IGBT等效電路模型IGBT Equivalent Circuit Model

全文

(1)

(2) . . IGBT An Equivalent Circuit Model for IGBT. .

(3)

(4) .

(5) 1.

(6) be reverse-biased and the p+-anode/n-drift junction will be forward-biased. Now if a positive gate bias of sufficient magnitude (above the threshold voltage) is applied with positive anode voltage, the surface of p+-cathode region under the MOS gate will invert and then the device conducts. Electrons can now flow from the MOS source through the MOS channel and are injected into the BJT finally. When the device turns on, the resistance of n-drift region is very low and the voltage drop is almost on the depletion region of n-drift/p+-cathode junction. The higher the anode voltage is applied, the larger the depletion region and the voltage across it become. Here we need to concern some phenomenon. The first is the different values of the MOS current gain lin, sat in linear and saturation region respectively. This results from the nonuniform doping of the MOS channel. So the MOS saturates at the condition:. Abstrat- A new circuit model for the IGBT is developed. This model uses the transmission-line circuit technique and is suitable for circuit simulators. The new circuit model for the IGBT is implemented into widely used SPICE3 and all the simulation results will be compared with the results of the TMA MEDICI device simulator. Introduction Up to present, many IGBT models have been proposed. Some of them are behavior models and some of them are physical models. The behavior model [1] may just describe device characteristics exactly in some particular conditions. It can not predict the device characteristics when there is a change in the physical parameters. The analytical models [2]-[5] can predict the IGBT behavior more precisely as the device structure and parameter are changed but they still have problems in the combination of the phenomenon of high-level injection, low-level injection and the wide base effect in the IGBT. In this paper, a new physical based circuit model of the IGBT has been developed. We apply the concept of the transmission-line method [6] to the proposed IGBT equivalent circuit model. All effects are physically included in the new model.. V DS = ( V GS − V T )( 1 −. 1 − β. sat. β. lin. ). (1). When the MOS saturates, the device also saturates. The second phenomenon is the change value of the MOS gate-drain capacitance Cgd. As the gate voltage is smaller than the voltage of the drain region, the region beneath the gate-drain overlap is depleted but for the condition the gate voltage is larger than the drain voltage, the overlap region becomes accumulated. The change of depletion region is shown in Fig.3. Besides, there are also some nonideal effects exist which is due to the structure in the device like the JFET effect and the latch-up effect. In order to reduce JFET effect, a high doping n-type region is placed below the MOS gate just as the. Device Physics A schematic of an IGBT unit cell is shown in Fig.1 and the simplified equivalent circuit is in Fig.2. When a positive voltage is applied to the anode, the n-drift/p+-cathode junction will 2.

