Millimeter-Wave Characterization and test-benches

Transcription

1 Millimeter-Wave Characterization and test-benches M. Spirito, L. Galatro, G. Gentile, S. Galbano, Electronic Research Laboratory, TU Delft Delft University of Technology Challenge the future

2 Outline Challenges in mm-wave wave probing Wafer-probes technology and limitation Calibration substrates technology and limitation Mm-wave modules Small-signal measurements Large signal measurements Mixed-signal active load-pull at V-Band Conclusions 2

3 Outline Challenges in mm-wave wave probing Wafer-probes technology and limitation Calibration substrates technology and limitation Mm-wave modules Small-signal measurements Large signal measurements Mixed-signal active load-pull at V-Band Conclusions 3

4 Challenges in mm-wave wave probing Wafer-probe technology Wafer probes represent the required transition to convert the waves travelling in a coaxial (e.g., 1 mm) or waveguide section into a coplanar field distribution with a GSG connection. Coaxial Waveguide * ** * ** *** Example (sub)mm-wave probes from Cascade Microtech*, GGB** and Dominion***. 4

5 Challenges in mm-wave wave probing Wafer-probe technology The transition to the coplanar (GSG) mode is performed in different ways from the different manufacturers. Micro-coax/MS/GSG Micro-coax/GSG WG/MS/GSG Top view [1] [2] [3] 5

6 Challenges in mm-wave wave probing Insertion loss versus frequency Wafer probe insertion loss increases dramatically with frequency. Reflection losses are double, i.e., when measuring S11 or S22. Insertion loss [db] Cascade Dominion GGB Approx data extracted from data sheet 100G 200G 300G 400G 500G 600G Frequency [Hz] 6

7 Challenges in mm-wave wave probing System DR versus frequency VNAs dynamic range decreases with frequency mostly due to limited output power available. Minimum Dynamic Range [db] R&S Vdi OML Min Dynamic Range from data sheet 0 100G 200G 300G 400G 500G Frequency [Hz] 7

8 Challenges in mm-wave wave probing Calibration accuracy at mm-wave Since the losses up to the wafer tips are higher and the DR of the system is lower, the calibration accuracy at mm-wave reduces. [4] 8

9 Challenges in mm-wave wave probing Calibration substrates technology and limitation The system limitations in terms of losses and DR are not the only ones reducing calibration accuracy, also electrically large structures and electrically thick substrates play an important role. [5] Higher order parallel-plate modes propagating in calibration substrate. 9

10 Challenges in mm-wave wave probing Calibration substrates technology and limitation Moreover, electrically thick materials (i.e., 254um) with high permittivity (i.e., ε r =9.9) can support (above cutoff) surface waves modes propagating inside the equivalent dielectric slab associated to the substrate. as an example a 254 um alumina substrate (ε r =9.9) the cutoff frequency of the first surface wave mode is ~197 GHz. 10

11 Challenges in mm-wave wave probing Calibration substrates technology and limitation The onset of higher order modes (i.e. parallel plate and surface waves) increases the losses of CPW structures and are not accounted by calibration. Fundamental (CPW) mode at 120 GHz on alumina (no backside ground). Fundamental plus higher (PPL) mode at 120 GHz (backside ground). 11

12 Challenges in mm-wave wave probing Calibration substrates technology and limitation The onset of higher order modes (i.e. parallel plate and surface waves) increases the losses of CPW structures and are not accounted by calibration. Fundamental (CPW) mode plus surface waves at 200 GHz on alumina substrate (no backside ground). 12

13 Challenges in mm-wave wave probing Calibration substrates technology and limitation When targeting (sub)-mm-wave calibrations different high quality substrates (i.e., quartz) with lower ε r, can be used to fabricate lines for on-wafer TRL cal. Providing surface wave cutoff at 465 GHz (height 200 um). Permittivity Frequency [GHz] x

14 Outline Challenges in mm-wave wave probing Wafer-probes technology and limitation Calibration substrates technology and limitation Mm-wave modules Small-signal measurements Large signal measurements Mixed-signal active load-pull at V-Band Conclusions 14

