Power amplifier design and load pull

Transcription

1 Power amplifier design and load pull Designing for RF Performance with Load-Pull Characterized Components Remi Tuijtelaars CTO BSW Test Systems and Consulting Herman Westra Technical Consultant Agilent Technologies Gustaaf Sutorius Application Engineer Agilent Technologies 1

2 Agenda : Introduction: Why is load-pull important : Review of non-linear RF device models : Collecting measurement data: Setup Part : Lunch : Collecting measurement data: Setup Part : Design, simulation and optimization with loadpull data-based models : Circuit demonstrator and live measurements : Closure & Drinks 2

3 Design, simulation and optimization with load-pull data-based models 3

4 The Load-Pull concept Characterize performance of a device resulting from applying varying load impedances with the objective to find the load that provides the 'optimum' performance. 4

5 The Load-Pull concept (cont'd) Traditionally measured at a single frequency for use with narrow-band designs Typical performances are Pout, Efficiency (PAE), IMD3, IP3, P1dB, Supply current, ACPR, etc. Results presented as contours of equal-valued performances Extra independents: Frequency, Vbias, Temp, Band, etc. LP data is terrain data: Performance = f(x,y) (RI or MA) LP data is often irregular ("scattered") in x and y. 5

6 The Load-Pull concept (cont'd) Typical contour plot resulting from Load-Pull measurements 6

7 Typical ADS Load-Pull use cases 1. Have measured load pull data use to design and verify matching networks 2. Have a nonlinear device model (A compact model or X- Parameters) use "LP-measurement-like" simulations to determine optimal load impedances 7

8 Use case 1a: You have measured Load-Pull data And you want to: a) See load pull performance contours b) Design/Optimize an output matching network 8

9 Start from Load Pull DesignGuide or Tutorial\Using_Meas_Load_Pull_Data example 9

10 Load pull example simulation with measured data The new DataBasedLoadPull controller reads-in measured Load-Pull data, (re-)grids it, runs an S-parameter simulation and outputs Performance data (e.g. Pout, Eff) to the dataset dependent on the load sensed at its "Load" terminal. Works with Maury *.lp, *.lp2, *.lp3, and *.spl files 10

11 Load pull example with measured data The Load_Tuner presents a series of impedances inside an area of the Smith chart 11

12 The Load_Tuner contains a two-dimensional (gridded) sweep for the real and imaginary value of the presented impedance 12

13 The LPDB controller's independent variables and performance parameters show LP data file content 13

14 The independent variables and performance parameters come from the LP data file Example file: Maury_Loadpull_data.lp 14

15 LPDB controller's analysis parameters The embedded S-parameter simulation can be set to simulate broader range of frequencies than available in the LP data file. In the example the single value (2 GHz) from the LP file is used. 15

16 Example plot of performance contours of interest Frequency dependent, but with sweep length of 1 (index=0) Best performance will be for load in this area 16

17 Example plots are based on equations that are on separate page. Specifically contour_ex() function: Contour_ex(). Automatically generates contours at round numbers at specified levels down from max. value or up from min. value. "POLAR" if to be plotted on Polar plot. "RECT" for rectangular plots. "RI" if input data is complex, "MA" if input data is Magnitude/Angle. 17

18 Use case 1b: You have measured Load-Pull data And you want to: a) See load pull performance contours b) Design/Optimize an output matching network 18

19 Example: simulate matching network over range of C and L values. Term2 allows for simulation of S21 of matching network. 19

20 Extent DBLP controller frequency sweep to simulate S21 values outside LP frequency range 20

21 Example simulation results for S21, Pout and Efficiency as function of C and L values 21

22 22

23 Use optimization to find best component values for 'optimum' match. An example. 23

24 Concurrent goals for S21, Pout and Efficiency 24

25 Special optimization setting in DBLP controller creates additional 'penalty' goal 25

26 After 250 iterations performance 'optimum' reached (although not all goals met.) 26

27 Optimized component values and corresponding reflection coefficient 27

28 Limitations of the DBLP Controller Linear (SP) simulation controller (can't be used in HB or CE simulation set-ups) Maury models only DBLP controller is not modifiable (can't open the underlying subcircuit) 28

