Méthodes avancées de caractérisation et de modélisation des transistors HEMT GaN

Transcription

1 Méthodes avancées de caractérisation et de modélisation des transistors HEMT GaN Jean Christophe NALLATAMBY, Julien COUVIDAT, Sylvain LAURENT, Raphaël SOMMET, Michel PRIGENT & Raymond QUERE XLIM, CNRS Université de Limoges mars 2018

2 INTRODUCTION Potential applications for the GaN HEMTS are numerous - For power amplifier of Radar T/R modules - For base stations PA - For multi-tone PA in space applications All those applications involve complex modulated signals - Pulsed, phase/frequency modulated for Radar Systems - Highly complex signal (OFDM, QAM,.) 2

3 Key challenges of GaN HEMT Technology Deep level states (Traps) in the bandgap of the material. Surfacetrapscouldbeformeddueto defects or impurities formed at the surface during crystal growth or in device fabrication process. AlGaN/GaN HEMTs grown on foreign substrates results in imperfect crystals with dislocations or defects. These defects cause the formation of deep level states within the GaN and AlGaN material. Physical location of traps in the device Consequences of Traps Gate lag, Drain lag and drain current collapse/ DC RFdispersion. DC RF dispersion limits the large signal performance of the device. Trapping effects also contributes to dynamic R ON, which affects the power switching applications. 3

4 Different ways for characterization Parasitic effects still limit the performances of PA : Thermal effects Trapping effects Trap investigation using : Pulsed I-V characterization: Gate lag and drain lag identification but no information on the dynamics of the traps Low frequency S-parameters measurement Low frequency noise characterization Deep level transient spectroscopy (I-DLTS) measurement Large signal characterization (in frequency and time domain) All these characteristics should be consistently modeled 4

5 Gate and Drain-Lag Mechanism Gate-Lag Drain-Lag 5

6 Ron & Idss measurements for thermal assessment Based on a new and simple method proposed by J. Joh and all Measurement of Channel Temperature in GaN High Electron Mobility Transistors : I ds Ron measurement. Idss measurement. I dss Vgs=0V R on =δv ds / δi ds Tools required : Pulsed I(V) Test bench measurement (PIV). (pulse width: 500 ns; duty cycle: 0.05%) Thermal chuck. (variation range: 25 C to 175 C) Thermal Resistance Determination Methodology in 2 steps: Thermal chuck sweep with fixed zero dissipated output power. Dissipated output power sweep with fixed thermal chuck control. V ds 6

7 Ron & Idss measurements for thermal assessment Pulsed I-V characteristics (V GS = 0 V) from zero power quiescent bias point (V DS0 = V GS0 = 0 V) : Pulsed I(V) characteristics (V GSi = 0 V) from various quiescent bias points (VGS0 = 0 V, VDS0 = V) with fixed thermal chuck at 25 C : V gs0 : gate bias point. V ds0 : drain bias point. V dsi : drain pulsed point. 7

8 Thermal Resistance ,5 420 y = 0,0323x + 3, Idss Ron 8 7,5 Idss (ma) Idss Ron 7 Linear (Ron) Linear (Idss) 6 5 y = -0,6888x + 449, Ta ( C) Ron (Ohm) Idss (ma) Power dissipation (W) 7 6,5 6 5,5 5 4,5 4 Ron (Ohm) Baseplate temperature sweep Dissipated power sweep dron RON ( T) RON ( T0 ) T dt dron R ( P ) R (0) P dp ON diss ON diss diss R TH T dr ON dron / Pdiss dpdiss dt 8

9 Trapping and Thermal Effects Positive Pulse T j = T chuck + R TH P diss Trapping P diss = V DS I DS T chuck =25ºC Trapping + Thermal Power limit Illustration of Trapping and Thermal Effects Difficult to disassociate trapping and thermal effects, especially at lower operating frequencies. Trapping and thermal time constants are almost of the same order. Negative Pulse 9

10 LF- Y 22 Parameter setup Study of the output admittance (Y 22 ) Low Frequency Y 22 analysis at several temperatures Output admittance very sensitive to trapping phenomena Easy extraction of time constant (Low Frequency S 22 measurement [10Hz to 10MHz]) 10

11 Trapping model AlGaN/GaN : 8x250 μm Ids=50mA and 100 C with In Low Frequency Cds is equal to open then The emission constants can be extracted either from peak values of Y22 imaginary part The number of traps is defined by the number of peaks of the Y22 imaginary part 11

12 LF Y 22 Parameter (8 75 µm Device) The frequency at the peak of Imag(Y 22 ) gives the emission time constant Arrhenius law extraction of activation energy (E a ) and capture cross section (σ n ) F I,peak f, f = e T σ A exp. E g KT A. e Traps Energy Level: E a = 0.4 ev e n trap emission rate σ n Cross section of traps N C effective density of electrons in the conduction band V th thermal velocity of electrons g degeneracy factor E a Trap activation energy 12

