Completing the Design Flow. A Short Course Covering Component-Level Modeling and Measurement, Circuit Design and Analysis and System Modeling

Transcription

1 Completing the Design Flow A Short Course Covering Component-Level Modeling and Measurement, Circuit Design and Analysis and System Modeling

2 Demystifying Device Characterization Abstract - Device characterization is an essential process in many aspects of research, development and testing of RF and microwave devices. To the inexperienced, this might seem intimidating or even scary. In this course, we will explore various interconnected topics of device characterization that form the amplifier design flow. Topics include pulsed IV and S-parameters for compact model extraction, load pull for model validation and measurement, amplifier design and IC stability analysis, X-parameter modeling and system-level simulations. Instructors from Maury Microwave, Agilent Technologies and AMCAD Engineering will provide instruction and demonstrations.

3 Demystifying Device Characterization Load Pull S-Parameters IV Curves Compact Models Amplifier Design Circuit Simulation Harmonic Balance Amplifier Stability X-Parameters What do they mean? Are they somehow related?

4 System Design from Compact Models Component level Circuit level System level Pulsed IV and RF measurements Compact FET model extraction VNA based load pull FET Compact model Validation & Refinement IC Design & Stability analysis IC X-Parameter model Simulation at system level TOOLS PIV PNA-X IVCAD IVCAD ICCAP TUNERS PNA-X IVCAD ADS ADS IVCAD STAN ADS (PNA-X) ADS

5 System Design from X-Parameters Component level Circuit level System level X-Parameters load pull IC Design & Stability analysis IC X-Parameter model Simulation at system level TOOLS TUNERS PNA-X NVNA IVCAD ADS IVCAD STAN ADS (PNA-X) ADS

6 System Design from Measurements Component level Circuit level System level VNA based load pull IC Design & Stability analysis IC X-Parameter model Simulation at system level TOOLS TUNERS PNA-X (or PSG/PSA) IVCAD ADS IVCAD STAN ADS (PNA-X) ADS

7 Design flow entry points Component level Circuit level System level Pulsed IV and RF measurements Compact FET model extraction VNA based load pull FET Compact model Validation & Refinement IC Design & Stability analysis IC X-Parameter model Simulation at system level VNA based load pull IC Design & Stability analysis IC X-Parameter model Simulation at system level

8 Instructor Introduction Pulsed IV/RF and Compact Modeling Load Pull Amplifier Design in ADS Stability Analysis X-Parameters System-level Simulations

9 Instructor Gary Simpson Gary Simpson received his Bachelor degree from DeVry Institute of Technology in 1972, and his Masters degree from Arizona State University in He has been involved with microwave measurements since 1973, starting with device characterization through manual load pull on microwave power transistors at his first job at Motorola. He has been with Maury Microwave since 1982, where he began developing components and fixtures for microwave measurements, including network analyzer calibration standards and techniques. Gary is a pioneer in device characterization systems; in 1987 he developed one of the earliest automated slide-screw tuners for advanced load pull measurements. Since then, he has been responsible for much of the on-going development of device characterization techniques, methodologies and systems. He is currently Chief Technical Officer at Maury Microwave Corp.

10 Instructor Ali Boudiaf Dr. Ali Boudiaf received his Master's degree in EE in 1987 at the Ecole Polytechnique of Algiers and his PhD at the University of Paris XI in He has an extensive experience in the microwave device characterization and modeling fields; started his career as an assistant professor at university of Marne-la-Vallée near Paris and then worked with ATN Microwave, Agilent Technologies, ACCO USA (Auriga Measurement Systems), Focus Microwaves and Maury Microwave. He holds three patents. He is a Member of IEEE and is active in the IEEE MTT Society and the Automatic RF Techniques Group (ARFTG). He is currently the Application Engineering Manager at Maury Microwave.

11 Instructor Carlos Nadal Carlos Nadal is a Senior Applications Engineer at Agilent Technologies Served in the Air Force as a crew member on board Air Force One as a radio operator/repairman during the Gerald Ford administration. Attended University of Arizona. Started with Hewlett Packard in Started in Manufacturing Engineering and over the years have worked in R&D, Marketing, Sales, and finally Applications Engineering. With HP/Agilent worked in New Jersey, Los Angeles, San Diego and Santa Clara. Hold a Commercial Pilots License and is a 3rd Degree Black Belt in Kung Fu San Soo.

12 Instructor Al Lorona Al Lorona is an Application Engineer who helps customers use SystemVue and other Agilent Technologies EDA products more effectively and creatively. With 24 years of experience at Hewlett-Packard and Agilent he is a seasoned presenter, teacher and sales team member. Al is based in southern California.

