Very Basic Q: Is L Drawn Length or Effective Length?

[Randy Yates]

When using a MOSFET device, is the length parameter in the properties
dialog box the drawn length or the effective length? Assume this is
a simple level-1 model.
--
% Randy Yates % "Maybe one day I'll feel her cold embrace,
%% Fuquay-Varina, NC % and kiss her interface,
%%% 919-577-9882 % til then, I'll leave her alone."
%%%% <yates@ieee.org> % 'Yours Truly, 2095', *Time*, ELO
http://www.digitalsignallabs.com

 

[Randy Yates]

Randy Yates <yates@ieee.org> writes:

When using a MOSFET device, is the length parameter in the properties
dialog box the drawn length or the effective length? Assume this is
a simple level-1 model.

If it is the drawn length (which I have assumed), can someone please
tell me why my simulated gain is 31.5 percent greater than the
theoretical in this problem?

http://galois.digitalsignallabs.com/hw4p4p14.pdf
--
% Randy Yates % "My Shangri-la has gone away, fading like
%% Fuquay-Varina, NC % the Beatles on 'Hey Jude'"
%%% 919-577-9882 %
%%%% <yates@ieee.org> % 'Shangri-La', *A New World Record*, ELO
http://www.digitalsignallabs.com

 

[Riad KACED]

Hi Randy,

This is my quick after-midnight comment. There could be many reasons
really but this is the first idea that sprang into my mind.

The GM sounds good for me and I'm rather suspecting the value of the
output impedance r_p, i.e RDS(M4) || RDS(M2); assuming R0=1G being too
large to account.
Your equations show RDS=1 / (lambda * ID) and lambda = Sigma/L. I
don't remember off the top of my head what Sigma stands for but it
looks to be a constant anyway. I'm just wondering whether you have
taken into account the VDS-VEFF in this Sigma. In real life (note the
level-1 model), the Lambda is inversely proportional to sqrt(VDS-
Veff). This is very likely to be the cause of the mismatch between
your hand-made an simulated data, Assuming your hand calculation does
not take into account the VDS-VEFF and your simulator does. In this
this case, the value of lambda would be lower than the theoretical
value, leading to a higher RDS value and thus higher Gain.
The best way for to check your simulation against the theory is to
print out the DCOP output parameters from your simulation. Spectre/
Eldo ... do provide the values of gm, rds for each of your
transistors. This way, you could check which parameters drifts from
the theory.
There have been some discussion in this forum about comparing hand-
made calculation against nowadays human-unreadable models ... I'll
keep out of arguing this time

Hope this helps!
Best Regards,

Riad.

 

[Andrew Beckett]

Riad KACED wrote, on 11/27/08 00:19:

Hi Randy,

This is my quick after-midnight comment. There could be many reasons
really but this is the first idea that sprang into my mind.

The GM sounds good for me and I'm rather suspecting the value of the
output impedance r_p, i.e RDS(M4) || RDS(M2); assuming R0=1G being too
large to account.
Your equations show RDS=1 / (lambda * ID) and lambda = Sigma/L. I
don't remember off the top of my head what Sigma stands for but it
looks to be a constant anyway. I'm just wondering whether you have
taken into account the VDS-VEFF in this Sigma. In real life (note the
level-1 model), the Lambda is inversely proportional to sqrt(VDS-
Veff). This is very likely to be the cause of the mismatch between
your hand-made an simulated data, Assuming your hand calculation does
not take into account the VDS-VEFF and your simulator does. In this
this case, the value of lambda would be lower than the theoretical
value, leading to a higher RDS value and thus higher Gain.
The best way for to check your simulation against the theory is to
print out the DCOP output parameters from your simulation. Spectre/
Eldo ... do provide the values of gm, rds for each of your
transistors. This way, you could check which parameters drifts from
the theory.
There have been some discussion in this forum about comparing hand-
made calculation against nowadays human-unreadable models ... I'll
keep out of arguing this time

Hope this helps!
Best Regards,

Riad.

Indeed, the GDS modelling in level-1 models (and many old models) is pretty
lousy, particularly around the region you tend to bias transistors. As a result
gain is often higher than it should be with such old models.

Years ago, I ended up writing code to compensate for the errors in modelling by
adding additional components around the transistors based on the operating point
to correct the GDS. Good thing is that most modern device models (e.g. bsim4,
psp) are _much_ better at modelling GDS.

Andrew.