High Current Trace Widths - PCB Design

[NickP]

I have designed a protection circuit for some outdoor equipment that
could be exposed to lightning strikes. The discrete components are PCB
mountable but I can't find any guidelines for sizing the traces for
high current from a strike. Any references for determining trace
widths able to handle 10kA for a few microseconds would be greatly
appreciated.
Nick

 

[FrankW]

You're kidding, right?

 

[Spehro Pefhany]

I have designed a protection circuit for some outdoor equipment that
could be exposed to lightning strikes. The discrete components are PCB
mountable but I can't find any guidelines for sizing the traces for
high current from a strike. Any references for determining trace
widths able to handle 10kA for a few microseconds would be greatly
appreciated.
Nick

My copy of the _Standard Handbook for Electrical Engineers_ contains
the following equation:
credited to I. M. Onderdonk.

I = A * sqrt( log10 ((Tm -Ta)/(234 + Ta) + 1) / 33*S)


where I = current in Amperes
A = area in circular mils
Tm = melting temperature of copper = 1083°C
Ta = ambient temperature in degrees C
S = time current is applied in seconds

But..

1) Who knows how applicable this is to a PCB trace with such
short duration rather than the relatively long durations and
round or roughly square conductors that it was designed for?

2) I have doubts about the size of your lighting stroke
being at all representative. The National Lightning Safety
Institute mentions 200,000A strikes with 300us duration.
That's more than 3 orders of magnitude larger..

Best regards,
Spehro Pefhany

 

[John Woodgate]

I read in sci.electronics.design that Spehro Pefhany <Spehro@Pefhany.?>

2) I have doubts about the size of your lighting stroke
being at all representative. The National Lightning Safety
Institute mentions 200,000A strikes with 300us duration.
That's more than 3 orders of magnitude larger..

That's a direct strike; only cost-no-object equipment will withstand
that.

10 kA is within the range of indirect lightning strike - where it misses
by a few metres to half a kilometre. The 10 kA comes from induced and
conducted surges. It is possible to protect against, but it may require
2 or three stages of protection. ITU-T K20, K21 and K22 give background
information on this subject. E-mail [email protected] for
information. You can download three ITU documents a year for no charge.

 

[Active8]

I read in sci.electronics.design that Spehro Pefhany <Spehro@Pefhany.?>


That's a direct strike; only cost-no-object equipment will withstand
that.

10 kA is within the range of indirect lightning strike - where it misses
by a few metres to half a kilometre. The 10 kA comes from induced and
conducted surges. It is possible to protect against, but it may require
2 or three stages of protection. ITU-T K20, K21 and K22 give background
information on this subject. E-mail [email protected] for
information. You can download three ITU documents a year for no charge.

Polyphaser's tech papers have some good info on lightning
protection from induced and conducted surges as well as tower
protection.

 

[Tim Shoppa]

I have designed a protection circuit for some outdoor equipment that
could be exposed to lightning strikes. The discrete components are PCB
mountable but I can't find any guidelines for sizing the traces for
high current from a strike. Any references for determining trace
widths able to handle 10kA for a few microseconds would be greatly
appreciated.

Generally all the "signal" and "power" lines go through surge suppression
at the point of entry (preferably all external connections go through
suppressors threaded into big thick solid metal plate which is hooked
to an equally impressive ground) and then they go to the PC board.

If you have any currents of 10kA flowing anywhere on the PC board,
you're already done for.

Tim.

 

[R.Legg]

I have designed a protection circuit for some outdoor equipment that
could be exposed to lightning strikes. The discrete components are PCB
mountable but I can't find any guidelines for sizing the traces for
high current from a strike. Any references for determining trace
widths able to handle 10kA for a few microseconds would be greatly
appreciated.
Nick

Size printed wiring conductors to have the same current capability as
is used in the protective component leadwire. Keep tracks short, loop
area small and place protection devices physically close to the board
input terminals.
Gas tubes for this service typically depend on 1mm dia (20AWG) plated
copper, when not located in sockets.

RL

 

[N. Thornton]

[email protected] (NickP) wrote in message news:<69fb664b.0401291144.

If you have any currents of 10kA flowing anywhere on the PC board,
you're already done for.

Indeed. You'd have to use a PCB with a copper layer several mm thick
at least. Otherwise you might want to count on conduction through the
vapourised copper than will hang around for your 3us.


