How to create a "fuse" in a PCB?

Can anyone share their experience(s) with designs that have failed due to
excessive heat caused by excessive current in their PDN? Iâ??ve been reading
IPC-2152 and Doug Brooksâ?? articles and wonder if you could create a power
delivery shape that would provide adequate voltage to the load(s) and yet
heat excessively. From what Iâ??ve read, the thermal conductivity of copper
and FR4 is such that, for any reasonable case, the heat generated would be
dissipated by adjacent copper and FR4. I suspect the active components
would have much more influence on the PCB temperature than the passive PDN
shapes, but donâ??t have any concrete information on that. Taken to its
extreme, I wonder what would happen if you had a 1-micron wide trace that
was only 1-micron long, with large planes on either side of that trace.
Its short length keeps it from being highly resistive, but the current
density is obviously very high in that 1-micron trace. While there may not
be much voltage drop, would the heat generated be dissipated by the
adjacent large copper shapes which have very low current density
(generating very little heat)?


Right now I suspect that most shapes which provide adequate voltage to the
loads also provide adequate thermal dissipation. I can envision some
scenarios where this might not be true â?? an outer layer thatâ??s been highly
perforated by BGA anti-pads, underneath a very hot part (or perhaps next to
some very hot parts), for instance. But, Iâ??d like to hear from others who
have had experience in this area. How have you created a fuse
(accidentally or on purpose)? What methods have you used to mitigate
excessive heat in your PDN, apart from using minimum trace width
constraints in IPC-2152?

I also welcome references to published works in this area.



---------------- ------------

| |

| |

| |

| |

| |

| |

big plane | | big plane

-------

short, narrow trace

-------

| |

| |

| |

| |

| |

| |

| |

---------------- ------------



Thanks in advance,

*Jeff Loyer*

Signal and Power Integrity Product Manager, Altium

jeff.loyer 5 years 2 months 22 days

6 answers


The best answer


You can select the best answer for current question!
Answered bym.mansfeld 5 years 2 months 22 days
On 27 Jan 2016 at 10:14, Jeff Loyer wrote:

Can anyone share their experience(s) with designs that have failed due to
excessive heat caused by excessive current in their PDN?

Exactly this topology with a thin thin trace between wonderful
planes... been there, done that (almost...) ML 4 layers, split planes
for 2..3 different supply voltages, everything perforated with many
vias, and thus (via clearances and plane shapes together) formed an
unintentional fuse.... Normal DRC and connectivity checks found
nothing, but luckily I saw it during final visual checks on final
gerbers...... I guess with 1 oz copper the "fuse" had eventually
blown by charge current of two really fat elcos in the
neighbourhood.......

Best regards
Matthias Mansfeld

PS.: I could tell also stories about intentionally made "fuse"
traces, which worked absolutely perfect for the intended purpose (to
get a more or less controlled, spectacular but encapsulated
self-murder on a frequency inverter instead of relying on a properly
rated fuse in the power distribution...)
Answered bykent.h.zhang 5 years 2 months 22 days
I reviewed a design a couple of years ago. At first one out of the 5-6 boards 
started to burn after a few hours of power. Then the 2nd, then the third. It
was root caused to be a 4 mil trace on the path of a few Amperes of current. It
was a design miss. The short net was not named in schematic and was not
assigned power property. I don't have evidence but I suspected that the trace
was heated, melted, continued to generate heat until the liquid copper had a
place to go, then stopped with an open.
Kent Zhang
Hardware Engineer, Intel

-----
Answered byjrbarnes 5 years 2 months 22 days
Jeff,
Doug Brooks just had an article published on this subject:
Brooks, Douglas G., "Empirical Results of FUSING TESTS," Printed
Circuit Design & Fab/Circuits Assembly, vol. 31 no. 1, pp. 22-25,
Jan. 2016.
which may be downloaded from
http://pcdandf.com/pcdesign/index.php/magazine/
10546-thermal-overload-1601

John R. Barnes (retired)
Lexington, KY
Answered byleeritchey 5 years 2 months 22 days
I saw a demo board from a noted IC manufacturer that set fire to itself because 
there was not enough copper in the planes under a high pin count BGA.


