SPICE Gummel-Poon (SGP) BJT Model • SPICE Gummel-Poon (SGP) model improves dc characterization of EM3 model by a unified theory. • The SGP unified model was developed to improve: – base-width modulation – high-injection effects – base-widening effect resulting in τF vs. IC. • The starting point of SGP model is: – EM2-model – two additional diodes in EM2 representing the extra component of IB for β roll-off at low IC. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #1 SGP BJT Model: Starting Point C qV B′C ′ C 4 I S (0) e nC kT − 1 Csub r'c C' CDC B r'b CjC IEC/βR B' CDE ICT = ICC - ICE CjE ICC/βF qV B′E ′ E' I CC = I S e kT − 1 r'e qV B′E′ qV B′C ′ C 2 I S (0) e nEkT − 1 I EC = I S e kT − 1 E The starting point of SGP model is EM2-model with two extra diodes to account for β roll-off at low current level. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #2 SGP BJT Model: Model Parameters • EM2-parameter set: dc (EM1) bulk-ohmic resistors charge storage effects • Extra model parameters: transistor sat. current low-current β roll-off forward Early voltage inverse Early voltage knee current in ln(IC) vs. VBE inverse knee current τF vs. IC model 5-Feb-04 βF, βR, Tref, Eg (re-define IS in SGP) r'c, r'e, r'b, CjE0, φE, mE,CjC0 , φC, mC, τF,τR Csub ISS (replacement of IS) C2, nE, C4, nC VA VB IK IKR B HO #10: ELEN 251 - SGP BJT Model Saha #3 Derivation of ISS Emitter Base Spacecharge layer Collector p (x ) ← ε (x ) Spacecharge layer n(x ) xjE xE xC xjC x • Assumptions: – one-dimensional current equations hold – npn-BJT with EB junction forward biased and BC reverse biased – depletion approximation, that is, no mobile charge inside the depletion region – BJT gain is high, that is IB ≅ 0. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #4 Derivation of ISS One-dimensional current equations (HO #2, slide #66) are: Jn = qµnn(x)ε(x) + qDn(dn(x)/dx) (1) Jp = qµpp(x)ε(x) − qDp(dp(x)/dx) (2) Since, we assume IB ≅ 0, ∴ Jp = hole current in base ≅ 0 and from (2) we get, or, qµpp(x)ε(x) − qDp(dp(x)/dx) = 0 D p 1 dp( x) kT 1 dp( x ) ∴ε (x ) = = q p ( x) dx µ p p ( x) dx (3) (4) here we used, Dn/µn = Dp/µp = kT/q The direction of the ε-field in (4) aids e- flow from E → C and retards e- flow from C → E. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #5 Derivation of ISS The e- flow between E and C is given by (1): Jn = qµnn(x)ε(x) + qDn(dn(x)/dx) (1) Using (4) in (1) we get: J n = kT µ n n (x ) dp(x ) dn(x ) + q Dn p (x ) dx dx (5) qDn dp( x) dn( x) + p( x ) ∴ J n = p ( x ) n( x ) dx dx q Dn d = or, J n p ( x) [n( x) p( x)] dx (6) We integrate (6) over the neutral base width WB from xE to xC . 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #6 Derivation of ISS Emitter Base Spacecharge layer Collector p (x ) ← ε (x ) Spacecharge layer n(x ) xjE xE ∴Jn xC XC ∫ p( x )dx = qDn XE ∴Jn = XC ∫ XE xjC d [n( x) p ( x) ]dx dx qDn [n( xC ) p( x C ) − n( x E ) p ( x E )] XC x (7) (8) ∫ p ( x)dx XE 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #7 Derivation of ISS From PN-junction analysis, we know that the pn-product at the edge of the depletion regions are: p( xC ) n( xC ) = n e qV B ′C ′ kT (9) p ( x E )n ( x E ) = n e qV B ′E ′ kT (10) 2 i 2 i Substituting (9) and (10) in (8) we get: q V B′C ′ qV B′E ′ q Dn n e kT − e kT ∴Jn = xC ∫ p( x)dx 2 i (11) xE 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #8 Derivation of ISS If A = cross-sectional area of the emitter, then from (11) we can show that: qV B′E ′ qV B′C ′ − q ADn n e kT − 1 − e kT − 1 In = xC 2 i (12) ∫ p( x)dx xE Where In = total dc minority injection current from E → B in the positive x-direction. We have shown in EM1-model: I CT ( EM1− model) ≡ I CC − I EC qV B ′C ′ qV B ′E ′ = I S e kT − 1 − e kT − 1 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model (13) Saha #9 Derivation of ISS At low level injection, p(x) ≅ NA(x) in the neutral base region where xE ≤ x ≤ xc. Then we can write (12) as: q AD n ni2 qV B′E′ qV B′C′ I CT ( low − level) = xC e kT − 1 − e kT − 1 ∫ N A ( x)dx (14) xE Since xE and xC depend on applied voltages, we define the fundamental constant, ISS @ VBE = 0 = VBC. I SS ≡ q AD n n2i xC 0 ∫ N A ( x) dx (15) xE0 Where xE0 and xC0 are the values of xE and xC with applied zero voltages. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #10 Derivation of ISS - Base Charge, QB Again, (14) can be expressed as: I CT xC 0 qA N A ( x)dx ∫ 2 qV B′E ′ qV B′C′ q AD n ni xE 0 = xC xC 0 e kT − 1 − e kT − 1 ∫x p ( x)dx qAx∫ N A ( x)dx E E0 xC 0 = q AD n n 2 i qA ∫ N A ( x) dx xC 0 xE 0 xC xE 0 xE qA ∫ N A ( x )dx ∫ p( x)dx qV B′E′ − 1 − qV B′C′ − 1 e kT e kT (16) 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #11 Derivation of ISS - Normalized Base Charge, qb Defining: Q B ≡ qA xC (V B′C′ ) ∫ p ( x)dx (17) x E (VB′E ′ ) xC 0 Q B 0 ≡ qA ∫ N A ( x )dx (18) xE 0 We get: I CT = I SS I SS = qb QB 0 QB qV B′E ′ − 1 − qV B ′C′ − 1 e kT e kT qV B ′E ′ − 1 − qV B′C′ − 1 e kT e kT where qb ≡ QB/QB0 5-Feb-04 (19) (20) HO #10: ELEN 251 - SGP BJT Model Saha #12 Saturation Current ISS - Summary I CT I = SS qb qV B′E′ qV B′C ′ e kT − 1 − e kT − 1 (19) • Eq. (19) is the generalized expression for current source at all injection levels. • I SS ≡ q ADn ni2 xC 0 ∫N A is a fundamental constant @ VBE = 0 = VBC. ( x) dx xE 0 • qb ≡ QB/QB0 is the normalized majority charge in the neutral base region and accurately models base width modulation: – QB0 = majority carrier concentration @ V BE = 0 = VBC – QB = majority carrier concentration under applied voltages. • qb is expressed as bias-dependent measurable parameters in SGP. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #13 Derivation of Base Charge, QB • In order to evaluate qb, we first determine the components of QB. • For the simplicity of QB analysis, we assume: – npn-BJT is in saturation, that is, VB'E' > 0 and VB'C' > 0, then ♦ minority carriers are injected into the Base both from Emitter and Collector ♦ from the charge neutrality, total increase in majority carriers in Base = total increase in minority concentration – superposition of carriers in different regions hold ♦ total excess majority carrier density = sum of the excess majority carrier density due to each junction separately ∴ excess majority carrier concentration in base = excess carriers due to forward voltage [VB'E' + VB'C'] − depletion approximation holds. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #14 Components of Base Charge, QB If pF(x) = majority carrier concentration in base @ VB'C' = 0 pR(x) = majority carrier concentration in base @ VB'E' = 0 NA(x) = base doping concentration QF pF (x) ↓ ↓ ↑ N A (x ) QE xjE xC(VB'E') xE0 5-Feb-04 p (x ) pR (x) ↓ QR QB0 Base Collector Emitter Then the excess majority carrier concentration in the base is given by: p'(x) = [pF(x) − NA(x)] + [pR(x) − NA(x)] (21) QC xC0 xC(VB'C' ) xjC HO #10: ELEN 251 - SGP BJT Model x Saha #15 Components of Base Charge, QB From (17), the total majority charge in the base is given by: QB ≡ xC (VB′C′ ) ∫ qAp( x)dx xE (VB′E′ ) = xC (VB′C′ ) ∫ qAN A ( x )dx + xE (VB′E′ ) xC (V B′C′ ) ∫ qAp′( x)dx (22) x E (VB′E′ ) equilibrium component excess component The equilibrium component of QB can be split into three-components: xC (VB′C ′ ) ∫ qAN x E (VB′E′ ) A xE0 ( x) dx = ∫ qAN A ( x ) dx + x E (VB′E′ ) ∫ qAN A ( x ) dx + xE 0 QE 