Parameter Extraction for a Physics-Based Circuit ... - Semantic Scholar
Parameter Extraction for a Physics-Based Circuit Simulator IGBT Model X. Kang, E. Santi, J.L. Hudgins, P.R. Palmer* and J.F. Donlon** Department of Electrical Engineering University of South Carolina Columbia, SC 29208, USA [email protected]
*Department of Engineering University of Cambridge Trumpington Street Cambridge CB2 1PZ, UK
Abstract- A practical parameter extraction method is presented for the Fourier-based-solution physics-based IGBT model. In the extraction procedure, only one simple clamped inductive load test is needed for the extraction of the eleven and thirteen parameters required for the NPT and PT IGBT models, respectively. Validation with experimental results from various structure IGBTs demonstrates the accuracy of the proposed IGBT model and the robustness of the parameter extraction method. * Keywords—power semiconductor modeling, IGBT, Model, Parameter extraction
I. INTRODUCTION In recent years, the characteristics of IGBTs have been improved greatly. For example, with the application of advanced lifetime killing technology and structure development [1], the 4th generation IGBTs (trench-gate punch-through), which have been available in the market for some time, exhibit fast turn-off speed and low saturation voltage. The application of light punch through technology [2] in the CSTBT (Carrier Stored Trench Bipolar Transistor) IGBT now offers the new 5th generation IGBT with superior characteristics. Field-Stop IGBTs [10] also exhibit tail-free turn-off current under high voltage conditions. These external characteristics, brought about by internal device design changes, can only be accurately modeled by a physicsbased (analytical) IGBT model. However, in spite of the accuracy of analytical IGBT models, either overly complex or inaccurate parameterization often discourages electronic engineers from attempting to use IGBT models in their system designs. Reviewing the IGBT circuit model parameter extraction methods described in the literature shows that further work must be done in the field. The parameter extraction provided for the accurate and comprehensive Hefner model [3] is so complex that it is not practical for electrical engineers. The extraction method proposed in [4] is only for the lumpedcharge IGBT model. In [5], the parameter extraction for Hefner model was further developed with seven very precise but complex experiments. The parameters extracted in [6] *
This work was supported by the U.S. Office of Naval Research under Grant N00014-00-1-0131.
**Powerex, Inc 200 Hillis Street Youngwood, PA 15697, USA [email protected]
and [7] are used for the IGBT behavioral models. In [8], the parameter extraction lacks detail. In this work, we will provide a practical parameter extraction methodology to extract eleven and thirteen parameters for the non-punch-through (NPT) and punchthrough (PT) physics-based IGBT model presented in previous work [9]-[11]. In the parameter extraction procedure, as in the model itself, the better trade-off between accuracy and simplicity is pursued. Therefore, without affecting the simulation accuracy, only one simple clamped inductive load switching experiment is needed to extract several key parameters, while most parameters are extracted directly from the manufacturer’s datasheet, or estimated according to empirical value ranges and relations with already known parameter values. With parameter values obtained through the extraction procedure mentioned above, the PSpice IGBT model accurately simulated the hard-switching behavior of IGBTs having various structures. Comparison between the simulated and experimental results validates the proposed circuit model and the parameter extraction method. II. IGBT PHYSICS-BASED CIRCUIT SIMULATOR IGBT MODEL A. Categorization of Analytical IGBT Models Being a conductivity-modulated power device, the behavior of an IGBT depends heavily on the carrier distribution in its wide drift region. Under high-level injection conditions, the Ambipolar carrier Diffusion Equation (ADE) (1) describes the carrier dynamics in the majority of this region,
D
∂ 2 p ( x , t ) p ( x , t ) ∂p ( x , t ) = + τ ∂t ∂x 2
(1)
where D is the ambipolar diffusion coefficient, τ is the highlevel carrier lifetime within the drift region and p(x,t) is the excess carrier concentration. Therefore, most physics-based IGBT circuit simulator modeling approaches focus on the simulation of the drift region, and consequently implement the modeling of the ADE. However, it would be difficult and would also lead to potential convergence problems to directly implement the 2nd order partial differential equation in the circuit simulator
without any simplification. Consequently, various mathematical simplifications have been proposed for the implementation of analytical IGBT models. Depending on the type of mathematical simplification technique used to solve the ADE, analytical models can be categorized into four groups, as listed in Table I for example models, in which the right column shows the simplified equations representing the mathematical method used in each modeling approach. Table I. Physics-Base IGBT Models IGBT Model
Simplification
Hefner Model (Hefner) [3]
dp ( x , t ) = dt
Fourier Based Solution Model (Leturcq-Palmer) [9]
p(x, t) = p0 (x, t) + ∑ gi (t) * fi (x)
Lumped Charge Model (Lauritzen-Ma) [4]
qA
Laplace Transformation Model (Strollo) [10]
D
∑ h (t ) * f ( x ) i
i
dp ∆Q → dx ∆x
d 2 p' = p ' (s + 1 / τ ) dx 2
p ' = p ( x , t ) − p ( x ,0 )
(2) (3)
(4)
Further mathematical simplification allows the n-drift region to be modeled as an RC equivalent circuit as shown in Fig.1. The details are given in [13]. The representation requires the width of the undepleted base region and the hole and electron currents at the boundaries of the region (x1 and x2), which corresponds to the gradients of the carrier concentrations, f(t) and g(t), at x1 and x2, respectively. The functions f(t) and g(t) are defined as follows: 1 I n1 I p1 ∂p ( x, t ) (6) = − f (t ) = ∂t x1 2 qA Dn D p
1 I n2 I p2 ∂p( x, t ) g (t ) = = (7) − ∂t x2 2qA Dn Dp where A is the active cross-sectional area of the device, Dn and Dp, the electron and hole diffusion coefficients, In1 and Ip1, the electron and hole currents at x = x1 (p+ side), and In2 and Ip2 the electron and hole currents at x = x2 (p-body side). Clearly, the success of the approach now depends solely upon developing the appropriate boundary conditions.
(5)
The Fourier-based-solution (FBS) model is adopted in this research. By modeling the non-quasi-static characteristic of the carrier distribution, this approach preserves the essentials of the distributed nature of charge dynamics within the drift region and offers reasonable simulation speed while preserving satisfactory accuracy. B. Fourier-Based-Solution IGBT Model By applying a Fourier transformation, the ADE is simplified to a set of first order discrete linear equations, as shown in equation (3).
