1. Introduction

Optical reflectometry with a narrow spatial resolution and a long measurement range is becoming increasingly important as a tool for optical fiber network maintenance. Optical time domain reflectometry (OTDR) [1] can provide a long measurement range but it suffers from a tradeoff between spatial resolution and sensitivity resulting in a resolution of greater than one meter. For applications such as the distributed measurement of polarization mode dispersion (PMD) [2] in previously installed fibers, the beat length caused by a high PMD section may be much less than one meter, making it nearly impossible to measure with OTDR.

Optical frequency domain reflectometry (OFDR) [3–5] is a promising technique for realizing a narrow spatial resolution and high sensitivity. Although sub-millimeter resolution is achievable, the distance range is limited to a few tens of meters, because it is restricted by the coherence length of the light source used. Several attempts [6,7] have been made to realize cm-level high-resolution OFDR measurements over long distances, and the maximum measurement range is about 2 km [6]. To extend the measurement range, we proposed phase-noise compensated optical frequency domain reflectometry (PNC-OFDR) [8], and successfully obtained a spatial resolution of ~10 cm over 20 km in a normal laboratory environment [9]. In fact, a longer measurement range is desired for diagnosing long-haul links. If we assume that the maximum length of a terrestrial link is 80 km, a range of at least 40 km is needed with both-end access. In our previous work [10], we found that, in a long range measurement, the dominant factor as regards resolution degradation is the environmental perturbation on the fiber under test (FUT), whereas the PNC algorithm well works even for distances of several tens of kilometers. If we are to extend the measurement distance while maintaining the spatial resolution, we must find a way to reduce the influence of such environmental perturbation.

In this paper, we present a bandwidth-division PNC-OFDR technique capable of reducing the influence of environmental perturbation via a fast sweep of the optical source frequency, thus decreasing the ratio of the acoustic noise band (~several kHz) to beat frequency. The bandwidth-division scheme lets us deal with high-frequency signals even when we employ a fast sweep of the optical source frequency at 3 THz/s. Meanwhile, the signal bandwidth is greatly reduced by using this scheme, enabling us to sample and process more data (corresponding to a full frequency sweep of 15 GHz), thus helping us to improve the spatial resolution with the sacrifice of the measurement time. By using this technique, we realize a sub-cm spatial resolution over 40 km in a normal laboratory environment and a 5 cm spatial resolution at 39.2 km in a field trial.

2. Principle

The theory of PNC-OFDR has already been described in detail [8], and we provide just a brief introduction here. In PNC-OFDR, an auxiliary interferometer is used to provide phase noise information about the optical source allowing us to resample the temporal signals of the measurement. The phase noise of the measurement signal is then compensated when the reflection events occur around the best compensation positions, and is totally eliminated at those positions. Figure 1 shows schematically that the FUT could be divided into N sections for compensation using different concatenation-generated phases (CGPs) X_N(t), which are calculated using the following expression:

 

Fig. 1 Reference signals used in each section of FUT.

Download Full Size | PDF

where X₁(t) is a phase term obtained from an auxiliary interferometer with a delay time τ_ref in one arm. The phase noise term Φ(t) is compensated from θ(t) – θ (t-τ_FUT) to the following term:

where τ_FUT is the time needed for a round-trip in the FUT. When τ_FUT is equal to Nτ_ref, the phase noise term is canceled out and the optimum compensation can be achieved.

At a long distance such as 40 km, even after phase noise compensation, environmental perturbation becomes a dominant factor degrading the spatial resolution. Therefore, as mentioned in the introduction, a fast sweep of the optical source frequency is needed to reduce the effect of environmental perturbation. However, a fast sweep generates a high beat frequency, which makes it difficult for later sampling or processing. To deal with this problem, we can adopt a bandwidth-division method. If we assume a maximum frequency of F, we need a sampling card with a sampling rate of greater than 2F. Figure 2 shows that we can divide the frequency into M sections, each with a bandwidth of F/M. If we down-convert the m-th (m = 2, 3,…, M) section so that it is within a frequency of 0 < f_base < F/M, we only need a sampling card with a sampling rate of 2F/M. Meanwhile, the amount of data decreases to 1/M compared with that required when using the conventional sampling method.

 

Fig. 2 Concept of bandwidth-division process.

Download Full Size | PDF

We should note that X_N(t) is no longer the correct reference to compensate for the phase noise in the down-converted signal. To obtain the correct reference, we should also down-convert the frequency of X_N(t) by the same value. This can be realized by changing X_N(t) as follows:

Figure 3 shows schematically how to compensate for every section of the FUT by using suitable reference signals. Another method for accomplishing the compensation correctly is to rebuild the original measurement signal before down-conversion, and use X_N(t) for compensation. Although the two methods are equivalent to each other and can be implemented digitally, we adopt the former because it consumes fewer calculation resources.

 

Fig. 3 Reference signals used in each section of FUT for different bandwidths.

Download Full Size | PDF

As a result, with bandwidth-division PNC-OFDR, the bandwidth is reduced to 1/M, thus enabling us to find a suitable sampling card even when the original signals have a very large electrical bandwidth. Moreover, for a personal computer with a limited physical memory, it helps us to process more divided section data, providing another way of improving the spatial resolution. The data-sampling number will be M times the previous number, accompanying the increase in electrical equipment. However, the most important thing is that we can reduce the influence of environmental perturbation to a 1/M level.

