1. Introduction

To satisfy ever-growing demands for high speed communication, wireless technologies and devices have become ubiquitous. Home networking application requires viable last mile technologies which can distribute such high data rates. According to new IEEE 802.11ad standard, radio frequency (RF) systems have moved to 60 GHz band. At these frequencies RF channels are spatially confined and have almost line of sight (LOS) propagation and their properties resemble an optical channel. For this reason, OWC emerges as suitable support since it provides a simpler alternative to RF systems. OWC offers abundance of unregulated bandwidth, immunity to electromagnetic interference and simplicity in design and installation. The challenges for Gbps wireless optical transmission are stated in [1]. Low receiver sensitivity emerges as the most important one. Receiver sensitivity determines the required transmission power levels which are limited by eye safety regulations. The low receiver sensitivity would naturally lead to the demand for a low path loss which is offered by highly directive transmitters and receivers. However, the demand for coverage must also be fulfilled. Non-directed optical links provide reasonable coverage, but in these configurations multipath propagation is possible. Multipath propagation results in inter-symbol-interference (ISI) which imposes major limitation to high data rates. Various configurations on system level have been investigated with the aim to overcome the abovementioned difficulties. Regarding the receiver architecture, development of angle diversity receivers (ADRs) considerably improves overall system performance by relaxing the opposed requirements for speed and coverage area. Both imaging [2, 3] and non-imaging ADRs [4] were mentioned in literature.

Use of a single-channel imaging receiver (SCIR) with self-orienting capability was proposed in [5]. A miniaturized positioning mechanism allows the photodetector to move along the focal plane of the lens searching for the best signal-to-noise ratio (SNR). Although only in the simulation stage, the SNR search algorithm together with narrow FOV of SCIR reduces the shot noise and influence of possible multipath propagation. One drawback was that when the signal link was completely blocked, the system had to search again for a connection. The demonstration of single-channel imaging receivers was done in [6], where a 2-axis actuator was used for positioning TIA with InGaAs photodiode. The authors of [6] used a 1550 nm optical fiber system as a transmitter with fixed beam width and proposed a steering mirror to change the direction of transmitter beam.

The single-channel versions of OWC receivers require rather large-diameter photodiodes in order to facilitate the tracking/alignment procedure. To improve sensitivity and reduce cost, an optoelectronic integrated circuit (OEIC) in standard Si-technology working as imaging receiver was presented in [7]. The transmitter had an adaptable beam and used beam steering with MEMS based mirror. The use of an aligned narrow beam and a receiver lens reduced the transmit power. The channel data rate was 3 Gbit/s and the transmission distance was 19 m. The OEIC was developed in BiCMOS technology with a PIN photodiode and reached a sensitivity of –24.3 dBm at 3 Gbit/s for the bit error ratio (BER) = 10⁻⁹. Similarly, in [8] an APD detector with large diameter was produced in high-voltage HV CMOS technology. The sensitivity of receiver reached –31.8 dBm at 1 Gbit/s for BER 10⁻⁹. Compared to [7], no optics was needed at the receiver side which resulted in a wide receiving angle. The system operated at 1 Gbit/s up to distances of 3.3 m.

The presented work uses a system setup that is different to those described in [7] and [8]. We use now a fixed transmitter beam with a better collimating optics so that longer distances can be achieved at a cost of higher demands for beam positioning precision. The main focus of this work will be the application for OWC of fully integrated APD receiver in 0.35 µm BiCMOS technology. The HV CMOS was previously used for integration of APD since APDs usually require high operating voltages. The used BiCMOS process ensures the proper operation of electrical circuitry down to a substrate potential of –35 V which is sufficient for low-breakdown-voltage APDs previously realized in the CMOS part of the same BiCMOS process [9]. The reduction in the necessary substrate voltage for BiCMOS process is a significant improvement compared to –65 V for HV CMOS [8, 10]. With APD as photodetector the sensitivity of the receiver is improved by 8 dB compared to [7]. This large difference in sensitivity is by far not only because of the higher data rate in [7]. The main reason is the high gain of APD receiver which improved the sensitivity by 7.6 dB compared to a PIN receiver at 2 Gbit/s [11]. Additionally, in [11] the comparison of the state of the art APD OEICs is provided where we see that using a bipolar technology also gives benefits compared to HV CMOS receiver used in [8] in terms of sensitivity and speed. Therefore, the integration of APD and electrical circuit on the same chip allowed highly sensitive receiver. Any kind of optic at receiver side is omitted relaxing the challenges for achieving highly sensitive compact receivers suitable for mass production for future OWC systems.

