< Back to previous page


CMOS Millimeter Wave Circuits for Dielectric Waveguide Communication

The capacity demand for data communication keeps on increasing for applications ranging from small hand-held electronic devices to enormous data centers. With 5G on the horizon, the common trend for high-speed wireless communication is to shift the operating frequency towards the mm-Wave spectrum (30-300 GHz) and exploit the highly available bandwidth. The continuous improvement of bulk CMOS and packaging techniques drives this shift in operating frequency, enabling the full integration of mm-Wave radios. On the other hand, high-speed wireline communication is introducing PAM-4 modulation and power-hungry equalization techniques for short distance 0m to 10m electrical links. The required cables, however, are complex, expensive and prone to electromagnetic interference. Optical communication benefits from low-loss glass optical fibers with enormous bandwidths available to cover transatlantic distances. This technology, however, requires micrometer alignment precision and costly optical-electrical conversion modules. Therefore, optical communication is less suitable for shorter distances. Dielectric waveguides have proved themselves in latest publications to be able to exploit the mm-Wave spectrum and provide robust communication for distances up to 20 m.

This communication technology relies on a dielectric fiber that guides the electromagnetic energy from transmitter to receiver. According to the frequency, the energy will be more confined within the core of the fiber or propagate through the material surrounding the core. The surrounding electromagnetic wave can be disturbed and, therefore, a cladding is essential. With increasing frequency, the cross section of the fiber core can scale down and becomes acceptable within the mm-Wave range. Combined with coupling structures and bulk CMOS mm-Wave radios, a full communication link is realized. This thesis presents the limitations involved with dielectric waveguide communication and the challenges for mm-Wave radios in nanoscale bulk CMOS technologies.

The different limitation mechanisms of dielectric waveguides are distinguished with a Matlab model. Channel measurements are combined with ideal transceivers where different modulation schemes can be evaluated. Besides the typical noise limitation, two distortion effects are discussed. The first distortion effect is related to a frequency dependent group delay, which causes pulse broadening and deteriorates the performance at high data rates. A first order approximation of this limitation is discussed and confirmed with link measurements. The second distortion effect is related to the frequency dependent magnitude of the channel, which disturbs the symmetry of the power spectral density and deteriorates performance over longer distances. Frequency shift keying turns out to be the most promising modulation scheme for dielectric waveguide communication in the mm-Wave spectrum.

The implementation of mm-Wave radios is discussed with the focus on frequency shift keying modulation for communication 100 GHz. Center frequency deviations due to process, voltage and temperature variation can severely affect the link performance. Therefore, a low-frequency feedback loop is introduced to track such deviations and feed this information back to a phase shifter. An active phase shifter is implemented at 120 GHz in 28nm bulk CMOS and shows state-of-the-art performance. Furthermore, a novel frequency shift keying receiver topology is proposed, designed and measured at 140 GHz in a 28nm bulk CMOS technology. Combined with a transmitter, a dielectric waveguide link is realized with a foam-cladded fiber and broadband CMOS-to-DWG transitions. Link measurements show state-of-the-art performance with the possibility of higher order modulation to increase the data rate.

Date:2 Jun 2014  →  27 May 2019
Keywords:CMOS, mm-wave circuits, polymer fibres
Disciplines:Nanotechnology, Design theories and methods
Project type:PhD project