13 Instructor Introduction Pulsed IV/RF and Compact Modeling Load Pull Amplifier Design in ADS Stability Analysis X-Parameters System-level Simulations

14 Large-Signal Transistor Models Convergence Operating range Physic model Compact model Behavioral model Extrapolation Accuracy Physical insight Easy modeling process Usability for Circuit design

15 Commercial compact FET models Mostly used models for GaN HEMTs FET models Number of parameters Electro-thermal effect Trapping Effects Original Device Context Curtice3 [1] 59 No No GaAs FET CFET [2] 53 Yes No HEMT EEHEMT1 [3] 71 No No HEMT Angelov [4] 80 Yes No HEMT/MESFET AMCAD HEMT1 [5] 65 Yes Yes GaN HEMT

16 Rs, Rd Idss Compact FET model extraction flow Rd y = x Rs 0.8 y = x T C y = x T C Small-Signal Non-linear capacitances IV Model Thermal model Trapping effects Rg Lg Cpg Ls Cpd Ld Rs Rd Ri Cds τ Gm Gd Cgs Cgd Rgd Cgs=f(Vgs) Cgd=f(Vgd) Dgs=f(Vgs) Dgd=f(Vgd) Ids=f(Vgs,Vds) Dgs=f(Vgs,T) Dgd=f(Vgd,T) Ids=f(Vgs,Vds,T) Rs=f(T) Rd=f(T) Ids=f(Vgs_trap,Vds,T) Various effects are successively added

17 Pulsed IV measurements Short pulse : Quasi-isothermal conditions Low duty cycle : Constant mean temperature Quiescent bias point : Thermal conditions fixed Several quiescent bias point

18 Pulsed IV measurements must be accurate from low to high voltage/current values Accurate IV data = Reliable current source Transconductance Leakage current Ideality factor schottky diode Pulsed IV measurements IVCAD

19 How to get accurate pulsed IV measurements? PIV system Pulsed IV measurements

20 Pulsed IV measurements How to get accurate pulsed IV measurements? +20V Gate 15 bits + sign Drain 250V 16 bits 15 bits + sign -20V 25V 16 bits Pulse shape monitoring 20ns time resolution

21 Pulsed IV measurements How to get accurate pulsed IV measurements? AM212 Gate access 1A 100mA 10mA 1mA 0mA 33µA 4mA 1A 3,3µA 400µA 100mA 330nA 40µA 10mA 33nA 4µA 1mA -20V 650µV 20mV 20V -2V 65µV 2mV 2V 0V Measurement Resolution Voltage Absolute Accuracy Voltage Range

22 Pulsed IV measurements How to get accurate pulsed IV measurements? 10A AM µA 20mA Drain access 1A 0A 22µA 2mA Measurement Resolution Voltage Absolute Accuracy Voltage Range 0,53mV 4,9mV 50mV 500mV 0V 25V 250V

23 Pulsed S-parameter measurements Pulsed S parameter measurements Bias Bias

24 Pulsed S-parameter measurements The first & most important point : Pulsed S parameter measurements must not be noisy Small S2P measurement variation = strong influence over the linear model extraction : optimization algorithm Requirements : IVCAD Dynamic range in pulsed mode > 90dB for Duty Cycle ~ 5%

25 Pulsed S-parameter measurements Pulsed S-parameter measurements must not be noisy at low duty cycle with narrow pulse width Pulse detection methods Wideband detection Narrowband detection Receiver samples IF filter Receiver samples IF filter No pulse desensitization Increased noise with narrow pulse width due to wider IF bandwidth Limited pulse width by maximum available IF bandwidth Narrower minimum pulse width than wideband pulse Reduced dynamic range with low duty cycle due to pulse desensitization by 20*log(duty cycle)

26 PNA/PNA-X Noise reduction techniques and performances Peak-to-peak noise with wideband detection at 10% duty cycle No averaging in calibration and measurements 0.03 db 0.04 db 0.06 db 0.09 db Pulse width (IFBW) 10 us (150 khz) 5 us (280 khz) 1 us (1.5 MHz) 500 ns (3 MHz) Averaging 20 times in calibration and measurements db db db db

27 PNA/PNA-X Noise reduction techniques and performances Dynamic range with wideband detection at 10% duty cycle with 10 us, 5 us, 1 us, 500 ns pulse width No averaging in calibration and measurements, 1% smoothing on Averaging 20 times in calibration and measurements, 1% smoothing on

28 PNA/PNA-X Noise reduction techniques and performances Peak-to-peak noise at 10% duty cycle Wideband detection with 20 times averaging in calibration and measurements Narrowband detection with no averaging in calibration and measurements Pulse width db db db db 10 us 5 us 1 us 500 ns db db db db

29 PNA/PNA-X Noise reduction techniques and performances Dynamic range with narrowband detection at 500 ns pulse width Hardware gating No Averaging, 1% smoothing on, 500 Hz IF bandwidth Crystal filter Software gating >100 db at 10% >100 db at 5% 90 db at 1% 85 db at 0.5% Spectral nulling

30 VNA performance comparisons E836x Legacy PNA N524xA PNA-X N522xA New PNA Pulse generator External Internal/External Internal/External Pulse modulator External Internal/External Internal/External Wideband detection Max BW/Min PW 35 khz / 50 us 15 MHz / 100 ns 15 MHz / 100 ns High level noise* dbrms dbrms to dbrms Dynamic range** 114 to 123 db 124 to 129 db 127 db Narrowband detection Min IF gate width 20 ns <20 ns <20 ns Dynamic range*** 85 db <105 db <105 db * Specified as trace noise magnitude, at 20 GHz, at 1 khz IF bandwidth ** Specified performance at 20 GHz, at 10 Hz IF bandwidth *** Measured performance at 10 GHz at 10 Hz IF bandwidth, 1% duty cycle

