Load Pull

Background

Characterisation of non-linear devices is not generally possible using a stand-alone network analyser.  There are a number of issues which prevent this from being possible.  Some of the more important ones are:

  • Power level
 The power level applied during the test must be appropriate.  In general, measurements must be performed at the power level at which the device will eventually operate.  For most non-linear devices this will be beyond the capability of a network analyser.
  • Test ports & biasing
 While measuring the input plane of a device, a network analyser applies the test signal to the input plane of the device.  In general, while operating in a non-linear region, this will result in the bias conditions of the device being influenced to some extent by the presence of the test signal.
While measuring the output plane of a device, a network analyser applies the test signal to the output plane of the device.  In general, this will not result in operation of the device at appropriate bias conditions.  The correct conditions may exist only when the device is driven by a signal at the input plane. 
  • Simultaneous matching
 The impedance presented to the output plane of a device will generally affect the impedance measured at the input plane.  While this is true (to some extent) of all devices, it is particularly important for non-linear devices where subsequent linear analysis of s-parameter data cannot be used to determine the simultaneous matching condition.

Techniques

The technique most commonly applied to overcome the above problems is known as Load Pull.  This technique places adjustable tuners at the device under test (DUT), which is then operated at bias and power levels resembling the intended final application.  The results obtained from a successful load pull test will be the impedance values at which the device operates with optimum performance.  There are essentially two forms of load pull analysis, single-point or multi-point:

  • Single-point
 This is the most simple form of load pull.  The tuners are adjusted to give optimum performance while monitoring parameters important to the intended application, e.g. power output, efficiency etc.  The single point analysis may be repeated at a number of test frequencies.
  • Multi-point
 This is a more rigorous approach, usually requiring automated or semi-automated techniques.  The tuners are adjusted across a range of impedance values while recording performance parameters at each point.  The data which is collected is used to plot contours of performance.  These contours reveal not only the optimum impedance value, but also the sensitivity of the performance to variations of impedance.  This allows for tolerance and statistical analysis, and for the option to trade performance parameters e.g. gain and linearity.

For both single-point and multi-point methods, a test configuration is required which allows the impedance presented to the device to be determined and recorded.  There are many different ways to do this, but all will involve the use of a vector network analyser:

  • Measure once
    (manual)
 Occasionally, a single measurement point is required.  In this case, following the load-pull tuner optimisation, the tuner positions are fixed and the DUT may be removed from its test fixture.  A network analyser is then used to directly probe the DUT mounting position to determine the impedance presented to the device.  This approach, though not very elegant, has the advantage of being very fast where a single measurement point is adequate.
  • Per-point
    (semi
    -automated)
 This method will place a high quality RF switch between the tuner and the DUT.  For each new setting of the tuner, the switch is used to toggle the tuner from the DUT to the network analyser.  It is necessary to understand the effect of any cables, adaptors or other devices that have been placed between the switch and the DUT.  Generally this requires a one-off measurement of such components, which are subsequently de-embedded from the load pull data.  This method provides semi-automated data collection for multi-point analysis.  Data collected in this way may be used to plot performance contours.
  • Pre-calibrate
    (fully
    -automated)
 This is the "Rolls-Royce" method, requiring automated tuners and complex software to control the data collection process.  Commercial systems are available at considerable expense from Focus Microwaves and Maury Microwave.
A rigorous calibration of the tuners and other components within the test system is followed by an automated sweep of pre-selected impedance points, while measurements of device performance are obtained.

An Example

This photograph shows a semi-automated load-pull configuration in action, characterising a GaAs power device at 2.4GHz.

The network analyser (port 2) is used as the test signal source (via the Litton power amplifier at right) and as the means to measure (port1) the actual impedance presented to the DUT (mounted behind the fan to the left).

A triple-stub tuner is used to match the input while a probe-tuner is used at the output.  The high-power attenuator and power meter head are at the extreme right.