15 Mm-wave modules Small-signal measurements Mm-wave S-parameter measurements are conventionally performed using T/R up-conversion waveguide modules. Low frequency inputs (i.e GHz) are provided by the VNA and multiplied up to the designed WG band. [10] 15

16 Mm-wave modules Small-signal measurements Dynamic Range and SFDR At low frequency: ALC Required Pav level (set by the DUT) mm-wave 16

17 Mm-wave modules Small-signal measurements Pout vs Attenuation 1,2 Manual Attenuator 5 0 X: 9.075e+010 Y: P RF NO Attenuation[dBm] P RF 20 db Attenuator [dbm] Power [dbm] X: 7.5e+010 Y: ~6 db Variation Internal block diagram ~9 db Variation X: 1.1e+011 Y: freq [Hz] x

18 Mm-wave modules Small-signal measurements Power level control (via SW) To implement at SW level requires: Knowledge of power Knowledge of chain gain/loss 18

19 Mm-wave modules Small-signal measurements Off-wafer steps 1. WG 2-port calibration (TRL, SOLT,..) 10 error-terms 19

20 Mm-wave modules Small-signal measurements Off-wafer steps 2. Absolute power calibration (*) *Power Sensor: Agilent W8486A (75GHz 110GHz) 20

21 Mm-wave modules Small-signal measurements Off-wafer steps 3. Power leveling PPP = PPP (1 Γ 2 ) 21

22 Mm-wave modules Small-signal measurements Off-wafer steps 3. Pout of Port 1 and Port 2 vs frequency and Pin Freq range: 75GHz 110 GHz; Input Power range (PNA): dbm PNA: Agilent E8361A; mm-wave modules: Agilent N

23 Mm-wave modules Small-signal measurements On-wafer step Probes TRL de-embedding 23

24 Mm-wave modules Small-signal measurements Power leveled vs frequency 5 X: 7.675e+010 Y: Anritsu 3740A-EW wg modules P RF Pav [dbm] X: 7.675e+010 Y: X: 7.675e+010 Y: X: 7.675e+010 Y: X: 7.675e+010 Y: X: 7.675e+010 Y: X: 7.675e+010 Y: Frequency [Hz] x DUT Input power controlled Large Signal Measurements 24

25 Mm-wave modules Small-signal measurements Example: S-parameters measurement Power leveled Norden Millimeter Amplifier Mod: N

26 Mm-wave modules Large-signal measurements Example: Large Signal measurements Norden Millimeter PA Mod: N

27 Mm-wave modules Example: Large Signal vs S-parameter measurement S parameters meas Large Signal meas Norden Millimeter Amplifier Mod: N

28 Mm-wave modules Large-signal measurements The same cal info allows also to directly perform electronic power sweeps to measure fast and accurately the large signal performance of the DUT. f 0 =84 GHz [12] 28

29 Mm-wave modules Spectral response The limiting factor for the low-power level control is given by the mixer LO feed through. 29

30 Mm-wave modules Spectral response Consider the output spectrum of a WR10 mm-wave module set at f=90ghz P RF =0dBm X: 9e+10 Y: (Note f RF =f/6, f LO =f/8, f LO *7=78.75GHz, f LO *9=101.25GHz). Power [dbm] X: 7.875e+10 Y: X: 1.013e+11 Y: Frequency [Hz] x

31 Mm-wave modules Spectral response Consider the output spectrum of a WR10 mm-wave module set at f=90ghz 0 P RF =-3dBm -10 X: 9e+10 Y: Power [dbm] X: 7.875e+10 Y: X: 1.013e+11 Y: Frequency [Hz] x

32 Mm-wave modules Spectral response Consider the output spectrum of a WR10 mm-wave module set at f=90ghz P RF =-5dBm X: 9e+10 Y: Power [dbm] X: 7.875e+10 Y: X: 1.013e+11 Y: Frequency [Hz] 10 32

33 Mm-wave modules Spectral response Consider the output spectrum of a WR10 mm-wave module set at f=90ghz 0 P RF =-10dBm Power [dbm] X: 7.875e+10 Y: X: 9e+10 Y: X: 1.013e+11 Y: Frequency [Hz] x