31 Internal or external master pulse (PULSE SYNC IN) Synchronized data acquisition (P0) Synchronized internal pulse generators (P1 P4) with independent delay and width Internal or external drive for modulators or receiver gates N1966A Pulse I/O adapter PNA/PNA-X internal pulse access Search on

32 Pulsed IV/RF parameter measurements How to get accurate pulsed IV measurements? Synchronisation between Pulse IV and pulse S parameters

33 Pulsed S-parameter measurements How to get accurate pulsed IV measurements? Synchronisation between Pulse IV and pulse S parameters

34

35 Instructor Introduction Pulsed IV/RF and Compact Modeling Load Pull Amplifier Design in ADS Stability Analysis X-Parameters System-level Simulations

36 Pout (dbm) and Gain (db) Pout (dbm) and Gain (db) Model Validation Large-signal Model validation of a 8x75 µm GaN HEMT with load-pull measurements performed at 6 GHz for optimum PAE load impedance in class-ab Model validation of a 8x400 µm GaN HEMT with load-pull measurements performed at 3 GHz for the optimum Pout load impedance in class-b meas. model Pout PAE gain PAE (%) meas. model Pout gain PAE PAE (%) Pin dbm Pin dbm -20

37 Model Validation With non optimal loads : Time domain load pull measurements Deembedding in the intrinsic reference plane Parasitic extrinsic elements must be accurately extracted by previous S parameter measurements

38 Model Validation What are the unique or specific requirements for load pull with regards to model validation? -Independent powers at each frequency (fo, 2fo ) -Independence of source impedance match

39 Introduction to load pull 1) Vary impedance presented to DUT (active device, transistor) Highest Pout 2) Measure Pout, Gain, Efficiency 3) Determine best matching impedance 4) Design matching network (EEsof ADS)

40 Impedances and impedance tuners VSWR α Gamma α 1/Ω 10:1 VSWR = Γ=0.82 = 5Ω 20:1 VSWR = Γ=0.9 = 2.5Ω Γ = a/b Probe Y X Mechanical Tuner Gamma comes from probe (slug) inserted into airline Airline X Y Probe Airline

41 Traditional load pull measurements (optional ) In Traditional Load Pull, delivered output power is calculated from Power Meter de-embedded through S-Parameter block and Impedance Tuner Available input power is calculated from gain lookup table created during power calibration or from input Power Meter and then de-embedded through S-Parameter block and Impedance Tuner

42 What about large-signal Zin? Large signal input impedance, Zin, changes as function of: - Drive power - Zload Traditional load pull matches source impedance at single power, not taking into account varying Zin during power sweep

43 What about large-signal Zin? Gain values look low because only Pin,available is used reflected power due to mismatch is not taken into account Traditional load pull only reports Transducer Gain

44 Vector-receiver (real-time) load pull Network Analyzer Low-loss Signal Source Amplifier Impedance Tuner Coupler 1 1 Pout b a b load 1 1 Pin, del a1 b1 a in Low-loss Coupler G p PAE P P 50Ω Load Impedance Tuner out b 2 a in, del 1 P out P P DC in, del load in

45 Vector-receiver (real-time) load pull Gain values look low because only Pin,available is used reflected power due to mismatch is not taken into account Knowing Zin allows us to calculate Power Gain, taking into account mismatch thereby showing true gain potential of device

46 Traditional LP Vector Receiver LP Pre-Characterization Required Recommended (not required) Number of Points Tuner De-embedding Vector-receiver (real-time) load pull More points = greater accuracy (even with interpolation) Critical! (Accuracy relies on de-embedding) Minimum points required (no impact on accuracy) No tuner de-embedding Power Meter Network Analyzer Spectrum Analyzer Power Sensor Power Low-loss Signal Source Amplifier Impedance Tuner Impedance Tuner Sensor Impedance Tuner Coupler Signal Source Amplifier Low-loss 50Ω Load Coupler Impedance Tuner

47 Vector-receiver (real-time) load pull Traditional LP Verification Procedure ΔG t complex conjugate matched verification Vector Receiver LP Zin vs. Zload comparison ΔG t complex conjugate matched verification

48 Vector-receiver (real-time) load pull

49 Active and hybrid-active load pull VSWR α Gamma α 1/Ω 10:1 VSWR = Γ=0.82 = 5Ω 20:1 VSWR = Γ=0.9 = 2.5Ω Γ = a/b Probe Airline X Y Probe Airline Mechanical Tuner Gamma comes from probe (slug) inserted into airline Γ<1 Active Tuner Gamma comes from signal generator and amplifier Γ=1 or Γ>1

50 Active and hybrid-active load pull Maximum Tuning Range (exaggerated for effect) Tuner Tuner + Cable Tuner + Cable + Probe 50 Losses of cables, probes, test fixtures reduces tuning range and cannot be overcome using traditional load pull methods

51 Active and hybrid-active load pull External Tuners For Harmonic Load Pull, Traditional Load Pull systems require one mechanical tuner per frequency per DUT side To tune Fo, 2Fo and 3Fo at the same time requires 3 tuners (using multiplexer or cascaded methods) It is possible to build 3 tuners in 1 box, but it becomes 2-3x longer and 2-3x more expensive