(7) region surrounded by the dash line in Fig.1. In this paper, all nonideal effects are neglected.. ∂ J ∂x. ∂ J ∂ x. n. − q. ∂ n ∂ t. =. qR. D. − N. A. ). +. q. ∂ p ∂ t. =. (4). − qR. The conventional drift-diffusion current model is applied to the current equations:. Proposed IGBT Model It is very complex to simulate the IGBT with two-dimensional simulation. Thus we use one-dimensional approximation to simplify the simulation. When the device is switched on, the resistance of the n-drift region is very low. Although the electron flux from the MOS spreads across the n-drift region, the n-drift region voltage drop due to the current is quite low. Therefore the anode voltage distribution from p+-anode to the edge of depletion region of n-drift region is horizontally identical. The voltage distribution is no more symmetrical at the distance d from the n-drift/p+-cathode junction. If the high n+ doping region is placed at y=0 to y=20 um, the distance d is 10 um. Therefore we assume the MOS electron current injected to the n-drift region at a certain distance d and the n+ region (within dash line in Fig.4) can be replaced as a resistor Rmos In conclusion, the IGBT device structure can be approximated by the equivalent structure shown in Fig.5 and then the BJT can be replaced by one-dimensional equivalent circuit using transmission-line method. The electrical properties in a semiconductor can be controlled by the Poisson equation, electron continuity equation and hole continuity equation: ∂ ∂ q ⋅( V) = − (p − n + N ∂x ∂x ε. p. J J. n. p. = − q ( µ nn = −q (µ. p. p. ∂ ∂ V − Dn n) ∂x ∂x ∂ V + D ∂x. p. (5) (6). ∂ p) ∂x. For one-dimensional device simulation, the device can be partitioned into many segments with area A . In the following, we will present the equivalent circuits for the Poisson equation and continuity equation. We assume identical division for each segment:i.e hi=hi-1=hi+1. A. The Poisson Equation The difference equation of the Poisson equation: CV. where I fp, i =. ∂ ∂ (V − Vi ) − CV (Vi − Vi −1 ) = − I fpi − I fni ∂t ∂t i + 1. Qp,i =qAh⋅ pi. ∂ Q ∂ t p, i. and. (7). and. Qn,i =−qAh⋅ ni . C V = Aε. I fn ,i =. ∂ Q ∂t n, i. h. ,. .This equation. can be represented by the equivalent circuit as shown in Fig.6. But the equivalent circuit in Fig.6 can’t perform dc analysis in the circuit simulator. Therefore we develop the second equivalent circuit of the Poisson equation. V i+1 − V i V i − V i −1 − = − I v ,i Rv Rv. where. Rv =. h Aε. (8) and. I v, i = qAh( pi − n i − N A + N D ) . Here, we treat. the physical charge as the pseudo electrical current in equation (8). The equivalent circuit of equation (8) is shown in Fig.7.. (2). (3). B. The Continuity Equations The difference equation of electron 3.

(8) continuity equation can be expressed as: I. n ,i +. 1 2. − I. n ,i−. + C. 1 2. where I. fn ,i. n. ∂ V n,i ∂t. = I. voltage source. Because the MOS saturates (9). R ,i. Cn ⋅ Vn,i = Qn,i. = Cn. ∂ V n, i ∂ Q n, i = ∂t ∂t. The values of. C. n. ,. Vn. ∗. Cn ⋅Vn = Qn (0) = −NDqAh. and ∗. Cn =−. N. D qAh V n∗. at. Vn ∗. V0 ,. Vm ,. offset voltage χ is the potential of n-drift region when there is no current in the IGBT. Variable capacitor Cgd is also added in the circuit. Because the parameter CGS0 in MOS model must be constant, we place the capacitor outside the MOS. The circuit technique is shown below: Then we obtain the expression:. is. C gd = (1 + k ) C gd 0. (10) If k is a function of Vgd, it would control Cgd. Simulation Results The proposed IGBT model is implemented into the circuit simulator: SPICE3, and is evaluated using different test circuits to examine the device behavior for the static and dynamic ranges in which the device operates. In order to verify the proposed circuit model, the simulation results are compared with the TMA MEDICI device simulator results.. V m− 1. respectively. In the circuit, Emos is a dependent voltage source that depends on. A. Static Simulation Fig.13 compares the gate voltage static I-V characteristic. The gate voltage offset due to threshold voltage is achieved by. the voltage VK that is away from the base/collector junction at the distance d . Imos is the current flows through Emos. Vbi1 ,. assigning the parameter of VT in the MOS. Vbi2 , Vbi3 , VPI , VPE , VNI and VNE are. model of the SPICE program. Fig.14 compares the low anode voltage static I-V characteristics. There is no current as the anode voltage is smaller than 0.5V because. boundary conditions. E MOS. ∗. the. E MOS ∗ to do behavior approximation. The. physical charge so the possion equation circuit representation can’t be connected to external circuit directly. For this reason, we isolate the external circuit and the second type possion equation circuit representation. We use the first type possion circuit representation in the first and the last segment of Fig.10. E1, E2, E3, E4 represent V1 ,. and. .. Fig.10. The element I v in the circuit is the. voltage. βlin). parameter β in usual circuit simulator can not be described as a function, we use. normalized to 1. The circuit representation of the electron continuity equation is shown in Fig.8. Similarly the circuit representation of the hole continuity equation is shown in Fig.9. Since the circuit representations of each equation have been presented, the circuit representation of an IGBT is presented in. the. βsat. ,. are selected to fit and. VDS =(VGS −VT )(1− 1−. in the circuit is a dependent 4.