Fig.2 Definition of variables and typical carrier distribution under forward conduction of the PT IGBT
Therefore, in order to establish the boundary conditions, hole and electron currents at the edges of the drift region are needed. Fig. 2 defines some of relevant variables used in the analysis for PT IGBT (no n+-buffer layer for the NPT IGBT). Since by current continuity
I A = I n1 + I p1 = I n 2 + I p 2
Fig.1 Equivalent circuit diagram modeling the storage charge of the n drift region
(8)
the sum (IA) of the hole (Ip1 & Ip2) and electron currents (In1 & In2) is known, it is sufficient to find one current component at each edge. At the right side of the drift region, WB, (the MOS-channel edge), the electron current In2 is the channel current and it can be found as a function of the gate-source voltage using the well-known MOSFET equation. Current Ip2 can then be calculated using (8). At the left side of the drift region, different equations are used for the NPT and PT IGBT due to the structural differences. The details of the respective boundary conditions are given in [10] and [11].
For the NPT IGBT, equation (9) is used to calculate the electron current element in the junction between the p+ layer and the n-drift region. I n1 = qAh p p L2 0
(9)
where hp is the hole recombination parameter at the p+ layer [14]. For the PT IGBT, equation (10) is used to obtain the hole current component at the junction between the n buffer layer and the n drift region. qADpH W I p1 = [PH 0 − PHW cosh( H )] + I QH (10) WH LpH LpH sinh( ) L pH where LPH and DPH are the diffusion length and mobility of the hole carrier in the buffer layer, respectively, and other variables are defined in Fig.2. Finally, substituting the results of (9) and (10) into (8) gives the boundary conditions for each case. III. PARAMETER EXTRACTION The Fourier-based-solution IGBT model requires eleven and thirteen non-silicon parameters for the NPT and PT structure IGBTs respectively. Table II lists the parameters, which are classified into three categories based on the related device part: MOS-gate, geometry, and collector body. The parameter extraction procedure presented below is also divided into three parts according to this division.
Generally, there are three kinds of parameter extraction methodologies, which include: 1) Simple estimation based on the empirical value range, 2) Extrapolation according to the manufacturer’s datasheet, and 3) Extraction with simple experiments. To make the extraction procedure practical, the three methods are often employed jointly. Most parameters listed in Table II can be obtained from the datasheet or calculated based on textbook equations, so there is no need to use measurement for their extraction. Extra measurement can be used to verify the mathematical results. But some parameters, like carrier lifetime, have to be determined by experiment, which also need to be simple. Accordingly, the clamped inductive load test experiment is used. A. Extraction of MOS-gate parameters Because a behavioral MOSFET model is used for the IGBT MOS gate, the relevant parameters can be extracted following the extraction procedure for a MOSFET. The information given in the manufacturers’ datasheets is enough to extract these five parameters. The first three parameters: MOS threshold voltage (Vth), Transconductance Coefficient (Kpl), and short channel parameter (λ), can be obtained from the known I-V characteristic curve, shown in Fig. 3.
Table II. IGBT Model Parameter List Part
Symbol Vth (V)
MOS threshold voltage
Kpl (A/V )
MOS transconductance coefficient
λ
Short-channel parameter
Fig. 3 Forward I-V characteristics of the IGBT
Cge (nF)
Gate-emitter capacitance
Cox (nF)
Oxide capacitance
A (cm2)
Effective die area
a_i
Ratio of Inter-cell to total die area
WB (cm)
N drift region width
Gate-emitter capacitance Cge can be directly obtained from the input capacitance Cies (measured gate-emitter capacitance when collector is shorted to emitter) provided in the datasheet. Since the input capacitance Cies is the sum of the Cge and the Miller capacitance, and the former is much larger than the latter at higher voltages, Cge can be chosen equal to Cies at 10V.
NB (cm-3)
Doping concentration of N drift region
τHL (µs)
Carrier high-level lifetime in N drift
hp
Hole recombination coefficient in emitter (NPT)
NH (cm-3)
Doping concentration of N buffer Layer (PT)
WH (cm)
Width of n+ buffer layer (PT)
τBF (µs)
Carrier lifetime in N buffer layer (PT)
2
MOS Gate
Geom.
Collector Body
Description
Fig. 4 IGBT capacitance diagram
As shown in Fig.4, the MOS oxide capacitance Cox, together with the series-connected depletion capacitance Cdep, determines the gate-collector capacitance Cgc, which is also called the reverse transfer capacitance Cres in datasheets and is commonly referred to as the Miller capacitance between gate and collector. Obviously, Cox is the maximum value of the Miller capacitance when the depletion region under the gate area has not formed (Cdep = 0). That point corresponds to the maximum point of the Cres curve in the datasheet. Therefore, Cox can be written as:
C ox = max( C res )
(11)
B. Extraction of geometry parameters Instead of employing the destructive visual measurement, which is commonly used, the active die area A can be extracted based on the empirical range of IGBT maximum current density J. Therefore, the active die area can be roughly estimated as: A =
I
CM
J
(12)
where ICM is the peak collector current from the RBSOA curve given in the datasheet. Practically, the ICM value is decided by a multitude of factors, not only the chip but also the package and heat extraction. Nevertheless, the range of current density J generally is 100 A/cm2-150 A/cm2.
In this case, we can assume that Cce is proportional to the area under the emitter terminal while Cgc is proportional to the area under the gate terminal. The constants of proportionality are assumed to be the same since the depletion layer is quite far out from the gate and emitter terminals, and has, by this time, merged into one simple 1-D depletion layer. Thus:
a _i =
C gc Ai ≈ A C gc + C ce
(13)
Data sheets provide the reverse capacitance, Cres, and the output capacitance, Coes, which is the measured capacitance between collector and emitter when the gate is shorted, and can be written as:
C oes = ( C gc + C ce )
(14)
Vge = 0
As discussed above, we have
C res = C gc
(15)
Therefore, the intercell ratio can be obtained as
a_i ≈
C res C oes
(16)
Eq. (16) is only valid under the condition that the collector-emitter voltage Vce is sufficiently high to allow for an approximately uniform space-charge layer capacitance over the whole crystal area. In this case, the minimum values of Coes and Cres in the datasheet should be chosen, as sufficiently high Vce is reached at that point. C. Extraction of BJT body parameters There are several different ways to extract the carrier doping concentration of the drift region NB. The first is the direct estimation based on the normal range of the drift region background doping, which is from 6 × 1013 to 2 × 1014 cm-3 for the IGBT device. The typical value, 1 × 1014 cm-3, is generally chosen as the doping concentration NB during the simulation. The second way is based on the relation between the doping concentration and the breakdown voltage [4] using equation (17).
Fig. 5 Capacitance distribution diagram for a trench gate IGBT
The ratio of inter-cell area to total die area a_i can be extracted based on the capacitances in the datasheet. From above figure of the capacitance distribution, it can be seen that the displacement current due to the variation of the depletion region flows through two branches: one is the collector-emitter capacitance, Cce, branch under the emitter terminal, while the other is the Miller capacitance branch, which includes the collector-gate capacitance, Cgc. When the gate is shorted to the emitter (Vge = 0), the latter branch only includes Cgc.
