Measurement and modeling of noise parameters for desensitized low noise amplifiers

It can be shown that devices with low noise resistance (R/sub n/) values can significantly relax the noise-input match trade-off in LNA design, resulting in desensitized wideband LNAs. In this paper, we show that the measurements of such devices are error-prone and can cause modeling and simulation inaccuracies. Using a CAD-oriented approach, an error-propagation analysis from measurements to a device-model is performed. We find that the errors in measurements affect the simulated values of R/sub n/, NF/sub min/ and B/sub opt/ the least and these parameters need to be calibrated well in any good device model. Values of G/sub opt/ are shown to propagate from measurements and errors are bigger for devices with a large transadmittance (/sub y21/). We suggest the use of scaled MOS models extracted from devices with high R/sub n/ to predict the values of G/sub opt/ for devices with low R/sub n/, using the other noise parameters as calibration points.