52 Active and hybrid-active load pull Γ=0.99 Γ=0.99 Tuner + Cable + Probe Gamma advantage of Active Load Pull Losses of cables, probes, test fixtures reduces tuning range, and can be overcome using larger amplifiers

53 Active and hybrid-active load pull Active Fo Load Pull

54 Active and hybrid-active load pull Hybrid-Active Fo Load Pull

55 Active and hybrid-active load pull Active Fo, 2Fo, 3Fo Load Pull

56 Active and hybrid-active load pull Hybrid Active Fo, 2Fo, 3Fo Load Pull

57 Active and hybrid-active load pull Measured Data Passive VS Active Excellent Agreement Traditional Load Pull Active Load Pull

58 Active and hybrid-active load pull Passive Fo Active 2Fo, 3Fo Γ 2Fo DUT on-wafer! One of many configurations of hybrid/active load pull

59 Vector-receiver (real-time) load pull Load pull for model validation Compact model ADS Load pull for measurements Measurements ADS

60

61 Instructor Introduction Pulsed IV/RF and Compact Modeling Load Pull Amplifier Design in ADS Stability Analysis X-Parameters System-level Simulations

62 Which Type Are You? Designers usually fall into one of two camps: Compact or X-parameter models Use any of the setups in the Load Pull Design Guide HB Can sweep Can optimize A wide variety of simulations possible; great data displays Measured LP data Must use a Data-based LP component S-parameter analysis Can sweep Can optimize Good for designing matching networks ADS is set up to handle any case.

63 Simple load pull introduction to concepts Which Impedance should I present the Device at the in- and output (over a broad frequency range to over the higher harmonics) to have a maximal Pdel, PAE and Gain with minimal distortion (XdB-compression, EVM, ACLR, etc.)?

64 Device performance due to Z l and Z s f1 f2 f3 freq External source (or previous stage) Input match. network Output match. network f1 f2 f3 freq External load (or next stage)

65 Fundamental load pull Why? Quick sanity check ; adjust sampled area f1 f2 f3 freq Source tuner Available source power constant f1 f2 f3 freq Load tuner Guess reasonable values for all variables. Adjust, if necessary.

66 Fundamental load pull with power sweep Why? See gain compression and constant power delivered data f1 f2 f3 freq Source tuner Load tuner Available source power swept freq f1 f2 f3 freq

67 Fundamental source pull Why? Source impedances affect gain primarily, but also PAE f1 f2 f3 freq Source tuner Available source power constant f1 f2 f3 Load tuner freq

68 Fundamental load pull with parameter sweep Sweep any parameter - source frequency, bias, stability network parameter values, etc. Why? Investigate device performance more thoroughly f1 f2 f3 freq Source tuner Load tuner Available source power constant freq f1 f2 f3 freq

69 Harmonic load phase sweep Why? Harmonic impedances matter, but usually want high reflection f1 f2 f3 freq Source tuner Load tuner Sweep input power to see constant power delivered data freq f1 f2 f3 freq

70 Source stimulus responses IMD from 2-tone source ACLR from modulated source Gain comp. curves from source power sweep

71 Amplifier design in ADS What is available for the non-linear device? Model run load pull simulations to determine optimal matching and biasing conditions for amplifier design Measured Load Pull Data analyze measured data and determine optimal matching and biasing conditions for amplifier design

72 Start with fast, simple load pull Most parameters are passed to tuner inside instrument subcircuit Device Model from Design Kit

73 Start with fast, simple load pull Available source power held constant Guess optimal Zsource and harmonic Zs Refine sample space Source Power = 5 dbm Source Power = 12 dbm

74 Load pull with power sweep

75 Pdel, dbm Select load for highest Pdel or highest PAE PAE

76 Contours versus swept parameter (frequency) 28 dbm contour at 750 MHz 28 dbm contour at 1.25 GHz

77 Dependency on phase of gamma at harmonic

78 Sweep Gate Bias Results with gate bias = 2.25V

79 Constant power del. load pull with two tones

80 Load pull with WCDMA signal Read modulated data from file. Scale signal amplitude by optimizing SFexp variable.

81 Maury measured data Examine contours and make trade-offs for optimal load condition Use measured data files directly in impedance matching network design and optimization

82 Performance contours from Load Pull Data 1) Reads LP data file 2) Simulates S-parameters of network 3) Gets corresponding performance data Tuner generates loads in region you specify

83 Indep. variables and performance parameters Frequency and input power constant

84 Plot performance contours from LP Data Load giving best performance Check the Contours, Rectangular or Circular Regions Frequency Slider PAE Pdel Gt

85 Using power sweep of Load Pull data Why sweep power? See gain compression data. Sweep values within range of those in file Sweep based on gamma_x, gamma_y values in file

86 Contours at specified gain compression Why do contours look strange? Measurements at some loads were not valid.

87 Pdel, dbm Choosing load: high efficiency or high power PAE

88 Choosing optimal load at 2.17 GHz

89 Use measured data directly in optimization This impedance should be the same as this.

90 Load Pull delivers the Impedance for the Matching Network Design Frequency Sweep

91 Matching Network Design Smith Chart Utility Design impedance matching network(s) using existing techniques, or optimization