(9) A new circuit model for the IGBT is developed. It has three main advantages. One is that physical effects like high-level injection, wide base BJT and early effect are physically included in the new model. The second is we can assign device parameters such as doping profile, etc to the model. The third is the model can be easily connected to external circuit. The new model has been used to study the static and dynamic characteristics of an IGBT under different conditions. The results match with the MEDICI results accurately.. the BJT is off at this time. The simulation results show the implementation of EMOS* is successful. In the saturation region, the anode current rises slowly due to the early effect of the BJT. The early effect is included in the proposed model. B. Dynamic Simulation Fig.15, Fig.17 show two basic test circuits used for dynamic simulation and the list of anode supply, gate voltage pulse amplitude, resistor and inductor. 1. Resistor Load Switching. REFERENCE. Consider the turn-off current transient in Fig.16. The current falls rapidly at the turn off and then decay slowly due to the recombination of excess carriers in the n-drift region of the device.. [1]. J.T. Hsu and Khai D.T. Ngo, “Behavior Modeling. of. the. IGBT. Using. the. Hammerstein Configuration”, IEEE Trans. on Power Electronics, vol.11, no.6 ,1996.. 2. Series Resistor-Inductor Load Switching Consider the transient responses of anode voltage and anode current shown in Fig.18 and Fig.19. The anode current 10 A is determined by external circuit elements because the IGBT device has low voltage drop as in the on-state. After VIN turns off at 35us, the anode voltage and the anode current maintain their original value for a period time until the gate voltage is too low to conduct enough current to maintain low voltage drop across the device. The time when anode voltage begins to rise or anode current begins to fall fast is approximately the time when Cgd changes to small value. At last, the overshoot of anode voltage during turn off results from the rising voltage drop across the inductor LL caused by fast anode current fall.. [2]. B.J. Baliga, “Analysis of Insulated Gate Transistor Turn-Off Char acte ristics”, IEEE Electron Devic e Letters, vol.EDL-6, no.2, 1985.. [3]. A.R.. HEFNER. and. DAVID. L.. BLACKBURN, “An Analytical Model for the Steady-State and Transient Characteristics Of the Power Insulated-Gate Bipolar Transistor”, Solid-State. Electronics,. vol.31,. no.10,. pp.1513-1532, 1988. [4]. D.S. Kuo and Chenming Hu, ”An Analytical Model for the Power Bipolar-Mos Transistor”, Solid-State. Electronics,. vol.29,. no.12,. pp.1229-1237, 1986. [5]. Y.Yue and J.J Liou, ”An Analytical Insulated Gate Bipolar Transistor (IGBT) Model for Steady-State. and. Transient. Applications. under All Free-Ca rrie r Injection Conditions”, Solid-State. Electronics,. vol.39,. no.9,. pp.1277-1282, 1996. [6]. Conclusion. C.T. Sah, “The Equivalent Circuit Model In Solid-State. Electronics”,. Solid-State. Electronics, vol.13, pp.1547-1575, 1970. 5.

