Spec in. Signed-off silicon out.

This is where the tool sizes the circuit for you. Type the numbers it must meet, then either generate a design from the specs (a physics-based starting point, refined by search) or optimize from the values already on the page. Every step of the search is a real ngspice run, so the SKY130 circuits take minutes, not seconds, and the winning sizing can be applied to the page with one click.

A sweep measures the amplifier being polite. This asks how fast it can actually move: the same circuit as a unity buffer, hit with a 0.3 V step, up and back down. One transient run, about twenty seconds.

The rest of the datasheet: how well the amplifier ignores what both inputs do together and what its supply does, and how far its input and output can travel before it stops working. Four copies of the circuit in one run, about half a minute.

A schematic pretends a wire is free and a transistor takes no room. This draws the real cell: matched pairs split into fingers around a shared centroid, dummies at the ends, a guard ring around each group, and two metal layers of routing. It reports what that costs in silicon, then extracts the wiring off the polygons: area, perimeter and overlap at the constants for the plane each piece of metal is actually on, coupling between two nets as a capacitor between those two nets, and a resistor wherever there is metal. Then it measures the circuit again with that network in the deck. Two simulations, run right here. The foundry's own deck needs the local app: it is a runset inside a gigabyte of Qt, and it is the one thing about this drawing that a browser cannot answer.

Two questions get answered. The geometry is checked against thirty-six rules read from the PDK, and the drawing is compared against the netlist that was simulated, so a wire on the wrong pin is caught rather than admired. Neither is sign-off: the real rule deck runs to thousands and a real layout versus schematic needs Netgen. The placer is a row and the router is one track a net, which is not what anyone would call placement or routing. The extraction says what it leaves out: coupling between two wires side by side on one plane is measured and reported and is not turned into a capacitance, because the technology file states it as a value and a distance and does not say what the distance is for.