Author: Luís Botelho Ribeiro

1. Introduction

Modern technology offers high-speed moderate-cost components for indoor optical communications. This means that within the next few years, indoor communication networks, e. g. local computer networks, are likely to pass through revolutionary transformation. This paper reports an experimental wireless optical network realisation and modelling, constituted by a few point-to point links.

The network was first designed to transport analogue voice conversations but is intended to accommodate digital traffic in future upgrades. In parallel, analytic models are being developed to estimate the system bit-error rate performance as a function of operational parameters like the link distances and the digital baud rate.

2. Experimental set-up

A ring network, as illustrated in Figure 1, was implemented. A total of nine nodes has been interconnected through point-to-point links using laser transmitters and PIN-detectors at the receivers. An addressing keypad is used to route the voice channel to the desired destination node. Over the voice channel a data link can then easily be established through a pair of modems, one at each end.



Figure 1 - Ring network architecture with nine nodes. Node features include voice communication, addressing and possibility for (low-baud) digital data transmission.

The architecture at each node involves a number of functional units, as depicted in Figure 2. The optical signal is detected and amplified in a moderate sensitivity receiver, followed by a processing stage including some filtering and demodulation facilities. The digital outputs are then fed into temporisation and digital processing circuits.



Figure 2 - Block diagram of main control/communication functions in a node.

Audio and data communication drivers are connected to the digital processing unit. A serial output with frequency multiplexed data and control information is driven into a laser which converts it to a 670 nm collimated beam pointed to the next node. A feedback control system ensures that the laser power is kept at a constant and safe average level. This is an important feature as low-cost lasers don’t normally include thermal stabilisation hardware, such as a thermo-electric cooler.

The laser used as a transmitter was the Sony SLD1122VS with a maximum output power of 5 mW in the red emission band (l=670 nm). The detectors for the receiver were supplied by Siemens. The PIN detector SFH217 has a maximum responsivity of 0.62 A/W (quantum efficiency h=0.89 @ 850 nm and h=0.7 @ 670 nm). At the laser output, a spherical collimating lens was included for minimum beam aperture. The assembled boards were then put at different positions over the laboratory and communication tests have been successfully performed.

3. Evaluation of theoretical performance limits

An analytical model for the operation of the system was developed in such a way that simple estimates of performance could be assessed. Thus, limits in distance range as well as transmission speed could be calculated. The model, still in development, relies on some basic assumptions. The received beam is assumed to be completely subtended by the focusing optics and also these optics are believed to deliver all the power to the active area of the PIN photodiode [1].

Due to high absorption in the atmosphere, less significant effects on the transmitted pulses (under these circumstances) like dispersion and non-linear Kerr effect have been neglected. A FET high-impedance amplifier was considered as the first amplification stage at the optical front-end. Then, Gaussian statistics were assumed for the noise affecting the digital symbol decision circuit. As a consequence of considering low-distance and low-dispersion, intersymbol-interference and telegraphic distortion at the decision stage were also neglected.

The total optical power incident at the receiver for binary symbol iÎ {0,1} is given by

(1)

where P_tx is the optical power at the laser transmitter output, a is the air optical attenuation coefficient and L is the Tx-Rx distance of the particular link. The average electric current at the decision instant referred to the FET input is given by

(2)

A certain amount P_r of residual power due to sun light or artificial illumination is considered. R_s is the detector responsivity and i_d takes into account its dark current. The noise is mainly due to the electric receiver thermal effects and quantum noise, arising from the quantum nature of light (photons).

(3)

The first term is symbol-independent, but the second changes with the received power, thus with the symbol as amplitude modulation was used. The high-frequency noise is filtered by an electric low-pass filter with an equivalent bandwidth equal to the binary data rate r. The total thermal noise variance due to the FET and the polarisation resistor R_L, has variance [2]

(4)

where K_B is the Boltzmann constant, T_a is the operating temperature, g_m and G_FET are, respectively, the FET transconductance gain and the noise factor, q is the electron charge, C_in is the resulting input capacitance from the PIN and the gate of the FET and finally the current i₀ is the result of . For simplicity, a situation was considered where the Personick I_n integrals [3] are equal to 1.

Under Gaussian statistics, as stated above, the bit error rate (BER) is then easily calculated after the signal to noise ratio, according to the following expression [4],

(5)

The Q-integral function is well known and can, without significant loss of accuracy, be approximated by the following

(6)

In the calculations, a temperature of 300 K was considered. At the laser, a symbol one is codified by a 5 mW power level and a zero by 0.4 mW, to limit the laser chirp and maximise commutation speed. The optical power loss was 20 dB/km. The detector dark current is 3 nA and the residual optical power equals 0.2 mW. The electrical front-end is characterised by g_m=40 mS, G_FET=1.78, C_in=1 nF and R_L=50 W.

Under the conditions above, calculations were performed to study the BER dependence on the link distance, for fixed baud rate values between 40 Mb/s and 360 Mb/s. Results are shown on Figure 3.



Figure 3 - BER variation with transmission range, for baud rates of 40, 100, 140 and 360 Mb/s.

Referring to the standard performance limit of BER=10^-9, one may conclude that a 40 Mb/s link can be established for as much as 1000 m without reaching that limit. For rates of 100 Mb/s and 140 Mb/s, respectively, the maxim ranges are at approximately 700 m and 600 m. Less than 500 m can be covered by a 360 Mb/s system, as an error rate of 10^-2 is observed for this value of L.

Now looking at the maximum baud rates attainable for fixed distances, a last set of calculations was realised with results as shown in Figure 4.



Figure 4 - BER variation with baud rate for transmission ranges of 100, 200, 500 and 1000m.

A 100 m system is then capable of supporting data channels up to 650 Mb/s. As distance increases, the maximum speed is naturally reduced. Thus, for L=200 m the maximum speed is now a little more than 450 Mb/s. For 500 m the limit is 200 Mb/s and for 1000 m it will stand below 50 Mb/s.

4. Conclusions