paper does not always describe real-life so I think I will be fair in saying that if I put 2 resistors in series no single resistor would see more than two thirds of the maximum applied voltage if I lay my board out properly which would give me a reasonable derating of the component. I suppose you could use an inductor but wouldn't the windings need to be rated for voltage? Surely if you have an inductor the voltage will simply jump from winding to winding rather than go round the winding.

I usually follow my input resistors with a capacitor or and a TVS diode I'm still trying to get my head around how much voltage and current you would actually get through the resistor considering the actual amount of power in a spike which at 250 V would not be enough to produce enough current through the resistor and or a 250 V voltage drop.

2/3rds is a good rule of thumb. 

[MK14]

I would have thought that if you are pushing resistors in series to increase their voltage rating would need to make sure they are placed in a straight line end to end increasing the distance between 2 points where the potential is applied.

They don't need to be all in a straight line, but you do see some very dumb designs, like a zig-zag layout which brings every second connection close together.

yes that is what I meant by head to tail I high voltages would just jump across the board from top to the bottom of the ladder or across sections. After all a a higher rated resistor is nothing more than the same device in a bigger package which increases the distance between the entry points so I can't see a lot against replicating that with multiple parts particularly where you are just adding a little safety margin and peace of mind for another 2p.

Where it is just the second resistor, and the first already meets the intended maximum voltage rating. I agree the 2p adds some extra protection and peace of mind.

Where resistor stacks somewhat badly fall to pieces, is when E.g.
It is a 10,000 Volt supply/transient and there are 10, 1000 Volt rated resistors.
Ignoring the PCB/inductance and PCB capacitance (to simplify the explanation). The variations between the different stray capacitances of the different resistors, can mean that the 1000 Volts x 10, are NOT evenly split during very rapid transients and/or switch on/off events.

Hence (I might be exaggerating a little bit), one of the resistors may get a disproportionately high portion of the voltage. So one might get 2000 Volts (they are ONLY rated for 1000V each), because it happens to have the lowest stray capacitance value. So in time these 2000V (I accept I may be way off here. But anything over 1000V is potentially bad datasheet wise anyway. Ultimately it depends on the stray capacitance variations and other stuff) transients, can break down the resistors insulation.
If you exceed the maximum voltage of a resistor, it can concentrate most of the voltage on a VERY small part of the insulator, and break it down. In time the broken part of the insulation (which now potentially conducts), GROWS. Eventually compromising the entire resistor e.g. making it somewhat conduct. Hence the other 9 resistors will get almost all the 10 KV, which in time will blow another of resistors. So in time all/most of the resistors can/will fail.

Because you are specifically talking about having only two resistors, and hopefully keeping well below the (incorrect) 2 x maximum working voltage. You should NOT have this problem.

[Simon]

Yes I agree I would not take the maximum voltage applied and divide it by the maximum voltage the resistor could take to get the amount of resistors in the network. If you are talking 10 resistors I would probably go with 2 to 3 times expected voltage for some safety margin. If you are starting to need so many resistors it is probably time to just use a higher rate resistor.

[T3sl4co1l]

I haven't hit this directly, but I have several instances of customers looking for alternatives to a chain of resistors, because they had problems with approvals. Two examples are German smart meters (energy meters in most countries use a chain of several SMD resistors), and a power quality meter that had issues with UL approvals when it used a chain of resistors.

As I recall, IEC 60384-1 says cumulative gaps are okay, so long as the single-point-failure total is equivalent.  That is, if one resistor has a rating and gap suitable for 250V, then you need N = 1 + Vnom / (250V) number of resistors.  So for 500V, you'd use 3 or more.

As others noted above, you need to make sure that's true at all frequencies of interest, either by filtering (attenuation) or bypassing (C divider in parallel).

And,

They don't need to be all in a straight line, but you do see some very dumb designs, like a zig-zag layout which brings every second connection close together.

if they're doing dumb stuff like this, it's their own fault really

Tim

[coppice]

As I recall, IEC 60384-1 says cumulative gaps are okay, so long as the single-point-failure total is equivalent.  That is, if one resistor has a rating and gap suitable for 250V, then you need N = 1 + Vnom / (250V) number of resistors.  So for 500V, you'd use 3 or more.

That's the rule most people use. A lost of approvals test programs require the resistors to be shorted one by one to demonstrate the design really supports that SPOF requirement.

[Cerebus]

As others noted above, you need to make sure that's true at all frequencies of interest, either by filtering (attenuation) or bypassing (C divider in parallel).

What people tend to forget is what I might call the "frequencies of non-interest". If someone is working on a low frequency circuit they tend to put all issues associated with higher frequencies out of mind. So you bumble around designing your circuit ignoring small parasitics of a few pF here, 20 nH there and so on because "It's only audio frequencies". Then your circuit encounters a 500V/us transient coming in via the mains lead...

[T3sl4co1l]

Heck, from what I've seen, people have a hard enough time making things work at DC...

Tim