Introduction

Device characterization is required for power amplifier design, and the ideal form of the device data is a large signal model. With a model, the performance can be analyzed for varying drive and impedance conditions, so complex or multi-stage circuits can be designed. A method of formulating a large signal model is to use a measurement-based behavioral approach, as with the PHD model. This is based on measurements of X-parameters, which are a superset of S-parameters for nonlinear components, and are measured using an NVNA (Non-linear Vector Network Analyzer).

Load Pull with X-Parameters

Load pull with NVNA measurements of X-parameters can be used directly by the PHD model over a wide impedance range. The operator of the combined load pull NVNA system can select an impedance range of interest, possibly over the entire Smith chart. The PHD model can then be used as a circuit element in a non-linear analysis with great confidence, since it is based on measurement at the actual operating conditions of the device. The load pull X-parameter measurement can include a complete sweep plan. Stimulus variables can include impedance, power drive, bias, and frequency, for example. This can extend the applicability of the PHD model over a much wider range of validity – over the range of actual applications for many high-power and multi-stage PA designs.

• The process has three steps:
  
  1) The load pull system measures the X-parameters at each impedance setting, like a standard load pull, with X-parameters added to the measurement data set. When the measurements are complete at all the impedances, the measured
  X-parameters are saved into a single file.
  
  2) An enhanced design kit available for use in the ADS non-linear simulator then reads the file saved by the load pull- NVNA system and creates a PHD component associated with the file. This is a very quick step.
  
  3) This component can then be dragged and dropped directly into a circuit schematic as a non-linear device, and analysis can start immediately.
  
  
  
  Comparison of simulated (blue) and independent measured (red) delivered power contours (left) and efficiency contours (right) from a typical packaged FET show extremely accurate agreement.