Cxy values reported by operating point information through spectre are derivied from partial derivatives of channel charge by terminal voltage as defined by BSIM equations. See section B.2.4 in this link { http://nikola.com/pdf/BSIM3APX-B.pdf } for the exact BSIM equations for the partial derivatives.

To be precise, this link { http://www.designers-guide.org/Forum/YaBB.pl?num=1087411058/1 } suggests that the partial derivates computed by BSIM model are further modified by spectre simulator before reporting as operating points. (I have never verified it.)

As you will find out, they are mathematical constructs and may or may not have any correspondence with actual physical capacitors mentioned in the textbooks.

MOS models have four "buckets" of charge; depending on the four terminal voltages (D, G, S, B) and hence the operating region, you get different amounts of charge in each bucket.

Cxy tells you how much the charge in bucket "x" (Qx) changes when you wiggle the voltage on terminal "y" (Vy).

You really can't expect that the change in Qd when wiggling Vg would be the same as the change in Qg when wiggling Vd.

However, charge is conserved, Qd+Qg+Qs+Qb=0, and there are only 3 independent branch voltages, so you get identities like Cgg=Cgs+Cgd+Cgb.  Cdd=Cdg+Cdb+Cds and you can also write Cdd=Cgd+Csd+Cbd.  (Note that Cxy = -dQx/dVy when x!=y.)

supermoment wrote on Apr 5^th, 2010, 5:38am:


They aren't "dc" values; they're mathematical expressions of the partial derivatives of mathematical equations where the voltages are the inputs.  There is no frequency dependence of these mathematical expressions.

However, if you now place those "capacitances" along with the small-signal conductances into small-signal equivalent circuit, you will see frequency dependence of a (complicated) RC network.  Note that the capacitances aren't 2-terminal capacitors -- and the conductances aren't 2-terminal resistors/conductors.

A 2-terminal capacitor is one whose charge depends only on the voltage across it (across its two terminals).  That's not what the capacitance is like in the MOS.

Nice Explanation from Geoffrey_Coram, esp the first reply...Thanks...