Professional Documents
Culture Documents
Bsim3 Manual PDF
Bsim3 Manual PDF
Chapter
Page
Page
Contents
Index
Vertical and lateral non-uniform doping Short channel effects Narrow channel effects
Mobility
Output Resistance Drain induced barrier lowering (DIBL) Channel length modulation (CLM) Substrate current induced body effect (SCBE)
5-1
Chapter
Chapter
Page
Page
Contents
Index
For a detailed description of these features please refer to the BSIM3 manual of Berkeley University. You can order this manual from Berkeley or you can get it over the Internet. Please see the last chapter of this manual for details. The BSIM3v3 Modeling Package provides the user with a complete extraction strategy for the model parameters of the BSIM3v3.2.2 model. The extraction routines are based on the BSIM3v3.2.2 device equations to ensure that the extracted model parameters represent as good as possible the original physical meaning. Therefore, no or only a minimum of optimization is needed to get a good fit between measured and simulated device behavior. The routines of this release refer to version 3.2 of the BSIM3 model which was released by University of California at Berkeley in June 1998.
5-2
Chapter
Chapter
Page
Page
Contents
Index
5-3
Chapter
Chapter
Page
Page
Contents
Index
Threshold Voltage
The threshold voltage is one of the most important parameters of deep submicron MOS transistors and is affected by many different effects when the devices are scaled down into the region of 0.1 microns. The complete equation of the threshold voltage in BSIM3v3.2 is given below. V th = V Tideal + V th ( 1 ) + V th ( 2 ) V th ( 3 ) V th ( 4 ) + V th ( 5 ) V th ( 6 ) (1)
Chapter
Chapter
Page
Page
Contents
Index
The different parts of this complex equation are expressed by the following sub-equations in more detail: V th = V th0 K 1 s T ox % T ox & +K 1 ------------ s V bseff K 2 # ------------V bseff$ T oxm ! T oxm " T ox Nlx & +K 1 ------------ % % 1 + --------- 1& s " T oxm ! ! L eff" D VT0 e
L eff& % D -------VT1 ! 2l t "
+ 2e
( V bi s )
(2) ( V bi s )
D VT0w e
+ 2e
+ 2e
(5)
Chapter
Chapter
Page
Page
Contents
Index
where:
Vthideal VFB s ni Eg0 = ideal threshold voltage = flatband voltage = surface potential = 1.45 1010 (Tnom/300.15)1.5 (21.5566 Eg0/2Vtmo) = 1.16 7.02 10-4 Tnom2/(Tnom + 1108)
This equation had been implemented into the first MOS simulation models assuming long and wide channels and uniform substrate doping. The following sections describe the effects which overlay this basic equation.
Figure 5-1. Vertical Doping Profile in the Channel It is usually higher near the silicon to silicon dioxide interface than deeper in the substrate. This higher doping concentration is used to adjust the threshold voltage of the device. The distribution of impurity atoms inside the substrate is approximately a half Gaussian distribution, which can be approximated by a step function with NCH for the peak concentration in the channel near the SiSiO2 interface and Nsub in the deep bulk. XT is the depth where the approximation of the implant profile switches from NCH to NSUB. The non-uniform vertical channel doping affects the threshold voltage when a bulk source voltage is applied to the device and is represented here as the part Vth(1) of the overall threshold voltage.
Chapter
Chapter
Page
Page
Contents
Index
T ox T ox V th ( 1 ) = K 1 ------------ s V bseff K 2 ------------V T oxm T oxm bseff K 1 = 2 2K 2 s V bm ( 1 2 ) ( s V bx s ) K 2 = ------------------------------------------------------------------------------2 s ( s V bx s ) + V bx 2q si N ch 1 = -------------------------C ox 2q si N sub 2 = ----------------------------C ox where:
Vbx = substrate bias voltage when the depletion width equals
(9)
(10)
qN ch X t2 X t = s -------------------2 si
Vbm Toxm Tox = maximum substrate bias voltage = gate oxide thickness at which parameters are extracted = default value of Toxm
Vbseff = V bc + 0,5 V bs V bc 1 +
( V bs V bc 1 ) 4 1 V bc
1
Vbc
= 0.001V
2 % K1 & = 0,9 # s ---------$ # 2$ 4K 2" !
Chapter
Chapter
Page
Page
Contents
Index
In BSIM3, either the model parameters K1 and K2 or NCH, NSUB, VBM or XT can be used to model this effect. Figure 5-2 shows the threshold voltage Vth as a function of the applied bulk voltage for a transistor with a large channel length and a wide channel width (LARGE).
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-3. Lateral Doping Profile in the Channel As the channel length becomes shorter, the lateral non-uniform doping will cause the threshold voltage to increase strongly because the average doping concentration in the channel becomes higher. This part of the threshold voltage is modeled with the parameter Nlx and is represented by Vth(2) as a part of the overall threshold voltage. T ox % % Nl x & & V th ( 2 ) = K 1 ------------ # # 1 + ---------$ 1$ s T oxm ! ! L eff" " where: Nlx = 2Lx(Nds Na)/Na (11)
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-4 shows the influence of the non-uniform lateral doping on the threshold voltage as a function of gate length.
Figure 5-4. Threshold Voltage as a Function of Gate Length Due to Lateral Non-Uniform Doping You can distinguish between the theoretical trace following equation (11) and the real world ones with the short channel effect described in the next chapter.
Chapter
Chapter
Page
Page
Contents
Index
+ 2e
( V bi s )
(12)
lr =
(13)
X dep = where:
Vbi
(14)
built-in voltage of the PN junction between the source/drain and the substrate
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-5. Influence of Short Channel Effects on the Threshold Voltage For short channel lengths together with small channel widths, the following additional expression Vth(4) is needed to formulate the threshold voltage: V th ( 4 ) = D VT0w e where: l tw = si T ox X dep ---------------------------- ( 1 + D VT2W V bseff ) sio2
W eff L eff& % D ! VT1w ------------------2l tw " W eff L eff& % D ! VT1w ------------------l tw "
+ 2e
( V bi s )
(15)
Chapter
Chapter
Page
Page
Contents
Index
T ox V th ( 5 ) = ( K 3 + K 3b V bseff ) ------------------------------- s (W + W )
eff 0
(16)
Chapter
Chapter
Page
Page
Contents
Index
V th ( 6 ) =
+ 2e
(17)
l t0 =
(18)
Chapter
Chapter
Page
Page
Contents
Index
(19)
MOBMOD=2:
0 eff = ------------------------------------------------------------------------------------------------------------------------------------2 1 + ( Ua + U c V bseff ) ( V gsteff T ox ) + U b ( V gsteff T ox )
(20)
MOBMOD=3:
0 eff = -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------2 1 + [ U a ( V gsteff + 2V th ) Tox + U b ( ( V gsteff + 2V th ) T ox ) ] ( 1 + U c Vbseff )
(21)
The influence of the mobility reduction parameters is demonstrated in Figure 5-7 where the simulated drain current with and without mobility reduction is shown.
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-8 shows the effective mobility as a function of gate voltage and bulk-source voltage.
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-9. Influence of Channel Length Reduction on the Drain Current The effective channel length is defined in BSIM3 as follows: L eff = L Designed 2dL (22)
The channel length reduction on one side of the channel consists of several empirical terms as shown below: Ll Lw L wl dL = L int + ------------ + --------------- + --------------------------Lln Lwn Lln Lwn L W L W (23)
The use of the model parameters LL, LLN, LWN, LW and LWL is very critical because they are only used for fitting purposes. On the other hand, they may be needed to achieve a good fit over a large area of channel lengths especially for processes with a minimum designed gate length of less
Chapter
Chapter
Page
Page
Contents
Index
than 0.25m. Figure 5-9 shows the influence of the geometrical channel length reduction LINT on the drain current of a short channel transistor while Figure 5-10 represents the channel length reduction according to equation (23).
Chapter
Chapter
Page
Page
Contents
Index
The channel width reduction on one side of the channel consists of several empirical terms as shown below: Wl Ww W wl dW = W int + -------------- + ---------------- + -----------------------------Wln Wwn Wln Wwn L W L W (25)
The use of the model parameters WL, WLN, WWN, WW and WWL is very critical because they are only used for fitting purposes. On the other hand, they may be needed to achieve a good fit over a large area of channel widths especially for processes with a minimum designed gate width of less than 0.25m. Figure 5-11 shows the influence of the geometrical channel width reduction WINT on the drain current of a narrow channel transistor while Figure 5-12 represents the channel width reduction according to equation (25).
Chapter
Chapter
Page
Page
Contents
Index
Chapter
Chapter
Page
Page
Contents
Index
Drain Current
Single Equation for Drain Current
In contrast to former implementations of the BSIM3 model, the drain current is represented through a single equation in all three areas of operation (subthreshold region, linear region and saturation region). Due to this single formula, all first order derivatives of the drain current are continuous, which is an important prerequisite for analog simulations. In the case that no parasitic drain/source resistance is given, the equation for the drain current is given below: V dseff % & V gsteff # 1 A bulk ---------------------------------------------$ V dseff 2 ( V gsteff + 2V tm )" ! W - ------------------------------------------------------------------------------------------------------------I ds0 = eff C ox ---L V dseff 1 + --------------E sat L
(26)
This equation is valid for all three regions of operation of the MOS transistor because the voltages at drain, gate and bulk are replaced by effective drain voltage Vdseff, the effective gate voltage Vgsteff and the effective bulk voltage Vbseff, which are all defined by the continuous equations below: Equation (27) shows the effective (Vgs - Vth) voltage, where the factor n is defined in equation (31). % % V gs V th& & 2n t ln # 1 + exp # -----------------------$ $ ! ! 2n t " " V gsteff = ---------------------------------------------------------------------------------------------------------------------2 s % V gs V th 2V off& 1 + 2nC ox ------------------------ exp # --------------------------------------------$ q ( si N ch ) 2n t ! "
(27)
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-13. Effective Voltage Vgs - Vth Figure 5-13 shows Vgsteff in logarithmic scale. Vgsteff fits a linear function for values of Vgs greater than Vth while the subthreshold area is covered by the fit of an exponential function. Through this equation the first derivative is continuous between both operational regions (subthreshold and linear) of the MOS transistor. Equation (28) shows the effective drain source voltage, Vdseff:
2 1% V dseff = V dsat -- V dsat V ds + ( V dsat V ds ) + 4 V dsat& " 2!
(28)
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-14 shows Vdseff in both the linear and the saturation region of operation of the MOS transistor. Vdseff models the transition between linear and saturation region without discontinuity in the first derivative of the drain current.
