Distortion losses of high speed single-photon avalanche diode receivers approaching quantum sensitivity

Abstract

Data traffic is growing exponentially, and the radio frequency (RF) spectrum is under pressure to meet these demands. Visible light communication (VLC) has hundreds of terahertz of unused and unregulated bandwidth and the widespread use of solid-state lighting makes it viable to supplement RF networks. Present optical receivers (RXs) use positive-intrinsic-negative (PIN) diodes or avalanche photodiodes (APDs) and amplification circuitry that impairs RX sensitivity. In this work, the extremely high gain of single-photon avalanche diodes (SPADs) is utilised to remove the need for an amplifier. This offers significantly improved sensitivity and allows the quantum limit of detection to be approached. A SPAD array integrated in 40 nm CMOS is used to determine the transient response of SPADs and investigate the effect of dead time after a photon is detected. A 130 nm CMOS SPAD array RX in this work achieves 500 Mb/s four-level pulse amplitude modulation and 350 Mb/s OFDM in a 450 nm laser diode-based VLC link within 15.2 dB of the quantum limit. However, SPAD dead time induces around 5.7 dB of transient distortion which restricts error performance and data rate an order of magnitude below that of APDs. This thesis builds a model of a discrete photon counting system which exhibits this nonlinear behaviour and compares it to practical measurements with the RX. A unipolar intensity-modulated optical signal is considered, as opposed to bipolar electric fields in conventional RF systems. Intermodulation is analysed, and the resulting degradation of signal-to-noise-and-distortion ratio and bit error rate is evaluated. The model is a tool for understanding distortion to ultimately allow rectification through RX architecture, modulation scheme, coding, and equalisation techniques. The thesis concludes that the SPAD RX is effective with very low optical power, allowing considerable improvements of two orders of magnitude in transmitter energy efficiency or one order of magnitude in link distance compared to present VLC systems – useful for underwater applications. This work proves that the high electrical power consumption disadvantage due to the SPAD bias can be alleviated by operating the RX in an optimum region determined in the model. Further savings and integration advantages are gained by using CMOS. This SPAD RX demonstrates the lowest power consumption and highest sensitivity to date. The need for narrow bandpass spectral filtering in bright ambient light conditions remains a limitation of the SPAD RX.