N
B
=
3
60 ( Eg ) 1 .1 V BR
3 / 2
4
10
16
(17)
where VBR is the breakdown voltage value from the manufacturer’s data sheet plus about 150V-200V typical margin for general IGBTs, and Eg is the energy band gap value of silicon. Besides the above simple but rough extraction method, the extrapolation based on the Coes and Cres versus collectoremitter voltage curves in the datasheet can also lead to a reasonable extraction value of NB. Assuming breakdown in the bulk, an abrupt junction, and no mobile carriers in the depletion region (no leakage current), the depletion layer width is given as:
2 ε V CE
x dep =
qN
(18) B
where x dep is the depletion width and Vce is the collectoremitter voltage. From the capacitance analysis given above, the collectoremitter capacitance Cce is due to the depletion region under the emitter region: ε A (1 − a _ i ) (19) C = ce
x dep
Substituting (18) into (19) leads to a linear relationship 1 between and Vce. ( C ce ) 2 1 ( C ce ) 2
=
2 V ε ( A (1 − a _ i )) 2 qN B ce
(20)
where ic (t ) is the collector current, I c is the DC collector current, and I k is the current value at the starting point of the current tail. The high level lifetime τ HL can be obtained from the curve of the decay rate − d ln( i c ( t )) versus the collector dt
Furthermore, substituting the relation Cce = Coes − Cres into (20) leads to
1 2 Vce = (Coes − Cres ) 2 ε ( A(1 − a _ i )) 2 qN B
occurs under a constant collector-emitter voltage after the collector-emitter voltage reaches the clamp voltage. Therefore, the two-stage turn-off characteristic makes it easy to identify the current tail interval from the current and voltage waveforms. Based on the theory in [3], the decay time constant of the IGBT current tail portion for a constant collector voltage is given by: d ln(ic (t )) I 1 (24) =− (1 + c ) dt τ HL Ik
(21)
current. Fig.6 shows the measured results of a Dynex NPT IGBT with a 90A current rating. Notice that the effective lifetime is approximately independent of the collector current, therefore the collector current level used to extract the highlevel lifetime τ HL is unimportant.
Eq. (21) reminds us that the capacitance versus voltage curves provided by manufacturers can be used to extract the doping concentration in the drift region NB. With a linear regression applied to the curve for 1 /(Coss − C rss ) 2 versus Vce
Vcslope = 2 /(qεN B (1 − a _ i) 2 A2 )
(22)
Therefore, NB can be obtained from the slope value equation (22). The extraction for the remaining parameters in this section is different for the NPT and PT IGBT cases.
WB.
2ε
14 12 10 8 6 4 2 Collector Current Ic (A) 0
Under the triangular shape of the electrical field distribution in the NPT IGBT, the standard breakdown voltage equation gives: V BR = qN B (W B )
16
0
a) Parameter Extraction for the NPT IGBT
2
18
Effective lifetime (us)
[13], we can get the slope of this curve, which we will call Vcslope.
20
(23)
Substituting NB into (23) will give the drift region width
Since the carrier high-level lifetime dominates the IGBT 2nd stage turn-off current, the turn-off tail current decay will be used to extract the high-level lifetime τ HL . However, the current tail of latest-generation IGBTs is becoming less and less obvious than previous generation IGBTs, due to the application of advanced lifetime killing and other semiconductor technologies. Compared with other test circuit setups, the clamped inductive load test circuit is more suitable for lifetime extraction, especially for the current IGBTs with smaller lifetime. During IGBT turn-off in the clamped inductive load circuit, the IGBT current tail
10
20
30
40
50
60
70
80
90
100
Fig. 6. High-level lifetime extraction for the NPT IGBT (The x-axis is collector current, 10A/div, and the y-axis is time, 2µs/div)
The classic value of the recombination coefficient hp [14] is given by: hp =
1 NA
Dn
τn
(25)
where N A is the acceptor carrier concentration of the high doped p+ collector region, and τ n is the minority carrier
lifetime. Generally, the recombination coefficient hp can be selected based on the empirical value 1~2 *10-14 cm4 s-1 for abrupt junction. b) Parameter Extraction for the PT IGBT Like the WB extraction in the NPT case, using NB previously known into (26), the equation of the trapezoidal
electric field at breakdown for the PT structure, will give the extraction value of the drift region WB. 1 qN (26) VBR = Ec.WB − . B .WB2 2 ε where Ec is the critical electrical field value for the silicon.
For the lifetime extraction of the PT IGBT, a method similar to the NPT case can be used, which is based on versus the collector current in the clamped d ln( I c ( t )) − dt
inductive load circuit. The difference is that the effective lifetime τ eff , instead of the high-level lifetime τ HL in the NPT case, is used in (24) for the PT case. The effective lifetime is function of the high-level lifetime in the drift region and low-level lifetime in the buffer layer τ BF . Moreover, τ eff is dependent on the clamp voltage, unlike
τ HL in the NPT case, which is independent of the voltage. Therefore, lifetime extraction for the PT IGBT needs to be performed under several clamp voltages. The effective lifetime extraction under clamp voltage condition is the same as in the NPT case. Fig. 7 shows the curve of the effective lifetime τ eff versus the clamp voltage based on the test results for a PT IGBT. Notice that Fig. 6 and Fig. 7 have different xaxes: in Fig. 7 the x-axis is the collector-emitter voltage, whereas in Fig.6 the x-axis is the collector current. The highlevel carrier lifetime τ HL corresponds to the low-voltage τ eff value, while the low-level carrier lifetime in the buffer layer τ BF is equal to the τ eff value at high clamp voltage since the drift region is depleted under that condition. 0.3
τ HL
Effective Lifetime (us)
0.25
0.2
0.15
value range. The typical PT IGBT buffer layer width W H is about 4~10 um. The normal range of the doping concentration N H is 1016~1017 cm-3. For the 5th generation CSTBT IGBT and the Field-stop IGBT, smaller W H and N H values should be chosen. IV. VALIDATION OF THE IGBT MODEL AND PARAMETER EXTRACTION
In order to validate the analytical IGBT model and its parameter extraction method discussed above, the IGBT model with extracted parameters was used to simulate the switching behavior of various structure IGBTs from different manufacturers. They are listed in Table III. The validation experiment with the 5th generation CSTBT IGBT is still in process and will be included in the presentation at the conference. Table III. IGBT Device Lists IGBT
