aspiceq
New Member
Offline
Posts: 2
UK
|
Hi, I have seen MULU0 and DELVTO as mos instance line parameters in (Hspice) BSIM3 models.
e.g. m1 d g s b ne l= W= M= MULU0= DELVTO= ...
I understand that they are ways to model device variation on an instance by instance basis MULU0 scales the mobility (of effectively the device width) and DELVTO the threshold voltage shift.
My questions are 1. Are these commonly used by modern PDKs? Or is there a better more standard way to model such variation? 2. Are they intended to model statistical variations of hard shifts in parameter? With statistical variations one would expect a "random walk" effect when the mos multiplier M is greater than zero - so the impact of the values would be reduced by 1/sqrt(M)? So in other words, are they in any way modified by the value of M present on the device line? Are simulators consistent in how they handle this? This gets complicated by implicit passing of M (i.e. M scaling entire subckts).
An example demonstrates my point. we have something like DELVTO='5m/sqrt(W*L*1e12)' but should it not be DELVTO='5m/sqrt(W*L*1e12*[effective M scaling of subckt and device])'
Regards, Dave Gee (aspiceq)
|