2. APD OEIC Receiver

A completely integrated optical receiver with avalanche photodiode was developed for OWC. The APD OEIC benefits from the intrinsic gain of the APD as well as the high gain and low noise of bipolar transistors. Compared to hybrid receivers, the integrated receiver gives many advantages such as: better immunity to electromagnetic disturbances, smaller parasitic capacitances, larger bandwidth, smaller size and cheaper mass production.

The 0.35 µm BiCMOS process was chosen because of its low doped epitaxial zone which results in thick absorption zone. The thick absorption zone allows an improvement of the quantum efficiency of the APD and lowers the photodiode’s capacitance. The APD with a diameter of 200 µm was used for the high-speed channel based on the results of previous related work [7, 8]. The structure of the APD is the same as in [9]; the cross section of APD together with the block diagram of the receiver is depicted in Fig. 1. The area between n + +/p-well serves as a multiplication zone while the absorption zone consists of the 12 µm thick epitaxial layer. The p-well is surrounded by an additional n-well ring to prevent edge breakdown. The highly doped n + + region represents cathode and the substrate, which can be biased down to –35 V, serves as an anode. As stated in [9], the epitaxial layer is completely depleted already for a substrate voltage of –18.4 V, while at –30 V the bandwidth of the APD is 1.15 GHz and its responsivity is 20.4 A/W (675 nm). The dependence of APD responsivity on wavelength can be found in [12].

 

Fig. 1 APD cross-section with OEIC receiver block diagram

Download Full Size | PDF

Besides the high-speed APD photodiode, four additional PIN photodiodes were placed for monitoring the difference in the illumination on each side of the APD, Fig. 1. Each of these outer PDs has an area of 0.023 mm². The output signal of the surrounding PIN photodiodes could be used for detecting the direction of the incoming optical beam and therefore facilitate its alignment procedure. For this purpose the signals from opposite situated PIN photodiodes were fed into TIAs optimized for high sensitivity and high transimpedance gain of 700 kΩ. The complete current consumption of the both differential TIAs attached to PIN photodiodes was 5 mA at 3.3 V.

The central APD serves for communication and it is followed by a high speed TIA optimized for low noise. The TIA has shunt-shunt feedback topology with common-emitter as the input stage. The dummy TIA and operational amplifier provide offset compensation and dc level for differential post-amplification. This topology of the receiver ensures immunity to power supply noise and cuts off the low frequency components of the optical signal such as ambient light. Total measured transimpedance for a single-ended output was 260 kΩ; the output of the receiver had a 50 Ω termination and a differential output swing of 1.1 V. The sensitivity of the APD OEIC was –32.2 dBm for BER below 10⁻⁹ at 2 Gbit/s. As can be seen in Fig. 2, the bandwidth (BW) of the receiver at –21 V reverse voltage is around 800 MHz, which is the limiting factor for achieving higher speeds. The –21 V voltage was chosen through exhaustive measurement trials in order to achieve the best sensitivity of APD OEIC, since the choice of the reverse voltage is a compromise between multiplication noise and bandwidth [11].

 

Fig. 2 Receiver frequency response at reverse voltage of –21 V

Download Full Size | PDF

3. Transmitter

The transmitter consists of a driver, laser diode and an optical system as depicted in Fig. 3(a). For transmitter PCB design, commercially available components where chosen. A high-speed laser driver MAX3740A was used to modulate a 680 nm single-mode (SM) vertical cavity surface emitting laser (VCSEL). The SM VCSEL offers a highly Gaussian beam shape and a low beam divergence of 7° (full width half maximum, FWHM). The transmitter’s beam was further collimated using a collimating lens from Roithner (MPG – 95), with a focal length of 9.5 mm and a numerical aperture NA = 0.29. The collimator was placed inside of laser diode optic mounting tube LDMT-56-10 and glued onto the PCB together with the VCSEL. Using this highly collimating optics together with the SM VCSEL resulted in a beam divergence of 0.5 mrad. The average received optical power depends strongly on this angle and the distance between transmitter and receiver. The transmitter PCB was mounted together with a gimbal-less two-axis scanning MEMS micro-mirror from Mirrorcle (S1911) as shown in Fig. 3(b). This construction allows positioning of the beam by steering it with a MEMS mirror by 21.6° in x- and y- direction each. More on the MEMS mirror and its controlling can be found in [7].

