Some Design Aspects on RF CMOS LNAs and Mixers

Stemnan A.K.

Tins thesis deal with two parts of a radio receiver. The low noise amplifier (LNA) and the mixer.A timed RP front-end amplifier in CMOS with an operating frequency around 1.8 to 2GHz has been investigated. A symmetric design has been used to improve linearity with preserved power consumption. In this case it has been a parallel connection of complementary devices. The symmetric solution allows the use of inductor pairs in transformer configuration to improve the gain of the circuit.The input impedance has been controlled by an inductive series feedback of a transconduct-ance. An extended investigation has been made to determine the effect of the capacitive shunt feedback, traditionally ignored, as well as the effects of other internal parasitics. Simulation results showrs that for a required input resistance and operating frequency a maximum trans conductance of the transistor can be found. A new approximation of the input resistance where found, valid for low frequencies and for small transconductances.The most important part of the amplifier is the first stage (the inverter) as it will affect both input impedance and noise. Analysis of the optimum noise figure both with the capacitive shunt feedback excluded and included are made to show that it is possible to repeat the calculations in the same manner both times and thereby being able to compare the results.The major conclusion of this paper is that the shunt feedback can not be ignored and that good design methods therefore require a more extended analysis.Imbalances in quadrature demodulators, based on a passive CMOS mixer with capacitive load, has been studied with respect to process spread.A mathematical model of the mixer has been presented which explains the statistical spread m the mixer. The mixer transfer function has been divided into a multiplier, an amplifier and a time-variant filter, wrhere the statistical spread will affect the filtering part of the mixerThe filter is tinie-invariant if the time constant is large enough compared with the local oscillator frequency. A first approximation of the mixer filter is therefore that it is time-invariant which simplifies the statistical analysis. The model does not require tune-consuming transient analysis, as it possible to use the time average of the transfer function, enabling the use of AC simulation instead.The time-invariant model gives a maximum of the minimum image rejection. The time-variance increases the spread and reduces the image rejection.The mam conclusion of the investigation is that the time average of the transfer function can be used as the small signal model for the mixer. The model can be used to analyse and simulate the statistical spread and thereby the imbalance and minimum image rejection of a quadrature demodulator.