Structure
Rating
Manufacturer
A
Trench PT
600V/600A
Powerex
B
DMOS PT
1200V/600A
Powerex
C
DMOS NPT
1200V/100A
Dynex
D
Trench Field Stop
1200V/80A
Infineon
In order to avoid the effects brought by the external circuit, the validation experiments were performed under a simple hard switching environment, including resistive and clamped inductive load test circuit. The current generation of IGBTs generally has very fast switching speeds and becomes sensitive to the circuit parameters, so these parameters, such as parasitic inductance and connection resistance, have to be precisely measured and accounted for in the simulation. The comparison between the experimental and simulated results for various IGBTs at turn-off under clamped inductive load circuit is seen in Figs. 8, 9, 10, and 11, in which the red and pink curves are experimental voltage and current, and the black and blue curves are simulation voltage and current. The time scale is 200ns per division. 600
650
500
550
0.1
τ BF 0.05
400
Collector-Emitter Voltage (V)
450
50
100
150
200
250
300
350
400
Fig. 7. Effective lifetime extraction under different clamped voltages for the PT IGBT (The x-axis is collector-emitter voltage, 50V/div, and the yaxis is time 0.05µs/div)
The accuracy of the Fourier-based-solution IGBT modeling approach critically depends on the accuracy of the drift region parameters, but it is less sensitive to other device parameters. Therefore, upon the consideration of simplicity and negligible loss of accuracy, the remaining parameters of the buffer layer can be obtained based on their empirical
Current (A)
0
300
350
Ic_Exp Ic_Sim Vce_Exp Vce_Sim
200
Vce
100
0 1.35E-05
1.40E-05
250
Voltage (V)
Ic
0
150
1.45E-05
1.50E-05
1.55E-05
50 1.60E-05
Time (S) -100
-50
Fig. 8. Comparison of experiment and simulation for the IGBT_A turn off transient under 400V/300A
300
900
250
750
Ic 600
Ic_exp Ic_Sim Vce_exp Vce_Sim
Current (A)
150
100
450
300
50
Voltage (V)
200
150
Vce 0 1.65E-05
1.70E-05
1.75E-05
1.80E-05
1.85E-05
1.90E-05
1.95E-05
0 2.00E-05
Time (s) -50
-150
Fig. 9. Comparison of experiment and simulation for the IGBT_B turns off transient under 600V/200A 100
500
80
400
VI. ACKNOWLEDGE The authors would like to thank Angus Bryant of the University of Cambridge for many interesting discussions and valuable suggestions.
Ic
Ic_exp Ic_sim Vce_exp Vce_sim
40
300
200
Vce
20
0 0.0E+00
Voltage(V)
Current(A)
60
4.0E-07
6.0E-07
8.0E-07
1.0E-06
1.2E-06
1.4E-06
1.6E-06
1.8E-06
0 2.0E-06
[3]
Time(Sec) -20
-100
Fig. 10. Comparison of experiment and simulation for the IGBT_C turns off transient under 400V/90A
[4] [5]
80
840
70
735
60
630
[7]
50
525
[8]
40
420
Ic_Exp Ic_Sim Vce_Exp Vce_Sim
30
315
20
210
10
105
[6]
Voltage (V)
Current (A)
REFERENCES [1] [2]
100
2.0E-07
parameter extraction method is provided for the Fourierbased-solution analytical IGBT model. Since the extraction procedure is general in nature, some methods used in the research are also suitable for the extraction of some parameters needed for other kinds of IGBT models in Table I. By jointly using three general parameter extraction methods empirical-value-based extraction, datasheetbased extrapolation, and simple-test-based extraction the total extraction procedure only needs a simple clamped inductive load test for the extraction of the eleven and thirteen parameters needed for the NPT and PT IGBT models, respectively. Finally, the validation with the experimental results from various structure IGBTs demonstrates the accuracy of the proposed IGBT model and the parameter extraction method. The validation with the 5th generation CSTBT IGBT, which is still in process, will be provided at a later time.
[9] [10] [11]
0 0 9.80E-06 1.00E-05 1.02E-05 1.04E-05 1.06E-05 1.08E-05 1.10E-05 1.12E-05 1.14E-05 1.16E-05 1.18E-05
-10
Time (S)
-105
Fig. 11. Comparison of experiment and simulation for the IGBT_D turns off transient under 600V/60A
[12] [13] [14]
V. CONCLUSION AND DISCUSSION After review and comparison of current analytical IGBT modeling approaches and their parameterization, a practical
[15]
“Powerex Product Selector Guide" Powerex, Inc Junji Yamada, Yoshiharu Yu, Y. Ishimura, Mitsubishi Electric corporation, JAPAN; John F. Donlon and Eric R. Motto, Powerex Incorporated “Low Turn-off Switching Energy 1200V IGBT Module”, IEEE APEC Annual Mtg. Rec., March 2002. A.R. Hefner “Modeling buffer layer IGBTs for the Circuit Simulation”, IEEE Trans.P.E , vol. 10, no. 3, pp. 111-123, 1995. Peter O. Lauritzen, “A Basic IGBT Model with Easy Parameter”, IEEE PESC Rec., June 2001 A. Claudio, M.Cotorogea and M.A.Rodriguz “Parameter Extraction for Physics-based IGBT models By Electrical Measurement”, IEEE PESC, Proceeding. 2002. S. Musumeci, A.Raciti and M.Sardo etc “PT-IGBT PSpice Model with New parameter Extraction for life-time and Epy Dependent Behavior Simulation,” IEEE PESC Annual Mtg. Rec., pp. 1682-1688, 1996. Antonio G.M. Strollo “A New IGBT Circuit Model for Spice Simulation” IEEE PESC Rec. 1997 J. Sigg, P. Turkes and R. Kraus, “Parameter Extraction Methodology and Validation for an Electro-Thermal Physics-Based NPT IGBT Model”, IEEE IAS Rec., Oct. 1997 P.R. Palmer, J.C. Joyce, P.Y. Eng, J.L. Hudgins, E. Santi, and R. Dougal, “Circuit simulator models for the diode and IGBT with full temperature dependent features,” IEEE PESC Rec., June 2001. X. Kang, A. Caiafa, E. Santi, J.L. Hudgins and P.R. Palmer “Low Temperature Characterization and Modeling of IGBTs,” IEEE PESC Annual Mtg. Rec., in press, June. 2002. X. Kang, A. Caiafa, E. Santi, J.L. Hudgins and P.R. Palmer “Characterization and Modeling of High-voltage Field-stop IGBTs,” IEEE IAS Annual Mtg. Rec., in press, Oct. 2002. Antonio G.M. Strollo “A New IGBT Circuit Model for Spice Simulation” IEEE PESC Rec. 1997 Ph. Leturcq, J-L. Debrie and M.O.Berraies, “A Distributed Model of IGBTs for Circuit Simulation”, pp.1494-1501, EPE’97 H.Schlangenotto and W. Gerlach “On the Effective Carrier Lifetime in p-s-n Rectifiers at High Injection Levels”, Solid-State Electronics. 1969 R.A. Kokosa and R.L. Davies, “Avalanche breakdown of diffused silicon p-n junctions,” IEEE Trans. ED, vol. 13, no. 12, pp. 874-881, 1966.
