I try to identify the inversion coefficient, IC, of a MOS devices, which is based on EKV complex model, but PDK uses the BSIM4 model to describe a behavior of the device. When I compare an extracted Ids(BSIM4) with the predisction of a drain current (weak trough strong inversion, no small-geometry effects) defined as:
Ids (WI to SI, no VS, no VFMR)=2*n*u0*Cox*Ut^2*(W/L)*[ln(1+exp(Veff/2*n*Ut))]^2
there is big error in weak and moderate inversion (Vgs ~ Vth ~ 0.45V)
Obviously, there are short-channel effects (technology 90nm), but:
- Velosity Saturation (VS): can be eliminated using a long channel, L >10u.
- Verical Field Mobility Reduction (VFMR): contrubution due to thin layer and inperfection of SiO2 (tox~2.1nm), but again at high, vertical electric field between the gate and the channel.
- Drain Induced Barier Lowering (DIBL): should be eliminated using long channel length.
- Channel Length Modulation (CLM): is always present, but the corection factor of drain current,lp(2-D analysis)=lamda*Lc*(1+(Vds-Vdssat)/Ve), is dominant for strong inversion.
where I am wrong?