The experimental performance of an end excited, internally tuned, multipolar electron cyclotron resonance plasma source is investigated versus several independent input variables: (1) incident power (185-350 W), (2) discharge pressure (1-7 mTorr), and (3) microwave matching (applicator length, L, varies from 65-170 mm). Plasma source performance in argon gas is determined from measurements of (1) absorbed microwave power, (2) the magnitude, and (3) the spatial variation of the impressed electric field, and the output (4) plasma density and (5) electron temperature. The nonlinear experimental relationships between variations in the input variables and performance variables are established. Two well-matched stable discharge operational regions or modes are identified as L is varied. Using the measured impressed electric field patterns, these two modes are identified as TM, Φ symmetric, λ, and λ/2 standing wave modes. As L, pressure and power vary, considerable hysteresis in absorbed power and density is observed. However, it is shown that this complex input/output discharge behavior can be approximately explained using simple global plasma models. Despite the multiple steady states and hystereses, it is shown that the discharge can be easily maintained in well-matched, high density (n>3×10/cm) stable operating conditions. The impressed electric field varies between 3 and 4.5 kV/m. An equivalent circuit of the plasma loaded cavity applicator is developed and provides important insights for the plasma matching process. The microwave coupling efficiencies are greater than 98% and the ion production costs vary between 300 and 450 eV/ion versus the experimental inputs. The data presented is useful for plasma source optimization and in developing intelligent plasma processing control strategies. © 1997 American Vacuum Society.