-----
Answered byjeff.loyer 5 years 2 months 22 days
I apologize â?? it seems I wasnâ??t clear. I was being sarcastic with the term
â??fuseâ? in the title. I donâ??t intentionally want to create a fuse in a PCB
design. I was wondering if folks had inadvertently created them (had
designs that burned either the dielectric material or their conductors).


Sorry for the confusion.

*Jeff Loyer*

Can anyone share their experience(s) with designs that have failed due to
excessive heat caused by excessive current in their PDN? Iâ??ve been reading
IPC-2152 and Doug Brooksâ?? articles and wonder if you could create a power
delivery shape that would provide adequate voltage to the load(s) and yet
heat excessively. From what Iâ??ve read, the thermal conductivity of copper
and FR4 is such that, for any reasonable case, the heat generated would be
dissipated by adjacent copper and FR4. I suspect the active components
would have much more influence on the PCB temperature than the passive PDN
shapes, but donâ??t have any concrete information on that. Taken to its
extreme, I wonder what would happen if you had a 1-micron wide trace that
was only 1-micron long, with large planes on either side of that trace.
Its short length keeps it from being highly resistive, but the current
density is obviously very high in that 1-micron trace. While there may not
be much voltage drop, would the heat generated be dissipated by the
adjacent large copper shapes which have very low current density
(generating very little heat)?



Right now I suspect that most shapes which provide adequate voltage to the
loads also provide adequate thermal dissipation. I can envision some
scenarios where this might not be true â?? an outer layer thatâ??s been highly
perforated by BGA anti-pads, underneath a very hot part (or perhaps next to
some very hot parts), for instance. But, Iâ??d like to hear from others who
have had experience in this area. How have you created a fuse
(accidentally or on purpose)? What methods have you used to mitigate
excessive heat in your PDN, apart from using minimum trace width
constraints in IPC-2152?

I also welcome references to published works in this area.



---------------- ------------

| |

| |

| |

| |

| |

| |

big plane | | big plane

-------

short, narrow trace

-------

| |

| |

| |

| |

| |

| |

| |

---------------- ------------



Thanks in advance,

*Jeff Loyer*

Signal and Power Integrity Product Manager, Altium

Answered bymovax 5 years 2 months 22 days
On Wed, Jan 27, 2016 at 10:14 AM, Jeff Loyer  wrote:
Can anyone share their experience(s) with designs that have failed due to
excessive heat caused by excessive current in their PDN? I’ve been reading
IPC-2152 and Doug Brooks’ articles and wonder if you could create a power
delivery shape that would provide adequate voltage to the load(s) and yet
heat excessively. From what I’ve read, the thermal conductivity of copper
and FR4 is such that, for any reasonable case, the heat generated would be
dissipated by adjacent copper and FR4. I suspect the active components
would have much more influence on the PCB temperature than the passive PDN
shapes, but don’t have any concrete information on that. Taken to its
extreme, I wonder what would happen if you had a 1-micron wide trace that
was only 1-micron long, with large planes on either side of that trace.
Its short length keeps it from being highly resistive, but the current
density is obviously very high in that 1-micron trace. While there may not
be much voltage drop, would the heat generated be dissipated by the
adjacent large copper shapes which have very low current density
(generating very little heat)?

Hi Jeff --

I've explored the idea in the past for one of our applications -- the
two primary analysis equations I used were from Preece and Onderdonk.
I believe tools like the Saturn PCB Toolkit implement Onderdonk's
equation to estimate fusing currents and est. burn times. I decided
not to go this route because of how dependent on manufacturing
tolerances this is, and in general the relative imprecise nature of
it.

It's a cool trick to pull-off, but for any real safety application a
real fuse is a must.

Krunal