[Magic-dev] Extracting cross coupled capacitance with sidewall

[Jeff Sondeen]

Hi, yeah that's the historic workaround.

think of the sidehalo value as the distance beyond which no side
capacitances are accounted for, but within which perhaps mulitple
capacitances are counted (since shielding isn't implemented).

however, note that the over-counted sidehalos are reduced by a factor
of their distance (1/S) so the "overcounting" converges to 0.

but the bigger problem with the magic capacitance model is that the
side (edge) capacitance goes only "down" to whatever layer is
immediately below.  there's no edge capacitance calculated to an upper
layer, nor to a lower layer shape even 1 lambda over.  but since such
edge capacitance factors are not modelled as closely with a 1/S
dependence anyway, a more complicated set of parameters would be
needed, such as a set of (distance, factor) pairs which would be
interpolated. also, the fact that edge and sidewall factors track each
other would also probably require the sidewall factor(s) to also be
(distance, factor) pairs.

another approach is to create different sets of extraction parameters
(for both the edge ("side") and sidewall capacitance factors) to match
the "density" of your layout, meaning the average widths and spacings
of wires, perhaps naming such sets "dense" (close, so small sidehalo),
"sparse" (far apart -- so large sidehalo), etc.

maybe magic could automatically pick the set of parameters with the
best matching "sidehalo" for the current cell's density ?

/jeff

Graham Petley writes:
 > Hallo,
 > 
 >    Back on Feb 18 this year I queried Magic extraction for cross coupling
 > capacitances when there is a shielding wire between.
...

 > Magic and the workaround is to make sure the value of sidehalo is less than
 > (2*track_pitch-wire_width) for the tightest pitch, probably poly.
 >