The Accurate Electro-Thermal Model of Merged SiC PiN Schottky Diodes

Transcription

1 The Accurate Electro-Thermal Model of Merged SiC PiN Schottky Diodes M. Zubert, Ł. Starzak, G. Jabłoński, M. Napieralska, M. Janicki, A. Napieralski

2 Outline The Classical Model of SiC MPS diode Measurement Equipment and Procedure The Static Diode Model The Parameter Estimation Procedure The Electro-Thermal Model The Full Electro-thermal simulation The Dynamic Model Summary 2

3 Why SiC? Excellent thermal properties High operating frequencies High power levels Si 4H-SiC 6H-SiC GaAs Thermal Conductance [W/(cm K)] Bandgap [ev] Breakdown Sat. Electron Drift Velocity [10 6 cm/s] Src. 3

4 The Classical Model of MPS diode The statistical analysis of the measurements shown that: The classical model is extremely not consistent with the MPS diode measurements under forward bias! The reverse electro-thermal characteristics also cannot be accurately simulated with the classical model *). *) The only exception here was the SDP04S60 diode for which the average and the maximal errors in the considered temperature range were 15% and 70% respectively. 4

5 Examined Elements: Silicon Carbide Schottky Diodes Infineon Technologies AG (thinq!): o SDP04S60-2 nd gen. (600V, 4A, 13nC) o SDP10S30-2 nd gen. (300V, 10A, 23nC) Cree Power (Zero Recovery Rectifier): o CSD nd gen. (600V, 4A, 9nC) o CSD nd gen. (300V, 10A, 11.5nC) o C3D nd gen. (600V, 4A, 8.5nC) 5

6 Measurement Equipment and Procedure for The Electric Domain Tektronix 576 Curve-Tracer System. Nikon Coolpix 3200 digital camera. The specially developed image processing software for the automatic data readout from the digital camera. Voltages ranging from 300 V to 1200 V. Currents from 2 A to 20 A. 6

7 Measurement Equipment and Procedure for The Thermal Domain The device temperature was stabilized with liquid cooling/heating system: The dual cold plate. Peltier thermo-electric modules (the stabilisation accuracy of less than 1 K). The liquid circulation is forced by the HAAKE DC5 thermostat. 7

8 Measurement Equipment and Procedure Procedure The primary measurements of the device (SDP04S60) were taken for the following case temperature values: -2C, 2.5C, 15C, 25C and 35C 120C with the step of 5C. The model was verified for the SDP10S30, CSD04060 and CSD10030 diodes for the case temperatures ranging from 25C to 150C with the step of 25C. 8

9 I [A] The Measurement was Validated Diode: 50 C 1.E+01 U [V] 1.E Equipment: Metex M-3860M Fluke E-01 1.E-02 1.E-03 1.E-04 Verification Curve-Tracer System 1.E-05 9

10 Why Not The Classical Diode Model? CSD10030 for the Forward Bias 10

11 Why Not The Classical Diode Model? CSD10030 for the Forward Bias 11

12 Why Not The Classical Diode Model? CSD10030 for the Reverse Bias 12

13 I [A] The Static Diode Model for The Forward Bias I V T fwd fwd intrsc s fwd fwd, fwd, exp V V T R I V T V r T 1E+1 for physical kierunek interpretation przewodzenia 1E+0 1E-1 1E-2 1E-3 1E C 120 C 100 C 105 C 115 C 95 C 85 C 90 C 70 C 75 C 60 C 80 C 65 C 50 C 55 C 45 C 40 C 35 C 25 C 15 C 2,5 C 2 C 1E-5 0,5 0,7 0,9 1,1 1,3 1,5 1,7 1,9 2,1 2,3 2,5 U [V] 13

14 The Static Diode Model for The Forward Bias I V T fwd fwd intrsc s fwd fwd, fwd, exp V V T R I V T V r T for physical interpretation V T V V T intrsc intrsc1 intrsc V T V (1 T T ) r 2 ref ref,1 ref,2 where: I fwd device currents [A]; V fwd voltages [V], V fwd 0; R s internal parasitic resistance [W]; V intrsc internal voltage drop [V]; V r empiric coefficients [V, V -1, A]; T case temperature [C]. 14

15 The Static Diode Model for The Reverse Bias, exp I V T T V T rev rev rev 1E-4 1E-5 25 Celsius degree 50 Celsius degree 75 Celsius degree 100 Celsius degree 125 Celsius degree 150 Celsius degree Irev [A] 1E-6 1E Vrev [V] 15

