Question about λ=1/KnL equation

[reltol1]

about MOST

So far I have been thinking that
λ=1/KnL, where λ is channel modulation constant, Kn is constant and L is channel length. So, although L changes, Kn does not.
However, by using a BSIM3_3 model parameters in spice3, Kn is not constant for different length L (i.e. 0.25um, 0.5um). I am wondering why??
Does anyone have any idea about that?
Thanks a lot.[/code]\]

 

[sutapanaki]

Re: about MOST

This relationship is not accurate anymore for short channel devices. The trend still remains, that longer transistors have higher ro, but you can not model it like this. Reason is the physicsof the transistor with sybmicron channel length is not the same as it used to be for 1u, 2u .. technologies.

 

[reltol1]

Re: about MOST

Then, how can I get ro??
Should I derive ro for each length?

 

[sutapanaki]

Re: about MOST

That is one possibility, but you hardly need it. You may do a simulation and plot ro over Vds, or Id or L, just to get a feel of it. You'll see that with accurate models it changes with Vds even if you keep W and L constant. Ro is unprecise parameter and you don't need it for design. It is true that gain is gm.ro but why would you need to know the gain precisely? If you design an opamp you just need a large enough gain to take advantage of the feed-back - opamps always work with feedback. But you don't need to calculate the exact value of the gain. However, you need to know how to increase or decrease the gain. I hope you do.