Chapter
Chapter
Page
Page
Contents
Index
(29)
2 & b = ( V gsteff + 2V tm ) % -- 1" + A bulk E sat L eff + 3A bulk R ds C ox W sat ( V gsteff + 2V tm ) ! c = E sat L eff ( V gsteff + 2V tm ) + 2R ds C ox W sat ( V gsteff + 2V tm )
2
The influence of the maximum carrier velocity VSAT on the drain current Ids and the conductance gds is demonstrated in Figure 5-15.
Figure 5-15. Influence of VSAT on Drain Current Ids and Conductance gds
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-16. Depletion Width along the Channel Length The bulk charge effect Abulk is modeled in BSIM3 with the parameters A0, AGS, B0, B1 and KETA as shown in the following equation (30). (30) % , B0 A 0 L eff # T ox ) + ------------------------------------------------------------------# - ) K 1 -----------W eff + B 1 L + 2 X X T oxm ) eff J dep # -* A bulk = # 1 + ----------------------------------L eff 2 s V bseff ) # % &2 -$ # ) 1 AGS V gsteff # -------------------------------------------# ! L eff + 2 X J X dep" ) ! ' -& )$ )$ )$ 1 ----------------------------+$ 1 )$ + K eta V bs )$ )$ ("
Chapter
Chapter
Page
Page
Contents
Index
Chapter
Chapter
Page
Page
Contents
Index
(31)
th = e
+ 2e
(32)
The influence of VOFF and NFACTOR on the drain current in the subthreshold region is shown in Figure 5-18.
Figure 5-18. Influence of VOFF and NFACTOR on Drain Current in the Subthreshold Region
Chapter
Chapter
Page
Page
Contents
Index
Parasitic Resistance
As MOS devices are scaled into the deep submicron region, both the conductance gm and the current of the device increase. Therefore the voltage drop across the source and drain series resistance becomes a non-negligible fraction of the applied drain source voltage. The resistance components associated with a MOSFET structure are shown in Figure 5-19. These include the contact resistance (Rcontact) between metallization and source/drain area, the diffusion sheet resistance (Rsheet) of the drain/source area, the spreading resistance (Rspread) which arises from the current spreading from the channel and the accumulation layer resistance (Raccum.).
Figure 5-19. Resistance Components of a MOS Device These components are put together to form the following equation in the BSIM3v3: R dsw [ 1 + P rwg V gsteff + P rwb ( s V bseff s ) ] R ds = -------------------------------------------------------------------------------------------------------------------------------Wr 6 ( 10 W eff ) (33)
The diagram in Figure 5-20 visualizes the equation of Rds. It should be noted that BSIM3 assumes, that the drain resistance is equal to the source resistance. This symmetrical approach may cause difficulties if a device with a nonsymmetrical drain source resistance, e.g. a DMOS power transistor, should be modeled. In this case, a scalable SPICE macro model should add the required behavior to BSIM3.
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-20. Drain Source Resistance Rds as a Function of Vg and Vb With this enhancement, equation (26)for the drain current can be rewritten: I ds0 I ds = -----------------------------------------------1 + R ds I ds0 V dseff (34)
The influence of the parasitic resistance on the drain current is demonstrated for a SHORT and a SMALL transistor in Figure 5-21.
Chapter
Chapter
Page
Page
Contents
Index
Output Resistance
a) Early Voltage
The drain current in the saturation region of submicron MOSFETs is influenced by the effects of channel length modulation (CLM), drain induced barrier lowering (DIBL) and substrate current induced body effect (SCBE). These effects can be seen clearly looking at the output resistance Rout of the device, which is defined as: V ds R out = ---------- I ds (35)
In Figure 5-22, the measured drain current and the output resistance of an n-type MOS transistor with a channel length of 0.5 m are shown.
Figure 5-22. Drain Current and Output Resistance in Linear and Saturation Region The leftmost region in Figure 5-22 is the linear region, in which carrier velocity is not saturated. The output resistance is small because the drain current has a strong dependence on the drain voltage. The other three regions belong to the saturation region. The three physical effects CLM, DIBL and SCBE can be seen in the saturation region and are discussed in the following subsections.
The Unified I-V Model of BSIM3v3 5-30
Chapter
Chapter
Page
Page
Contents
Index
With the output resistance, the equation for the drain current (34) is enhanced by two additional terms and can be rewritten as: V ds V dseff& I ds0 V ds V dseff& % % I ds = ------------------------------------------------ # 1 + ------------------------------$ # 1 + ------------------------------$ VA V 1 + R ds I ds0 V dseff ! "! ASCBE " (36)
The behavior of the output resistance is modeled in BSIM3 in the same way as the Early voltage of a bipolar transistor is modeled in the Gummel-Poon model. The Early voltage is divided in two parts, VA due to DIBL and CLM and VASCBE due to SCBE. VA is given by: P vag V gsteff& 1 % 1 1 V A = V Asat + # 1 + ------------------------------$ % ------------------- + ------------------------& E sat L eff " ! V ACLM V ADIBLC" ! where VAsat is the Early voltage at Vdsat: (38) A bulk V dsat & % -$ E sat L eff + V dsat + 2R ds sat C ox W eff V gsteff # 1 --------------------------------------------2 ( V gsteff + 2V tm )" ! V Asat = ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------2 1 + A bulk R ds sat C ox W eff (37)
Chapter
Chapter
Page
Page
Contents
Index
Chapter
Chapter
Page
Page
Contents
Index
simulated by a device simulator. The symmetrical profiles correspond to Vds = 0 and the asymmetrical profiles to Vds > 0. In the figure the simulated potential barrier near the source is observed to decrease with increasing drain bias which indicates the origin of the DIBL effect.
Figure 5-24. Band Diagram at Si-SiO2 Interface of a 0.1 m MOSFET The DIBL effect is modeled in BSIM3v3 with the following equations: ( V gsteff + 2V tm ) A bulk V dsat % & -$ V ADIBLC = ------------------------------------------------------------------ 1 # ----------------------------------------------------------------------- rout ( 1 + P DIBLC V bseff ) ! A bulk V dsat + V gsteff + 2V tm" with % D rout L eff& % D rout L eff& rout ( L ) = P DIBLC1 exp # -------------------------$ + 2 exp # -------------------------$ + P DIBLC2 2l t0 " l t0 ! ! " (41) (40)
Figure 5-25 shows the influence of the DIBL effect on the output resistance of a short channel transistor.
Chapter
Chapter
Page
Page
Contents
Index
The increase of the total drain current through hot electrons will be described by the part VASCBE of the Early voltage which results in a lowering of the output resistance for high drain voltage (Figure 5-26). P SCBE1 l & 1 P SCBE2 % V ASCBE = ------------------- exp # ---------------------------------$ L ! ( V ds V dsat )" (43)
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-26.
Chapter
Chapter
Page
Page
Contents
Index
Substrate Current
In a n-channel MOSFET, electrons in the channel experience a very large field near the drain. In this high field, some electrons coming from the source will be energetic enough to cause impact ionization, and additional electrons and holes are generated by avalanche multiplication. The high energy electrons are referred as "hot" electrons. The generated electrons are attracted to the drain, adding to the channel current, while holes are collected by the substrate contact, resulting in a substrate current, which is shown in Figure 5-27.
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-28. Substrate Current Ibs Paramaterized by Vg The substrate current is described in BSIM3 by the following equation:
0 + 1 L eff 0 V ds V dseff& & I % 1 + --------------------------- ---------------------------I sub = ----------------------------- ( V ds V dseff ) exp % -" ds ! V ds V dseff" ! L eff VA
(44)
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-29. pn-junction diode The drain/bulk and the source/bulk pn-junctions can be used as diodes in CMOS designs. BSIM3v3 offers a simple DC model for the current Ibs or Ibd flowing through these diodes.
V bs & , ------------" % % & ) ! NV tm # ) I sbs e 1$ + G MIN V bs # $ ) ! " I bs = * ) IJTH + I sbs ) IJTH + ---------------------------- ( V bs V jsm ) + G MIN V bs ) NV tm '
(45)
where NJ is the emission coefficient of the source junction and the saturation current Isbs is calculated as: I sbs = A S J S + P S (46)
where JS is the saturation current density of the source/bulk diode, AS is the area of the source junction, JSSW is the sidewall saturation current density of the source/bulk diode and PS is the perimeter of the source junction. JS and JSSW are functions of the temperature and can be described as:
Chapter
Chapter
Page
Page
Contents
Index
J S = J S0 e
E g0 Eg T && % ---------- -------- + X TI ln % ---------! T nom" $ # V tm0 V tm # ------------------------------------------------------------------$ NJ # $ ! " E g0 Eg T && % ---------- -------- + X TI ln % ----------" ! Tnom # V tm0 V tm $ # ------------------------------------------------------------------$ NJ # $ ! "
(47)
(48)
JS0 is the saturation current density (default is 10-4 A/m2) JS0SW is the sidewall saturation current density (default is 0) NVtm = NJ (KbT/q) Vjsm = NVtm ln (ijth/Isbs + 1)
Chapter
Chapter
Page
Page
Contents
Index
The current Ibs through the diode is shown in the following diagram:
Chapter
Chapter
Page
Page
Contents
Index
Capacitance Model
Please use the model bsim3_tutor_cv.mdl provided with the BSIM3v3 Modeling Package to visualize the capacitance model parameters. Load the file into IC-CAP and run the different macros to see how certain parameters affect the device behavior of a deep submicron MOS transistor. The capacitance in a MOS transistor can be divided into three different parts: Junction capacitance CJunc between source/drain and the bulk region Capacitance of the extrinsic MOS transistor which consists of: The outer fringing capacitance CF between polysilicon gate and the source/drain The overlap capacitance CGDO between the gate and the heavily doped source/drain regions The overlap capacitance CGDOL between the gate and the lightly doped source/drain regions Capacitance of the intrinsic MOS transistor in the region between the metallurgical source and drain junction when the gate is at flat band voltage. These different parts of the capacitance of a MOS transistors are shown in Figure 5-31 below. The following three subchapters explain each type of capacitance and its implementation in the BSIM3v3 model.
