Oscillators are ubiquitous to radio frequency circuits, where frequency translations and channel selection play a central role in the analogue communications channel. Oscillators also form part of digital systems as a time reference. Typical heterodyne receivers require an intermediate frequency channel. The associated oscillators and variable filters can only be centred perfectly at a single frequency, and degrade performance at the boundaries of the channel. These circuits also require image-rejecting filters and phase-locked loops in order to enable down-conversion. The penalties for these components are increased circuit area and power consumption. A direct down-conversion circuit will reduce the number of components in the system. A requirement added by the structural change is a passive sub-harmonic mixer. Quadrature oscillators may be achieved by cross-coupling two nominally identical LC differential voltage-controlled oscillators. Because of the widespread use of voltage-controlled oscillators in wireless communication systems, the development of comprehensive nonlinear analysis is pertinent in theory and applications. A key characteristic that defines the performance of an oscillator is the phase noise measurement. The voltage-controlled oscillator is also a key component in phase-locked loops, as it contributes to most of the out-of-band phase noise, as well as a significant portion of in-band noise. Current state-of-the-art modulation techniques, implemented at 60 GHz, such as quadrature amplitude modulation, and orthogonal frequency domain multiplexing, require phase noise specifications superior to 90 dBc/Hz at a 1 MHz offset. It has been shown that owing to the timing of the current injection, the Colpitts oscillator tends to outperform other oscillator structures in terms of phase noise performance. The Colpitts oscillator has a major flaw in that the start-up gain must be relatively high in comparison to the cross-coupled oscillator. The oscillation amplitude cannot be extended as in the cross-coupled case. The oscillator’s bias current generally limits the oscillation amplitude. The phase noise is defined by a stochastic differential equation, which can be used to predict the system’s phase noise performance. The characteristics of the oscillator can then be defined using the trajectory. The model projects the noise components of the oscillator onto the trajectory, and then translates the noise into the resulting phase and amplitude shift. The phase noise performance of an oscillator may be improved by altering the shape of the trajectory. The trajectory of the oscillator is separated into slow and fast transients. Improving the shape of the oscillator’s slow manifold may improve its phase noise performance, and improving the loaded quality factor of the tank circuit may be shown to directly improve upon close-in phase noise.
The approach followed describes oscillator behaviour from a circuit-level analysis. The derived equations do not have a closed form solution, but are reformulated using harmonic balance techniques to yield approximate solutions. The results from this closed form approximation are very close to both the numerical solutions of the differential equations, as well as the Simulation Program with Integrated Circuit Emphasis solutions for the same circuits. The derived equations are able to predict the amplitude and frequency in the single-phase example accurately, and are extended to provide a numerical platform for defining the amplitude and frequency of a multiphase oscillator. The analysis identifies various circuit components that influence the oscillator’s phase noise performance. A circuit-level modification is then identified, enabling the decoupling of some of the factors and their interactions. This study demonstrates that the phase noise performance of a Colpitts oscillator may be significantly improved by making the proposed changes to the oscillator. The oscillator’s figure of merit is improved even further. When a given oscillator is set at its optimum phase noise level, the collector current will account for approximately 85% of the phase noise; with the approach in this work, the average collector current is reduced and phase noise performance is improved. The key focus of the work was to identify circuit level changes to an oscillator’s structure that could be improved or changed to achieve better phase noise performance. The objective was not to improve passive components, but rather to identify how the noise-to-phase noise transfer function could be improved. The work successfully determines what can be altered in an oscillator that will yield improved phase noise performance by altering the phase noise transfer function.
The concept is introduced on a differential oscillator and then extended to the multiphase oscillator. The impulse sensitivity function of the modified multiphase oscillator is improved by altering the typical feedback structure of the oscillator. The multiphase oscillator in this work is improved from -106 dBc/Hz to -113 dBc/Hz when considering the phase noise contribution from the tank circuits’ bias current alone. This is achieved by uniquely altering the feedback method of the oscillator. This change alters the noise-to-phase noise properties of the oscillator, reducing phase noise. The improvement in the phase noise does not account for further improvements the modification would incorporate in the oscillator’s limit cycle. For a given tank circuit, supply current and voltage, compared to an optimised Colpitts oscillator, the modifications to the feedback structure proposed in this work would further improve the figure of merit by 9 dB. This is not considering the change in the power consumption, which would yield a further improvement in the figure of merit by 7 dB. This is achieved by relaxing the required start-up current of the oscillator and effecting an improvement in the impulse sensitivity function. Future research could include further modelling of the phase shift in the feedback network, including the transmission lines in the feedback networks using the harmonic balance technique in a numerical form. The feedback technique can also be modified to be applicable to single and differential oscillators.