5-Feb-04 xC 0 xC ( VB′C ′ ) ∫ qAN A ( x ) dx xC 0 QB0 HO #10: ELEN 251 - SGP BJT Model QC Saha #16 Components of Base Charge, QB So that: ∫ qAN A ( x) dx + x E (VB′E′ ) xC 0 ∫ qAN A ( x ) dx + xC (VB′C ′ ) ∫ qAN xE0 QE ∴ Q B = QE + QB 0 + QC + xC (VB′C ′ ) A ( x ) dx + xC 0 QB0 xC ( VB′C ′ ) ∫ qAp′( x)dx xE (VB′E′ ) QC ∫ qAp′( x)dx (23) Emitter x E (VB′E′ ) QF pF (x) ↓ ↓ ↑ N A (x ) QE xjE xC(VB'E') xE0 5-Feb-04 p (x ) pR (x) ↓ QR QB0 Base Collector QB = xE 0 QC xC0 xC(VB'C' ) xjC HO #10: ELEN 251 - SGP BJT Model x Saha #17 Components of Base Charge, QB From (21) and (23), QB = QE + QB 0 + QC + QF xC (VB′C′ ) ∫ qA[ p F ( x) − N A ( x)]dx xE (VB′E′ ) ∫ qA[ p R ( x) − N A ( x)]dx xE (VB′E′ ) QR Emitter ∴ Q B = QE + QB 0 + QC + QF + QR QF pF (x) ↓ ↓ ↑ N A (x ) QE xjE xC(VB'E') xE0 5-Feb-04 p (x ) pR (x) ↓ QR QB0 Base (24) Collector + xC (VB′C ′ ) QC xC0 xC(VB'C' ) xjC HO #10: ELEN 251 - SGP BJT Model x Saha #18 QF pF (x) ↓ p (x ) pR (x) ↓ ↓ ↑ N A (x ) QE xjE xC(VB'E') xE0 QR QB0 Base Collector Emitter Components of Base Charge, QB QC xC0 xC(VB'C' ) xjC x QB0 = charge in the neutral base under VB'E' = 0 = VB'C' . QE = increase in QB under V B'E' and is only a mathematical entity. QC = increase in Q B under V B'C' and is only a mathematical entity. QF = excess majority charge in the biased-device with VB'C' = 0. It is only a mathematical entity and important under high level injection. QR = excess majority charge in the biased-device with VB'E' = 0. It is only a mathematical entity and important under high level injection. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #19 Components of Normalized Base Charge, qb From Eq. (24), we get the normalized components of base charge: QB Q Q Q Q Q = E + B0 + C + F + R QB 0 QB 0 QB 0 QB 0 QB 0 QB 0 qb = qe + 1 + qc + q f + qr ∴ qb = 1 + qe + qc + q f + qr Where, qe ≡ QE QB 0 qc ≡ QC QB 0 qf ≡ QF QB 0 qr ≡ QR QB 0 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model (25) Saha #20 Evaluation of qb: Component qe We defined, QE = increase in the majority charge due to VB'E'. ∴ QE = VB′E ′ ∫C jE (V )dV (26) 0 1 and, qe = QB 0 VB′E′ ∫C jE (V )dV (27) 0 Assume, C jE = constant ≡ C jE = average value of C jE then, qe = C jE VB′E ′ QB 0 ≡ VB′E ′ VB where VB = inverse Early voltage defined as : QB 0 QB 0 VB ≡ = VB′E′ C jE 1 C ( V ) dV jE VB′E′ ∫0 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model (28) (29) Saha #21 Evaluation of qe • VB in Eq. (29) models base-width modulation due to the variation of E-B junction depletion layer under VB'E'. • VB due to VB'E' is the inverse of Early voltage due to VB'C'. • In Eq (29), VB = constant ⇒ CjE = constant independent of VB'E'. • Constant CjE is justified as: since Q E << QB0 ∴ qe << 1 and is not a dominant components of qb. • VB = constant may cause large error in qe, especially, @ VB'E' > 0. • The error in (29) due to qe for VB'E' > 0 can be eliminated by: – integrating CjE over the bias range – extracting VB from the slope of ln(IC) vs. qVB'E'/kT plot. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #22 Evaluation of qe • In order to determine the effect of qe accurately, we set: qc = qr = qf = 0 ⇒ qb = 1 + qe, • Then from (19) in the normal active region, we have: I SS qV B ′E ′ IC = e kT − 1 (1 + qe ) (30) • Thus, the slope of IC vs. VB'E' plot is given by: dI C qI kT C jE (VB′E ′ ) = C 1 − dV B′E′ kT q (1 + qe )QB 0 kT C jE (VB′E ′ ) kT 1 dI C d (ln (I C )) = 1 − = q I C dVB′E′ q (1 + qe )QB 0 qV d B′E ′ kT (31) 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #23 Evaluation of qe • From (31), the slope of ln(IC) vs. qVB'E'/kT plot is given by: 1 kT 1 dI C ≡ nE q I C dVB′E ′ ∴ nE = VB′C′=0 kT