*Department of Engineering University of Cambridge Trumpington Street Cambridge CB2 1PZ, UK
Abstract- A practical parameter extraction method is presented for the Fourier-based-solution physics-based IGBT model. In the extraction procedure, only one simple clamped inductive load test is needed for the extraction of the eleven and thirteen parameters required for the NPT and PT IGBT models, respectively. Validation with experimental results from various structure IGBTs demonstrates the accuracy of the proposed IGBT model and the robustness of the parameter extraction method. * Keywords—power semiconductor modeling, IGBT, Model, Parameter extraction
I. INTRODUCTION In recent years, the characteristics of IGBTs have been improved greatly. For example, with the application of advanced lifetime killing technology and structure development [1], the 4th generation IGBTs (trench-gate punch-through), which have been available in the market for some time, exhibit fast turn-off speed and low saturation voltage. The application of light punch through technology [2] in the CSTBT (Carrier Stored Trench Bipolar Transistor) IGBT now offers the new 5th generation IGBT with superior characteristics. Field-Stop IGBTs [10] also exhibit tail-free turn-off current under high voltage conditions. These external characteristics, brought about by internal device design changes, can only be accurately modeled by a physicsbased (analytical) IGBT model. However, in spite of the accuracy of analytical IGBT models, either overly complex or inaccurate parameterization often discourages electronic engineers from attempting to use IGBT models in their system designs. Reviewing the IGBT circuit model parameter extraction methods described in the literature shows that further work must be done in the field. The parameter extraction provided for the accurate and comprehensive Hefner model [3] is so complex that it is not practical for electrical engineers. The extraction method proposed in [4] is only for the lumpedcharge IGBT model. In [5], the parameter extraction for Hefner model was further developed with seven very precise but complex experiments. The parameters extracted in [6] *
This work was supported by the U.S. Office of Naval Research under Grant N00014-00-1-0131.
**Powerex, Inc 200 Hillis Street Youngwood, PA 15697, USA [email protected]
and [7] are used for the IGBT behavioral models. In [8], the parameter extraction lacks detail. In this work, we will provide a practical parameter extraction methodology to extract eleven and thirteen parameters for the non-punch-through (NPT) and punchthrough (PT) physics-based IGBT model presented in previous work [9]-[11]. In the parameter extraction procedure, as in the model itself, the better trade-off between accuracy and simplicity is pursued. Therefore, without affecting the simulation accuracy, only one simple clamped inductive load switching experiment is needed to extract several key parameters, while most parameters are extracted directly from the manufacturer’s datasheet, or estimated according to empirical value ranges and relations with already known parameter values. With parameter values obtained through the extraction procedure mentioned above, the PSpice IGBT model accurately simulated the hard-switching behavior of IGBTs having various structures. Comparison between the simulated and experimental results validates the proposed circuit model and the parameter extraction method. II. IGBT PHYSICS-BASED CIRCUIT SIMULATOR IGBT MODEL A. Categorization of Analytical IGBT Models Being a conductivity-modulated power device, the behavior of an IGBT depends heavily on the carrier distribution in its wide drift region. Under high-level injection conditions, the Ambipolar carrier Diffusion Equation (ADE) (1) describes the carrier dynamics in the majority of this region,
D
∂ 2 p ( x , t ) p ( x , t ) ∂p ( x , t ) = + τ ∂t ∂x 2
(1)
where D is the ambipolar diffusion coefficient, τ is the highlevel carrier lifetime within the drift region and p(x,t) is the excess carrier concentration. Therefore, most physics-based IGBT circuit simulator modeling approaches focus on the simulation of the drift region, and consequently implement the modeling of the ADE. However, it would be difficult and would also lead to potential convergence problems to directly implement the 2nd order partial differential equation in the circuit simulator
without any simplification. Consequently, various mathematical simplifications have been proposed for the implementation of analytical IGBT models. Depending on the type of mathematical simplification technique used to solve the ADE, analytical models can be categorized into four groups, as listed in Table I for example models, in which the right column shows the simplified equations representing the mathematical method used in each modeling approach. Table I. Physics-Base IGBT Models IGBT Model
Simplification
Hefner Model (Hefner) [3]
dp ( x , t ) = dt
Fourier Based Solution Model (Leturcq-Palmer) [9]
p(x, t) = p0 (x, t) + ∑ gi (t) * fi (x)
Lumped Charge Model (Lauritzen-Ma) [4]
qA
Laplace Transformation Model (Strollo) [10]
D
∑ h (t ) * f ( x ) i
i
dp ∆Q → dx ∆x
d 2 p' = p ' (s + 1 / τ ) dx 2
p ' = p ( x , t ) − p ( x ,0 )
(2) (3)
(4)
Further mathematical simplification allows the n-drift region to be modeled as an RC equivalent circuit as shown in Fig.1. The details are given in [13]. The representation requires the width of the undepleted base region and the hole and electron currents at the boundaries of the region (x1 and x2), which corresponds to the gradients of the carrier concentrations, f(t) and g(t), at x1 and x2, respectively. The functions f(t) and g(t) are defined as follows: 1 I n1 I p1 ∂p ( x, t ) (6) = − f (t ) = ∂t x1 2 qA Dn D p
1 I n2 I p2 ∂p( x, t ) g (t ) = = (7) − ∂t x2 2qA Dn Dp where A is the active cross-sectional area of the device, Dn and Dp, the electron and hole diffusion coefficients, In1 and Ip1, the electron and hole currents at x = x1 (p+ side), and In2 and Ip2 the electron and hole currents at x = x2 (p-body side). Clearly, the success of the approach now depends solely upon developing the appropriate boundary conditions.
(5)
The Fourier-based-solution (FBS) model is adopted in this research. By modeling the non-quasi-static characteristic of the carrier distribution, this approach preserves the essentials of the distributed nature of charge dynamics within the drift region and offers reasonable simulation speed while preserving satisfactory accuracy. B. Fourier-Based-Solution IGBT Model By applying a Fourier transformation, the ADE is simplified to a set of first order discrete linear equations, as shown in equation (3).
Fig.2 Definition of variables and typical carrier distribution under forward conduction of the PT IGBT
Therefore, in order to establish the boundary conditions, hole and electron currents at the edges of the drift region are needed. Fig. 2 defines some of relevant variables used in the analysis for PT IGBT (no n+-buffer layer for the NPT IGBT). Since by current continuity
I A = I n1 + I p1 = I n 2 + I p 2
Fig.1 Equivalent circuit diagram modeling the storage charge of the n drift region
(8)
the sum (IA) of the hole (Ip1 & Ip2) and electron currents (In1 & In2) is known, it is sufficient to find one current component at each edge. At the right side of the drift region, WB, (the MOS-channel edge), the electron current In2 is the channel current and it can be found as a function of the gate-source voltage using the well-known MOSFET equation. Current Ip2 can then be calculated using (8). At the left side of the drift region, different equations are used for the NPT and PT IGBT due to the structural differences. The details of the respective boundary conditions are given in [10] and [11].
