[SOLVED] question about gm vs gm/Id if L increases

[melkord]

I characteritzed my device and got the result like the picture below.
Here I also copy some text from the book by Binkley.
In Chapter 4, the book said that if L increases, gm nearly unchanges.
in CHapter 5, it is said that low IC and long channel lengths maximizes gm.

I think it is contradictive with the result I got.
I really appreciate it if someone can shed some light here.

 

[arthorios]

You are probably not doing this simulation with a fixed inversion coefficient. If you change L, you must change W accordingly so as to keep a constant IC (assuming a fixed drain current). This way both gm and gm/id should not change too much with L.

 

[melkord]

You are probably not doing this simulation with a fixed inversion coefficient. If you change L, you must change W accordingly so as to keep a constant IC (assuming a fixed drain current). This way both gm and gm/id should not change too much with L.

Thank you.
Yes, you are right.
Actually I use gm/Id method to characterize my device, not this IC method introduced in the book.
I use the book to study the optimization.