|
Flicker Noise in Orbit 2um N-well CMOS
|
Flicker Noise in Orbit 2um N-well CMOSResulting HSPICE BSIM model file
Introduction:The SPICE parameters provided by MOSIS for the Orbit 2um N-well CMOS process do not include noise parameters, so the HSPICE default of no flicker noise is used if the MOSIS model is dropped into one's SPICE deck. This makes designing a low noise amplifier (LNA) for use at low frequencies difficult. To determine the flicker noise in this process, a LNA was first designed without knowledge of the flicker noise contribution. The LNA was designed for open loop operation with a voltage gain of 10 to amplify signals from a piezoresistive sensor. The amplifier and sensor were fabricated on the October 25, 1995 run, named N5AK. The differential output was fed into a Burr-Brown INA103 Instrumentation Amplifier that has an input voltage noise of 2nV/sqrt(Hz) at 10 Hz, operating with a gain of 100. Measurements:The measurements were made with a HP 3588A Spectrum Analyzer using its built-in noise measurement function and a resolution bandwidth of 1.1 Hz. The circuits were constructed on a perf board with a copper ground using point-to-point wiring and soldered junctions. The board was enclosed in a oxygen-free copper box for thermal stability and to reduce EM interference. Connections were made through panel-mount BNC connectors and RG-58/U coaxial cables. Simulation:The LNA was simulated in HSPICE using the PMOS and NMOS models available from MOSIS for this run with the flicker noise terms added and the INA103 was modelled with the INA103E.MOD macro model available from Burr-Brown. The S-Edit schematic for the simulated circuit is shown below. A noise analysis was performed and the noise parameters were adjusted until there was a reasonable match between the simulation results and the measured data. The match results in simulated results with less than 12% mismatch from the measured data from 10 Hz to 200 kHz.
Comments:The test circuit only contained PMOS transistors, so it is not known how well these values will work for NMOS transistors. Furthermore, additional devices and runs need to be measured to determine the variance in these parameters across a lot and between runs; however, these numbers will give much closer results than the default of zero flicker noise.
If you have comments or suggestions, email me at warneke@eecs.berkeley.edu This page created with Netscape Navigator Gold |
|
Brett Warneke - last edited July 25, 2000 |