What are the implications of having very long channel transistor?

[ASIC_Norway]

Hi people,

This is my first post here! I do simulations to understand behaviour of MOSFET as resistors. I am simulating a very long nmos transistor. My simulation setup is as follows:

W/L=1u/100u

Source is either 0V or 0.3V (To see effect of Vsb)
Drain is varied from 0V up to 3.3V
Gate is either 1.6 or 1.7V.

Tech file is BSIM 0.35um I found it online so I am not sure how correct the process is modelled.

My observations so far are:

1) Length size definitely affects resistance. Longer length higher resistance
2) Gate voltage affects the resistance.
3) Non-zero vsb affects the resistance which all agrees with the theory.

Here is a screenshot of the simulation.



But I have some questions,
1) what really happens when the VSB is non zero.
2)How can we explain the drop behaviour in resistance?
3) What are the ultimate limits of transistor gate length? (I know most of the foundries do not measure very long transistors, but in case we use them how can we model them correctly, what parameters to change?)
4) Least but not least, is there any rules of thumb for maximum length? Like 100 time minimum length ?

Thanks in advance,
I hope this can be an interesting exercise for everybody

 

[beetwee]

1) You say that the effect you observe in case of a nonzero vsb corresponds with the theory. What does this theory say about what really happens?
2) Try to simulate with more accuracy and see whether the drop still occurs
3) The limits are set by the foundries, often max 50u in .13 or .18 processes. A longer channel still works, but is not verified by the foundry as nobody will use it. You can also use a cascode structure to get a larger output impedance. Often foundries provide verifications graphs of models vs measurements up to L=5um or L=10um (.13/.18 processes).
4) Normally L=2..4* minimum length is advised for analog design as a trade-off between speed and output impedance. Remember that a large L means lower speed!

 

[MahmoudHassan]

also typically W is larger than L

- - - Updated - - -

Cons and pros of long channels
Advantages
1- long channels are nearly free from DIBL problems so for long channels ID is closely constant with VDS so it is very good as current source circuits
2- if you are going to drive a load of 1 mA long channels will need higher biasing (VGS) than short channels

Disadvantages
1 long channel is slower than short channel as the Cox = (Epslon * Area / L ) for higher L capacitor becomes small and negligible