92 Matching Network Design Matching Utility (Broad Band) ADS Impedance Matching Utility Low-pass, high-pass, and band-pass, lumped element matching Multi-section quarter-wave matching Tapered-line impedance matching Single-stub impedance matching Several others

93 Using optimization to adjust parameter values Preliminary output matching network to be optimized

94 Impedance optimization at 3 frequencies Output matching network to be optimized Goal impedance values:

95 Testing performance of completed amplifier One-tone harmonic balance frequency and power sweep Two-tone harmonic balance frequency and power sweep

96 Testing performance of completed amplifier

97 Verification of the of the Layout EM Cosim Run EM to obtain more accurate results Input Output EM Model Analytical Model

98 PA Design Workflow 1) Run load pull simulation on the active device model or load pull measured data a. 1-tone, 1 input power load pull b. Power sweep to see gain compression c. Frequency or bias sweep d. Harmonic load phase sweep e. Constant output power with swept var f. Source pull g. 2-tones to see IMD h. Modulated signal to see ACLR 1) Choose optimal load impedances across frequency band 2) Use Smith Chart Utility or favorite matching tool to design preliminary matching network 3) Use optimization to adjust values 4) Use EM simulation and/or optimization to obtain more accurate results 5) Repeat steps 1-5 for to design source matching network 6) Test final design, including matching networks

99

100 Instructor Introduction Pulsed IV/RF and Compact Modeling Load Pull Amplifier Design in ADS Stability Analysis X-Parameters System-level Simulations

101 Stability analysis is a critical step of RF design flow Classical methods are either not complete or too complex Stability analysis need to be efficient (especially in large signal) - Rigorous - Fast - User-friendly - Compatible with commercial CAD softwares Introduction to stability analysis

102 Existing methods Linear analysis small signal K factor Normalized Determinant Function (NDF) Stability envelope Non-linear analysis large signal Nyquist criterion NDF Bolcato, Di Paolo & Leuzzi, Mochizuki,

103 Existing methods linear analysis Widely used: K factor (also µ and µ now) - K>1 & <1: unconditional stability of two port network - K<1: conditional stability stability circles Unconditional stability Conditional stability Unconditional instability Limitations: Only indicates that a stable circuit will continue to be stable when loading it with passive external loads at the input or output Do not guarantee the internal stability of the circuit!

104 Source Existing methods linear analysis Potentially instable architectures for which K factor is not enough Multi-stage power amplifier Multi-fingers transistor IN OUT Gate Drain

105 phase(zsond) H (º) db(zsond) H (db) Pole-zero identification principle H( j ) R G H( s) Frequency domain Identification techniques n i 1 p j 1 ( s z ) ( s ) i j Im (GHz) E9 4.0E9 6.0E9 8.0E9 1.0E10 1.2E10 Freq frequency (GHz) Pole-zero plot poles zeros Re (GHz) f 0, P in v out Node n (i,f ) in s R L Complex conjugate poles with positive real part -> start-up of an oscillation Oscillation frequency = Module of the imaginary part

106 New stability analysis method - STAN Suitable for both linear and non-linear stability analysis Very easy to use with any CAD tool Very easy to analyze results Relative stability information delivered Oscillation mode knowledge -> Help to find the suitable stabilization strategy Parametric Analysis implemented Monte-Carlo Analysis

107 STAN Integration with CAD GENERATOR Perturbation introduction node CIRCUIT LOAD in out Var VAR Eqn VAR1 fin=9.65 GHz Pin=12 Input frequency Input power P_1Tone cmp1198 Num=1 Z=50 Ohm P=polar(dbmtow(Pin),0) Freq=fin ampli X1 Term Term1 Num=1 Z=50 Ohm HARMONIC BALANCE HarmonicBalance HB1 Freq[1]=fin Order[1]=10 SS_MixerMode=yes SS_Start=f1 SS_Stop=f2 UseAllSS_Freqs=yes MergeSS_Freqs=yes Var VAR Eqn VAR3 f1=fstart+fin e9 f2=fend+fin Meas MeasEqn Eqn meas1 Zsond=mix(v_sond,{-1,1})/mix(I_sond.i,{-1,1}) frequency=ssfreq-fin Var VAR Eqn VAR2 fstart=4.325 GHz fend=5.325 GHz n_point=101 I_Probe I_sond v_sond I_1Tone SRC1 I_LSB=polar(0.0001,0) Start sweep frequency Stop sweep frequency Number of frequency points Nonlinear stability analysis template EDA Tool Templates for Agilent ADS AC simulation for linear HB simulation for non-linear STAN tool integrated in IVCAD software User-friendly GUIs

108 Example 1 (Linear analysis) STAN Integration with CAD Low frequency instability of a medium power 1.2 GHz FET amplifier built in hybrid microstrip technology L hole C in V gg L cable L in L cable V dd C out L hole Low frequency oscillation 14MHz C imn GaAs FET L out C omn R s L imn L omn R L L hole L hole

109 STAN Integration with CAD ADS schematic v out i in Note: It can also be done with a voltage probe connected in series at a circuit branch and observing the total admittance

