Figure 2-7. COUPLING LOSS (dB) COUPLING LOSS (dB) Coupling loss for a glass plate beam splitter as a function of angle of incidence for different degrees of polarization of the incident beam. the fraction of the laser diode power which is polarized in the direction perpendicular to the plane defined by the normal to the plate surface and the incoming beam propagation vector. In the calculation of the coupling loss for this device we have taken into account reflections from both the front and back surfaces. The Fresnel equations [10] were used to determine the reflectivity as a function of angle of incidence. A prism may also be used in the same manner. If space is a consideration, the beam splitter illustrated in figure 2-5(c) is another possibility. The laser diode, detector and test fiber can be connected together by appropriate pigtails. Since numerical apertures of fibers are independent of their diameters, the diode can couple into the launch fiber with a relatively high efficiency in the forward direction. The coupling loss is calculated from branching ratios determined from the relative fiber cross sections presented to the backward traveling wave. This arrangement apparently has not been described in the literature. We have employed the Glan prism of figure 2-5(d) in the current work. This beam splitter provides good isolation (-37 dB) to Fresnel reflections from the input end of the fiber, and has a relatively low coupling loss. One disadvantage is the fact that the prism material is dispersive; the backscattered radiation is returned at different angles as the wavelength is changed. Use of different laser sources causes misalignment at the detector. The Foster or beam-splitting Glan-Thompson prism shown in figure 2-5(e) corrects this problem, but at considerable expense. Polarizing beam splitters do not have this difficulty, but the thin films that comprise the beam splitting surface do not maintain their polarizing properties over a wide spectral interval. The taper coupler (fig. 2-5(f)) has been described by Barnoski et al. [1]. Other types of beam splitters have been described in the literature, including bifurcated couplers [11], and a patented coupling cell filled with index-matching liquid and a beam splitter [12]. Optical fiber directional couplers are also available commercially, but have not been evaluated by our laboratories. 2.5 Boxcar Averager The signal averaging device used in the present work was a PAR Model 162 boxcar integrator. This particular instrument has the desirable feature that ratios may be taken of the scanned backscatter signal and an unscanned reference signal. The latter is obtained from a power divider after the first stage of amplification. This type of signal processing minimizes the effect of drift in the laser diode source and detector. The adjustable delay (fig. 2-1) allows the boxcar scan to begin at any desired point in Since time maps into distance along the fiber, this is equivalent to starting the scan at any desired fiber location. time. The time scan of the boxcar averager may be calibrated with markers which are output from a pulse generator. Time intervals are measured with an electronic timer/counter. The display time for a full scale scan is then accurate to within an estimated 0.02 percent. 2.6 Launcher A fiber launcher which we have found to be convenient is shown in figure 2-8. This is an adaptation of a device originally described by Dakin et al. [13]. It consists of a precision-bore capillary tube [14] which is flared at one end for easy insertion of the fiber. The usual capillary diameter of 0.152 mm will hold 0.125 mm o.d. fibers without creep once the fibers are inserted. The launcher is filled with index-matching oil which also serves as a mode stripper. Once aligned, fibers may be removed and reinserted without significant realignment. However, care must be exercised that dust or foreign matter does not contaminate the bore or launch end. The launcher is demountable so that it can be cleaned if desired. O RING OIL FILL .1 mm MICROSCOPE COVER SLIP PRECISION BORE GLASS CAPILLARY Figure 2-8. Demountable fiber launcher. It is not necessary to use antireflection coatings on the glass cover slip on the input end if polarizing optics are used for the beam splitter. The reflection off the front surface is polarized and therefore largely rejected at the Glan prism coupler. 2.7 Apertures The aperture labeled No. 1 in figure 2-1 is a laser-drilled pinhole and is used to control the spot size on the fiber. In certain cases, for example, with concentric-core fibers, it may be desirable to restrict the set of excited modes which are launched at the input end of the fiber. Alignment is facilitated if the optics have a magnification of one. Here the desired spot size is the same as the pinhole size and is independent of the launch NA. The aperture labeled No. 2 in the same figure is used to reduce the magnitude of the Fresnel reflection from the input end of the fiber. However, in many of the measurements reported on here, both apertures were removed in order to maximize the backscatter signal. 3. SIGNAL AND NOISE CONSIDERATIONS 3.1 Loss Budget The ultimate limitation on the OTDR system for obtaining useful information from optical fibers will be associated with the problem of extracting useful signals in the presence. of noise. The backscatter power levels are inherently small, and for fibers that are many kilometers in length, the signal-to-noise ratio (SNR) of the returning signal may be degraded to an unacceptable level. We will present here a numerical example of the backscatter signal levels to be expected in a somewhat idealized OTDR system for a special case. We consider a pure silica fiber whose only loss is due to the intrinsic Rayleigh scattering, and a laser diode whose radiation is 100 percent polarized. Other conditions are assumed as follows: From eq (1-1) we can calculate the maximum backscatter power in the fiber, at t=0, to be about -20.6 dBm, or considering the beamsplitter cooling loss and optical loss, a power of about -26.8 dBm at the detector. From design data presented elsewhere [15], we can infer that, for a given SNR, a typical APD requires a power level of approximately This implies a loss margin of about 42 dB, so that a fiber may have a one way loss of 21 dB and still have an observable backscatter feature. For the silica fiber considered here this represents a length of about 14 km. In most cases the loss of commercially available fibers is much greaer than 1.5 dB/km at 850 nm so that the critical length is reduced correspondingly. Figure 3-1 illustrates the backscatter signal levels expected from eq (1) as a function of fiber length for several probe pulse widths (at constant peak power) for silica at 850 We have converted time to length according to the usual prescription 2L=tvg. Figure 3-1. LOG BACKSCATTER POWER (dB) BACKSCATTER POWER (dB) Backscatter power levels for silica at 850 nm. To a first approximation this power is proportional to the probe duration as shown here. Other conditions are given in the text. Figure 3-2. Relative backscatter power as a function of the parameter X (see text). |