@User 2x5/1 means 2 identical transistors drawn with width=5 and length=1 and connected in parallel. 1x10/1 means one single transistor with W=10 and L=1. The difference is in second order effects, parasitic capacitances, and so on. Usually they are similar but do not have identical electrical parameters. These differences are more important at small geometries.

@Stefan Schippers @Tim Edwards that is my feeling that such a fundamental change now would break a lot. the display of the NF parameter is really helpful, thanks! can you think of a backwards-compatible way of making clear that the displayed width is the total width? an alternative is to display

Wfinger

instead of

nf

, but I like

nf

more as it is more useful I think.

@Harald Pretl: I would assert that the people who came up with the BSIM definition treated

W

in a way that is not natural for most people to think of, and is inconsistent with the way

W

and

M

are used, for example. In the parameterized cell for magic, I just let

W

be the width of the finger and

nf

be the number of fingers. The point is moot in magic, anyway, since magic doesn't extract

nf

(although the cell could be set up to extract

nf

, and I probably should do that to satisfy the people who are distraught that magic doesn't know the difference between fingers and multiplicity (which is not actually easy to define in a consistent way). But one thing that is easy to do in the magic parameterized cell is just to put both "W(total)" and "W(finger)" in the display. I do that with resistors; the total W is just a derived quantity and you can specify a routine that calculates it on the fly.

@Tim Edwards putting “W(total)” and “W(finger)” in the display would be a cool first step. usually I am not too concerned about the extraction of

nf

and

M

and check in the LVS for this, as long as total width is properly checked. One exception is for RF circuits, here you want transistor parasitics as accurately as possible, so getting this properly extracted is key to good modelling and simulation accuracy. Eventually people (like me) may want to build RF circuits, but my plan of attack would be to carefully build a fixed RF MOST layout cell, extract and model it accurately, and then just blackbox it from the usual extraction (if that is even possible in magic, don’t know).