For the NPT IGBT, equation (9) is used to calculate the electron current element in the junction between the p+ layer and the n-drift region. I n1 = qAh p p L2 0
(9)
where hp is the hole recombination parameter at the p+ layer [14]. For the PT IGBT, equation (10) is used to obtain the hole current component at the junction between the n buffer layer and the n drift region. qADpH W I p1 = [PH 0 − PHW cosh( H )] + I QH (10) WH LpH LpH sinh( ) L pH where LPH and DPH are the diffusion length and mobility of the hole carrier in the buffer layer, respectively, and other variables are defined in Fig.2. Finally, substituting the results of (9) and (10) into (8) gives the boundary conditions for each case. III. PARAMETER EXTRACTION The Fourier-based-solution IGBT model requires eleven and thirteen non-silicon parameters for the NPT and PT structure IGBTs respectively. Table II lists the parameters, which are classified into three categories based on the related device part: MOS-gate, geometry, and collector body. The parameter extraction procedure presented below is also divided into three parts according to this division.
Generally, there are three kinds of parameter extraction methodologies, which include: 1) Simple estimation based on the empirical value range, 2) Extrapolation according to the manufacturer’s datasheet, and 3) Extraction with simple experiments. To make the extraction procedure practical, the three methods are often employed jointly. Most parameters listed in Table II can be obtained from the datasheet or calculated based on textbook equations, so there is no need to use measurement for their extraction. Extra measurement can be used to verify the mathematical results. But some parameters, like carrier lifetime, have to be determined by experiment, which also need to be simple. Accordingly, the clamped inductive load test experiment is used. A. Extraction of MOS-gate parameters Because a behavioral MOSFET model is used for the IGBT MOS gate, the relevant parameters can be extracted following the extraction procedure for a MOSFET. The information given in the manufacturers’ datasheets is enough to extract these five parameters. The first three parameters: MOS threshold voltage (Vth), Transconductance Coefficient (Kpl), and short channel parameter (λ), can be obtained from the known I-V characteristic curve, shown in Fig. 3.
Table II. IGBT Model Parameter List Part
Symbol Vth (V)
MOS threshold voltage
Kpl (A/V )
MOS transconductance coefficient
λ
Short-channel parameter
Fig. 3 Forward I-V characteristics of the IGBT
Cge (nF)
Gate-emitter capacitance
Cox (nF)
Oxide capacitance
A (cm2)
Effective die area
a_i
Ratio of Inter-cell to total die area
WB (cm)
N drift region width
Gate-emitter capacitance Cge can be directly obtained from the input capacitance Cies (measured gate-emitter capacitance when collector is shorted to emitter) provided in the datasheet. Since the input capacitance Cies is the sum of the Cge and the Miller capacitance, and the former is much larger than the latter at higher voltages, Cge can be chosen equal to Cies at 10V.
NB (cm-3)
Doping concentration of N drift region
τHL (µs)
Carrier high-level lifetime in N drift
hp
Hole recombination coefficient in emitter (NPT)
NH (cm-3)
Doping concentration of N buffer Layer (PT)
WH (cm)
Width of n+ buffer layer (PT)
τBF (µs)
Carrier lifetime in N buffer layer (PT)
2
MOS Gate
Geom.
Collector Body
Description
Fig. 4 IGBT capacitance diagram
As shown in Fig.4, the MOS oxide capacitance Cox, together with the series-connected depletion capacitance Cdep, determines the gate-collector capacitance Cgc, which is also called the reverse transfer capacitance Cres in datasheets and is commonly referred to as the Miller capacitance between gate and collector. Obviously, Cox is the maximum value of the Miller capacitance when the depletion region under the gate area has not formed (Cdep = 0). That point corresponds to the maximum point of the Cres curve in the datasheet. Therefore, Cox can be written as:
C ox = max( C res )
(11)
B. Extraction of geometry parameters Instead of employing the destructive visual measurement, which is commonly used, the active die area A can be extracted based on the empirical range of IGBT maximum current density J. Therefore, the active die area can be roughly estimated as: A =
I
CM
J
(12)
where ICM is the peak collector current from the RBSOA curve given in the datasheet. Practically, the ICM value is decided by a multitude of factors, not only the chip but also the package and heat extraction. Nevertheless, the range of current density J generally is 100 A/cm2-150 A/cm2.
In this case, we can assume that Cce is proportional to the area under the emitter terminal while Cgc is proportional to the area under the gate terminal. The constants of proportionality are assumed to be the same since the depletion layer is quite far out from the gate and emitter terminals, and has, by this time, merged into one simple 1-D depletion layer. Thus:
a _i =
C gc Ai ≈ A C gc + C ce
(13)
Data sheets provide the reverse capacitance, Cres, and the output capacitance, Coes, which is the measured capacitance between collector and emitter when the gate is shorted, and can be written as:
C oes = ( C gc + C ce )
(14)
Vge = 0
As discussed above, we have
C res = C gc
(15)
Therefore, the intercell ratio can be obtained as
a_i ≈
C res C oes
(16)
Eq. (16) is only valid under the condition that the collector-emitter voltage Vce is sufficiently high to allow for an approximately uniform space-charge layer capacitance over the whole crystal area. In this case, the minimum values of Coes and Cres in the datasheet should be chosen, as sufficiently high Vce is reached at that point. C. Extraction of BJT body parameters There are several different ways to extract the carrier doping concentration of the drift region NB. The first is the direct estimation based on the normal range of the drift region background doping, which is from 6 × 1013 to 2 × 1014 cm-3 for the IGBT device. The typical value, 1 × 1014 cm-3, is generally chosen as the doping concentration NB during the simulation. The second way is based on the relation between the doping concentration and the breakdown voltage [4] using equation (17).
Fig. 5 Capacitance distribution diagram for a trench gate IGBT
The ratio of inter-cell area to total die area a_i can be extracted based on the capacitances in the datasheet. From above figure of the capacitance distribution, it can be seen that the displacement current due to the variation of the depletion region flows through two branches: one is the collector-emitter capacitance, Cce, branch under the emitter terminal, while the other is the Miller capacitance branch, which includes the collector-gate capacitance, Cgc. When the gate is shorted to the emitter (Vge = 0), the latter branch only includes Cgc.