34 Mm-wave modules Spectral response, impact on S-parameters To analyze if the spectral spurious in the waveguide band can influence the measurement let s consider a BJT modeled with HiCUM. 34

35 Mm-wave modules Spectral response, impact on S-parameters To analyze the potential impact of the spectral spurious on the measurement let s consider a BJT modeled with HiCUM. Ideal CW 30dB below P1dB for smallsignal operation Real signal 35

36 Mm-wave modules Spectral response, impact on S-parameters When also the spurious are in the small-signal range of operation, no effect of the spurious can be noticed, since the device is truly linear. Spurious level -51 dbm (real value), 35 db B.O. P1dB. 36

37 Mm-wave modules Spectral response, impact on S-parameters When the spurious are closer to the P1dB of the device we start making errors, this requires attention since they are often not seen (VNA measures narrow-band). Spurious level 15 db B.O. P1dB. 37

38 Mm-wave modules Spectral response, impact on S-parameters When the spurious are closer to the P1dB of the device we start making errors, this requires attention since they are often not seen (VNA measures narrow-band). Spurious level 10 db B.O. P1dB. 38

39 Outline Challenges in mm-wave wave probing Wafer-probes technology and limitation Calibration substrates technology and limitation Mm-wave modules Small-signal measurements Large signal measurements Mixed-signal active load-pull at V-Band Conclusions 39

40 Mixed-signal active load-pull at V-Band Waveguide (WR15) implementation Two operation modes: 1. Load-Pull 2. Characterization under UWB [13] 40

41 System Architecture Active Load Pull mode IF coupled waves acquired using wideband ADCs (100 MS/s) Down-conversion LO signal by low frequency synthesizer and x3 multiplier coaxial distribution IQ modulation on input and output injection signal I and Q baseband signals generated by ARBs (100 MS/s) Up-conversion signal generated by low frequency synthesizer and x4 multiplier 41

42 System Implementation Input branch DUT driving capabilities up to 13 dbm 42

43 System Implementation Output branch Output injection signal up to 22 dbm 43

44 Waveguide vs. Coaxial Waveguide components provide drastic reduction of the interconnect losses, when compared to coaxial lines. In active load-pull setups it is important to analyze the impact of losses on: 1. Accuracy and stability of the calibration 2. The capability to provide high reflection conditions to un-matched devices 44

45 Waveguide vs. Coaxial Comparison between 3 topologies b1 a1 a2 b2 1. Waveguide test-set (this work): front-end is realized with WR15 waveguides Source LO LO LO DUT LO Load 2. VNA with external test-set: external amplifiers and coaxial couplers used in close proximity to the DUT Source To Mixer/Samplers DUT To Mixer/Samplers Load 3. VNA-internal setup: using internal instrument sources and reflectometers Port 1 Port 2 45

46 Waveguide vs. Coaxial Stability comparison Agilent ADS simulation of a SOLT calibration using commercial coaxial and WG couplers and cables performances. 46

47 Waveguide vs. Coaxial Stability comparison Nominal component 60 GHz Component Insertion Losses Directivity VSWR Coaxial 1.85 mm 6 db/m WR-15 WG db/m Coaxial coupler 3.2 db 8 db 1.9 WG coupler 2 db 40 db 1.2 Improvement of calibration error terms Topology e d e s e t Coaxial -8 db db db WG -40 db db -5.5 db 47

48 Waveguide vs. Coaxial Stability comparison Experimental data Measurement of amplitude and phase of cables in GHz and GHz bands Fixed movement of 5 mm every 5 minutes for 2 hours (on probe station) Extraction of mean standard deviation (STD) for each band σ = db, φ(σ) = within GHz, times 3 (analog mult.) σ = 0.08 db, φ(σ) = 0.57 within GHz 48

49 Γ [db] Waveguide vs. Coaxial Stability comparison Montecarlo Simulation Coaxial test-set WG test-set Iterations phase [degrees] Simulation on a data item Γ = j Resulting standard deviation Coaxial: σ = db φ(σ) = 1.22 WG: Improved stability performances with WG reflectometer test-set σ = db φ(σ) =