C jE (VB′E′ ) = 1 − q (1 + qe )QB 0 1 kT C jE (VB′E ′ ) 1 − q (1 + qe )QB 0 (32) • When CjE = constant ≡ <CjE>, then from (28), qe = VB'E'/B; and from (29), <CjE> = QB0/VB therefore, from (32): 1 nE ≅ kT 1 (33) 1 − q (VB + VB′E′ ) 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #24 Evaluation of qe: Summary • For most transistors: – qe = VB'E'/VB – VB = QB0/<CjE> is the inverse of Early voltage = constant. • VB = constant ⇒ CjE = constant independent of VB'E'. • Constant CjE is justified: since Q E << QB0 ∴ qe << 1 and is not a dominant component of qb • VB = constant may cause large error in qe, especially, @ VB'E' > 0. • The error due to qe for VB'E' > 0 can be eliminated by: – integrating CjE over the bias range – extracting VB from the slope of ln(IC) vs. qVB'E'/kT plot. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #25 Evaluation of qb: Component qc • qc models base-width modulation and therefore, Early voltage by changing depletion layer-width due to VB'C' at low current level. • Using the procedure used for qe, we can show: 1 qc = QB 0 V B′C′ ∫C jC (V ) dV (34) 0 Assume, C jC = constant ≡ C jC = average value of C jC then, qc = C jC VB′C ′ QB 0 ≡ VB′C ′ VA where VA = Early voltage defined by QB 0 QB 0 VA ≡ = V C jC 1 B′C′ C ( V ) dV jE VB′C′ ∫0 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model (35) (36) Saha #26 Evaluation of qc • VA in Eq. (36) models base-width modulation due to the variation of C-B junction depletion layer under VB'C'. • In Eq (36), VA = constant ⇒ CjC = constant independent of VB'C'. • Constant CjC is justified in the normal active region when B-E junction is reversed biased, that is, VC'B' < 1. • VA = constant may cause large error in qc, when B-C junction is forward biased, that is, the device is in: – inverse region – saturation region. • A more accurate expression for qc is required for accurate modeling of Early voltage in the: – inverse region – saturation region. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #27 Effect of qc on Ic • The effect of qc on BJT device characteristics in the normal active region is finite output conductance g0. • In order to determine g0, we set: qe = qr = qf = 0 ⇒ qb = 1 + qc, • Then neglecting bulk-ohmic resistors, we get: I SS qV BE IC = e kT − 1 (1 + qc ) qV BE I SS = e kT − 1 VBC 1 + VA • From (37), we can show: g0 = dI C I (0) ≅ C dVCE V = constant VA BE 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model (37) (38) Saha #28 Evaluation of qb: Component qf • qf can be considered as the normalized excess carrier density in the base with E-B bias VB'E' and models high level injection. • From charge neutrality: total excess majority carriers = total excess minority carriers. Therefore, for an npn transistor with |VB'E' | > 0; VB'C' = 0 ni2 QF = ∫ qA[ pF ( x) − N A ( x) ]dx = ∫ qA nF ( x ) − dx N A ( x) xE xE xC xC (39) • QF in (39) is shown as QB in EM2 model (HO #7) and is given by: QF = QBEM2 = τBICC (40) τ B I CC τ B I SS ∴q f = = QB 0 QB 0 qb 5-Feb-04 qVkTB′E ′ − 1 e HO #10: ELEN 251 - SGP BJT Model (41) Saha #29 Evaluation of qb: Component qr • qr can be considered as the normalized excess carrier density in the base with C-B bias VB'C' and models high level injection. • From charge neutrality: total excess majority carriers = total excess minority carriers. Therefore, for an npn transistor with |VB'C'| > 0; VB'E' = 0 ni2 QR = ∫ qA[ pR ( x ) − N A ( x )]dx = ∫ qAnR ( x ) − dx N A ( x) xE xE xC xC (42) • QR in (42) is shown as QBR in EM2 model and is given by: QR = QBREM2 = τBRIEC (43) τ BR I EC τ BR I SS ∴ qr = = QB 0 QB 0 q b 5-Feb-04 qVkTB ′C ′ − 1 e HO #10: ELEN 251 - SGP BJT Model (44) Saha #30 Solution for qb • We know: qb = 1 + qe + qc + q f + qr (25) • Then substituting (28), (35), (41), (44) in (25) we get: VB′E′ VB′C ′ τ Bdc qb = 1 + + + VB V A QB 0 q1 τ BRdc + QB 0 I SS qV B′E′ e kT − 1 qb I SS qV B′C′ e kT − 1 qb (45) q2/qb Where τBdc ≡ modified forward base transit time due to mobile charge in the depletion region (i.e., without depletion approximation). τBRdc ≡ modified reverse base transit time due to mobile charge in the depletion region (i.e., without depletion approximation). 