13 LF Noise Measurement (6 75 μm Device) Measurement Setup Equivalent Circuit =. _ Voltage Amplifier: e ati Voltage noise source i ati Current noise source Z ati Impedance S V_meas Measured voltage noise spectral density S ir Thermal noise spectral density Z = Zeq R 13

14 LF Noise Measurement (6 75 μm Device) Identified Traps Physical Properties: Extracted physical properties of traps are in agreement with literature reported data for E a1 = 0.57 ev; σ n = cm 2 Fe-doped GaN devices. E a2 = 0.51 ev; σ n = cm 2 14

15 I DLTS Test Bench (Vg Pulse) Pulse generator on the gate DC voltage supply on the drain Pulses qualified (dc path bandwidth = 20MHz) bias tees One scope acquires the range 10 7 to 10 2 sec, the other for 10 3 to 10 2 sec Sampling first scope=100 9 sec second scope=1 6 sec An AWG is used to generate pulse width and triggered start point measurement Probe station with thermal chuck Example : Vgs_pulse = 8V Vds0 = 10V Id = 50mA/mm Tfilling = 10µs, 1ms et 100ms Température = 25 C, 75 C, 100 C et 125 C Equation of fitting curve for one temperature : GPIB AlGaN/GaN 2x100 um 50 Ω Pulse Generator 8114A RF AWG AF3252 TRIG DC Bias Tee RF+DC DUT DC Supply E3641A Gauss Probe Gauss Probe DC RF+DC Bias Tee RF 50 Ω High Sampling Scope DPO7054 Slow Sampling Scope DPO7054 Id normalisé 1,2 1 0,8 0,6 25 C 50 C 0,4 75 C 0,2 Tfilling=1ms 100 C 125 C 0 1,0E 06 1,0E 05 1,0E 04 1,0E 03 1,0E 02 1,0E 01 1,0E+00 1,0E+01 Temps (s) Thermal Chuck 15

16 I-DLTS Test Bench (Vd Pulse) Pulse generator on the drain DC voltage supply on the gate Pulses qualified (dc path bandwidth = 20MHz) bias tees One scope acquires the range 10-7 to 10-2 sec, the other for 10-3 to 10 2 sec Sampling first scope=100-9 sec second scope=1-6 sec An AWG is used to generate pulse width and triggered start point measurement Probe station with thermal chuck AWG AF3252 TRIG GPIB Example : Vds_pulse = 20V Vds0 = 10V Id = 50mA/mm Tfilling = 10μs, 1ms et 100ms Température = 25 C, 75 C, 100 C et 125 C Equation of fitting curve for one temperature : AlGaN/GaN 2x100 um 50 Ω DC Supply E3641A RF DC Bias Tee RF+DC DUT Pulse Generator 8114A Gauss Probe Gauss Probe DC RF+DC Bias Tee RF 50 Ω High Sampling Scope DPO7054 Slow Sampling Scope DPO7054 Id normalisé 1,2 1 Tfilling=1ms 0,8 0,6 25 C 75 C 0,4 100 C 0,2 125 C 0 1,0E 06 1,0E 05 1,0E 04 1,0E 03 1,0E 02 1,0E 01 1,0E+00 1,0E+01 Temps (s) Thermal Chuck 16

17 Identification of traps ln(τt²) Trap 1 Fit (Trap1) y = 0,4522x 11,321 Vg Pulse E a = 0,4522 ev σ = 1,43E-16 cm² To determine the activation energy E a and the capture cross section σ n by using the Arrhenius equation : With and /KT ln ln ln(τt²) y = 0,2517x 0, Trap 1 2 y = 0,4806x 11,303 Trap 2 Fit (Trap1) 1 Fit (Trap2) /KT Vd Pulse E a = 0,2517 ev and 0,4806 ev σ = 2,50E-21 cm² and 1,40E-16 cm² g = 1 ; T² = K² Nc = 3,13E 18 cm 3 Vth = 2,92E 7 cm/s 17

