Detectors employing superconducting microwave kinetic inductance detectors (MKIDs) can be read out by
measuring changes in either the resonator frequency or dissipation. We will discuss the pros and cons of both
methods, in particular, the readout method strategies being explored for the Multiwavelength Sub/millimeter
Inductance Camera (MUSIC) to be commissioned at the CSO in 2010. As predicted theoretically and observed
experimentally, the frequency responsivity is larger than the dissipation responsivity, by a factor of 2-4 under
typical conditions. In the absence of any other noise contributions, it should be easier to overcome amplifier
noise by simply using frequency readout. The resonators, however, exhibit excess frequency noise which has been
ascribed to a surface distribution of two-level fluctuators sensitive to specific device geometries and fabrication
techniques. Impressive dark noise performance has been achieved using modified resonator geometries employing
interdigitated capacitors (IDCs). To date, our noise measurement and modeling efforts have assumed an onresonance
readout, with the carrier power set well below the nonlinear regime. Several experimental indicators
suggested to us that the optimal readout technique may in fact require a higher readout power, with the carrier
tuned somewhat off resonance, and that a careful systematic study of the optimal readout conditions was needed.
We will present the results of such a study, and discuss the optimum readout conditions as well as the performance
that can be achieved relative to BLIP.