Layer parasitics setup

[Blebowski]

Hi,

I am trying to bring-up ORFS on a proprietary sub 100 nm PDK. I am following the guide in:
https://openroad-flow-scripts.readthedocs.io/en/latest/contrib/PlatformBringUp.html

I am wondering about the "setRC Configuration" part. .

I have following questions:

• The manual says that "liberty units needs to be used". Does that mean that all resistance values shall be scaled by the Resistance unit given by pulling_resistance_unit ?
• The resistance is claimed to be "per-unit length". The .lib file does not contain length units (obviously), so is that assumed to be always um , or is it some other unit ?
• The PDK documentation I have does not define any per-layer capacitances between the metals (nor any dielectric constants) to compute such "per-unit" capacitance.
• Is the capacitance / resistance really a "per-unit-length" value ? Isn't it rather "per-square-micron" value ?

Thanks in advance for any feedback.

[maliberty]

Yes pulling_resistance_unit (poorly named)
Liberty does have distance_unit to specify the units (apparently not so obvious). They default to um if not specified
If the PDK doesn't give you the information then what information do you have? You must get parasitics somehow.
Yes it really is per-unit-length - the assumption is for min width.

[Blebowski]

Thanks for quick reply.

I had no clue liberty has distance_unit, thought no "physical layout" information was present in this file. Anyway, my
libs do not contain these, so it is um.

Well, we did outsource PNR, and now we are making some attempts with OR, so this is first time we are actually
setting up PNR. Therefore we did not do parasitics extraction so far. And we are just receiving information from the
foundry step-by-step so to say. I will try to get this information again.

[maliberty]

Often the design rule manual will have some typical res/cap values that you can use.

[Blebowski]

I need to dig deeper, since the ERC document I have available, contains only resistances for metals.

Also, what is the "minimum length" ? Is it simply what is in the WIDTH attribute for the given metal layer
in the tech-lef file ?

Btw. how does it work with the capacitances ? For a given layer, the actual capacitance heavily depends on
"what is below and above" right ? Are such numbers simply assuming the layers N+1 and N-1 are fully filled
with a metal with opposite potential (therefore e.g. as-if having a microstrip transmission line) ? The metal
layers +-1 have perpendicular routing directions, therefore the "material above and below" will alternate
between dielectric and conductor. Is this somehow reflected in the estimated capacitance ?

[maliberty]

These values are average values since we don't have detailed routes when using these numbers. They should reflect reasonable assumptions about density. Once you have detailed routes we use OpenRCX and its tech file.

There is a more advanced methodology where you PnR a bunch of designs, get the final cap values, and then get the layer rc values by curve fitting. You still need some initial values to start the process.