110 phase(h) db(h) STAN Integration with CAD Simulated frequency response Z(j ) 50 freq MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz MHz mag(z) phase(z) freq, GHz freq, GHz

111 Identification results STAN Integration with CAD

112 Centered on 20 MHz STAN Integration with CAD

113 STAN - Selecting the node All nodes are equal, but some nodes are more equal than others SISO transfer function exact pole/zero cancellations are possible Pole/zero cancellations are associated with the lack of controllability and/ or observability in the system

114 STAN - Selecting the node Example: let us consider a complex circuit in which the oscillation is taking place in a part that is totally isolated from the node selected to perform the analysis 2 e 9 v out Node n (i,f ) in s 2 e 9

115 STAN - Selecting the node However, the oscillation can be predicted if the node is connected in the part of the circuit that is not isolated v out (i,f ) in s Node m 2 e 9 2 e 9

116 STAN - Selecting the node The analysis is performed at a node that is not completely isolated from the part where the oscillation is taking place, i.e. there is a weak electrical link between parts 2 e 9 v out Node n (i,f ) in s 2 e 9 Quasi-cancellation poles and zeroes almost cancelled low degree of controllability/observability in the selected node we are (electrically) far from the place where the oscillation is being generated useful information for circuit stabilization

117 Recommendations STAN - Selecting the node In simple circuits with a clear feedback structure any node should serve for the analysis Multistage power amplifiers At least one analysis per stage

118 Relevant information about the nature of the oscillation and the place in which it is being generated can be extracted extremely useful for circuit stabilization V bias_1 STAN - Selecting the node V bias _ m V bias _ n Im (GHz) Im (GHz) clear -2 quasi-cancellation -2 not observable -4-4 Im (GHz) Re (GHz) Re (GHz) Re (GHz)

119 STAN - Selecting the node Node n v out (i,f ) in s A B FET 2 FET 1 FET 3 A- No oscillation detected in the common node FET 4 FET 5 FET 6 B- Oscillation detected in the transistor node Odd mode (parametric frequency division) will determine the stabilization strategy

120 STAN Multi parameters Analysis with swept parameter(s) Verification for various conditions (Pin, Zload, ) Checking of critical resonances Optimization of stabilization networks R G P IN f 0, v out (i,f ) in s Z load R stab We might increase the stability margin of critical resonances when part of the system dynamics is not correctly modeled (or is likely to change).

121 P out (dbm) P out (dbm) Power splitter Power splitter Power splitter Power combiner STAN Multi parameters Example 2 (HB analysis) Q1 Two-stage X-Band MMIC power amplifier built in HBT technology based on AsGa/GaInP process R S Q2 Q3 R L f in,p in Q4 Parametric frequency division measured for P in =13.8 dbm and f in =9.65 GHz Frecuencia (GHz) Frecuencia (GHz) f in (GHz) f in (GHz)

122 ADS schematic (parametric analysis on P in ) STAN Multi parameters

123 STAN Multi parameters Identification results: evolution of critical poles Pin

124 STAN Multi parameters Example 3: RC network in series at the base of the transistors trying to increase resistance at f0/2 without significantly degrading performances at f0 Large signal stability analysis to find suitable values for RC Guaranteeing sufficient stability margin considering technological dispersion Selected circuit: R=15Ω C=2.5pF

125 polar(inestables_hb1..mod,inestables_hb1..phase) S(1,1) STAN Multi parameters Example: 3-stage LDMOS DPA for SDR applications Application requires absence of spurious for a wide range of operating conditions Multivariable large-signal stability analysis versus input frequency, input power and real and imaginary parts of load termination Z L. Frequency division (f in /2) detected Unstable loads freq (1.000GHz to 1.000GHz) Stable mod and (0.693 unstable to 0.990) regions in the L plane for fin=500 MHz and Pin=17.1 dbm Stable loads

126 Imaginary Axis (MHz) Imaginary Axis (MHz) STAN Monte Carlo Example: L-Band medium power FET amplifier Low frequency instability related to the input bias network Stabilization by the inclusion of a gate-bias resistor R STAB Monte Carlo sensitivity analysis for different R STAB (5 % dispersion in all circuit parameters) R STAB = R STAB = Real Axis (MHz) Real Axis (MHz)

127 Example: Ku-Band MMIC PA for active space antenna Stable original circuit STAN Performance optimization Inter-branch stabilization resistances RF in RF out RC stabilization networks Natanael Ayllón Rozas Développement des méthodes de stabilisation pour la conception des circuits hyperfréquences : Application à l optimisation d un amplificateur de puissance spatial., PhD Thesis, February 2011.