16 The Static Diode Model for The Reverse Bias, exp I V T T V T rev rev rev T T T T exp T 1 2 where: I rev device currents [A]; V rev voltages [V], V rev 0; α, β empiric coefficients [V, V -1, A]; T case temperature [C]. 16

17 The Parameter Estimation Procedure All parameters in this paper were estimated using the Weighted Least Square (WLS) method, using the following objective function T 1 J z R z where z are the estimated measurement residuals. The measurement vector (z) are constructed in following way z z n 1 V, I,, V, I,, V, I fwd,1 fwd,1 fwd, i fwd, i fwd, n fwd, n where Vfwd, i, I fwd, i, Ti is i-th measurement pair (the voltage, current and temperature value). T 17

18 The Parameter Estimation Procedure The estimated parameters are associated with equation written in more convenient form without explicit exponent function: V ( p T p ) p log I p T fwd, i 2 i 1 3 fwd, i 4 i The nonlinearity of equation is overcome by Gauss-Newton method applying to Taylor series expansion lead to the iterative solution of the so-called Normal Equation (NE): where the non-vanishing elements of jacobian h(x) are calculated using following rules: Vfwd, i Vfwd, i H 1, H V, H x H I p 1 p T T p T T fwd, i 5 6 i nom 7 i nom i, n1 i, n7 p1 p7 T 1 T 1 H R H x H R z,, H, NEW x : x x 2 i, i 2i1,1 fwd, i 2 i,1 i 2 18

19 The Parameter Estimation Procedure The convergence of WLS formulation after 6 7 iteration steps. The other classical formulations of WLS method (e.g. Hybrid QR, Hatchel) lead to illconditioning and singularity. The χ 2 -test shows that all models are consistent with the measurements at the confidence level (CL) of 0.99 except for one of the devices (CL 0.95). 19

20 Parameters Estimation Procedure 2 nd generation of SiC MPS diodes 3 rd generation The model The model verification design Diode SDP04S60 SDP10S30 CSD CSD10030 C3D04060 Forward bias: V intrsc 1 V intrsc E E E E E-3 2 V ref α ref, E E E E E-3 α ref, E E E E-6 0 Comm ent Measurement s 5098; C; CL 0.99 Measurement s 420; C; CL 0.95 Measure ments 235; C ; CL 0.99 Measurement s 366; C;CL 0.99 Measurement s 1825; C;CL

21 Parameters Estimation Procedure 2 nd generation of SiC MPS diodes 3 rd generation The model The model verification design Diode SDP04S60 SDP10S30 CSD0406 CSD10030 C3D R s Reverse bias: α E E E-07 0 α Equ is E E E E-05 used α instead of β E E E E-12 equ. β Comm ent C; CL C; CL C C; CL C 21

22 The Static Model for The Forward Bias 22

23 The Static Model for The Forward Bias 23

24 The Static Model for The Forward Bias 24

25 The Static Model for The Reverse Bias SDP04S60 CSD

26 The Static Model for The Reverse Bias 26

27 The Electro-Thermal Model 27

28 Thermal Parameters The 2 nd generation of SiC MPS diodes The model The model verification design Diode SDP04S60 SDP10S30 CSD0406 CSD Thermal domain: R Th1, 1.11, 8.20E , 7.07E-04 C Th1 R Th2, 1.77, 7.53E , 1.64E-03 C Th2 R Th3, 0.924, , C Th3 R Th4, C Th , , rd gen. C3D

29 The Full Electro-thermal SPICE Simulation Ifwd [A] P [W] θ [ C] DMCS INFINEON 29 Vfwd [V]

30 The Full Electro-thermal SPICE Simulation Irev [ua] INFINEON DMCS 30 - Vrev [V]

31 The Dynamic Model The dynamic behaviour has been obtained by the junction charging model 1 Cj Cj0 V 511 VJ MJ 31

32 The Dynamic Model The dynamic behaviour has been obtained by the junction charging model 1 C C V 511 VJ MJ j j0 SDP04S60 SDP10S30 CSD04060 CSD10030 C3D04060 Dynamic behaviour (based on datasheets, VJ estimated for T=300K) Qj=Cj0*PWR(VJ,MJ)*PWR(V(511)+VJ,1-MJ)/(1-MJ), cj E E E E E-12 vj mj

33 Dynamic simulated diode turnoff (C3D04060). 33

34 The final verification (10 C)

35 The final verification (10 C)

36 Summary The novel electro-thermal dynamic HSPICE model of SiC merged PiN Schottky diode (2-nd and 3-rd generation) has been presented. The main advantage of the proposed model is that it is given in a closed form. The manufactures models are not agree with the real device behavior. 36