Capacitance Model
5-41
Chapter
Chapter
Page
Page
Contents
Index
Junction Capacitance
The source/drain - bulk junction capacitance can be divided into three components as shown in Figure 5-32. The calculation is shown for the drain - bulk junction capacitance. The source - bulk capacitance is calculated in the same way with the same model parameters. The overall junction capacitance Cjdb is given by: , C AREA + C SW + C SWG C jdbs = * ' C AREA + C SW where: CAREA is the bottom area capacitance CSW is the sidewall or peripheral capacitance along the three sides of the junctions field oxide CSWG is the sidewall or peripheral capacitance along the gate oxide side of the junction if PS > Weff + if PS < Weff ( (49)
Figure 5-32. Dimensions of Drain/Source Region and Different Capacitance Parts Bottom area capacitance CAREA
Capacitance Model
5-42
Chapter
Chapter
Page
Page
Contents
Index
C AREA = AD C jbd where: AD Cjbd For Vbs < 0: V bs& M j % C jbs = C j # 1 --------$ Pb " ! For Vbs 0: V bs& % C jbs = C j # 1 + M j --------$ Pb " ! area of bottom side of pn junction, given as SPICE model parameter capacitance per unit area of the drain-bulk junction
(50)
Cjbd is calculated according to the following equation and is shown in Figure 5-33. (51)
Capacitance Model
5-43
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-33. Bottom Area Capacitance Cjbd as a Function of Vg Peripheral sidewall capacitance CSW along the field oxide C SW = ( PD W eff ) C jbdsw where: PD Weff Cjbdsw total perimeter of pn junction, given as SPICE model parameter effective gate width of transistor, calculated in SPICE capacitance per unit length (52)
Cjbdsw is calculated according to the following equation and is shown in Figure 5-34: For Vbs < 0: V bs & M jsw % C jbdsw = C jsw # 1 -----------$ P bsw" ! (53)
Capacitance Model
5-44
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-34. Sidewall Capacitance Cjbdsw as a Function of Vg Peripheral sidewall capacitance CSWG along the gate oxide C SWG = W eff C jbdswg where: Weff Cjbdswg effective gate width of transistor, calculated in SPICE capacitance per unit length (54)
Cjbdswg is calculated according to the following equation and is shown in Figure 5-35.
Capacitance Model
5-45
Chapter
Chapter
Page
Page
Contents
Index
For Vbs < 0: V bs & Mjswg % C jbs = C jswg # 1 --------------$ P bswg" ! For Vbs 0: V bs & % C jbs = C jswg # 1 + M jswg --------------$ P bswg" ! (55)
Figure 5-35. Sidewall Capacitance Cjbdswg Along the Gate Oxide as a Function of Vg
Capacitance Model
5-46
Chapter
Chapter
Page
Page
Contents
Index
Extrinsic Capacitance
As mentioned in the introduction to this chapter, the extrinsic capacitance of a MOS transistor consists of the following three components: the outer fringing capacitance CF between polysilicon gate and the source/drain the overlap capacitance CGDO between the gate and the heavily doped source/drain regions the overlap capacitance CGDOL between the gate and the lightly doped source/drain regions The contribution of these different components to the overall extrinsic capacitance is demonstrated in the following diagrams in Figure 5-36 and Figure 5-37.
Capacitance Model
5-47
Chapter
Chapter
Page
Page
Contents
Index
a) Fringing Capacitance
The fringing capacitance of a MOS transistor consists of a bias independent outer fringing capacitance and a bias dependent inner fringing capacitance. In the present release of the BSIM3v3 model, only the bias independent outer fringing capacitance is implemented. Experimentally, it is virtually impossible to separate this capacitance with the overlap capacitance. Nonetheless if the model parameter CF is not given, the outer fringing capacitance can be calculated with the following equation:
7 2 si02 % 4 10 & - ln # 1 + ----------------CF = --------------$ T ox " !
(56)
b) Overlap Capacitance
In BSIM3v3 an accurate model for the overlap capacitance is implemented. In old capacitance models this capacitance is assumed to be bias independent. However, experimental data show that the overlap capacitance changes with gate to source and gate to drain biases. In a single drain structure or the heavily doped S/D to gate overlap region in a LDD structure the bias dependence is the result of depleting the surface of the source and drain regions. Since the modulation is expected to be very small this region can be modeled with a constant capacitance. However in
Capacitance Model
5-48
Chapter
Chapter
Page
Page
Contents
Index
LDD MOSFETs a substantial portion of the LDD region can be depleted, both in the vertical and lateral directions. This can lead to a large reduction of overlap capacitance. This LDD region can be in accumulation or depletion. In BSIM3v3, a single equation is implemented for both regions by using such smoothing parameters as Vgsoverlap and Vgdoverlap for the source and drain side, respectively. Unlike the case with the intrinsic capacitance, the overlap capacitances are reciprocal. In other words, Cgsoverlap = Csgoverlap and Cgdoverlap = Cdgoverlap. The model equations for the overlap capacitance are shown for the drain overlap capacitance and are identical for the source overlap capacitance: Overlap charge per gate width:
Q overlap 4V gd, overlap& , CKAPPA % ------------------ = CGDOV gs + CGDL * V gd V gd, overlap ------------------------ 1 + 1 ----------------------------! W eff CKAPPA " + 2 ' (
(57)
(58)
for the measurement and simulation conditions given in Figure 5-36, this results in the overlap capacitance: Q overlap C gd, overlap = ----------------------- V gs The model parameter CGDO in equation 57 can be calculated by the following equation: CGDO = ( DLC C OX ) CGDL (60) (59)
where DLC represents the channel length reduction in the BSIM3v3 capacitance model. Please see the next sub-chapter for more details about DLC:
Capacitance Model
5-49
Chapter
Chapter
Page
Page
Contents
Index
Intrinsic Capacitance
a) Geometry for Capacitance Model
The BSIM3v3 model uses different expressions for the effective channel length Leff and the effective channel width Weff for the I-V and the C-V parts of the model. The geometry dependence for the intrinsic capacitance part is given as the following: WW WWL WL - + ------------------- + ---------------------------------- W = DWC + --------------WLN WWN WWN WLN W L L W LW LWL LL - + ----------------- + ------------------------------- L = DLC + ------------LLN LWN LWN LLN W L W L (61)
(62)
Lactive and Wactive are the effective length and width of the intrinsic device for capacitance calculations. The parameter L is equal to the source/drain to gate overlap length plus the difference between drawn and actual poly gate length due to processing (gate printing, etching and oxidation) on one side. The Lactive parameter extracted from the capacitance method is a close representation of the metallurgical junction length (physical length). W active = W Drawn 2 W L active = L Drawn 2 L (63) (64)
Capacitance Model
5-50
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-38. Dimensions of a MOSFET While the authors of the BSIM3v3 model suggest to use a parameter LINT for the I-V model which is different from DLC, other literature sources [3] propose that LINT should have the same value as DLC. This approach is also implemented in the BSIM3v3 Modeling Package to ensure that the extracted values of the channel length reduction are very close to the real device physics. Therefore, the channel length reduction LINT for the I-V model will be set to DLC from the C-V model extracted from capacitance measurements.
Capacitance Model
5-51
Chapter
Chapter
Page
Page
Contents
Index
To ensure charge conservation, terminal charges instead of the terminal voltages are used as state variables. The terminal charges Qg, Qb, Qs, and Qd are the charges associated with the gate, bulk, source, and drain. The gate charge is comprised of mirror charges from 3 components: The channel minority (inversion) charge (Qinv) The channel majority (accumulation) charge (Qacc) The substrate fixed charge (Qsub) The accumulation charge and the substrate charge are associated with the substrate node while the channel charge comes from the source and drain nodes: Q g = ( Q sub + Q inv + Q acc ) Q b = Q sub + Q acc Q inv = Q s + Q d (65)
The inversion charges are supplied from the source and drain electrodes. The ratio of Qd and Qs is the charge partitioning ratio. Existing charge partitioning schemes are 0/100, 50/50 and 40/60 (given by the model parameter XPART = 0, 0.5 and 1) which are the ratios of Qd to Qs in the saturation region. From these four terminal charges, 9 transcapacitances C(terminal,voltage) are calculated inside the BSIM3 model as partial derivatives with respect to the voltages Vgb, Vdb and Vsb. The abbreviation can be interpreted as: Cggb ..... partial derivative of Qg with respect to Vgb
Capacitance Model
5-52
Chapter
Chapter
Page
Page
Contents
Index
Partial derivatives of Qg: Qg C ggb = ----------- V gb Qg C gdb = ----------- V db Qg C gsb = ---------- V sb Partial derivatives of Qd: Qd C dgb = ----------- V gb Qd C ddb = ----------- V db Qd C dsb = ---------- V sb Partial derivatives of Qb: Qb C bgb = ----------- V gb Qb C bdb = ----------- V db Qb C bsb = ---------- V sb (68) (67) (66)
Capacitance Model
5-53
Chapter
Chapter
Page
Page
Contents
Index
The 9 transcapacitances introduced above are shown in the following three plots for a simulation setup as shown below:
Capacitance Model
5-54
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-41. Partial derivatives of Qg, Qb and Qd with respect to Vdb, Vgb and Vsb
Capacitance Model
5-55
Chapter
Chapter
Page
Page
Contents
Index
Capacitance Model
5-56
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-43. Different Parts of Overlap Capacitance C_Gate_SDB The overlap capacitance C_Gate_SDB consists of: V bs& % C jbs = C j # 1 + M j --------$ Pb " ! where: Cggb Cgd,overlap Cgs,overlap Cgb,overlap intrinsic capacitance overlap capacitance between gate and drain overlap capacitance between gate and source overlap capacitance between gate and bulk (69)
Other capacitances can be calculated in the same way. Please refer to the BSIM3 manual for more details.
Capacitance Model
5-57
Chapter
Chapter
Page
Page
Contents
Index
Chapter
Chapter
Page
Page
Contents
Index
The model itself is implemented as macro model as shown in Figure 5-44 and no changes are done in the BSIM3v3.2 model code itself. This is the ultimate pre-condition for its use in a commercial circuit simulator which includes the BSIM3v3.2 model and makes it available to circuit design engineers. The BSIM3v3.2 model already consists of a non-quasi-static model and an accurate capacitance model which makes it the ideal basis for RF simulations. However, the description of the resistance behavior of a transistor is very poor. In the BSIM3v3.2 model itself, no gate resistance is included. Due to the nature of the MOS transistor, such a resistance cannot be seen in the DC operation region. However, looking at the real existing poly silicon gates of modern MOS devices, there is a resistance which cannot be neglected in AC simulations. This resistance Rgate has a major influence on the reflections S11 of an input signal to the MOS transistor as demonstrated in Figure 5-45. It should be noted, that the parameter Rgate in this high frequency model is used to fit the input reflection of the MOS transistor. Therefore it is very likely that Rgate has a different value than the measured sheet resistance of the poly-Si gate during process characterization on PCMs using for instance a van-der-Pauw test structure. The second enhancement in the RF BSIM3v3.2 macro model is a resistance network for the substrate resistance which is described by the three resistors Rbulk1, Rbulk2 and Rbulk3 [7]. The substrate resistance can be seen in the back-reflection S22 at the output of the transistor. Together with the resistance network, the internal drain-bulk junction diode and source-bulk junction diodes of the BSIM3v3.2 model are replaced by the external elements Djdb_area, Djdb_perim, Djsb_area and Djsb_perim. The de-coupling diodes account for the same voltage dependant values of the bottom and the sidewall capacity as the internal junction capacitances. This replacement is the prerequisite for a correct modeling of the substrate resistance.