 

Fig. 3 a) Laser diode and collimating optical system b) transmitter PCB mounted together with the MEMS beam steering mirror

Download Full Size | PDF

4. OWC system performance verification

The suitability of APD and OEIC performance for optical wireless communication was verified through a series of measurements. First, the range for error free data (BER < 10⁻⁹) was determined by measuring receiver’s BER as the distance between transmitter and receiver was varied. Secondly, the receiving angle was determined by measuring BER with respect to the incidence angle of the optical beam onto the receiver’s surface. For the case of normal incidence when the received signal power is highest, the immunity to background light was verified by measuring BER versus background illuminance. The block diagram of measurement setup can be found in [13]. The measurement procedures are described in the following sections.

4.1 BER vs. distance

The system setup was placed within our lab facilities with the usual background lighting of 500 lux. The transmitter was connected to a bit pattern generator and modulated with 2 Gbit/s applying a pseudo random binary sequence (PRBS) of 2³¹–1. The emitted average laser power was adjusted to 0.85 mW and the extinction ratio (ER) was set to 8. One output of the receiver was connected to a Tektronix TDS6124C oscilloscope to record eye diagrams. The other output was connected to a bit error analyzer. Without any optics at the receiver side, we were able to reach a 2 Gbit/s data transfer over a distance of up to 6.5 m with a BER lower than 10⁻⁹. Compared to [8], the maximum achievable transmission distance increased by a factor of 2 for a twice as large data rate. The diameter of a focused spot at a distance of 6.5 m was 6 mm, which results in a power density of 27.89 µW/mm² at that distance. This power density multiplied with the high-speed photodiode’s area (0.0314 mm²) gives the received optical power of –30.57 dBm. This value agrees quite well with measured OEIC sensitivity of –32.2 dBm. As the distance between the transmitter and receiver is increased, the beam diameter expands, power density decreases and as a result BER increases. In Fig. 4. BER vs. distance curve is shown also for 1 Gbit/s where 11 m of error free distance is achieved. The sensitivity of OEIC at 1 Gbit/s was –35.5 dBm for BER of 10⁻⁹ [11].

 

Fig. 4 BER versus distance between transmitter and receiver at 1 Gbit/s and 2 Gbit/s

Download Full Size | PDF

4.2 BER vs. angle

For this measurement, receiver and transmitter were placed at a distance of 6 m in order to obtain BER < 10⁻⁹. The receiver was mounted on a rotatable platform while the transmitter remained in a fixed position. The transmitter was again modulated with a 2 Gbit/s PRBS of 2³¹–1. By rotating the receiver for an angle α between the receiver’s surface normal and the direction of the incident light, the collected optical power decreases since effective signal collection area A_coll and physical area of detector A_d are related by A_coll = A_d cos(α). The BER curve, Fig. 5, depends strongly on this angular power penalty (P_in ~cos(α)), but there are also additional angle dependent oscillation effects due to optical interference in oxide and passivation layers on top of the APD. For angles smaller than 11°, we can obtain BER < 10⁻⁹, so the respective receiving angle equals 22°. Compared to [7], we more than doubled the receiving angle by omitting any kind of collecting optical system at the receiver side due to implementing an APD.

 

Fig. 5 BER versus incidence angle of the transmitted beam onto the receiver at a data rate of 2 Gbit/s at distance of 6 m

Download Full Size | PDF

4.3 BER vs. background light

For the experiment, a cold-light source (Euromex) was used as a source of ambient light. It was placed at the receiver side and directed in parallel with the transmitter’s beam onto the receiver. Transmitter and receiver where again placed at a distance of 6 m where the BER was well below 10⁻⁹. The intensity of background light was varied and the illuminance was recorded with a luxmeter (Testo 545). The average power of this background radiation generates shot noise in the detector. The additional shot noise lowers the BER of the receiver. From Fig. 6 we can see that error free data transmission (BER < 10⁻⁹) can be obtained even up to 6000 lux of background light illuminance. This is a promising result for indoor OWC, since common indoor light levels are in the range of 300 to 500 lux.

 

Fig. 6 BER versus background illuminance at distance of 6 m

Download Full Size | PDF

5. Conclusion

Relying on ground work described in [7, 8] an OWC demonstration system was developed which uses a highly collimated steerable transmitter beam. Another main focus of the work was in the receiver domain, and for this purpose a monolithically integrated optoelectronic receiver with 200 μm diameter APD was developed in standard 0.35 µm BiCMOS technology. To the best of the authors’ knowledge, this is the most sensitive large-diameter APD OEIC receiver operating at 2 Gbit/s. The performance gains were demonstrated through a series of measurement tests. Substantial progress has been made in terms of speed, transmission distance, receiving angle and background light immunity.