Regards, NT

 

[Stefan Heinzmann]

Indeed. You'd have to use a PCB with a copper layer several mm thick
at least. Otherwise you might want to count on conduction through the
vapourised copper than will hang around for your 3us.

I was going to try figuring out how much copper cross-section area you'd
need to heat the copper 100 degrees above ambient with a 10kA pulse of
3µs duration, using the thermal mass of copper alone. But then I thought
that the frequencies involved would invoke the skin effect and hence
probably vaporize the fringes of the track anyway, no matter how wide
(or thick) it is. Is that a reasonable assumption?

 

[Leon Heller]

I have designed a protection circuit for some outdoor equipment that
could be exposed to lightning strikes. The discrete components are PCB
mountable but I can't find any guidelines for sizing the traces for
high current from a strike. Any references for determining trace
widths able to handle 10kA for a few microseconds would be greatly
appreciated.
Nick

Surely some form of external protection would be better - several
suitable devices are available.

Leon

 

[Fred Bartoli]

I was going to try figuring out how much copper cross-section area you'd
need to heat the copper 100 degrees above ambient with a 10kA pulse of
3µs duration, using the thermal mass of copper alone. But then I thought
that the frequencies involved would invoke the skin effect and hence
probably vaporize the fringes of the track anyway, no matter how wide
(or thick) it is. Is that a reasonable assumption?

I've done calculations, just ouf of curiosity, instead of staying on a gut
feel :
For a 10us, 10kA pulse and a standard 35um copper, considering that the
process is strictly adiabatic for such a low pulse duration, we have a 37cm
wide for a 10K temp rise and a *huge* asumption : uniform current density.
Not too bad.

Using a 105um, your 3us pulse, and a 40K rise that gives a small 1cm wide.

That's, of course for a *single* pulse.

I'll be more concerned about mechanical stress : this kind of current is the
one that violently "shakes" the wires when power installations fails
short-circuited. The pulse length is much longer though, but also the wire
mass, so...


thanks,
Fred.

 

[Spehro Pefhany]

I've done calculations, just ouf of curiosity, instead of staying on a gut
feel :
For a 10us, 10kA pulse and a standard 35um copper, considering that the
process is strictly adiabatic for such a low pulse duration, we have a 37cm
wide for a 10K temp rise and a *huge* asumption : uniform current density.
Not too bad.

Using a 105um, your 3us pulse, and a 40K rise that gives a small 1cm wide.

That's, of course for a *single* pulse.

I'll be more concerned about mechanical stress : this kind of current is the
one that violently "shakes" the wires when power installations fails
short-circuited. The pulse length is much longer though, but also the wire
mass, so...

Plugging the numbers into I. M. Onderdonk's equation, we get a
required cross-sectional area of only 100 circular mils for actual
melting of the copper at 25°C ambient and a 5usec 10kA pulse. That's
73 square miles, and with 1-oz copper (0.0014" (?)) then the width
would be 52 mils. If that's the actual requirement, and if the
equation holds up, then a 125 mil trace on 1-oz copper ought to do it
with some margin. BTW, lightning damage is interesting, I've seen it
damage components without touching relatively narrow traces, unlike
shorts to the mains, which tend to vaporize as much of the trace as
the juice can arc over.

For those FAQ on designing a crude fuse to be etched into the copper
of a PCB, the equation ought to be sufficient, at least to get to the
testing point.

BTW, Onderdonk's work on determining fusing current is referenced in
ASTM "F855-97e1 Standard Specifications for Temporary Protective
Grounds to Be Used on De-energized Electric Power Lines and
Equipment".

You can also find some *steady-state* fusing currents for various
metal wires on my page at http://www.speff.com. I'll add the Onderdonk
equation when I get a chance.


Best regards,
Spehro Pefhany

 

[N. Thornton]

I was going to try figuring out how much copper cross-section area you'd
need to heat the copper 100 degrees above ambient with a 10kA pulse of
3µs duration, using the thermal mass of copper alone. But then I thought
that the frequencies involved would invoke the skin effect and hence
probably vaporize the fringes of the track anyway, no matter how wide
(or thick) it is. Is that a reasonable assumption?

Hmm... the top and bottom are skin too, so I wouldnt expect that
effect per se to kill it. 6kA can blow 2.5mm^2 mains cable out of the
walls, so if Fred's right I guess it must have been far longer lasting
than 3uS. I'd be surprised to see a PCB handle that though.


Regards, NT