Electromigration in VLSI design

[praveen1984]

Hi,
I have a doubt regarding electromigration on nets in VLSI design.
In any SoC design, there are only three nets: the Power/Ground net, Signal net, and Clock net.
My question is which of the net is more prone to electromigration and why?

Please correct me if my question or net assumptions were wrong.

Thank You

 

[dick_freebird]

You can't say one is more prone without considering the
construction and the burden.

However all things being equal except the application (i.e.
equal RMS current density), the supply net will be the worst
because its current is unidirectional (so no partial reversal
of material migration, per cycle, because no cycle).

Next worse would be the clock because while AC, it's highest
cycle rate. Logic is generally a don't-care. We would apply a
2X DC Jmax rule to periodic, reversing currents. For outputs
with signigicant pulsed load, unidirectional legs would be
criticized against the DC value for time averaged current
density, and a 10X DC limit to the pulse amplitude (these
were locally developed rules in a HiRel products organization;
unlikely you'd see the same exactly, but someone should have
published foundry*application-space EM groundrules for you.

 

[praveen1984]

You can't say one is more prone without considering the
construction and the burden.

However all things being equal except the application (i.e.
equal RMS current density), the supply net will be the worst
because its current is unidirectional (so no partial reversal
of material migration, per cycle, because no cycle).

Next worse would be the clock because while AC, it's highest
cycle rate. Logic is generally a don't-care. We would apply a
2X DC Jmax rule to periodic, reversing currents. For outputs
with signigicant pulsed load, unidirectional legs would be
criticized against the DC value for time averaged current
density, and a 10X DC limit to the pulse amplitude (these
were locally developed rules in a HiRel products organization;
unlikely you'd see the same exactly, but someone should have
published foundry*application-space EM groundrules for you.

Thank you Dick for the brief explanation.