[Moved]: how to get the lamda(Va) from the Ids_Vds curves?

[shanmei]

As shown in the bellowing figure, when can set different Vgs conditions, and then sweep Vds to get Ids_Vds curves with various Vgs. Then extending those curves and crossing at Va point, then we know the value of Va, then lamda=1/Va.


In real life, the curves are not so "straight", so how can we know which location is the start point the extend the Ids_Vds curve?
For example, in the bellowing figure, we can choose node A or B or C to extend the lines, and each one will result in different Va value.



My question is how should we set the start point to extend the lines to get Va? Is this method still effective considering real cmos model is indeed much complex. If this method is not working, how can we know Va? Thanks

 

[dick_freebird]

Re: how to get the lamda(Va) from the Ids_Vds curves?

Yes, the "ugliness" of the MOS curves (especially in
more advanced, or aggressively applied nodes) adds
an element of judgement to the extraction.

Biasing at low Vgs and taking your two points from
the straight-line region is best. Avoid the impact
ionization "curl" and stay away from the linear-to-
saturation "knee", both.

Now the number you get, may not be useful in itself
(other than filling in simple equations which may
themselves be rough at best, dating from days when
"short channel" was more than a micron). But it can
serve as a good "fit-to" or a validation of the model
fit.