N
B
=
3
60 ( Eg ) 1 .1 V BR
3 / 2
4
10
16
(17)
where VBR is the breakdown voltage value from the manufacturer’s data sheet plus about 150V-200V typical margin for general IGBTs, and Eg is the energy band gap value of silicon. Besides the above simple but rough extraction method, the extrapolation based on the Coes and Cres versus collectoremitter voltage curves in the datasheet can also lead to a reasonable extraction value of NB. Assuming breakdown in the bulk, an abrupt junction, and no mobile carriers in the depletion region (no leakage current), the depletion layer width is given as:
2 ε V CE
x dep =
qN
(18) B
where x dep is the depletion width and Vce is the collectoremitter voltage. From the capacitance analysis given above, the collectoremitter capacitance Cce is due to the depletion region under the emitter region: ε A (1 − a _ i ) (19) C = ce
x dep
Substituting (18) into (19) leads to a linear relationship 1 between and Vce. ( C ce ) 2 1 ( C ce ) 2
=
2 V ε ( A (1 − a _ i )) 2 qN B ce
(20)
where ic (t ) is the collector current, I c is the DC collector current, and I k is the current value at the starting point of the current tail. The high level lifetime τ HL can be obtained from the curve of the decay rate − d ln( i c ( t )) versus the collector dt
Furthermore, substituting the relation Cce = Coes − Cres into (20) leads to
1 2 Vce = (Coes − Cres ) 2 ε ( A(1 − a _ i )) 2 qN B
occurs under a constant collector-emitter voltage after the collector-emitter voltage reaches the clamp voltage. Therefore, the two-stage turn-off characteristic makes it easy to identify the current tail interval from the current and voltage waveforms. Based on the theory in [3], the decay time constant of the IGBT current tail portion for a constant collector voltage is given by: d ln(ic (t )) I 1 (24) =− (1 + c ) dt τ HL Ik
(21)
current. Fig.6 shows the measured results of a Dynex NPT IGBT with a 90A current rating. Notice that the effective lifetime is approximately independent of the collector current, therefore the collector current level used to extract the highlevel lifetime τ HL is unimportant.
Eq. (21) reminds us that the capacitance versus voltage curves provided by manufacturers can be used to extract the doping concentration in the drift region NB. With a linear regression applied to the curve for 1 /(Coss − C rss ) 2 versus Vce
Vcslope = 2 /(qεN B (1 − a _ i) 2 A2 )
(22)
Therefore, NB can be obtained from the slope value equation (22). The extraction for the remaining parameters in this section is different for the NPT and PT IGBT cases.
WB.
2ε
14 12 10 8 6 4 2 Collector Current Ic (A) 0
Under the triangular shape of the electrical field distribution in the NPT IGBT, the standard breakdown voltage equation gives: V BR = qN B (W B )
16
0
a) Parameter Extraction for the NPT IGBT
2
18
Effective lifetime (us)
[13], we can get the slope of this curve, which we will call Vcslope.
20
(23)
Substituting NB into (23) will give the drift region width
Since the carrier high-level lifetime dominates the IGBT 2nd stage turn-off current, the turn-off tail current decay will be used to extract the high-level lifetime τ HL . However, the current tail of latest-generation IGBTs is becoming less and less obvious than previous generation IGBTs, due to the application of advanced lifetime killing and other semiconductor technologies. Compared with other test circuit setups, the clamped inductive load test circuit is more suitable for lifetime extraction, especially for the current IGBTs with smaller lifetime. During IGBT turn-off in the clamped inductive load circuit, the IGBT current tail
10
20
30
40
50
60
70
80
90
100
Fig. 6. High-level lifetime extraction for the NPT IGBT (The x-axis is collector current, 10A/div, and the y-axis is time, 2µs/div)
The classic value of the recombination coefficient hp [14] is given by: hp =
1 NA
Dn
τn
(25)
where N A is the acceptor carrier concentration of the high doped p+ collector region, and τ n is the minority carrier
lifetime. Generally, the recombination coefficient hp can be selected based on the empirical value 1~2 *10-14 cm4 s-1 for abrupt junction. b) Parameter Extraction for the PT IGBT Like the WB extraction in the NPT case, using NB previously known into (26), the equation of the trapezoidal
electric field at breakdown for the PT structure, will give the extraction value of the drift region WB. 1 qN (26) VBR = Ec.WB − . B .WB2 2 ε where Ec is the critical electrical field value for the silicon.
For the lifetime extraction of the PT IGBT, a method similar to the NPT case can be used, which is based on versus the collector current in the clamped d ln( I c ( t )) − dt
inductive load circuit. The difference is that the effective lifetime τ eff , instead of the high-level lifetime τ HL in the NPT case, is used in (24) for the PT case. The effective lifetime is function of the high-level lifetime in the drift region and low-level lifetime in the buffer layer τ BF . Moreover, τ eff is dependent on the clamp voltage, unlike
τ HL in the NPT case, which is independent of the voltage. Therefore, lifetime extraction for the PT IGBT needs to be performed under several clamp voltages. The effective lifetime extraction under clamp voltage condition is the same as in the NPT case. Fig. 7 shows the curve of the effective lifetime τ eff versus the clamp voltage based on the test results for a PT IGBT. Notice that Fig. 6 and Fig. 7 have different xaxes: in Fig. 7 the x-axis is the collector-emitter voltage, whereas in Fig.6 the x-axis is the collector current. The highlevel carrier lifetime τ HL corresponds to the low-voltage τ eff value, while the low-level carrier lifetime in the buffer layer τ BF is equal to the τ eff value at high clamp voltage since the drift region is depleted under that condition. 0.3
τ HL
Effective Lifetime (us)
0.25
0.2
0.15
value range. The typical PT IGBT buffer layer width W H is about 4~10 um. The normal range of the doping concentration N H is 1016~1017 cm-3. For the 5th generation CSTBT IGBT and the Field-stop IGBT, smaller W H and N H values should be chosen. IV. VALIDATION OF THE IGBT MODEL AND PARAMETER EXTRACTION
In order to validate the analytical IGBT model and its parameter extraction method discussed above, the IGBT model with extracted parameters was used to simulate the switching behavior of various structure IGBTs from different manufacturers. They are listed in Table III. The validation experiment with the 5th generation CSTBT IGBT is still in process and will be included in the presentation at the conference. Table III. IGBT Device Lists IGBT
Structure
Rating
Manufacturer
A
Trench PT
600V/600A
Powerex
B
DMOS PT
1200V/600A
Powerex
C
DMOS NPT
1200V/100A
Dynex
D
Trench Field Stop
1200V/80A
Infineon
In order to avoid the effects brought by the external circuit, the validation experiments were performed under a simple hard switching environment, including resistive and clamped inductive load test circuit. The current generation of IGBTs generally has very fast switching speeds and becomes sensitive to the circuit parameters, so these parameters, such as parasitic inductance and connection resistance, have to be precisely measured and accounted for in the simulation. The comparison between the experimental and simulated results for various IGBTs at turn-off under clamped inductive load circuit is seen in Figs. 8, 9, 10, and 11, in which the red and pink curves are experimental voltage and current, and the black and blue curves are simulation voltage and current. The time scale is 200ns per division. 600
650
500
550
0.1
τ BF 0.05
400
Collector-Emitter Voltage (V)
450
50
100
150
200
250
300
350
400
Fig. 7. Effective lifetime extraction under different clamped voltages for the PT IGBT (The x-axis is collector-emitter voltage, 50V/div, and the yaxis is time 0.05µs/div)
The accuracy of the Fourier-based-solution IGBT modeling approach critically depends on the accuracy of the drift region parameters, but it is less sensitive to other device parameters. Therefore, upon the consideration of simplicity and negligible loss of accuracy, the remaining parameters of the buffer layer can be obtained based on their empirical
Current (A)
0
300
350
Ic_Exp Ic_Sim Vce_Exp Vce_Sim
200
Vce
100
0 1.35E-05
1.40E-05
250
Voltage (V)
Ic
0
150
1.45E-05
1.50E-05
1.55E-05
50 1.60E-05
Time (S) -100
-50
Fig. 8. Comparison of experiment and simulation for the IGBT_A turn off transient under 400V/300A
300
900
250
750
Ic 600
Ic_exp Ic_Sim Vce_exp Vce_Sim
Current (A)
150
100
450
300
50
Voltage (V)
200
150
Vce 0 1.65E-05
1.70E-05
1.75E-05
1.80E-05
1.85E-05
1.90E-05
1.95E-05
0 2.00E-05
Time (s) -50
-150
Fig. 9. Comparison of experiment and simulation for the IGBT_B turns off transient under 600V/200A 100
500
80
400
VI. ACKNOWLEDGE The authors would like to thank Angus Bryant of the University of Cambridge for many interesting discussions and valuable suggestions.
