Load pull

Load-pull is the colloquial term applied to the process of systematically varying the impedance presented to a DUT, most often a transistor, to assess its performance and the associated conditions to deliver that performance in a network. While load-pull itself implies impedance variation at the load port, impedance can also be varied at any of the ports of the DUT, most often at the source.

Load-pull is necessary when superposition breaks down, which occurs, for example, under large-signal operating conditions associated with substantial harmonic generation or other forms of nonlinearity. The etymology of load-pull derives from oscillator characterization whereupon variation of the load impedance pulls the oscillation center frequency away from nominal. Source-pull is also used for noise characterization, which although linear, requires multiple impedances to be presented at the source to enable simultaneous solution of an over-determined system yielding the four standard noise parameters.

Load-pull is the most common method globally for RF and MW power amplifier (PA) design, transistor characterization, semiconductor process development, and ruggedness analysis. A central theme of load-pull is management of nonlinearity versus analysis of nonlinearity, the latter being the domain of advanced mathematics that often yields little physical insight to nonlinear phenomena and suffers from an inability to accurately render actual behavior embedded in a network with significant parasitic and distributed effects.

While there are in theory no physical limits on the frequency of which load-pull can be performed, most load-pull systems are based on passive distributed networks using either slab transmission lines or rectangular waveguide. Lumped tuners can be made for HF and VHF frequencies whereas active load-pull is ideal for on-wafer mm-wave environments with substantial loss between the tuner and DUT reference-plane.