There are three different PDN definitions for sky1...
# sky130
a
There are three different PDN definitions for sky130 across open_pdks, caravel and OpenRoad-flow-scripts:
Copy code
source    width  spacing  pitch
open_pdks   1.6      1.7  153.6
Caravel     3.1     15.5  180.0
ORFS        1.6      ???   27.2
The PDN pitch in ORFS is very small, and significantly reduces the available routing resources (especially met5). I want to change it, but to what? The caravel PDN settings are likely the most tested, so should we change both open_pdks and ORFS to match?
@Tim Edwards @jeffdi any thoughts on this? I'm leaning towards just making everything match the caravel PDN settings. @Matt Liberty points out we should do some IR drop analysis if we haven't.
m
I don't know how the caravel PDN setting were tested. Is there some data to see?
t
@Anton Blanchard: In my opinion, they're all wrong. The ORFS pitch is overkill, and the Caravel pitch is too sparse. The one in open_pdks is likely what is being used by default now and is the best of the three, but I would prefer to see bus width of no less than 3um and pitch no greater than about 75um or so. FYI, we don't know of any open source tool that does proper IR drop analysis. I generally work with rules of thumb and back-of-the-envelope calculations, and lots of margin.
m
@Tim Edwards what does OR IR drop analysis need to be 'proper' in your opinion?
t
@Matt Liberty: I think some evidence that it knows the difference between average and instantaneous current draw, and that it has access to the entire power grid and knows where power is being applied. I just don't know much about what OR uses for IR drop and how it's implemented.