applies to a uniform reciprocal fiber. In the more general sorption coefficients as well as capture fractions may be case, both scattering and abfunction of length. In this report, we will document various experimental techniques for measuring the parameters which occur in eq (1-1). We will also present examples of the observed characteristic backscatter responses, (t), for fibers which do not possess uniform properties. These backscatter "signatures" will be presented for fibers as received from the manufacturers as well as for fibers which have aberrations induced by external means. We will attempt to correlate observed signatures with changes in the physical properties of the perturbed fibers. Values of scattering coefficients, capture fractions, and other parameters appearing in eq (1-1) can be useful in the identification, characterization and location of anomalies occuring in secure optical fiber communication systems. Also, this information may be helpful in OTDR design work, computer modeling and military fiber and cable specification. Figure 2-1 shows a block diagram of the main elements of the OTDR system used to generate the data described in this report. The system was mounted on a bench top and no effort was made toward making the device compact or transportable for field use. The main components will be described separately below. Figure 2-2 illustrates the nature of the unprocessed oscilloscope display for a graded-index fiber which demonstrates a scattering anomaly. Figure 2-3 represents the corresponding display from the boxcar integrator, and figure 2-4 the boxcar output on a logarithmic scale. 2.2 Laser Diode Sources For OTDR applications it is desirable to have optical sources capable of producing high peak radiance and high repetition rate with pulse durations in the general range of 5 to 100 ns. If backscatter responses are to be used to estimate loss at the communication wavelength of interest, then the source should also emit radiation at that wavelength. In the spectral region around 850 nm, a suitable pulsed source is the single heterojunction GaAlAs laser diode which is available from a number of manufacturers [5]. The diode used in the present experiments was a RCA type C30012 [6] which was rated at 4 W peak power at a 1 kHz repetition rate. With this source and 1:1 imaging optics it was possible to get a maximum of about 1 W peak power coupled into most fibers. Many RCA GaAlAs laser diodes are partially polarized. This property can help reduce the coupling loss if a Glan-Thompson prism or polarizing beam splitter is used. found that the degree of polarization depends on the current drive level and also varies from diode to diode. A good sample can have about 80 percent of the output radiation polarized in one direction. LOG BACKSCATTER POWER BACKSCATTER POWER Figure 2-3. Backscatter signal, boxcar averager display. Fiber H. Figure 2-4. Backscatter signal, boxcar averager display, logarithmic scale. Fiber H. Higher average output powers are possible with thermoelectric cooling of the laser diode [7]. No cooling devices were used in the present work, however. 2.3 Detectors In view of the low signal levels involved in backscatter systems, the detector usually has internal gain. This can be either an avalanche photodiode (APD) or photomultiplier tube (PMT). In our somewhat limited experience with both of these types of detectors, we have found the APD to be preferable from several standpoints. In the present case we used an RCA type C30818E silicon APD [8]. For accurate measurements, the linearity of the detector at the normal operating gain is an important consideration. We checked the linearity in the following manner. A calibrated neutral density filter (NDF) with approximately 3 dB attenuation (neutral density 0.3) was inserted in the backscatter beam and the signal level noted. Comparison between the unattenuated and observed signal levels yields a quantitative measure of the detector linearity. When using this method, the NDF should always be inserted in the beam where the rays are parallel and allowance should be made for the fact that the focused beam may be displaced slightly. If the detector response is not uniform across its sensitive area, the apparent detector responsivity may change. This problem may to a large extent be obviated by mounting the detector on a XYZ translator and maximizing all signals prior to reading. Figure 2-1 shows an appropriate location for insertion of the NDF. Linearity may also be checked over a wider dynamic range with this method by using higher density filters. 2.4 Beamsplitters Several types of possible OTDR beamsplitters and some of their characteristics are illustrated in figure 2-5. The coupling loss which is listed in the figure is based on the assumption that the backscattered radiation is unpolarized, since few fibers maintain polarization for more than a few meters [9]. The coupling loss is then defined as Coupling loss = 10 log (dB) 82 (2-1) where 1 is the collimated laser diode output power and 2 is the power that is directed toward the detector after having undergone a lossless reflection. For polarizing beamsplitters this coupling loss will depend on the initial state and degree of source polarizaThe minimum loss is for 100 percent polarization in the direction indicated by the arrows in figure 2-5. Most commercially available 50:50 (nominal) beamsplitters do not have an exact 50:50 beamsplitter ratio, and this ratio may be a function of wavelength. Figure 2-6 demonstrates how the coupling loss changes with reflectivity (we assume that transmission + reflection = 1). Figures 2-5(b) and 2-7 indicate that a fairly efficient beamsplitter can be constructed from a glass plate if it is used at a high angle of incidence. The parameter P represents |