Channel length modulation

[mike_bihan]

Hi, gurus:

I have a question on channel length modulation equation.

Rout = 1 / (lamda*I)
Lamda will decide mos output resistance.

According to some books, 1/Lamda = Constact*Leff

Then, if channel length is long, Leff~~L. Then 1/Lamda should be propotional to L.

However, it is not true.

Can anyone recommend any papers on that?

 

[mike_bihan]

Please look at the curve, it is not propotional to Length.

1/lamda does not double when Length double.

 

[srivatsan]

Hi mike_bihan,
recently i had the question of how to calculate the channel length modulation for a given technology. well, since i am using 0.5um tech, the bsim3v3.2 gives PCLM as the model parameter. but it involves a little complex equation. so i exeuted my test transistor alone in spice and sw the I-V curve for expected VGS and ramp VDS from vss to vdd. this can be obtained in hardcopy and get approximate lambda or (this is what i did) go and check the analog circuit design forum and look for channel length modulation: spice problem. may be this would help you.
srivatsa

 

[cetc1525]

if L is large, the para. cap.is also large.

 

[qutang]

I can use parallel device to reduce cap.

 

[ue]

possibaly,

channal lenth modulation occur due to lateral elecric field in device gettign stronger and important as channel leth reduces and so lambda is correction factor for the assumption about L considered in physics of device.

so i suppose the relation derived for lambda and ids remain valid in shot channel only and not when L=Leff.

correct me if i am wrong.
/ue

 

[aryajur]

Lambda which is defined for the saturation current is defined taking an approximation.

Leff = L - dL

1/Leff ~ (1+dL/L) / L

and dL/L is defined as Lambda*Vds an approximate model

So the approximation is true only of dL/L << 1

Also Rout = 1/(Lambda *I )

is also an approximate expression

Rout = (1+Lambda*Vds) / (Lambda * I)

This is true when the Lambda approximation holds, so I don't think you should compare these results to accurate plots from a simulator since Lambda is just a parameter useful to model Channel Lendth modulation for hand calculations, it does not even appear as a separate parameter in the BSIM3 model.

 

[mike_bihan]

THe issue is that when I used calculated parameters in simulation, the mismatch in lamda kills the gain of the amplifier.

Lamda vary so much between different Vds/Vgs/I, and I lost control over my circuits' output impedance.

Any good idea from design considerations?

 

[aryajur]

This is interesting. I think the hand calculations should give u a result that is reasonable approximation to the simulated output, i think about 10%. Could you give a circuit example for this I think that would be helpful.

 

[mike_bihan]

You can see attached waveform.

The transistor is in saturation region always. (how deep is a problem)

When current through the transistor is almost constant, the Rds change much.

 

[v_naren]

Message is unavailable.