Croaker wrote on Dec 1^st, 2006, 4:44am:


Hi Marc,

Let me recast your statement a little differently. The "body effect" does not matter. As you very rightly deduced from my previous posts, one can split a single MOS into as many sub-devices as one wishes, each with its own source and drain and each haveing its source at a different potential and hence seeing different amounts of body effect so to speak, from a source-oriented model perspective.

You have tempted me to try to explain something which I am not sure I understand well enough. So please excuse the various errors that I am source to make,  but here goes:

Looking at the above MOS structure from a charge perspective, you have the gate, bulk and the source and drain which are for all purposes
identical.

First, there is no voltage at either source/drain and you raise gate with respect to bulk (we are talking NMOS), you will induce a charge throughout
the channel. At any point, the charge in the channel is proportional to V(G) - V(x), where x is the position in channel between source and drain.

So, Q(x) ~ Cox( V(G) - V(x) -Vt0 )

Applying a finite +ve potential at Drain or Source will change V(x) at various points in the channel, reducing the Q(x) at various points in
response to increase in V(S) and/or V(D).

When V(S) goes up wrt V(B), the channel pinches off at Source, when V(D) goes up wrt V(B), the channel pinches off at drain. This identical phenomenon (in my opinion) are recast as "channel turn-off" and "drain pinchoff" when talking in source-oriented terms.

The fact that the channel charge at every single point is determined by V(G), V(x) wrt V(B) automatically should tell you that it is not a question of the "body effect" being absorbed anywhere, it is always there at each point in the channel. If it causes some of the transistors in your series arrangement to be "cut-off", then this does not mean that you have turned off the structure, only that the top ones behave like the pinched off end of a single MOS.

Since the channel potential is a strictly monotonic function of V(D) and V(S), we are used to oversimplification, and always think of the "body effect" as belonging to the source and to the source end alone.

Perhaps that is the problem. As to your other comment, yes it is unfortunate that such a simple and widely used structure cannot be simulate correctly. I would strongly advise you to look at the book by Tsividis. There is an excellent chapter on benchmarking models where he explains how to find out if your models are wirth using or not. I believe it is last chapter on modelling.

Regards
Vivek