can get the value of lambda (channel modulation) from IV curve?

[aindejeje]

Hi.
Can we estimate the value of lambda from IV curve?
Is the slope of the IV curve represent lambda?
Thanks.

 

[Braski]

yes, u can easily estimate it by the slope of the I-Vds curve in saturation region. However, if u will try to estimate the lambda parameter in different conditions, for different device sizes, you will notice that the value is not constant, but changes (more or less slightly). This is due to the channel length modulation effect, which in higher level models is described in a more complicated (and accurate) way than simply 1+lambda*Vds. However, the estimation is a good way to try some calculation, especially for not very short channels, to be refined by the simulator.

 

[aindejeje]

but the slope is:
del(I)/del(V) - mho (transconductance)

whereas lambda = V (power -1)
lambda = del(L) / VE x L

but why slope is used to estimate lambda?
i don't get it.

 

[Braski]

you are partially right.

the MOS current is:

Id=unCox/2*W/L*(VGS-VTH)^2*(1+lambda*VDS)

so, if u differentiate Id with respect to VDS, you obtain lambda*Id.
the derivative of Id(Vds) to VDS is the slope of Id(VDS). You divide it by Id and you have lambda for a given bias point.

 

[aindejeje]

let:
Idm = Id with channel modulation effect
Id = Id without channel modulation effect

therefore
del(Idm)/del(VDS) = slope = lambda x Id
right?

to get the lambda, divide slope with Id,
where Id is the value when VDS=VGS-VT
right?

 

[Braski]

yes, it seems you are right!

 

[sevak112]

There is another way you can find lambada value.If you use cadence.Do ac analysis and if you see model parameters their should be lambada value.

 

[Braski]

There is another way you can find lambada value.If you use cadence.Do ac analysis and if you see model parameters their should be lambada value.

i think this is only for low level models. in higher complexity models, lambda is not present, and the channel length modulation is described in more complicated and accurated ways.

 

[sevak112]

i think this is only for low level models. in higher complexity models, lambda is not present, and the channel length modulation is described in more complicated and accurated ways.

How you differentiate low level and higher complexity model regarding lambda??

 

[dgnani]

Hi.
Can we estimate the value of lambda from IV curve?
Is the slope of the IV curve represent lambda?
Thanks.

Assuming you can use a spice level 1 model in a 130nm design (L has to be far from minimum) then lambda is the inverse of the Early voltage, which in turns is defined as the (absolute) value where the tangent to the IV vs Vds curve in the saturation region crosses the Vds axis

Given a linear fit of the IV curve in saturation, say, Id=m Vds+q (m is the slope/derivative in saturation) the Early voltage is
Ve=-q/m
that is
lamba=-m/q

The attached drawing should help clarifying: the red lines converge (according to the spice 1 model) to the same point on voltage axis, that point is the Early voltage.



---------- Post added at 19:48 ---------- Previous post was at 19:45 ----------

How you differentiate low level and higher complexity model regarding lambda??

In ultra deep submicron processes, other second order effects such as DIBL and velocity saturation will dominate over channel length modulation. A better model to do hand calculation in those cases is gm/Id

 

[Braski]

How you differentiate low level and higher complexity model regarding lambda??

check bsim3v3 or mos9 manuals for example.

 

[sevak112]

check bsim3v3 or mos9 manuals for example.

got it you are right but you can find it my way too.i tried it.but might be hand calculation is better.Thank you.