50 Waveguide vs. Coaxial Driving power capability Maximum saturated power for commercially available amplifiers at 60 GHz ~22 dbm Consider a commercially available SiGe power cell Z OUT = 2.25 Ω P SAT = 9 dbm Only WG test set achieves Γ L >

51 Experimental Results Calibration accuracy LRM calibration before probes 51

52 Experimental Results Calibration accuracy LRM calibration at probe tips on calibration substrate 52

53 Experimental Results Load Pull on a thru standard Γ LOAD control over the entire Smith chart at 60 Pavs=10 dbm 53

54 Experimental Results Load Pull on a thru standard Γ LOAD control vs GHZ on a set Γ LOAD = Γ LOAD control vs avs =10 dbm on a set Γ LOAD =

55 Acknowledgments Arturo Santaniello, Yi Zhao, Atef Akhnoukh from TU-Delft Mauro Marchetti and Michele Squillante from Anteverta-MW. This research work is supported by the EU-Catrene RF2THz project. 55

56 References 1) Campbell, R.L.; Andrews, M.; Lesher, T.; Wai, C.;, "220 GHz wafer probe membrane tips and waveguide-to-coax transitions," Microwave Conference, 2005 European, vol.2, no., pp. 4 pp., 4-6 Oct. 2005, 2) Fung, A.; Samoska, L.; Pukala, D.; Dawson, D.; Kangaslahti, P.; Varonen, M.; Gaier, T.; Lawrence, C.; Boll, G.; Lai, R.; Mei, X.B.;, "On-Wafer S-Parameter Measurements in the GHz Band," Terahertz Science and Technology, IEEE Transactions on, vol.2, no.2, pp , March ) Reck, T.J.; Lihan Chen; Chunhu Zhang; Arsenovic, A.; Groppi, C.; Lichtenberger, A.W.; Weikle, R.M.; Barker, N.S.;, "Micromachined Probes for Submillimeter-Wave On-Wafer Measurements Part I: Mechanical Design and Characterization," Terahertz Science and Technology, IEEE Transactions on, vol.1, no.2, pp , Nov ) Douglas K. Rytting, Network Analyzer Error Models and Calibration Methods, 62nd ARFTG Conference Short Course Notes, December 2-5, 2003, Boulder, CO. 5) Schmiickle, F.J.; Doerner, R.; Phung, G.N.; Heinrich, W.; Williams, D.; Arz, U.;, "Radiation, multimode propagation, and substrate modes in W-band CPW calibrations," Microwave Conference (EuMC), st European, vol., no., pp , Oct ) Yi Zhao; Long, J.R.; Spirito, M.;, "Compact mm-wave power combiners in 65nm CMOS- SOI," Silicon Monolithic Integrated Circuits in RF Systems (SiRF), 2011 IEEE 11th Topical Meeting on, vol., no., pp.33-36, Jan ) Making Accurate and Reliable 4-Port On-Wafer Measurements, (2012) Cascade Microtech website. [Online]. Available 56

57 8) K. Silvonen, "LMR 16-a self-calibration procedure for a leaky network analyzer," Microwave Theory and Techniques, IEEE Transactions on, vol.45, no.7, pp , Jul ) V. Teppati, A. Ferrero, "On-wafer calibration algorithm for partially leaky multiport vector network analyzers," Microwave Theory and Techniques, IEEE Transactions on, vol.53, no.11, pp , Nov ) Millimeter Wave Vna Module Brochure, (2012) OML Inc. website. [Online]. Available 11) A. Ferrero and U. Pisani, An improved calibration technique for onwafer large-signal transistor characterization, IEEE Transaction on Instrumentation and Measurement, vol. 42, pp , Apr ) Y. Zhao, J.R. Long, M. Spirito, and A.B. Akhnoukh, A + 18dBm, GHz bandwidth power amplifier in 0.13 um SiGe-BiCMOS, Proceedings 2011 IEEE Bipolar/BiCMOS Circuits and Technology Meeting (BCTM) (pp ). Piscataway, NJ, USA. 13) L. Galatro, M.Marchetti, M.Spirito, A 60 GHz Mixed Signal Active Load-Pull System for Millimeter Wave Devices Characterization, Proceedings of the 80 th ARFTG Conference, November 2012, San Diego, CA. 57