Chapter
Chapter
Page
Page
Contents
Index
With this approach, the model is valid for both the DC and the RF behaviour of the transistor.
Figure 5-45. Influence of gate resistance on input reflection S11 This macro model approach results in a subcircuit for an RF MOS transistor, which is shown in part on the following page.
Chapter
Chapter
Page
Page
Contents
Index
* ******************************************************* * Subcircuit of the BSIM3 RF macro model created by the * 'BSIM3v3 Modeling Package' * * The subcircuit contains the ideal MOS transistor modeled with the * UCB BSIM3v3.2 model and additional peripheric elements to enable * high frequency simulations * * Advanced Modeling Solutions, October 1999 * * ******************************************************* .SUBCKT nmos_0p25_tran1_bsim3_hf 1 2 3 * .MODEL bsim_mos_transistor NMOS + LEVEL=8 VERSION=3.2 MOBMOD=1 CAPMOD=3 ... NOIC=-1.4E-12 * .MODEL bsim_diode_area D + CJO=0.0005 VJ=1 M=0.5 IS=9.5E-05 N=1.05 * .MODEL bsim_diode_perim D + CJO=5E-10 VJ=1 M=0.33 IS=8.7E-11 N=1.05 * ******************************************************* * Subcircuit of the BSIM3 RF macro model for n-type MOS transistors * ******************************************************* *.SUBCKT bsim_ac_pel 1 2 3 4 * * --------- Gate network -----------------------------Lgate 2 20 4.534E-11 Cgd_ext 20 11 1.648E-14 Cgs_ext 20 31 0 Rgate 20 21 13.97 * --------- Drain network ----------------------------Ldrain 1 11 4.299E-11 * --------- Source network ----------------------------Lsource 3 31 4.299E-11 * --------- Substrate network ------------------------* Diodes are for n-type MOS transistors * Djdb_area 12 11 bsim_diode_area AREA=3E-11 Djdb_perim 12 11 bsim_diode_perim AREA=6.6E-05 * Djsb_area 32 31 bsim_diode_area AREA=4E-11
Chapter
Chapter
Page
Page
Contents
Index
Djsb_perim 32 31 bsim_diode_perim AREA=8.8E-05 * Rsub1 40 4 937.6 Rsub2 12 40 100 Rsub3 32 40 100 * --------- Ideal mos transistor ---------------------MAIN 11 21 31 40 + bsim_mos_transistor L=3.5E-07 W=6E-05 AD=0 AS=0 PD=0 PS=0 * .ENDS
Chapter
Chapter
Page
Page
Contents
Index
Modeling Strategy
Modeling the AC behaviour of a MOS device with the BSIM3v3 model heavily depends on the accurate modeling of the DC curves and the capacitances at low frequencies, e.g. 10kHz to 1MHz. However, more and more applications, especially in the telecommunication industry, require the modeling of MOS transistors for the use in a frequency range of 1 to 10GHz. Therefore, S-parameter measurements have to be done (see also the chapter about test structures for Sparameters) to cover this frequency range by a proper device model. As is pointed out, using the BSIM3v3 model for high frequency applications requires some special attention in the model strategy. We found, that the following procedure gives the most accurate results: Measurement of DC and CV curves Extraction of the BSIM3v3 model parameters from DC and CV measurements with a special emphasis on a physically based extraction strategy. Here, model parameters should not be used for fitting purposes, they should have a correct physical meaning The modeling of the output characteristic Id=f(Vds) and the output resistance Rout=f(Vds) is very important for further S-parameter measurements (see Figure 5-48 and Figure 5-47) Performing S-parameter measurements and proper de-embedding of parasitics The starting points of the S-parameter curves at the lowest frequency can be modeled by fitting the curves with DC and capacitance parameters. The following diagrams describe this influence on the high frequency behavior
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-47. Influence of Incorrectly Modeled Output Characteristic on S21 Extraction of the gate resistance from the input reflection S11 (see Figure 5-48) Verification of the gate - drain overlap capacitance for higher frequencies Extraction of the substrate resistance network parameters from S22 (see Figure 5-48) If a good fitting could not be found, additional peripheral elements like inductances at drain, gate or source should be added in a further sub-circuit
Chapter
Chapter
Page
Page
Contents
Index
Chapter
Chapter
Page
Page
Contents
Index
Temperature Dependence
Please use the model 'bsim3_tutor_temp.mdl' provided with the BSIM3v3 Modeling Package to visualize the temperature model parameters. Load the file into IC-CAP and run the different macros to see how certain parameters affect the device behavior of a deep submicron MOS transistor.
(70)
N i = 2,6310 T
(71)
Unfortunately, the surface potential S, which is a very important model parameter from a physical point of view is not temperature dependent in BSIM3. % N ch & s = 2V tm ( T nom ) ln # ------------------------$ ! N i ( T nom )" (72)
Chapter
Chapter
Page
Page
Contents
Index
Temperature Effects
In addition to the built-in temperature dependencies, the following temperature related effects are modeled in BSIM3. They are related to threshold voltage, mobility, saturation of carrier velocity, drain-source resistance, and the saturation current of the drain/source bulk diodes. a) Threshold Voltage KT 1 L % & T - + KT 2 V bseff$ % ------------ 1& V th ( T ) = V th ( T nom ) + # KT 1 + -------------! " L eff ! " T nom The behavior of the threshold voltage for a large and a short device is shown in Figure 5-50. (73)
Figure 5-50. Threshold Voltage Vth=f(T) of a Large (Left) and a Short (Right) Device
Chapter
Chapter
Page
Page
Contents
Index
b) Carrier Mobility
All four model parameters of the carrier mobility are implemented in BSIM3 with a temperature dependence: T & UTE 0 ( T ) = U0 % -----------!T " nom T U A ( T ) = UA + UA1 % ------------ 1& !T "
nom
T & U B ( T ) = UB + UB1 % -----------! T - 1" nom T & U C ( T ) = UC + UC1 % -----------! T - 1" nom The following two diagrams show the effect of a temperature dependent mobility on the transconductance of a large transistor.
Chapter
Chapter
Page
Page
Contents
Index
(78)
Chapter
Chapter
Page
Page
Contents
Index
(79)
Chapter
Chapter
Page
Page
Contents
Index
J S ( T ) = JS e
(80)
The influence of XTI on diode current and saturation current density JS is shown below.
Chapter
Chapter
Page
Page
Contents
Index
Noise Model
There are two noise models implemented in BSIM3 - a conventional noise model which is named Spice2 model and a new formulated noise model, which is referred to as BSIM3v3 noise model. The following equations and diagrams should give insight into these two noise formulations. Please use the model 'bsim3_tutor_ac_noise.mdl' provided with the BSIM3v3 Modeling Package to visualize the model parameters. Load the file into IC-CAP and run the different macros to see how certain parameters affect the device behavior of a deep submicron MOS transistor.
(81)
Noise Model
5-72
Chapter
Chapter
Page
Page
Contents
Index
(83)
where: No is the charge density at the source given by: C ox ( V gs V th ) No = -------------------------------------q and Nl is the charge density at the drain given by: C ox ( V gs V th V ds ) Nl = ------------------------------------------------------q The channel thermal noise is given by: 4kT eff - Q inv V noise, eff = ------------------2 L eff with: A bulk % & - V dseff$ Q inv = W eff L eff C ox V gsteff # 1 --------------------------------------------2 ( V gsteff + 2V tm ) ! " (87) (86) (85) (84)
Noise Model
5-73
Chapter
Chapter
Page
Page
Contents
Index
5-74
Chapter
Chapter
Page
Page
Contents
Index
First coefficient of narrow-channel 2.2 effect on VTH Second coefficient of narrowchannel effect on VTH Body-bias coefficient of narrowchannel effect on VTH DIBL coefficient in the subthreshold region Body-bias for the subthreshold DIBL effect DIBL coefficient in subthreshold region 5.3E6 -0.032 0.08 -0.07 DROUT
Mobility U0 UA UB UC Mobility First-order mobility degradation coefficient Second-order mobility degradation coefficient Body-effect of mobility degradation 670 / 250 2.25E-9 5.87E-19 -4.65E-11 cm2/ (Vs) m/V Y Y1)
5-75
Chapter
Chapter
Page
Page
Contents
Index
Subthreshold region VOFF NFACTOR CIT CDSC CDSCB CDSCD Offset voltage in the subthreshold -0.11 region Subthreshold swing factor Interface trap density 1.0 0 V F/m2 F/m2 Y Y N Y
Drain-Source to channel coupling 2.4E-4 capacitance Body-bias coefficient of CDSC Drain-bias coefficient of CDSC 0 0
F/Vm2 Y F/Vm2 Y
Drain-source resistance RDSW WR PRWB PRWG Parasitic resistance per unit width 0 Width offset from Weff for RDS calculation Body effect coefficient of RDSW Gate bias effect coefficient of RDSW 1.0 0 0 m V-0.5 1/V Y Y Y Y
5-76
Chapter
Chapter
Page
Page
Contents
Index
Channel geometry WINT WL WLN WW WWN WWL LINT LL LLN LW LWN LWL DWG DWB Channel width reduction on one side Coeff. of length dependence for width offset Power of length dependence for width offset Coeff. of width dependence for width offset Power of width dependence for width offset Coeff. of length and width cross term for width offset Channel length reduction on one side Coeff. of length dependence for length offset Power of length dependence for length offset Coeff. of width dependence for length offset Power of width dependence for length offset Coeff. of length and width cross term for length offset Coefficient of Weff's gate dependence Coefficient of Weff's substrate dependence 0 0 1 0 1 0 0 0 1 0 1 0 0 0 m m m m m m m m m/V Y Y Y Y Y Y Y Y Y Y Y Y N
m/V0.5 N
5-77
Chapter
Chapter
Page
Page
Contents
Index
Output resistance PCLM PDIBLC1 PDIBLC2 PDIBLCB DROUT PSCBE1 PSCBE2 PVAG ALPHA0 ALPHA1 BETA0 Channel length modulation coefficient Second output resistance DIBL effect Body effect coefficient of output resistance DIBL effect L dependent coefficient of the DIBL effect in output resistance 1.3 1/V Y Y Y Y Y V/m m/V m/V 1/V Y Y Y Y Y Y
First substrate current body-effect 4.24E8 coefficient Second substrate current bodyeffect coefficient The first parameter of impact ionization Length dependent substrate current parameter The second parameter of impact ionization 1.0E-5
Diode characteristic JS JSSW NJ Source drain junction saturation density Side wall saturation current density Emission coefficient of junction 1E-4 0 1 A/m2 A/m Y Y Y
5-78
Chapter
Chapter
Page
Page
Contents
Index
Capacitance CJ CJSW CJSWG MJ MJSW MJSWG PB PBSW PBSWG CGSO CGDO GGBO CGSL CGDL Source/drain bottom junction capacitance per unit area Source/drain side junction capacitance per unit length Source/drain gate side junction capacitance per unit length Bottom junction capacitance grading coefficient Source/drain side junction capacitance grading coefficient Source/drain gate side junction cap. grading coefficient 5.0E-4 5.0E-10 CJSW 0.5 0.33 MJSW F/m2 F/m F/m V V V F/m F/m F/m F/m F/m Y Y Y Y Y Y Y Y Y Y Y N Y Y
Bottom junction built-in potential 1.0 Source/drain side junction built-in 1.0 potential Source/drain gate side junction built-in potential Gate-source overlap capacitance per unit W PBSW XJ*COX/2