18 Mesure Modele x75 POLAR Vgs= V, Vds= V, Id = A Vgs= V Vgs= +0.00nV Vgs= V Vgs= V Vgs= V Vgs= V Vgs= V Vgs= V Vgs= V Vgs= V Vds en Volts Cgs_1D Vgs 6,00E-13 5,00E-13 4,00E-13 3,00E-13 2,00E-13 1,00E-13 0,00E+00 Mesure Modele Cgd_1D Vgd 1,40E-13 1,20E-13 1,00E-13 8,00E-14 6,00E-14 4,00E-14 2,00E-14 0,00E+00 1,6 1,55 1,5 1,45 1,4 1,35 1,3 1,25 1,2 y = -0,0024x + 1, ,50E-14 3,00E-14 2,50E-14 2,00E-14 1,50E-14 1,00E-14 5,00E-15 0,00E+00 y= 1,6E-16+1,04973E-16*EXP(T/26,3157) Steps of the modeling process Étapes de modélisation db (S(1,2)) db (S(2,1)) Modèle petit-signal Modèle I-V Capacités NL Modèle thermique Modèles de pièges S(1,1) ; S(2,2) freq, GHz Phase (S(1,2)) Phase (S(2,1)) freq, GHz freq, GHz freq, GHz Ids (A) Vds (V) Id en Amperes Cgs (F) Cgd (F) Idss Is_gs 6.00E E E E E E E E E E E E E-0 k Vds + k + Vgs_int Rfill Vgs Vds Rempty C Vds freq (2.000GHz to 40.00GHz) Rg Lg Cpg Ls Cpd Ld Rs Rd Ri Cds τ Gm Gd Cgs Cgd Rgd Dgs=f(Vgs) Dgd=f(Vgd) Ids=f(Vgs,Vds) Cgs=f(Vgs) Cgd=f(Vgd) Dgs=f(Vgs) Dgd=f(Vgd) Ids=f(Vgs,Vds,T) Rs=f(T) Rd=f(T) Rgd=f(T) Ids=f(Vgs_pièges,Vds,T) Various parasitics effects are successively added 18

19 Non linear electrothermal model with trapping effects model (fast-trap only) meas. Bias 0V/0V, V DS,max = 15V meas. Bias 0V/0V, V DS,max = 30V We call these traps "Fast Emitting Traps". Highlighted in low frequency [S] meas. Knee walk-out for Vds,max > 15V Slow traps activated and not modeled 19

20 Large signal validation : DC drain current drop due to both fast and slow traps New trap model : Investigation of Fast and Slow Charge Trapping Mechanisms of GaN/AlGaN HEMTs through Pulsed I-V Measurements and the Associated New Trap Model Presentation by Julien Couvidat at IMS Philadelphie

21 Linearity Measurement with Multi-Tones Signal PA Linearity Performances hugely depend on driven signals To measure linearity with Complex Modulated Signals like real signals (LTE,..) the need to demodulate signal Emulation with an 8-ton generic signal with same statistical properties than Complex Modulated Signals (PAPR, PDF, ) No need to demodulate Signal composed of 8 tones such as IM3, IM5 non overlap with themselves and 8-tones signal Computation of a C/I by summing all InBand (between the 1st and last signal tone) IM3 and IM5 powers 10. Definition of Figure Of Merite (by analogy with EVM) % 10 21

22 Power Building of the 8-tones Signal f 1 f 2... f k... f 8 ( k 1). f k FFT grid : 2 17 Frequencies separated by 1. with 1 k n. 1 st frequency rank of frequency. distance between tons. offset frequency.. are Unequally Spaced p k 0,1, 3,3,3,3,3,3 guarantees non overlap between IM3, IM5 and 8 tones 38 22

23 Example of simulated PA Output Signal Non overlap between IM3, IM5 and 8-tones Signals 23

24 Unequally Spaced Multi Tons test bench up to 6 GHz IF 160MHz IF BW : 10 MHz Dynamic range: 65~70 db % 10 24

25 Multi-tone LP measurement AlGaN/GaN 8x75 m V DS0 =30V, f 0 = 2 GHz 1-tone to 8-tones characteristics Linearity Assessment Power characteristics are obtained for each frequency in the same way as for classical LP systems C/IM3 is the total power for carriers / total IM power FOM is an EVM like measure (can be set equal to the EVM in some particular cases) 25

26 Spectrum results and Linearity measurement AML26P2401 Bias : 15V 190mA Power sweep : -33dBm to 4dBm 8 tones around 2GHz With >2MHz of bandwidth Input Output Pin=-33dBm Pin=4dBm Linearity Characterization of RF Circuits through an Unequally Spaced Multi-Tone Signal Sylvain Laurent et al. Microwave Measurement Conference (ARFTG), th ARFTG 26

27 Unequally Spaced Multi Tons test bench Under development up to 50 GHz IF bandwidth : 38MHz Typical dynamic: 90dB A fully calibrated NVNA set-up for linearity characterization of RF power devices using Unequally Spaced Multi-Tone signal through IM3 & IM5 measurements» Presentation by V. Gillet at 91st ARFTG Microwave Measurement Symposium June 15, 2018, Philadelphia PA 27

28 Conclusion GaN HEMTs modeling still remains a challenging task Impact of Low Frequency parasitic on large signal performances must be evaluated Dynamics of traps play a major role A tentative model able to consistently take into account those dynamics has been proposed Special nonlinear measurement set-ups are mandatory to assess all the effects 28

29 Acknowledgements Thanks to III-V lab and UMS Foundry for supplying the device wafers. Thanks to the Agence Nationale de la Recherche (ANR) and Direction Générale de l'armement (DGA) France. This work was supported by COMPACT project under contract ANR-17- ASTR and VeGaN project, France, under Contract FUI 18 Thanks to PhD Students of XLIM for providing measurement characteristics 29

30 30