how to get the Channel length modulation coefficient?

[billchen]

channel length modulation cadence

how to get the Channel length modulation coefficient?
i am confused.
i can't get it from the models.
get it from experience?
anyone can tell me how to get it accurately.

 

[v_naren]

bsim4 channel length modulation coefficient

I had the same problem with my BSIM3.3.2 models!!!!

here the beautiful page that saved me......

**broken link removed**

maybe u might wanna see these too...

**broken link removed**

 

[billchen]

thx for u

 

[billchen]

maybe the results are rough
thx aynway
anybody have the more accurate ways

 

[billchen]

use that way
i found that lambda is more different when the MOSFET is biased on different voltage level
does it mean Early voltage i different when the MOSFET is biased on different voltage level(the relation between lambda and Early voltage can be founded in Analysis and design of analog integrated circuits by PAUL R.GRAY)
i dont understand

if above is right
so it means if i want to get lambda of a MOSFET,i need simulat the MOSFET which is based on the same voltage level

 

[v_naren]

please understand that in submicron processes mosfets...I mean if your mosfet's length is less than 1um then the whole device model equations break down ...WHATEVER IS GIVEN IN GRAY AND MEYER DOES NOT WORK ANYMORE ACCURATELY ENOUGH for the submic mosfet!!!

furthermore please understand that if your mosfet model is BSIM3 based...which I am quite sure is based anyway....there is nothing like a fixed "lamba"

On fact BSIM 3 and other such complex models dont use lamba unlike our good old nice level 1 and level 2 models!!!

The Id vs Vds output charactreristics is highly non linear

I am quite sure your lambda will not shift so much in the saturation region...I am working in a 0.18um process and I dont find my lambda shift so tremendously...and if ur circuit is so sensitive to lamba shifts during design process then u can find some average value for lamba for the region of Vds and Id swings that u are going to expect in your circuit....I hope u understand bill

 

[billchen]

The attanchment is the waveforms,when the biased current is 12u i can get lambda equal 0.008,and when it is baised on 62u the lambda equal 0.0049,so i get the conclusion that lambda is related to the MOSFET's bias.

so if u want to get more accurate results ,the best way is that u simulate the MOSFETs under the same bias(in u practical circuits).

 

[Puppet1]

if you really want to go nuts

read tsividis 80's paper on mos cad modelling and better
yet get his new book.

i tried to request it here, seems no one has it !

also liu's bok on mosfet modelling using bsim3v3 and bsim4
is good but again no one has it here

it talks about all that is wrong in deep submicron models !