Ic
Ic_exp Ic_sim Vce_exp Vce_sim
40
300
200
Vce
20
0 0.0E+00
Voltage(V)
Current(A)
60
4.0E-07
6.0E-07
8.0E-07
1.0E-06
1.2E-06
1.4E-06
1.6E-06
1.8E-06
0 2.0E-06
[3]
Time(Sec) -20
-100
Fig. 10. Comparison of experiment and simulation for the IGBT_C turns off transient under 400V/90A
[4] [5]
80
840
70
735
60
630
[7]
50
525
[8]
40
420
Ic_Exp Ic_Sim Vce_Exp Vce_Sim
30
315
20
210
10
105
[6]
Voltage (V)
Current (A)
REFERENCES [1] [2]
100
2.0E-07
parameter extraction method is provided for the Fourierbased-solution analytical IGBT model. Since the extraction procedure is general in nature, some methods used in the research are also suitable for the extraction of some parameters needed for other kinds of IGBT models in Table I. By jointly using three general parameter extraction methods empirical-value-based extraction, datasheetbased extrapolation, and simple-test-based extraction the total extraction procedure only needs a simple clamped inductive load test for the extraction of the eleven and thirteen parameters needed for the NPT and PT IGBT models, respectively. Finally, the validation with the experimental results from various structure IGBTs demonstrates the accuracy of the proposed IGBT model and the parameter extraction method. The validation with the 5th generation CSTBT IGBT, which is still in process, will be provided at a later time.
[9] [10] [11]
0 0 9.80E-06 1.00E-05 1.02E-05 1.04E-05 1.06E-05 1.08E-05 1.10E-05 1.12E-05 1.14E-05 1.16E-05 1.18E-05
-10
Time (S)
-105
Fig. 11. Comparison of experiment and simulation for the IGBT_D turns off transient under 600V/60A
[12] [13] [14]
V. CONCLUSION AND DISCUSSION After review and comparison of current analytical IGBT modeling approaches and their parameterization, a practical
[15]
“Powerex Product Selector Guide" Powerex, Inc Junji Yamada, Yoshiharu Yu, Y. Ishimura, Mitsubishi Electric corporation, JAPAN; John F. Donlon and Eric R. Motto, Powerex Incorporated “Low Turn-off Switching Energy 1200V IGBT Module”, IEEE APEC Annual Mtg. Rec., March 2002. A.R. Hefner “Modeling buffer layer IGBTs for the Circuit Simulation”, IEEE Trans.P.E , vol. 10, no. 3, pp. 111-123, 1995. Peter O. Lauritzen, “A Basic IGBT Model with Easy Parameter”, IEEE PESC Rec., June 2001 A. Claudio, M.Cotorogea and M.A.Rodriguz “Parameter Extraction for Physics-based IGBT models By Electrical Measurement”, IEEE PESC, Proceeding. 2002. S. Musumeci, A.Raciti and M.Sardo etc “PT-IGBT PSpice Model with New parameter Extraction for life-time and Epy Dependent Behavior Simulation,” IEEE PESC Annual Mtg. Rec., pp. 1682-1688, 1996. Antonio G.M. Strollo “A New IGBT Circuit Model for Spice Simulation” IEEE PESC Rec. 1997 J. Sigg, P. Turkes and R. Kraus, “Parameter Extraction Methodology and Validation for an Electro-Thermal Physics-Based NPT IGBT Model”, IEEE IAS Rec., Oct. 1997 P.R. Palmer, J.C. Joyce, P.Y. Eng, J.L. Hudgins, E. Santi, and R. Dougal, “Circuit simulator models for the diode and IGBT with full temperature dependent features,” IEEE PESC Rec., June 2001. X. Kang, A. Caiafa, E. Santi, J.L. Hudgins and P.R. Palmer “Low Temperature Characterization and Modeling of IGBTs,” IEEE PESC Annual Mtg. Rec., in press, June. 2002. X. Kang, A. Caiafa, E. Santi, J.L. Hudgins and P.R. Palmer “Characterization and Modeling of High-voltage Field-stop IGBTs,” IEEE IAS Annual Mtg. Rec., in press, Oct. 2002. Antonio G.M. Strollo “A New IGBT Circuit Model for Spice Simulation” IEEE PESC Rec. 1997 Ph. Leturcq, J-L. Debrie and M.O.Berraies, “A Distributed Model of IGBTs for Circuit Simulation”, pp.1494-1501, EPE’97 H.Schlangenotto and W. Gerlach “On the Effective Carrier Lifetime in p-s-n Rectifiers at High Injection Levels”, Solid-State Electronics. 1969 R.A. Kokosa and R.L. Davies, “Avalanche breakdown of diffused silicon p-n junctions,” IEEE Trans. ED, vol. 13, no. 12, pp. 874-881, 1966.
