Hints, Tips and Solutions

Q: How can I improve the BSIM3v3 model fit for narrow width and small geometry's?

A: The BSIM3v3 model has greatly improved the modeling of narrow width and small (narrow width and short channel) devices by introducing new model parameters. The BSIM3v2 model lacked these additional parameters and therefore the modeling of such geometry's were always a challenge.

The BSIM3v3 model parameters which improve the modeling of narrow devices are :

In the Linear Region

• K3B- Body effect coefficient of K3
• WR- Width offset from Weff for RDS calculation
• DWG- Weff gate dependence coefficient
• DWB - Weff substrate bias dependence coefficient

In the Saturation Region

• B0- Bulk charge effect coefficient for channel width
• B1- Bulk charge effect width offset

When modeling narrow devices, the above listed parameters should be used in addition to the existing parameters W0, K3 and WINT `(`DW in BSIM3v2). If the parameter values are selected carefully, it is possible to obtain good fits for the linear and zero body-bias saturation characteristics. However, the high body-bias saturation characteristics are still not modeled properly. In such cases the binning parameters can be used to improve the fit.

The `BSIM3v3` model parameters which improve the modeling of small devices are:

For threshold voltage calculations:

• DVT0W -First coefficient of narrow width effect on
• VTH for short channel length
• DVT1W-Second coefficient of narrow width effect on
• VTH for short channel length
• DVT2W - Body-bias coefficient of narrow width effect for small channel length

For Leff and Weff calculations (The most effective parameters for small geometry's are listed only. Please refer to the Berkeley BSIM3v3 manual for the complete set of parameters):

• WWL-Length and width cross term coefficient for width offset
• LWL - Length and width cross term coefficient for length offset

In BSIM3v2 there are no parameters to model the small geometry effects. This deficiency in BSIM3v2 forced users to generate separate models for small devices. The above listed parameters in BSIM3v3 have greatly improved the fits for small devices but some problem areas still remain. For example, DW and DL effects can be modified for small devices using WWL and LWL but these parameters can not model the high gate bias portion of the IDS/VGS characteristics (Figure 1). Once more, the binning parameters can be used to improve the fit in this region.

Figure 1. IDS/VGS characteristics of a small device (W=0.7 mm and W=0.65 mm) simulated with BSIM3v3 model (binning parameters are not used).

Q: What are the "binning parameters" and what is the benefit of using them in BSIM3v3?

A: The binning parameters are the geometry sensitivity factors of existing BSIM3v3 model parameters. The binning parameters are generated by adding the characters "L" ( for length sensitivity ),"W" for width sensitivity ) and "P" ( for width and length sensitivity ) at the start of the model parameter name. For example, the sensitivity factors of VTH0 can be written as: LVTH0, WVTH0 and PVTH0.

The complete list of the model parameters which can be binned is given in the Appendix D of the Berkeley BSIM3v3 manual. The implementation of the binning parameters in BSIM3v3 is as follows:

`P = P0 + PL/Leff + PW/Weff + PP/(LeffxWeff)`

As it can be seen this is an empirical implementation. In order to keep the physical nature of the BSIM3v3 model the binning parameters should be used with caution. Even though the binning option in BSIM3v3 is not recommended, in certain regions the existing BSIM3v3 parameters fail to provide sufficient model fit. In such regions the binning parameters may be used to complement the existing model parameters. For example, the body-bias effect on saturation characteristics is modeled using the BSIM3v3 model parameter KETA (Body bias coefficient of bulk charge effect) for all geometry's. A single parameter KETA does not provide sufficient fit for a wide range of channel lengths and channel widths (Figure 2).

Figure 2. IDS/VDS characteristics of a narrow device (W=1.2mm and L=5mm) simulated using BSIM3v3 model parameter KETA.

Therefore the binning parameters LKETA and WKETA can be used to improve the fits (Figure 3).

Figure 3. IDS/VDS characteristics of a narrow device (W=1.2mm and L=5mm) simulated using BSIM3v3 model parameter KETA and binning parameter WKETA = -.02.

Figure 4. IDS/VGS characteristics of a small device (W=0.7 mm and W=0.65 mm) simulated with BSIM3v3 model using the binning parameter PRDSW=-220.

For small geometry's, the existing BSIM3v3 model parameters can not model the high VGS region of the ID/VG characteristics very accurately. For the same operating region, it is possible to obtain good fits for the short channel and narrow width devices by using the parameters RDSW and WR respectively. However, with same RDSW and WR values, simulation underestimates the current for small devices (Figure 1). Introducing a binning parameter PRDSW enhances the model behavior in this particular region (Figure 4).

The binning parameters are not listed in the default sets of BSIM3v3 model parameters. The UTMOST user can add the binning parameter names to the parameter table by following these steps:

1. Open the system screen and press the EditNames button until it reads "Enable"
2. Open the parameters screen and increase the number of parameters.

The parameter table will be expanded and the user can enter the new binning parameter names to the expanded portion of the parameter table.

The effects of binning parameters can be studied using the UTMOST multi-plot rubberband tool. The multi-plot rubberband option is a new feature in UTMOST. More information about the rubberband tool can be found in this month's Simulation Standard.