128 All stabilization networks removed resistances maintained for topological reasons STAN Performance optimization Example: Ku-Band MMIC PA for active space antenna RF in RF out Parametric frequency division /2 instability

129 Optimized version resistances maintained for topological reasons STAN Performance optimization Example: Ku-Band MMIC PA for active space antenna RF in RF out No oscillation detected, especially around F0/2 Stabilization resistances

130 STAN Performance optimization Example: Ku-Band MMIC PA for active space antenna Results comparison Original Optimized

131

132 Instructor Introduction Pulsed IV/RF and Compact Modeling Load Pull Amplifier Design in ADS Stability Analysis X-Parameters System-level Simulations

133 Non-linear Vector Network Analyzer Test Procedure and Methodology

134 Scattering Parameters Linear Systems Linear Describing Parameters Linear S-parameters by definition require that the S-parameters of the device do not change during measurement. x() t y() t 0 a 1 10 e 1 1 a 1 S 21 1 b 2 01 e 2 0 b 2 0 b 1 00 e 1 01 e 1 11 e 1 1 b 1 S S S 12 1 a 2 11 e 2 10 e 2 00 e 2 0 a 2 b S a S a b S a S a b1 S11 S12 a1 b S S a S-Parameter Definition To solve VNA s traditionally use a forward and reverse sweep (2 port error correction). 134

135 Scattering Parameters Linear Systems Linear Describing Parameters If the S-parameters change when sweeping in the forward and reverse directions when performing 2 port error correction then by definition the resulting computation of the S- parameters becomes invalid. b S a S a b S a S a b1 S11 S12 a1 b S S a f r f r b1 b 1 S11 S12 a1 a 1 f r f r b S 2 b2 21 S 22 a2 a2 Hot S22( a1 ) x() t y() t This is often why people are asking for Hot S22 because the match is changing versus input drive power and frequency (Nonlinear phenomena). Hot S22 traditionally measured at a frequency slightly offset from the large input drive signal. 135

136 Measurements on Nonlinear Components

137 Lets see what really happens with hot S22: Start by driving a signal into the amplifier A 1,1 B 2,1 B 2,2 B 2,3 P1 P2 f 1 a=incident b=scattered A 1,1 Harmonic Index Port number f 1 f 2 f 3 137

138 Lets see what really happens with hot S22 A 1,1 B 2,1 P1 P2 B 2,2 B 2,3 f 1 f 1 f 2 f 3 We call the B response X F ( A 11 ) It is the output response to an input, as a function of the input amplitude. 138

139 Lets see what really happens with hot S22 Now, lets make the input signal bigger A 1,1 B 2,1 P1 P2 B 2,2 B 2,3 f 1 f 1 f 2 f 3 The harmonics increase faster than the fundamental, generally, by their order number (2 nd = twice as fast) 139

140 Lets see what really happens with hot S22 Let s add a small signal incident on port 2, offset in freq B 2,1 A 1,1 b 2,1 B b 2,2 B 2,1 * 2,3 P1 P2 a 2,1 f 1 f 1 f 2 f 3 Normally, we would say S22=b2/a2 What term describes b2*? b2* is Transposed on the other side of B2 Hint 140

141 Doing Hot S22 using Source from Port 4 rear panel J11 J10 J9 J8 J7 J4 J3 J2 J1 Source 1 OUT 1 OUT 2 Pulse modulator Source 2 (standard) OUT 1 OUT 2 Pulse modulator LO To receivers Pulse generators R1 R3 R4 R2 A C D B Test port 1 Test port 3 Test port 4 Test port 2 Going Beyond S-Parameters

142 We can see this real time on the PNA-X B2 b2* Hi Pwr b2 a2 (Port 2 open) Port 2 to Amplifier out 142

143 Lets see what really happens with hot S22 Let s change the frequency of a21 B 2,1 A 1,1 b 2,1 * b 2,1 P1 P2 B 2,2 B 2,3 a 2,1 f 1 f 1 f 2 f 3 When a2 goes up in frequency, b2* goes down in frequency 143

144 Lets see what really happens with hot S22 Let s change the phase of a21 B 2,1 A 1,1 b 2,1 b 2,1 * B 2,2 B 2,3 P1 P2 a 2,1 f 1 f 1 f 2 f 3 b2* What moves do as we the call conjugate a term that of changes a2 phase. with opposite phase? a CONJUGATE term! b2*=t22. a2* Now we can see that T22 = b2*/a2* And it is CLEAR that HOT S22 is insufficient to describe the reflection behavior 144

145 Lets see what really happens with hot S22 Now what happens if a2 is not offset in freq? B 2,1 A 1,1 b 2,1 b 2,1 * B 2,2 B 2,3 P1 P2 a 2,1 f 1 a2,1* f 1 f 2 f 3 The total response of the b2 scattered signal, due to a2 is a combination of b2 and b2*: b2=s22 x a2 + T22 x a2* b2= X S 22( A 11 )a2+ X T 22( A 11 )a2*; b2=x 22 ( A 11 )a2 145

146 Lets see what really happens with hot S22 Now what happens if a2 is not offset in freq? B 2,1 A 1,1 b 2,1 b 2,1 * B 2,2 B 2,3 P1 P2 a 2,1 f 1 f 1 f 2 f 3 Changing the phase of a2 changes the total magnitude of b2 wave, but not the magnitude of individual parts, b2 and b2* 146

147 Lets see what really happens with hot S22 Now what happens if a2 is not offset in freq? B 2,1 A 1,1 b 2,1 b 2,1 * B 2,2 B 2,3 P1 P2 a 2,1 f 1 f 1 f 2 f 3 Changing the phase of a2 changes the total magnitude of b2 wave, but not the magnitude of individual parts, b2 and b2* 147

