The output conductance , respectively the channel resistance are subject to a number of influences that must be carefully separated. Special care has to be taken when comparing DC- and RF- . DC- values obtained from the output characteristics can appear to be negative. A first effect to be separated is the effect of self-heating and the temperature dependence of the parasitic resistances, especially the gate resistance . Self-heating leads to a rise of and thus to a reduction of the internally applied bias, so self-heating causes an internal bias shift . A second effect is the influence of carrier generation/recombination. As reported especially for GaN HEMTs , but also for pseudomorphic HEMTs , the occurrence of traps leads to a hysteresis in the DC-output characteristics, an effect, which is not seen in RF-parameters. The RF-value of extracted of measured data is therefore always positive.
In a compact current model extending one-dimensional charge models, a saturated velocity model leads to a constant DC-output conductance as a function of in the so-called saturation region . Introducing first order compact approaches  for impact ionization leads to an exponential increase of the drain current as a function of bias.