The receiver performance is characterized by measuring the BER as a function of the average optical power received. The average optical power corresponding to a BER of 10-9 is a measure of receiver sensitivity. The figure below shows the receiver sensitivity measured in various transmission experiments by sending a long sequence of pseudorandom bits (typical sequence length 215 - 1) over a single-mode fiber and then detecting it by using either a p-i-n or an APD receiver. The experiments were performed at the 1.3- or 1.55-μm wavelength, and the bit rate varied from 100 MHz to 10 GHz. The theoretical quantum limit at these two wavelengths is also shown in the figure.
A direct comparison shows that the measured receiver sensitivities are worse by 20 dB or more compared with the quantum limit. Most of the degradation is due to the thermal noise that is unavoidable at room temperature and generally dominates the shot noise. Some degradation is due to fiber dispersion, which leads to power penalties; sources of such penalties are discussed in another tutorial.
The dispersion-induced sensitivity degradation depends on both the bit rate B and the fiber length L and increases with BL. This is the reason why the sensitivity degradation from the quantum limit is larger (25-30 dB) for systems operating at high bit rates. The receiver sensitivity at 10 Gb/s is typically worse than -25 dBm. It can be improved by 5-6 dB by using APD receivers. In terms of the number photons/bit, APD receivers require nearly 1000 photons/bit compared with the quantum limit of 10 photons/bit. The receiver performance is generally better for shorter wavelengths in the region near 0.85μm, where silicon APDs can be used; they perform satisfactorily with about 400 photons/bit; an experiment in 1976 achieved a sensitivity of only 187 photons/bit. It is possible to improve the receiver sensitivity by using coding schemes. A sensitivity of 180 photons/bit was realized in a 1.55-μm system experiment after 305 km of transmission at 140 Mb/s.
It is possible to isolate the extent of sensitivity degradation occurring as a result of signal propagation inside the optical fiber. The common procedure is to perform a separate measurement of the receiver sensitivity by connecting the transmitter and receiver directly, without the intermediate fiber. The figure below shows the results of such a measurement for a 1.55-μm field experiment in which the RZ-format signal consisting of a pseudorandom bit stream in the form of solitons (sequence length 223 - 1) was propagated over more than 2000 km of fiber. In the absence of fiber (0-km curve), a BER of 10-9 is realized for -29.5 dBm of received power. However, the launched signal is degraded considerably during transmission, resulting in about a 3-dB penalty for a 2040-km fiber link. The power penalty increases rapidly with further propagation. In fact, the increasing curvature of BER curves indicates that the BER of 10-9 would be unreachable after a distance of 2600 km. This behavior is typical of most lightwave systems. The eye diagram is related to the use of the RZ format.
The performance of an optical receiver in actual lightwave systems may change with time. Since it is not possible to measure the BER directly for a system in operation, an alternative is needed to monitor system performance. As discussed in another tutorial, the eye diagram is best suited for this purpose; closing of the eye is a measure of degradation in receiver performance and is associated with a corresponding increase in the BER. The eye is wide open in the absence of optical fiber but becomes partially closed when the signal is transmitted through a long fiber link. Close of the eye is due to amplifier noise, fiber dispersion, and various nonlinear effects, all of which lead to considerable distortion of optical pulses as they propagate through the fiber. The continuous monitoring of the eye pattern is common in actual systems as a measure of receiver performance.
The performance of optical receivers operating in the wavelength range 1.3-1.6 μm is severely limited by thermal noise, as seen clearly from the data in the first figure above. The use of APD receivers improves the situation, but to a limited extent only, because of the excess noise factor associated with InGaAs APDs. Most receivers operate away from the quantum limit by 20 dB or more. The effect of thermal noise can be considerably reduced by using coherent-detection techniques in which the receiver signal is mixed coherently with the output of a narrow-linewidth laser. The receiver performance can also be improved by amplifying the optical signal before it is incident on the photodetector.