1. NEWCOM WPR3 Meeting – Uppsala 21 09 05 RF & Microwave Electronics Group Power Amplifier for Wireless Links: System Level Models Daniel Bustos Marco Pirola Giovanni Ghione Simona Donati Dipartimento di Elettronica Politecnico di Torino Microwave & RF Electronics Group

2. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Outline Behavioral nonlinear model motivations; The implemented models: The cooperation between Polito and Chalmers Units; Simulatotion Tools; Model and extraction procedure description; Some simulation examples: (IM3, BER, …); A case study: a 812.11a WLAN Power amplifier: Model the PHEMT ATF-54143 Transistor; Circuit level and system level simulations; Model behaviour comparison (VSS VS IT++); Conclusions and future works;

3. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Envelope or base-band model I Oriented to high level (system) simulations (eye diagram, BER) rather than circuit ones; Identified through input/output system observation-> intrinsically behavioural (black-box); System level modelling is oriented to the prediction of the system envelope behaviour; System evaluation through RF simulation possible but: simulation circuit level far from the system level layer; system level model identification directly focused on system; system level simulation unbearably slow.

4. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Envelope or base-band model II x(t) y(t)

5. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Motivations I The models used to simulate the PA at circuit level are capable to represent nonlinearities with a high degree of accuracy Memory effects in principle included, although difficulties are related to the model extraction (from measured data or physics-based simulations) and to the simulation techniques used (HB); System level simulation unsuitable for fast simulation in presence of complex modulation scheme-; Classical system level models include nonlinearities in a too simplistic way -> lost of accuracy for complex modulation schemes; More sofisticated models: need complex identification procedures; increase the computation time; Find a model trade off between accuracy, simple model parameter identification, reasonable computation time

6. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Motivation II Chalmers unit develops and maintains a system level simulator (IT++) adopted within the NEWCOM network; IT++ is able to implement a complete communication links at system level; IT++ did not account for the effects of PA non idealities on the system level performances; Polito units has a consolidated activities on the PA field at circuit and system level (e.g. activities within TARGET network). Implement within IT++ the Polito NL models

7. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Circuit Level – System Level Link

8. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Simulation framework Circuit level simulations carried out with AWR MWOFFICE, used as the reference model virtual; System level identification through practicable non-linear experimental data Automated extraction implemented at the moment within Matlab; System level simulation: AWR VSS automatically extracted from MWO; Classical AM-AM AM-PM model implemented in IT++; Advanced model with memory implemented in IT++; Comparisons on a case study power amplifier for WLAN 802.11a application;

9. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 The implemented models U(k) Linear Part with Memory AM -AM AM - PM Static Nonlinear Part W(k)y(k) AM -AM AM - PM Static Nonlinear Part y(k) U(k) Classical AM-AM, AM-PM Advanced Wiener Scheme Model

10. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Memory Model Extraction The linear part of the model extracted from two tone excitation varying the tone spacing; FIR filter implemented through ARX approach; Static nonlinear part, extracted for single and two tones excitation as a function of tone power; Non linear static behaviour approximated with a suitable degree polynomial

11. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Case of study: A 802.11a WLAN Driver Amplifier using PHEMT ATF-54143 Transistor The frequency range includes USA U-NII lower band 5.125 – 5.250 GHz. MWO tools used to simulate a circuit implementation. For the nonlinear analysis, a harmonic – balanced (HB) simulation was used The Non-linear transistor model used in the simulation is based on the work of Curtice (Advanced Curtice2 Model)

12. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 MWO Diagram of Driver Amplifier

13. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Example: WLAN 802.11a Driver Amplifier Single Tone AM/AM at 5 GHz

14. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Single Tone AM/PM at 5 GHz

15. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Two-tone simulation at 0 dBm input power

16. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Two-tone simulation at 0 dBm input power

17. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Two-tone simulation at 0 dBm input power

18. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Two-tone simulation at 0 dBm input power

19. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 OFDM Mod. scheme: BER simulation at several PA compression levels, dB

20. RF & Microwave Electronics Group NEWCOM WPR3 Meeting – Uppsala 21 09 05 Conclusions Collaboration with Chalmers Univeristy through PhD student exchange (maybe to be renewed); Methodology for extraction of system oriented models from measured data or standard HB (multitone) circuit simulations; Model validation on virtual experimental data; Extractor implemented in Matlab; Model implemented within IT++; Future possible developments: model validation on true experimental non linear data; further model refinements and improvements (e.g. three box model) testing and implementation of other models (e.g. neural, volterra/wiener series appproach, …)