(10) one-dimensional approximation x (um) y=0. 10 um n+. y=20. (NA =1019). d. (ND=1017). 80 um. (ND=1014). Fig.5 The representation of the IGBT after one-dimensional approximation Vi− 1. (NA =1018). y=100. Vi. CV. CV Vi+ 1. 10 um. y (um). I. I. fp , i. fn , i. Fig.1 The cross section of an IGBT unit cell Fig .6 The first type circuit representation of the Poisson equation. CatCathode Gate. Vi −1. Vi. Rv. R v V +1. Iv,i. Anode. Fig.2 The simplified equivalent circuit MOS channel. Fig.7 The second type circuit representation of the Poisson equation. MOS channel. Dedepletion region accumulation. Dedepletion region. Vn,i −1. depletion. n-drift region. n-drift region. I. n,i−. 1. I fn,i. Fig.3 (a) Fig.3 (b) Fig.3 Figure (a) is at the condition VGVD and figure (b) is at the condition VGVD. I. Vn,i. 2. n,i+. 1. Vn,i +1. 2. IR,i. Cn. Fig.8 The circuit representation of the electron continuity equation. Vp,i −1. d R m os. I fp,i. I. p,i−. 1 2. Vp,i. Cp. I. p,i+. 1 2. Vp,i +1. IR,i. Fig.9 The circuit representation of the hole. Fig.4 The figure to explain the 6.

(11) continuity equation C gd. RMOS I MOS V n,0. V n,1. Vn,2. In,1. Vn,K. In,2. x * VD. EMOS. V n,K+1. V n,m-2. I n,K+1. V n, m-1 In,m- 1. VG x. x VD E*MOS. GATE. x VS. Vn,m In,m. NI. 2. NE. Cn,1. Cn,2. IR,0 V p,0. IR,1. V p,1. C n,K+1. IR,2. I R,K. Vp,2. Ip,1. V p,K. Ip,2 Cp,1. NI. Cn,K. Cn,m -2. Cn,m- 1. IR,K+1 V p,K+1. V p,m-2. I p,i+1 Cp,2. Cp,K. IR,m-2. IR,m -1. V p,m-1 Ip,m- 1. C p,K+1. Cp,m -2. I R,m. Vp,m Ip,m. Cp,m- 1. bi 1. V 0 C V,1. Ifp,0. E1. I fn,0. E. (VGS − VT ) ⋅ 1 −. βsat. Rv,1. MOS. ∗. E2. V1. Iv,1. Rv,2. V2. VK. Iv,2. Iv,K. R v,K+1. V K+1. I v,K+ 1. V m-2. Iv,m- 2. Rv,m-1 V m-1. Rv,m. C V,m V m E3. Iv,m-1. E4 I fp,m. CAT HODE. If n,m. Fig.10 The equivalent circuit representation for an IGBT Fig.13 The static characteristic of the gate voltage vs. the anode current . βlin 0. . . V. χ. (VGS. Fig.11. D. ∗. −V. S. ∗. E MOS. . .

(12) . . − V T )(1 − 1 − β sat β ) lin.

(13) .

(14) . ANO DE. function. . C gd0. Iref. . C gd.

(15) . . . . . . .

(16)

(17) . . VIN(vol). . . . . . . . . . . . . . . . . Fig.14 The static characteristic of the anode voltage vs. the anode current. . . .

(18) . Fig .12 The circuit technique for Cgd. . . 10. .

(19) . 0. 7. 10. Time(us).

(20) conditions: Rg=2k,3k,4k Fig.15 The input voltage waveform and test circuit with resistor load. . . . .

(21) . . .

(22) . . element value VDD 10V Rg 100 RL 2 . .

(23). . Simulation M E DICI. . element value VDD 300V RL 30 LL 80H. . . . . . . . . . . . . . . . . . . . . .

(24) . . VI N ( vo l) 20 20. 5. Ti me ( us) Ti me ( us). 35. Fig.17 The input voltage waveform and test circuit with series resistor-inductor load. .

(25) .

(26) . . . . . element value VDD 300V RL 30 LL 80H. . . . . . . . . . . . . Fig.19 Anode voltage waveforms for three conditions: Rg=2k,3k,4k. VI N ( vo l). 35. .

(27) . Fig.16 The turn-off current waveforms for the resistor load. 5. . .

(28) . Fig.18 Anode current waveforms for three 8. .

(29) 9.

(30)