<@U03L9SW4F28> I think orientation is expecially i...
# magic
s
@Zexi Liu I think orientation is expecially important if devices need to be matched. If you have differential pairs / current mirrors and similar structures, then all mos transistors should have the same orientation. Rotating the whole set of transistors (so they all will have the same orient) looks to me not that critical. As an example if you have a diode connected mos transistor used as a current mirror and you wire the gate voltage far away the destination transistor should have same orientation, even if located in different regions of the layout. For very deep sub-micrometer processes the design rules are more stringent, as geometries are smaller and lot of OPC corrections are done on the masks, in some cases rotating transistors is not allowed at all.
πŸ‘ 1
🌍 1
z
Great great answer. Thank you Stefan!
s
Effect of rotation is more noticeable on minimum size devices (0.15 channel length for sky130). Transistors used in analog circuits are usually bigger (unless if used as switches/passgates), however keeping the same orientation is always a good recommendation. If there is a small mask misalignment the resulting changes on the transistors (threshold shift, transconductance variation, capacitance, leakage) will be very similar to all devices in the same orientation , so they will remain (more or less) matched.
πŸ‘ 1
t
Generally, the only processes that will prohibit rotated devices as a DRC error are sub-45nm. Above that, it is not a DRC error, but it is a matching problem. If you want transistors to be well-matched, they should be in the same orientation.
πŸ‘ 1