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #31 Solution for qb We can simplify (45) to get: q2 qb = q1 + qb (46) Here VB′E ′ VB′C′ q1 ≡ 1 + + VB VA (47) qV B′C ′ qV B ′E ′ τ Bdc I SS e kT − 1 + τ BRdc I SS e kT − 1 q2 ≡ QB 0 (48) Where q1 in (47) models the base width modulation q2 in (48) models the effect of high level injection. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #32 Solution for qb • From (45) we get, qb2 − qbq1 − q2 = 0 (49) 2 q q ∴ qb = 1 + 1 + q2 2 2 (50) since qb > 0, (50) is obtained talking the positive solution only. • Equation (50): – offers solution for IC – defines injection level ♦ if q 2 << q12/4, q b = q1, then q f = qr = 0 ⇒ low level injection ♦ if q 2 >> q12/4 qb = (q1)1/2 ∴ 5-Feb-04 ⇒ high level injection HO #10: ELEN 251 - SGP BJT Model (51) (52) Saha #33 Solution for qb: High Level Injection For simplicity, we assume VB'C' = 0 (i.e., qr = 0). Then from (48) and (52), for high level injection in the forward active region: qb = q2 ≅ = τ Bdc I SS QB 0 τ Bdc I SS QB 0 e qVB′E′ kT e qVB′E′ 2kT (53) ∴ Substituting for qb in (19), we get for high level injection @ VB'C' = 0 and V B'E' >> kT/q: qV B′E ′ I SS e kT − 1 B′E′ QB 0 I SS qV2kT (54) I C ≅ I CT = ≅ e qVB′E ′ τ Bdc τ Bdc I SS 2 kT e QB 0 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #34 High Level Injection: Knee Current, IK • For low level injection, if we assume, qe = qc = 0, then qb = 0: • From (19), we get for low level injection @ V B'C' = 0 and VB'E' >> kT/q: qV B′E′ (55) I C ( low−level ) ≅ I SS e kT • The intersection of high current and low current asymptote is given by (IK,VK) where IC(high-level) = IC(low-level). • Therefore, from (54) and (55): QB 0 I SS IK = τ Bdc I K = I SS e K qV e 2 kT qVK kT 5-Feb-04 (56) (57) HO #10: ELEN 251 - SGP BJT Model Saha #35 High Level Injection: Knee Current, IK • From (56) and (57), we get: QB 0 IK = τ Bdc ln (I K ) • Similarly, for inverse region: I KR QB 0 = τ BRdc slope = 1/2 ln(IC) (58) (59) • Model parameters: (VA, VB, ISS, IK, IKR) are extracted from – ln(IC) vs. V B'E' plot – ln(IE) vs. VB'C' plot. 5-Feb-04 slope = 1 VBC = 0 qVK/kT HO #10: ELEN 251 - SGP BJT Model qVB'E'/kT Saha #36 SGP Model: Summary I CT I = SS qb qV B′E′ qV B′C′ e kT − 1 − e kT − 1 where (19) 2 q1 q1 ∴ qb = + + q2 2 2 V V q1 ≡1 + B′E′ + B′C ′ VB VA (50) (47) From (58) and (59) q2 ≡ τ Bdc qV B′E′ τ BRdc qV B ′C ′ I SS e kT − 1 + I SS e kT − 1 QB 0 QB 0 I SS qV B′E′ I SS qV B′C ′ = e kT − 1 + e kT − 1 I KR IK 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model (60) Saha #37 SGP Model: Summary • q1 models base-width modulation • q2 models high level injection – the condition for high level injection: 2 q1 q2 ≥ 4 • The model parameters: (VA, VB, ISS, IK, IKR) are extracted from – ln(IC) vs. V B'E' plot in the normal mode of BJT operation – ln(IE) vs. VB'C' plot in the inverse mode of BJT operation – IC vs. VCE characteristics in the normal mode of BJT operation – IE vs. V EC characteristics in the inverse mode of BJT operation. • A parameter, B is used to model base widening (base push-out) effect at high currents. 5-Feb-04 HO #10: ELEN 251 - SGP BJT Model Saha #38