29 Typical ADS Load-Pull use cases 1. Have measured load pull data use to design and verify matching networks 2. Have a nonlinear device model (including X-Parameters) use "LP-measurement-like" simulations to determine optimal load impedances 29

30 Start from Load Pull DesignGuide 30

31 Use of instrument subcircuits simplifies setup Most parameters are passed to tuner inside instrument subcircuit More easily see settings of all parameters, at a glance. Quickly switch from fundamental to harmonic load pull, while using same schematic. 31

32 Load pull instrument subcircuit May modify, if desired 32

33 Load pull tuner component Generates reflection coefficients ( s) within arbitrary circle All loads at harmonics are specified independently, as either Zs or s May sweep load at fund., 2 nd or 3 rd harmonic Only in Load Pull DG and Using_Meas_Load_Pull_Data example. If =1, use S_Load_*; If =0, use Z_Load_* May specify arbitrary Z0 enables finer sampling near edge of Smith chart 33

34 Using a non-50 Ohm reference impedance S_Load_Center_Fund= 0.6*exp(j*0.85*pi) S_Load_Radius=0.3 S_Load_Center_Fund=0 S_Load_Radius=0.6 34

35 35

36 The need for an alternative data-based LP simulation technique Problem statement - Need a simulatable component with performance based on measured LP data - The LP data files are in some non-maury format. - Need a true output signal (voltage and current) for use in lineup simulations including other non-linear devices (e.g. an electronically controlled matching network) 36

37 MGA Datasheet only provides numbers for P1dB, Gain and OIP3 when terminated with 50 Ohm. Device is characterized at different loads with a custom (non-maury) LP set-up 37

38 Solution: build our own component based on two special ADS components: FDD and DAC. 38

39 How does it work? The FDD ("Frequency Defined Device") senses the load impedance seen at its output terminal. The power associated with this load impedance is looked-up (using the sensed load values) from the MDIF formatted loadpull data DAC (Data Access Component). The FDD's output voltage is then set to deliver the looked-up power to the load. 39

40 Under the hood Z_sensed V / I 40

41 Under the hood Z_sensed V I 41

42 Under the hood Z_sensed V I 42

43 Under the hood Z_sensed V I 43

44 Requirements for the load-pull data to be usable by this model The load-pull data should be structured as a generic MDIF file. The independent variables in the load-pull data (typically the VSWR and phase of the load) should be monotonic and have the same range for all inner independents. Additional independent variables e.g. Frequency, Temperature, Vbias, Pin, etc. are supported as long as they are monotonic and have the same range for all inner independents. 45

45 Generic MDIF format Generic MDIF consists of blocks with formatted data. var1name, var2name, VarNName are the independents as well as bvar1name (the most inner -fastest changing- independent) bvar2name, bvar3name are the dependents. 46

46 Section of Generic MDIF file mga31189_lp_data.mdf 47

47 Example mga31189.lpd LP data files that needs reformatting and merging. Join files and make Frequency a VAR Not monotonous, needs re-ordering Break into blocks of equal Gamma values and variable Phase 48

48 File reformatted to generic MDIF (use a script or Excel and ascii text editor) 49

49 Use new LP_data_model to create contour plots Load tuner represents (harmonic) impedance on Smith chart. FDD requires HB simulator Sweeps arranged to create VSWR circles with load tuner. 50

50 Pout and OIP3 contours at 1900 MHz Pout optimum is quite different from OIP3 optimum. Discontinuity caused by extrapolation on sparse data around -180 degrees in LP file. 51

51 Long extrapolation required near -180 degrees due to limited number of data points. > > 52

52 Copy data from +180 degrees area to -180 area 53

53 Discontinuity removed by augmenting data in -180 degrees area. 54