Gate-drain overlap capacitance per XJ*COX/2 unit W Gate-bulk overlap capacitance per 0 unit W Light doped source-gate region overlap capacitance Light doped drain-gate region overlap capacitance 0.0 0.0
5-79
Chapter
Chapter
Page
Page
Contents
Index
5-80
Chapter
Chapter
Page
Page
Contents
Index
5-81
Chapter
Chapter
Page
Page
Contents
Index
(m/V)2 Y
5-82
Chapter
Chapter
Page
Page
Contents
Index
5-83
Chapter
Chapter
Page
Page
Contents
Index
5-84
Chapter
Chapter
Page
Page
Contents
Index
Notes on Extractions
N ....Parameter extraction will be implemented in a future release Y ....Extraction for this model parameter is implemented in the BSIM3v3 Modeling Package C ....The model parameter is calculated in the BSIM3v3 Modeling Package from other parameters
5-85
Chapter
Chapter
Page
Page
Contents
Index
5-86
Chapter
Chapter
Page
Page
Contents
Index
For a proper extraction of the basic model parameters, the short and narrow channel effects should not affect the large device extraction. Also the drain-source-resistance parameters should not have an influence on the simulated behavior of the large device. For a typical 0.5 micron CMOS process with a gate oxide thickness of 11 nm a large device with channel length of 10 microns and channel width of 10 microns was found to meet these requirements. You can check this prerequisite if you only extract the parameters in the idvg/Large setup and then perform a simulation of the setup idvg/Large_m. After that simulation, perform the other geometry extractions and re-simulate the idvg/Large setup again. Now, the curve ID = f(Vgs) should not change more than roughly 5% compared to the first simulation. If the difference is bigger, a larger device should be taken to enable a good extraction of the basic model parameters. Narrow For the DUT 'Narrow_m' you should use a device with the smallest designed gate width of your process. Using more narrow devices will increase the number of parameters which can be extracted and will lead to a better fit of the curves over the range of different channel widths. Short For the DUT 'Short_m' you should use a device with the shortest designed gate length of your process. Using more short devices will increase the number of parameters which can be extracted and will lead to a better fit of the curves over the range of different channel lengths. Small For the DUT 'Small_m' you should use a device with the shortest designed gate length and the smallest designed gate width of your process. This small device will incorporate all short and narrow channel effects and will be an indicator how good your parameter extractions are. In general It is recommended to use the designed gate lengths and widths. Effects due to under diffusion or decrease of poly-Si gate length are sufficiently covered by the extraction routines and the model itself.
5-87
Chapter
Chapter
Page
Page
Contents
Index
Diode_Area_m
C_Perim_m (pn-junction)
5-88
Chapter
Chapter
Page
Page
Contents
Index
A large number of parallel switched LDD MOS transistors (e.g. 200 transistors with L=0.25m, W=10.0m) or multi-finger transistors (see shape)
A large number of parallel switched LDD MOS transistors (e.g. 200 transistors with L=0.25m, W=10.0m) or multi-finger transistors (see shape)
5-89
Chapter
Chapter
Page
Page
Contents
Index
______
Testchips
You will find an example for a test chip design, which meets most of the requirements of the extraction of BSIM3v3 model parameter, in the JESSI Report AC 41 94-3 "Description of parametrized European Mini Test Chip." Please check also the test chip design of the Fabless Semiconductor Association in the U.S. (http://www.fsa.org).
Chapter
Chapter
Page
Page
Contents
Index
The test devices must drive enough current for correct measurement results They should fulfill the specifications of the high frequency probes Additional structures should be available for the measurement of parasitic elements to deembed them from the measurements on the test device A principle layout of such a test structure is shown in Figure 5-57 below [4].
Figure 5-57. Layout of a Test Structure for a MOS Transistor The MOS transistor is designed as a finger structure with four common gates, three source areas and two drain areas. In summary, this compact layout results in a very wide gate width, which can drive a high current Ids. BSIM3v3 can handle a wide range of different test structures for S-parameter measurements of MOS transistors. The commonly used devices are explained below and a short description is given how to use them in the '__Define_DUTs' macro. The setup routine automatically calculates the overall gate width, drain and source area and the drain and source perimeter for SPICE simulations from the input in the '__Define_DUTs' macro. The probes are connected in a Ground-Signal-Ground scheme according to the recommendations in [4]. As it is shown above, the calibration plane of the network analyzer is at the end of the probe head. That means, the transmission lines which connect the DUT with the probe head must be
5-91
Chapter
Chapter
Page
Page
Contents
Index
modeled and their effect must be de-embedded from the measured data of the DUT.This can be done by measuring an OPEN and a SHORT test device without a DUT and using these measurements to de-embed the parasitic influence of the pads. The following two figures show the design of these OPEN and SHORT test structures. Both of these test structures will be used for a simple and effective de-embedding procedure (OPEN_SHORT) as will be shown later. Additional test devices, like a THROUGH device can be used to verify the de-embedding strategy. In general, the complexity of the de-embedding procedure depends on the frequency range of the measurements and the design of the test structures. However, a proper de-embedding is the absolute pre-requisite for an accurate AC modeling of the MOS transistor.
Figure 5-58. OPEN, SHORT and THROUGH structure without MOS transistor
5-92
Chapter
Chapter
Page
Page
Contents
Index
n parallel transistors
No gate: No drain: No source: L: W: Area drain: Area source: Per. drain: Per. source:
5-93
Chapter
Chapter
Page
Page
Contents
Index
b) De-embedding procedures
The DUT 'Deembedding' contains three different setups with different de-embedding methods: OPEN OPEN_SHORT USER_DEFINED They can be selected depending on the availability of test structures and the frequency range of measurements:
1.OPEN:
This the simpliest way of de-embedding and is often used for freqency ranges up to 10GHz. It is assumed, that the parasitics can be modeled using the following equivalent circuit:
Figure 5-59. Equivalent circuit for the parasitic elements (including MOS-Transistor) The OPEN device is measured and the S-parameters of the DUT are calculated as shown next:
5-94
Chapter
Chapter
Page
Page
Contents
Index
Ydut -> Sdut where: Stotal -> measured S-parameters of the DUT including parasitics Sopen -> measured S-parameters of the OPEN test structure Sdut -> S-parameters of the DUT without influence of the parasitics Yxxx -> transformed Y-parameters with 'Ytotal = TwoPort(Stotal,"S","Y")' The typical behaviour of this test structure is shown in Figure 5-60.
5-95
Chapter
Chapter
Page
Page
Contents
Index
2.OPEN_SHORT:
This is a very fast and effective way of de-embedding from measurements of an OPEN and a SHORT device. It is useful for frequencies above roughly 10GHz if the accuracy of the OPEN method is not satisfying. This method is described in detail in the IC-CAP demo_features. (See the file: $ICCAP_ROOT/ examples/demo_features/4extraction/deemb_short_open.mdl) It is assumed, that the parasitics can be modeled using the following equivalent circuit:
Figure 5-61. Detailed equivalent circuit of MOS-Transistor The transistor is located between nodes: Gate = 222, Drain = 111, Source,Bulk = 333 Regarding the two test structures OPEN and SHORT and their equivalent circuits, it is assumed that there are ONLY parallel parasitics followed by serial parasitics. If this pre-requisite is valid, the measured data of the SHORT device and the measured data from the DUT have to be deembedded from the outer parallel parasitic elements first (after a conversion of S to Y parameters): Zdut_without_open = Z(Ytotal - Yopen) Zshort_without_open = Z(Yshort - Yopen)
5-96
Chapter
Chapter
Page
Page
Contents
Index
The subsequent step is to de-embed the measured data of the DUT from the serial parasitic elements and convert them back to S-parameters: Sdut = S(Zdut_without_open - Zshort_without_open) The typical behaviour of the OPEN_SHORT structure is shown below:
3.USER_DEFINED:
This setup can be used to implement user-specific de-embedding procedures with other test structures than OPEN and SHORT or to achieve a higher quality in de-embedding. Please have a look at the transform 'deembed_all' and you will find the entry point for your specific de-embedding procedure. The ultimate tool for de-embedding with IC-CAP is the 'De-embedding Tool-kit' where a large number of ready-to-go solutions together with the theoretical background can be found. Please contact Dr. Franz Sischka from HP EEsof (franz_sischka@hp.com) for more details
c) Verification procedures
The BSIM3v3 Modeling Package provides a method to verify the de-embedding. It uses a THROUGH dummy test device. After a correct de-embedding of the parasitic components, the Sparameters of the THROUGH should show the behavior of an ideal, matched transmission line with Z0=50 Ohm and a TD which represents the electrical length of the through line in the THROUGH dummy device.
5-97
Chapter
Chapter
Page
Page
Contents
Index
The S11 and S22 curve should be concentrated in the center of the Smith chart, while S21 and S12 should both begin at (1+j*0) and turn clockwise on the unity circle. If this pre-assumptions are not given, the following items should be checked: Is the calibration o.k.? If the OPEN method is used, consider to enhance the de-embedding quality by using the OPEN_SHORT method, which removes the inductive parasitics in the measured data. If the OPEN_SHORT method is used and the frequency is very high (>30GHz), it should be checked, whether the assumptions for using OPEN_SHORT are still given. The easiest way to do this, is to model the OPEN and the SHORT device using the equivalent circuits given in Test_open and Test_short. More about de-embedding If you want to use another than the implemented OPEN or SHORT-OPEN de-embedding method, please contact Advanced Modeling Solutions. Through the open architecture of the BSIM3v3 Modeling Package an integration of new features can be done according to your specifications.
5-98
Chapter
Chapter
Page
Page
Contents
Index
5-99
Chapter
Chapter
Page
Page
Contents
Index
5-100
Chapter
Chapter
Page
Page
Contents
Index
Extraction Strategy
This section describes two aspects of the extraction strategy: a group extraction and a physically oriented model parameter extraction.