148 X-Parameter Experiment Design & Identification Ideal Experiment Design B X ( A ) P X ( A ) P a X ( A ) P a ( F ) k ( S ) k l ( T ) k l * ik ik 11 ik, jl 11 jl ik, jl 11 jl Perform 3 independent experiments with fixed A 11 using orthogonal phases of a 21 input A jl output B ik Im Re Im Re B X A P (0) ( F) ik ik 11 k 11, 11, 11 B X A P X A P A X A P A (1) ( F ) k ( S) k l (1) ( T ) k l (1)* ik ik ik jl jl ik jl jl 11, 11, 11 B X A Pk X A P A X A P A (2) ( F ) ( S) k l (2) ( T ) k l (2)* ik ik ik jl jl ik jl jl For output port i, output harmonic k; input port j, input harmonic l 148

149 Consider an amplifier measurement with a power sweep (Gain = 11.6; Output Power Scale=11.6 times Input Power Scale) LinMag The power sweeps match exactly the same at lower power, until compression occurs

150 Let s look at this in a Polar Plot Magnitude is represented by distance from center

151 Now turn on Port 4 power (a2) to create an Active Load Pull at 0 dbm input power

152 Now turn on Port 4 power (a2) to create an Active Increase the input power and watch B2 output power

153 B X-parameters Reduce to S-parameters ( A ) X ( A ) P X ( A ) A X ( A ) P A ( F ) ( S) ( T ) 2 * , , db 40 X / A ( F ) [ X ( A )] s ( F ) A 0 21 A X ( S ) 21,21 X ( T ) 21,21 X ( A ) s X ( S ) 21,21 11 A 0 22 ( A ) 0 ( T ) 21, A A 11 (dbm) Reduces to (linear) S-parameters in the appropriate limit 153

154 Independent Phase Control Phase and amplitude control between internal sources Supports Active-Load, Differential Measurements, Phase Sweeps

155 Doing Active Load Pull using Source from Port 4 rear panel J11 J10 J9 J8 J7 J4 J3 J2 J1 Source 1 OUT 1 OUT 2 Pulse modulator Source 2 (standard) OUT 1 OUT 2 Pulse modulator LO To receivers Pulse generators R1 R3 R4 R2 A C D B Test port 1 Test port 3 Test port 4 Test port 2 Going Beyond S-Parameters

156 Introduction to X-Parameters X-parameters PNA-X ADS (Plus software options) It s easy and relatively cheap to turn a 4-port PNA-X into an NVNA

157 Introduction to X-Parameters MXG Power sensor (for amplitude CAL) E-CAL (for vector CAL) Comb generator 1 (to give stable phase) PNA-X with NVNA opt Comb generator 2 (for phase calibration)

158 Introduction to X-Parameters With NVNA only, X-parameters are valid for mismatched conditions near center of Smith chart (near 50 ohms) With NVNA and tuners, X-parameters can also be measured in highly mismatched conditions Useful for high-power, multi-stage amplifiers, and power transistors that are designed to work far from a 50-ohm environment Load-dependent X-parameters includes magnitude and phase of all harmonics as functions of power, device bias, and load impedances Data can be immediately used in a nonlinear simulator as a large signal model for complex microwave circuit analysis and design

159 Introduction to X-Parameters 1) Calibrate PNA-X NVNA 2) Vary impedances/power /bias presented to DUT (active device, transistor) no change! 3) Measure Pout, Gain, Efficiency no change! 4) Import X-Parameter model into ADS and simulate circuits

160 Introduction to X-Parameters Architecture of modern VNAs: well suited for RF component test With a single connection, measure: Gain and delay Port matches Gain compression versus frequency Intermodulation distortion Noise figure Fast measurements High accuracy with advanced calibration methods X-parameters go beyond S-parameters Complete non-linear behavioral model Easily measured with VNA hardware and accessories

161 Introduction to X-Parameters X-Parameter Summary Non-Linear Generalization of S-Parameters S-Parameters are special case where distortion = 0 Frequency Grid Includes: All Harmonics All Intermodulations Baseband Parameters at Each Freq Parameters Between Each Freq Combination Data covers one Large Signal Operating Point

162 Introduction to X-Parameters X-Parameters can be generated from circuits inside of ADS X-Parameters can be measured using PNA-X with NVNA

163 YouTube videos on X-parameters: X-Parameters from circuits Part 1 of 4, Generating X-Parameter Models from Circuits: Part 2 of 4, Generating Load-Dependent X-Parameter Models: vid=fp54ll8c2rq&v=xspklbneedy Part 3 of 4, Generating Two-Tone X-Parameter Models: vid=xspklbneedy&v=x0fjmqzzejk Part 4 of 4, Using X-Parameter Models in ADS for Wireless Verification: vid=x0fjmqzzejk&v=jiti6plullo

164

165 Instructor Introduction Pulsed IV/RF and Compact Modeling Load Pull Amplifier Design in ADS Stability Analysis X-Parameters System-level Simulations

166 System-level schematic

167 System-level simulations

168 System-level simulations

169 System level simulations YouTube videos on system-level design using X-parameters: Using Analog/RF X-Parameter Models in System-Level Design: