Bipolar Fiber Optical Receivers with SPADs

Common analogous optical receivers with PIN or avalanche photodiodes need a rather high optical input power - corresponding often to more than 1000 photons per bit because of electronic noise (shot noise, thermal noise) and excess noise (if avalanche photodiodes in the linear mode are used). Photon statistics however sets the physical limit (quantum limit) of only about 20 photons (in average). It was shown very recently with CMOS technologies that receivers with Single-Photon Avalanche Diodes (SPADs), which achieve amplification factors of more than a million, can eliminate electronic and excess noise. The gap to the quantum limit was reduced to 12.7dB, however, at only 20Mbit/s. At 100Mbit/s, the distance to the quantum limit increased to 21dB. It was also found that afterpulsing of integrated SPADs is the main limiting factor for approaching to the quantum limit and to higher data rates. This project therefore will investigate ways to reduce afterpulsing of integrated SPADs, using faster bipolar active quenching and gating circuits and using discrete SPADs, which use cleaner and optimized fabrication processes (silicon technologies are optimized for transistor and circuit performance). Bipolar transistors enable much faster and higher amplifying comparators for the quenching circuits and gaters for higher data rates than CMOS circuits. Additionally bipolar transistors match much better than MOS transistors and therefore allow the usage of much smaller offset voltages and in turn a lower detection threshold for the avalanche event than MOS comparators, which reduces the avalanche charge and afterpulsing. It is intended to achieve by a factor of 4 to 10 higher data rates with bipolar SPAD receivers than reported for CMOS receivers. Furthermore, the gap to the quantum limit shall be reduced below 10dB also above 100Mbit/s. In addition, for the first time, receivers for the telcom wavelengths 1.3 and 1.54m with discrete InGaAs SPADs will be investigated. All SPAD receivers will be verified in data transmission experiments via optical fibers by determining bit-error rates. For the first time, a model for the bit-error rate of gating SPAD receivers will be derived. Bipolar receiver circuits will be designed and fabricated as Application Specific Integrated Circuits (ASICs) to verify the innovative approaches towards digital SPAD optical receivers. Together with cheap high-data- rate LEDs, which possess a high extinction ratio, SPAD receivers will enable to reduce the optical input power compared to analogous optical receivers by possibly more than a factor of 100. Summarising, innovative bipolar SPAD receivers are investigated to verify a new generation of robust optical receivers and sensors with considerably improved data rate and sensitivity experimentally.

Transimpedance-amplifier based analog optical receivers need a relatively high optical input power due to thermal noise to reduce the bit error ratio below 10^-3. Often more than 10,000 photons are necessary in a bit - whereas the photon statistics sets the quantum limit to 7 photons. In modern electronics many analog circuit blocks, however, were replaced by digital signal processing. Single-photon avalanche diodes (SPADs) in optoelectronic integrated ICs allow because of their very high gain to realize a kind of digital optical receivers. SPADs possess, however, parasitic properties: thermal generation of charge carriers leads to dark counts, interaction with traps causes so-called afterpulsing and optical crosstalk occurs. These parasitics cause in one SPAD and with only one photon per bit a too high bit error ratio. Several detected photons per bit are necessary to reduce the bit error ratio sufficiently. Afterpulsing of integrated SPADs is the most important limiting factor for approaching the quantum limit and for achieving higher data rates. This project investigated therefore ways for reduction of afterpulsing of integrated SPADs: usage of faster bipolar active quenching and gating circuits. Bipolar transistors enable much faster comparators with higher gain for quenching and gating circuits for higher data rates than CMOS circuits. Furthermore, bipolar transistors match much better than MOS transistors and lead to smaller offset voltages and in turn to a lower detection threshold for the avalanche event than MOS comparators, whereby the avalanche charge and afterpulsing are reduced. For elimination of optical crosstalk, 4 separate bipolar SPAD receiver chips were investigated by using an optical 1-to-4 fiber splitter. Receivers with 169 SPADs and active quenchers with bipolar transistors were designed for use with a diffractive optical element to increase the data rate over that of CMOS receivers. Quenchers were characterized and a SPAD receiver was verified by determination of the bit error ratio in data transfer experiments with an optical fiber. In addition, for the first time a model for the bit error ratio of gated SPAD receivers was developed. It turned out, however, that due to a rather high power consumption not all MOS transistors in SPAD receivers can be replaced by bipolar transistors. Bipolar receiver circuits were designed and fabricated as application specific integrated circuits (ASICs). Summarizing, innovative bipolar SPAD receivers were investigated to verify a new generation of robust optical receivers and sensors with considerably improved data rate and sensitivity experimentally.