5-101
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-63 shows the principle procedure of model parameter extraction as it was used in older models like the MOS Level 3 model. The model parameter Px is determined from the measured electrical behavior of one single test transistor. The measured data is transformed in such a way that Px can be determined with regression methods.
Figure 5-63. Model Parameter Extraction from Single Devices In contrast, the group extraction strategy, which is shown in Figure 5-64 uses the measured electrical behavior of several test transistors with different gate lengths and gate widths.
Figure 5-64. Group Extraction Strategy In a first step, intermediate values like the threshold voltage Vth are determined and stored in a new data array as a function of gate length. In the next step, this new data array is transformed in such a way that the model parameters Px can be determined with regression methods. Parameters extracted with this method describe the behavior of the devices in a wide range of channel lengths and channel widths very good.
5-102
Chapter
Chapter
Page
Page
Contents
Index
5-103
Chapter
Chapter
Page
Page
Contents
Index
from this action are possible. In the first case, no data point is available for the extraction and the user is informed with a warning message. This may occur for instance after measurement errors or with old CMOS processes which do not show a short channel effect. As a further possible result, only one usable data point is returned. From this data point one model parameter can be determined while the second one has to be set to its default value.
Figure 5-65. Short Channel Effects in Vth as a Function of Gate Length and Vbs
5-104
Chapter
Chapter
Page
Page
Contents
Index
In the normal case, a group of usable data points can be identified and transformed in such a way that DVT0 and DVT1 can be extracted through linear regression methods.
5-105
Chapter
Chapter
Page
Page
Contents
Index
Figure 5-67. Setup vth_L with all Transforms used for the Model Parameter Extraction a) Control Flow The extraction of the four model parameters DVT0, DVT1, DVT2 and NLX is invoked in the macro Extract_all which controls all parameter extraction. There is a transform extract_all in every setup, which is responsible for the local control flow and for details of the model parameter extraction. The main part of this transform is shown next.
5-106
Chapter
Chapter
Page
Page
Contents
Index
UPDATE_MANUAL ! if DC_NO_SHORT > 0 then ! check for short devices L=/bsim_pel/Short_m.MAIN.L W=/bsim_pel/Short_m.MAIN.W !--- Initialize reset=0 iccap_func("failure","execute") ! failure check reset=1 ! reset failure condition !--- Start of extraction print "Start of extraction of parameters from short devices" print "----------------------------------------------------" print !--- calling aux.routines iccap_func ("Ldes","execute") ! start preparation of data iccap_func ("id_L","execute") iccap_func ("vg_L","execute") iccap_func ("vb_L","execute") iccap_func ("VTH_L","execute") ! Vth for all used devices iccap_func ("ident_curve_form","execute") ! identification ! of effects of Vth !--- calling extraction routines !--- NLX, check conditions for extraction, tuning, optimization if Extr_conf/Method_Conf/act_extr[35]==1 then iccap_func("extr_NLX","execute") ! invoke extraction end if if Extr_conf/Method_Conf/act_tune[35]==1 then iccap_func("tune_NLX_DVTx","execute") ! invoke tuning end if if Extr_conf/Method_Conf/act_opt[35]==1 then iccap_func("opt_NLX_DVTx","execute") ! invoke optimization end if : : iccap_func ( "failure", "execute" ) ! check failure conditions ! and write them to a logfile
In a first step, the failure conditions will be reset to zero. Then the data which is necessary for the parameter extraction is prepared as described in the section, Using IC-CAP Files for DC, CV and Diode Modeling. This step is followed by the extraction of the threshold voltage for all short
Extraction of Model Parameters 5-107
Chapter
Chapter
Page
Page
Contents
Index
devices and an identification of several short channel effects in the resulting vector Vth = f(Ldes). After this step, the first model parameter NLX is extracted. Three methods for the parameter determination are implemented, the physical based extraction, the parameter tuning and the optimization. Each of these options can be set as shown in Figure 5-68 in the configuration macro Select Extraction.
Figure 5-68. Part of the Configuration Macro Select_extraction After the extraction of all four model parameters, the failures which may have occurred are summarized and written to the failure log file log_fail. Automatic Extraction Dependent on the Device Physics All four model parameters NLX, DVT0, DVT1and DVT2 are extracted now from the vector Vth = f(Ldes). The strategy behind these procedure is to isolate every parameter or group of parameters in that region of operation where they dominate the device behavior.
5-108
Chapter
Chapter
Page
Page
Contents
Index
Tuning Approach The second possibility of the determination of a model parameter is to use the tuner function of ICCAP. Changing a model parameter with the tuner tune_NLX_DVT012 invokes the function visualize_VTH_L shown below. The tuner feature is a very intuitive method for model parameter extraction and uses the human brain which is much more flexible than any other optimization or pattern recognition algorithm to fit the simulated curve to the measured one.
5-109
Chapter
Chapter
Page
Page
Contents
Index
Optimization
The last method to determine the model parameters DVT0, DVT1, DVT2 and NLX is to use the optimization algorithm of IC-CAP. Here, the values of Vth = f(Ldes) must be extracted before the optimization can start.
5-110
Chapter
Chapter
Page
Page
Contents
Index
Using this strategy, the modeling task can be divided into two parts: The first model file bsim3_meas contains DUTs with the measured and simulated data and the definition of plots. It is used to perform measurements and to keep the fitted curves after the extraction. No extractions or optimizations are included in this file. All DUT names have the extension _m. Macros for an automatic setup of test and measurement conditions are provided to make the measurements fast and efficient. The second model file bsim_pel consists of two sections: again a data section with the actual measured and simulated data (DUTs with the extension _m) plus a section containing the extraction routines in DUTs like Large, Narrow, C_Area or Temp. The main interaction between both types of files is as follows: Perform the measurements using the model bsim3_meas Copy the data of the model bsim3_meas into bsim_pels data section (macro ___Import_data) Perform extractions (and optimization) in file bsim_pel (macro ___Extract_all) Export the final model parameter set back to the file bsim3_meas (macro ___Export_parameters)
Chapter
Chapter
Page
Page
Contents
Index
The Measurement File, bsim3_meas The model file bsim3_meas contains four kinds of DUTs, as shown in Figure 5-71.
DUTs for the configuration of the DUTs, the measurements and different temperature-dependent measurements
Chapter
Chapter
Page
Page
Contents
Index
The model file bsim3_meas contains four different kinds of macros, as shown in Figure 5-72
Macros for the configuration of the DUTs, the measurements and different temperature dependent measurements
Utilities
Chapter
Chapter
Page
Page
Contents
Index
The Extraction File, bsim_pel The model file bsim_pel contains four different kinds of DUTs, as shown in Figure 5-73.
DUT for the configuration of the parameter extraction Do not delete above this line !! DUTs with configuration information and user-defined DUTs with measured DC and CV data (with ending _m) imported from bsim3_meas
Chapter
Chapter
Page
Page
Contents
Index
The model file bsim3_pel contains four different kinds of macros, as shown in Figure 5-74.
Macros for the import and export of measured data and extracted model parameters
Utilities
Chapter
Chapter
Page
Page
Contents
Index
Chapter
Chapter
Page
Page
Contents
Index
Chapter
Chapter
Page
Page
Contents
Index
In item #3, you can specify the different voltages for DC measurements. Please note that you always specify the values for n-type transistors; they will automatically be transformed to negative values if you have specified TYPE = -1 (PMOS) for a p-type transistor.
The same procedure is done with CV(item #4) and diode(item #5) measurement conditions.
Chapter
Chapter
Page
Page
Contents
Index
Chapter
Chapter
Page
Page
Contents
Index
a) Modelname
Execute the macro Define_DUTs in the bsim3_meas model and configure your DC and CV DUTs. You will be prompted for the name of your model. If you would like to change the name, do it within this macro. Do not change the model name by simply renaming the model in the IC-CAP main window; this will disable the subcircuit simulation concept.!
b) CV Measurements
The procedure for setting up CV measurements is similar to the setting of DC measurements. You will first be prompted whether you want to perform CV measurements in general:
Chapter
Chapter
Page
Page
Contents
Index
The following menu allows a selection of different numbers of CV test devices. Specify them if necessary and continue. A warning window will inform you that the following operation may take some seconds. During this operation, DUTs will be deleted, copied or renamed. If you interrupt this procedure, the program runs into an undefined state and you will have to repair it manually. This warning will be shown in different steps of the program.
Chapter
Chapter
Page
Page
Contents
Index
The following choices allow you to disable the measurement of certain devices.
Chapter
Chapter
Page
Page
Contents
Index
The settings for the CV DUTs are displayed in the following window:
You can change each DUT by typing in the DUT number and then changing the values for this DUT. Please use only blanks as separators for different values.
Chapter
Chapter
Page
Page
Contents
Index
c) DC Measurements
In the first step you will be prompted for a definition of the numbers of different devices.
Specify them if necessary and continue. A warning window will inform you that the following operation may take some seconds. During this operation, DUTs will be deleted, copied or renamed. If you interrupt this procedure, the program runs into an undefined state and you will have to repair it manually. This warning will be shown in different steps of the program.
Chapter
Chapter
Page
Page
Contents
Index
The actual setting for the DC DUTs will be shown in the following window:
You can change the data of every single DUT by specifying the number of the DUT and you will be prompted for the following input:
You can change the name, the dimensions or the matrix port connections now. Different values must be separated by blanks. Comma or semicolon as separators are not allowed.
Chapter
Chapter
Page
Page
Contents
Index
If you are using more than the minimum set of devices, the following conventions must be maintained (example for different short devices):
DUT Short_m Short_m1 Short_m2 : Large_m 10 large device channel length [micron] 0.5 0.8 1.2 comment shortest designed channel length according to the design rules of the process
Example: Short_m is the DUT of the smallest device, Short_m1 is the DUT of the next larger device etc. The same convention applies to the series of Narrow and Small devices. If you have done the configuration, the channel lengths and channel widths of your test transistors are displayed in the following diagram. You should check again to ensure that the conventions explained above are maintained. If not, please re-run the Define_DUTs macro.
The program will now create the necessary DUTs for the DC test. Do not interrupt this procedure. If you want to change or correct a single parameter of a test device, e.g. the gate length of a certain transistor, you can also use the macro Change_DUTs which keeps the number of test devices and only changes certain parameters.
Using IC-CAP Files for DC, CV and Diode Modeling 5-126
Chapter
Chapter
Page
Page
Contents
Index
d) Parasitic Diodes
The setup of measurements for the drain/bulk and source/bulk diodes is done in the same way as DC and CV setups.
Chapter
Chapter
Page
Page
Contents
Index
Perform Measurements
You can perform the measurements by selecting the "Measure" button for the different DUTs manually or you can invoke the macros DC_measurements or CV_measurements. In this case, the following menu will be displayed and you can select the current device to measure. The menu is nearly the same for CV measurements. Here you may first want to calibrate your measurement equipment with the IC-CAP calibrate command. For the use of a HP switch matrix, the macros DC_drive_matrix and CV_drive_matrix are supplied. Please check if the commands in this macro are the correct ones in case you are using another matrix.
If you are measuring at different temperatures, please heat up or cool down your devices to the given temperatures and run the macro DC_measurements in the correspondent model. You can either save all models in separate files or in a common file.
Chapter
Chapter
Page
Page
Contents
Index
Documentation of Test
The whole test and measurement setup can be printed for documentation. Please use the macro Print to select the different reports to print.
Chapter
Chapter
Page
Page
Contents
Index
In the first step, you can select whether you want to use all measured devices for parameter extraction. This can be helpful, if for instance an error in the measurement data was not recognized during the measurements and you do not want to re-measure all the devices. For this purpose call the macro Select_devices and you can make your selection in the window shown below:
In the next step, it is possible to select a certain parameter extraction strategy. If you call the macro Select_Extraction you can choose between three different methods for extracting a model parameter.
Using IC-CAP Files for DC, CV and Diode Modeling 5-130
Chapter
Chapter
Page
Page
Contents
Index
While the parameter extraction routines in the BSIM3v3 Modeling Package are all physically oriented, it is very common that some things must be adapted to the CMOS process that is being characterized. For instance, this can be the fact that you are using a 3 micron process, where no short channel effects can be observed and therefore these parameters cannot be extracted or you have special processes with special effects which are not covered by the BSIM3v3 model. However, in most cases it is possible to find a work-around for these problems by adapting the parameter extraction strategy. The first one of three different extraction methods is the recommended physically oriented one, where the model parameters are extracted on the base of device equations (here called Extraction). The second one is the manual parameter tuning feature of IC-CAP. For every model parameter, such a tuning feature has been implemented and the user can manually try to fit the simulated curves to the measured ones. This may be very useful compared to the third method, the optimization. Experience has shown that the human brain is better able to determine a good fit than the relatively mechanistic optimization function which has the RMS error as a target function. If you do not want to extract a certain parameter, please de-select all three methods. Please try and compare all three methods with a simple example, e.g. the threshold voltage VTH0.
Chapter
Chapter
Page
Page
Contents
Index
Chapter
Chapter
Page
Page
Contents
Index
Naturally, parameter extraction is not a linear process and you will need some iteration steps to find the best way in extracting the BSIM3v3.2 model parameters for your process. Before you start with the model parameter extraction, please run the macro Init_parameters and then run the extractions with Extract_all. You can watch the intermediate results of the parameter extractions in the status window. Any failures that occur during the extractions will be written to the file log_fail, which will be displayed after the extractions as in the example below:
Now you can verify your parameter extraction by simulating all measured DUTs using the macro Simulate_all. Using the macro Extract_all is the simplest and fastest way to perform model parameter extractions. However, it is also possible to do the extractions more manually. Before you start with them, please run the macro Init_parameters to reset the model parameters and perform a first simulation. Then you can run the different extract_all transforms described in the section, Extraction of Model Parameters manually. It is strongly recommended to use these extract_all transforms because they invoke all the necessary data preparation functions for a proper model parameter extraction. Using this method, you can also create your specific model parameter extraction strategy which really fits your special needs.
Chapter
Chapter
Page
Page
Contents
Index
In some cases, it may be necessary to change some boundaries for model parameter extractions. As an example, the lowest usable value for the drain current for the extraction of subthreshold parameters heavily depends on the quality of the measurements in this region of operation of a transistor. Therefore, this lowest limit can be changed manually to adapt the extractions to the available measured data with the macro Set_extraction_variables as shown below:
Chapter
Chapter
Page
Page
Contents
Index
Naming Conventions
To keep the extraction routines as simple and fast as possible, no overhead for checking the semantics and syntax of the names of parameters, setups and DUTs is implemented. Therefore the following naming conventions must be strictly maintained. (Changing names will lead to errors and will disturb the extraction flow.) Do not delete the first twelve DUTs in bsim_pel: Large, Narrow, Short, Small, C_Area, C_Perim, C_Oxide, C_Overlap, C_Perim_Gate, Diode, Temp and Extr_conf. They include the extraction and optimization strategy. Do not change the model name bsim_pel. This would affect the path names in the macros (ICCAP_FUNC statements). For the names of the geometry parameters use capital letters like L, W, AD as proposed. The names of DUTs and setups for the measured data in the model file bsim_pel must be in the following style and the DUTs must at least contain the following setups (see original file set that you have been provided with):
DUT Setup Inputs/Outputs
Large_m
: : :
If you like to use more than the minimum set of devices, the following conventions must be maintained (example for different short devices):
Chapter
Chapter
Page
Page
Contents
Index
DUT
Comment
Short_m
Example: Short_m is the DUT of the smallest device, Short_m1 is the DUT of the next larger device etc. The same convention applies to the series of Narrow devices.
Chapter
Chapter
Page
Page
Contents
Index
Chapter
Chapter
Page
Page
Contents
Index
You can define the number of different temperatures and the program creates the correspondent models:
Please take care that you always re-run this macro if you change your DC DUTs or your DC measurement conditions in the model bsim3_meas. These conditions are directly copied to the temperature dependent models. Please heat up or cool down your devices to the given temperatures and run the macros DC_measurements or Diode_measurements in the correspondent model. You can either save all models in separate files or in a common file.
The models bsim3_meas and bsim3_meas_400K contain measured data at nominal temperature and at T=400K. The strategy of temperature parameter extraction is to extract first a set of model parameters at nominal temperature T=300K (=27C). For all other temperatures, a subset of
Chapter
Chapter
Page
Page
Contents
Index
temperature dependent DC model parameters, such as U0 or RDSW, is extracted. Both tasks are done by bsim_pel. From this subset of parameters, the temperature parameters like UTE or PRT are generated in a final step. The model bsim_temp controls the whole extraction and calculates the final model parameter set with all temperature model parameters from the single parameter sets at different temperatures. Run the macro Extract_temp in the bsim_temp model to start the extraction of temperature dependant model parameters according to the described strategy. You must restart the macro after the extraction of every model parameter set for a certain temperature.
Chapter
Chapter
Page
Page
Contents
Index
Perform the different steps, starting with the initialization as shown below:
Continue with the model parameter extraction at nominal temperature and different temperatures as shown in the IC-CAP window. In the final step, the temperature modeling parameters are created and stored in the resulting model parameter set. After a succesful parameter extraction, you can export your model parameters to the bsim3_meas model or the related temperature dependant models and store them together with the measured data. Use the macro Export_parameters for this task.
Chapter
Chapter
Page
Page
Contents
Index
bsim_ac_pel bsim3_ac_meas
Using this strategy, the modeling task can be divided into two parts: The first model file bsim3_ac_meas contains DUTs with the measured data. It is used to perform measurements. No extractions or optimizations are included in this file. All DUT names have the extension _m. Macros for an automatic setup of test and measurement conditions are provided to make the measurements fast and efficient. In addition, this model file contains all the measurements and procedures for the de-embedding of the test devices. The second model file bsim_ac_pel consists of two sections: again a data section with the actual measured and simulated data (DUTs with the extension _m) plus a section containing the extraction routines in DUTs like S_parameter. The main interaction between both types of files is as follows: Perform the measurements using the model bsim3_ac_meas De-embedding of the measured device behaviour in bsim3_ac_meas Copy the data of the model bsim3_ac_meas into the data section of bsim_ac_pel's (macro___Import_data) Perform extractions (and optimization) in file bsim_ac_pel (macro ___Extract_all) Create a final sub-circuit for the high frequency macro model in file bsim_ac_pel
Chapter
Chapter
Page
Page
Contents
Index
Chapter
Chapter
Page
Page
Contents
Index
In the next step, you can specify the different voltages for the DC and AC measurements. Please take care that you always specify the values for a n-type transistor! They will automatically be transformed to negative values if you have specified TYPE = -1 (PMOS) for a p-type transistor.
Chapter
Chapter
Page
Page
Contents
Index
Change the number if necessary and continue. A warning window will inform you that the following operation may take some seconds. During this operation, DUTs will be deleted, copied or renamed. If you interrupt this procedure, the program runs into an undefined state and you will have to repair it manually. This warning will be shown in different steps of the program.
Chapter
Chapter
Page
Page
Contents
Index
The actual setting for the AC DUTs will be shown in the following window:
You can change the data of every single DUT by specifying the number of the DUT and you will be prompted for the following input:
Now you can change the name, the dimensions or the comment. Different values must be separated by blanks. Commas or semicolons as separators are not allowed. Finally, the macro will generate the necessary DUTs for the AC test. Do not interrupt this procedure. If you want to change or correct a single parameter of a test device, e.g. the gate length of a certain transistor, you should use the macro Change_DUTs which keeps the number of test devices together with the measured data and changes certain parameters only.
Perform Measurements
You can perform the measurements by selecting the "Measure" button for the different DUTs manually or you can invoke the macros AC_DUT_measure or AC_OPEN_measure. The first macro is used for the DUTs and the second one for the OPEN test devices. Please note that the macro assumes one separate OPEN device existing for every DUT.
Using the IC-CAP Files for AC Modeling 5-145
Chapter
Chapter
Page
Page
Contents
Index
In both cases the following menu will be displayed reminding you about the connections of the SMUs and network analyzer ports to the DUTs.
In the next step, you can select the device to be measured from the following table:
Chapter
Chapter
Page
Page
Contents
Index
Please note, that 'MASTER_AC_PEL.mdl' is a macro model which contains four different single models 'bsim_ac_pel', 'bsim_mos_transistor', 'bsim_diode_perim' and 'bsim_diode_area'. The last three models (bsim_mos_transistor, ..., bsim_diode_area) contain single models for the MOS transistors and the junction diodes, while the model bsim_ac_pel contains the sub-circuit description of the RF BSIM3v3 model, the extraction routines and all other control macros.
Chapter
Chapter
Page
Page
Contents
Index
Please see the following flow-chart for an explanation of the data flow between the different model files.
To import the measured data from 'bsim3_ac_meas' simply run the macro 'Prepare_data' in 'bsim_ac_pel'. You will be prompted to define a project directory, where the .mdm files and the resulting SPICE subcircuits are stored. The name of this directory is /home/iccap/bsim3/project in our current example. In addition, the name of the model parameter set which contains the DC and CV model parameters already extracted in a former session using the bsim_pel model has to be specified (see the following window). After all these specifications, the macro writes the measured data in separate .mdm files for every DUT in the project directory. The device specific information, like gate length or gate width etc. is copied to the 'bsim_ac_pel' model. This structure of data storing has been choosen because Sparameter measurements usually generate a huge amount of data which can slow down IC-CAP when it is permanently present in one big model file.
Chapter
Chapter
Page
Page
Contents
Index
Now, the pre-requisites for extracting model parameters are given and the 'bsim3_ac_meas' model can be closed and removed from the main window.
Chapter
Chapter
Page
Page
Contents
Index
Using the macro 'Extract_all' gives you all the information to perform model parameter extractions. In contrast to the extraction of DC and CV model parameters, the high frequency modeling for the RF BSIM3v3 macro model requires more user interaction. Please use the functions which are described below to perform this modeling.
In general, two major steps are required for successful high frequency modeling - a correct representation of the DC and CV behaviour and a fit of the external parasitic elements of the macro model. The physical background of the modeling strategy is explained in detail in the section High Frequency Behavior.
Chapter
Chapter
Page
Page
Contents
Index
You can select the device to extract. Before you start with the model parameter extraction, please run the macro 'Init_parameters' to reset the model parameters and calculate some constants.
Chapter
Chapter
Page
Page
Contents
Index
The DC extractions provide you the possibility to re-adjust some selected DC model parameters to enhance the fitting quality of the output curve and the output resistance of the AC test devices. Please see the selection of functions in the window below.
Chapter
Chapter
Page
Page
Contents
Index
The main modeling part is done with the extraction functions in DUT/Setup Extract/S_parameter. Please see the README transform for more information.
Chapter
Chapter
Page
Page
Contents
Index
To use the extracted macro model in a SPICE like simulator, the macro 'Create_subcircuit' generates the necessary subcircuit:
The subcircuit is generated for the general UC Berkeley SPICE input format, which is used e.g. in SPICE3e2. Please see the example below for the format of the sub-circuit. The generation routine can easily be customized to add additional features of commercial simulators.
Chapter
Chapter
Page
Page
Contents
Index
* ************************************************************************** * * Subcircuit of the BSIM3 RF macro model * * * created by the 'BSIM3v3 Modeling Package' * * * * * * The subcircuit contains the ideal MOS transistor modeled with the UCB * * * BSIM3v3.2 model and additional peripheric elements to enable high * * * frequency simulations * * * * * * Advanced Modeling Solutions, October 1999 * * ************************************************************************** .SUBCKT Device_1_bsim3_hf 1 2 3 * .MODEL bsim_mos_transistor NMOS + LEVEL = 8 VERSION = 3.2 MOBMOD = 1 CAPMOD = 3 ..... NOIC = -1.4E-12 * .MODEL bsim_diode_area D + CJ0 = 0.0005 VJ = 1 M = 0.5 IS = 9.5E-05 N = 1.05 * .MODEL bsim_diode_perim D + CJ0 = 5E-10 VJ = 1 M = 0.33 IS = 8.7E-11 N = 1.05 * * ***************************************************** * * Subcircuit of the BSIM3 RF macro model * * * for n-type MOS transistors * * ***************************************************** *.SUBCKT bsim_ac_pel 1 2 3 * --------- Gate network -----------------------------Lgate 2 20 1E-12 Cgd_ext 20 11 1E-16 Cgs_ext 20 30 0 Rgate 20 21 1 * --------- Drain network ----------------------------Ldrain 1 11 1E-12 * --------- Substrate network ------------------------* Diodes are for n-type MOS transistors * Djdb_area 12 11 bsim_diode_area AREA = 3E-11 Djdb_perim 12 11 bsim_diode_perim AREA = 4E-11 * Djsb_area 31 30 bsim_diode_area AREA = 6.6E-05 Djsb_perim 31 30 bsim_diode_perim AREA = 8.8E-05 * Rsub2 12 40 1 Rsub3 31 40 1 Rsub1 40 30 1 * --------- Ideal mos transistor ---------------------MAIN 11 21 30 40 + bsim_mos_transistor L = 3.5E-07 W = 6E-05 AD = 0 AS = 0 PD = 0 PS = 0 * RGND 3 30 0.001 * .ENDS
Chapter
Chapter
Page
Page
Contents
Index
Variable _General_settings TYPE POLARITY OPEN_RES SIMULATOR COX esi esio2 q kb Eg0 Vtm PHI NI TEMP TNOM gen_fail fail_file VBM NSUB NCH
Setting 1 NMOS 1.0E12 spice3 0.00627545 1.04E-10 3.4515E-11 1.60219E-19 1.38062E-23 1.17 0.0344683 0.684466 7.04291E+12 27 27 1 11 2 1.77434E+17 1.44486E+17
Comment device type " constants for simulation " constants for extraction " " " " " " " " defined by user " failure indicator file descriptor for failure file model parameters not included in the circuit "
5-156
Chapter
Chapter
Page
Page
Contents
Index
Variable XT _Data NOM_DATASOURCE DATASOURCE MODELNAME __DC_variables tmp_pind tmp_ping tmp_pins tmp_pinb DC_VGSTART_VD DC_VGSTEP_VD DC_VDSTART DC_VDSTOP DC_VDSTEP DC_VGSTART
Comment " model name of measured data like above at TEMP 27 DO NOT CHANGE pin assignment for switch matrix " " settings for DC measurements " " " " "
Variable DC_VGSTOP DC_VGSTEP DC_VBSTART DC_VBSTOP DC_VBSTEP DC_VGSTART_VD DC_NO_ADD DC_NO_LARGE DC_NO_NARROW DC_NO_SHORT
Comment " " " " " " number of DC DUTs " " "
5-157
Chapter
Chapter
Page
Page
Contents
Index
Variable DC_NO_SMALL DC_NO_TRAN DC_L[0] DC_L[1] DC_L[2] DC_L DC_W[0] DC_W[1] DC_W[2] DC_W DC_DUT[0] DC_DUT[1] DC_DUT[2] DC_DUT __CV_variables CV_DUT CV_L CV_W CV_n CV_AREA CV_PER CV_TEST CV_VGSTART CV_VGSTEP CV_VBSTART CV_VBSTOP CV_VBSTEP CV_NO_TRAN __Temp_variables
Setting 1 3 1E-05 2.5E-07 2.5E-07 ICCAP_ARRAY[3] 1E-05 1E-05 3.2E-07 ICCAP_ARRAY[3] Large_m Short_m Small_m ICCAP_ARRAY[3] ICCAP_ARRAY[0] ICCAP_ARRAY[0] ICCAP_ARRAY[0] ICCAP_ARRAY[0] ICCAP_ARRAY[0] ICCAP_ARRAY[0] ICCAP_ARRAY[0] 2.5 0.1 -2 0.2 0.1 0
Comment " " geometry of DC devices " " " " " " " names of meas. DC DUTs " " " names of meas. CV DUTs geometry of CV devices " " " " flag for measured CV data settings for CV measurements " " " " number of CV Duts
5-158
Chapter
Chapter
Page
Page
Contents
Index
Comment number of measured temp. different temperatures " " " flag for temperature meas.
Variable DI_DUT[0] DI_DUT[1] DI_DUT DI_AREA[0] DI_AREA[1] DI_AREA DI_PER[0] DI_PER[1] DI_PER DI_TEST[0] DI_TEST[1] DI_TEST DI_NO_TRAN DI_VBSTART DI_VBSTOP DI_VBSTEP __Extraction_settings idmin_voff
Setting Diode_Area_m Diode_Perim_m ICCAP_ARRAY[2] 4E-10 5E-10 ICCAP_ARRAY[2] 8E-05 0.001 ICCAP_ARRAY[2] y y ICCAP_ARRAY[2] 2 -0.6 0 20m 1e-9
Comment names of meas. diode DUTs " " geometry of diode devices " " " " " flag for measured diode data " " number of diode DUTs settings for diode measurements " " user definable settings for
5-159
Chapter
Chapter
Page
Page
Contents
Index
Variable AREA_CLM_DIBL idmin_js idmin_cds PR_CMD mins[0] mins maxs[0] maxs params[0] params inits[0] inits scale[0] scale CV_NO_BASIC CV_NO_OVER CV_NO_INT CV_VGSTART_VD CV_VGSTOP_VD CV_VGSTEP_VD CV_VDSTART CV_VDSTOP CV_VDSTEP vgd_extr vs_extr MESSAGE
Setting 2 1e-10 1e-11 lp -dlaser -0.921219 ICCAP_ARRAY[1] 0.5 ICCAP_ARRAY[1] PRWB ICCAP_ARRAY[1] -0.01224 ICCAP_ARRAY[1] 0 ICCAP_ARRAY[1] 4 2 2 0 2.5 0.5 0 2.5 0.1 1.2 1.2 1
Comment parameter extraction set in macro 'Set_extraction_variables' variables for tuning " " " " " " " " " number of basic CV DUTs number of overlap DUTs number of DUTs for intr. cap. measurement conditions for intrinsic capacitance measurement " " " Bias point for extraction of CLC, CLE switch of input windows in extraction transforms (1=on / 0 = off)
5-160
Chapter
Chapter
Page
Page
Contents
Index
References
[1] "BSIM3 Manual," University of California at Berkeley, March 1997 [2] "Characterisation System for Submicron CMOS Technologies," JESSI Reports AC41 94-1 through 94-6 [3] Peter Klein, "A consistent parameter extraction method for deep submicron MOSFETs," Proc. 27th European Solid-State Device Research Conferenc, Stuttgart, 1997 [4] "Layout Rules for GHz Probing", Application Note Cascade Microtech [5] F. Sischka, "Deembedding Toolkit," Hewlett-Packard GmbH, Bblingen, Germany [6] File: "deemb_short_open.mdl" in IC-CAP examples, Hewlett-Packard EEsof [7] W. Liu et al., "R.F. MOSFET Modeling Accounting for Distributed Substrate and Channel Resistances with Emphasis on the BSIM3v3 SPICE Model," Proc. IEEE IEDM, 1997 How to get the BSIM3v3 manual from University of Berkeley/California: University of Berkeley/California provides a comfortable way to get a free copy of the BSIM3v3 manual and the BSIM3v3 source code from their world-wide web home page: http://www-device.EECS.Berkeley.EDU/~bsim3/
References
5-161
Chapter
Chapter
Page
Page
Contents
Index
Other useful internet addresses: Advanced Modeling Solutions: http://www.admos.de Compact Model Council in the United States (standardization of the BSIM3v3 model). http://www.eia.org/eig/CMC/ Agilent EEsof homepage http://www.tm.agilent.com/tmo/hpeesof
Copyright
BSIM3 is developed by the Device Research Group of the Department of Electrical Engineering and Computer Science, University of California, Berkeley and copyrighted by the University of California.
References
5-162