3. Experimental setup

In our experimental setup shown in Fig. 4 , the light source is a fiber laser (Koheras Adjustik^TM) and the linewidth measured using the self-delay heterodyne method with a 100 km delay fiber is 4 kHz. If we assume that the lineshape is Lorentzian, the measurement range of regular OFDR can be estimated to be c/2nπΔf as 8.0 km while considering the round-trip path, where c is the speed of light in vacuum, n is the refractive index of FUT, Δf is the linewidth. A single sideband with a suppressed carrier (SSB-SC) modulator and a frequency swept RF synthesizer are used for external frequency sweeping. The sweep rate is set at 3 THz/s, which is 7.5 times faster than that used previously [10], with a full sweep frequency of 15 GHz (limited by available bandwidth of the modulator) for a 5-ms acquisition time. The ratio of beat frequency to distance is determined by the sweep rate to be 2nγ/c as 300 Hz/cm, where γ is the sweep rate. If we consider the calculation process, after applying Eq. (1) and the round-off for later averaging, the useful part is 12.5 GHz (4.17 ms acquisition time), corresponding to a theoretical spatial resolution of 8 mm (240 Hz in frequency domain). A Mach-Zehnder interferometer with a 5-km delay fiber in one arm is used as an auxiliary interferometer for compensation, and is placed in a soundproof box to insulate it from acoustic noise. The main interferometer consists of a local arm and a measurement arm, which is equipped with a circulator for launching the light wave into the FUT and receiving the reflected signal. A polarization controller is used in the local arm to control the power of the local light so that it is split evenly by the polarization splitters, which are important elements of a polarization diversity scheme, adopted to remove the influence of the polarization effect. The signals from both the auxiliary and main interferometers are detected by balanced photodetectors (BPDs). Then, the signals from the auxiliary interferometer are filtered by a low-pass filter (LPF), sampled by using an analog-to-digital card (ADC), and collected by a computer. On the contrary, the signals from the main interferometers must undergo bandwidth-division processing before being sampled by the ADCs. The details of the process are given in the next paragraph.

 

Fig. 4 Experimental setup. SSB, single sideband; DFL, delay fiber loop; PC, polarization controller; PBS, polarization beam splitter; BPD, balanced photodetector; LPF, low-pass filter; ADC, analog to digital card; FUT, fiber under test.

Download Full Size | PDF

For this experiment, we choose an M value of 3 for the purpose of preliminary theoretical confirmation. For an FUT of ~40 km, the maximum frequency F of the signals from the main interferometers is ~1200 MHz. Figure 5 illustrates the electrical process after one BPD. Two electrical switches are used to select the route with the help of synchronization equipment. When the 1st route is selected, there is no processing within the route. When the 2nd or the 3rd route is selected, the signals are filtered by a band-pass filter (BPF) with a bandwidth of 400–800 MHz (BPF1), or 800–1200 MHz (BPF2). Then the signals are down-converted by using a mixer beat with sinusoidal signals whose frequency is 408 MHz (route 2) or 800 MHz (route 3), generated from an arbitrary waveform generator. For route 2, 400 MHz is not used for mixing since the mixer is not ideal and this signal will still be present after mixing, resulting in a spurious peak. Therefore, it is expected that a larger frequency will be used here since the spurious peak does not influence the original results. The signals are then divided into three sections of 0–408, 408–800, and 800–1200 MHz. After the mixer, the signals are filtered by an LPF with a cut-off frequency of 400 MHz, sampled by using an ADC with a sampling rate of 1 GS/s, and collected by a computer, to undergo a numerical compensation process.

 

Fig. 5 Electrical processing after the signals having been received by one BPD. The bandwidth of BPF1 and BPF2 is 400–800 MHz and 800–1200 MHz, respectively, and the LPF cutoff bandwidth is 400 MHz.

Download Full Size | PDF

The compensation is performed with a high-end personal computer with an Intel Core i7-980X CPU and a 24 GB cache memory. To reduce Rayleigh speckle noise, we take the average intensity of the results obtained with different probe wavelengths by using a laser tuning range of over 60 GHz. The full measurement time is about 2 min, and this is limited by the laser tuning speed. About 30 min is needed to compensate the data of one bandwidth (0–14.0, 14.0–27.5, or 27.5–41.25 km) with a 50-wavelength average.

4. Experimental results

The FUT is composed of a 40 km fiber spool connected to a 1.25 km fiber spool equipped with an angled physical contact (APC) connector. Figure 6 shows the reflectivity of backscattered/reflected light wave, which is a combination of three different divisions. Due to the abrupt attenuation around the cut-off frequency of the electronic filters, these signals with frequencies outside the bandwidth, such as signals around 400–408 MHz, are also highly attenuated. In fact, a more carefully designed filter should be adopted to alleviate this attenuation. Meanwhile, since the BPD used in our setup has a cut-off frequency of 800 MHz, which is less than the original frequencies (800–1200 MHz) generated at the third division, the results at longer distances are greatly attenuated. This is the widest bandwidth of our available equipment, and in fact a BPD with a wider bandwidth should be adopted to avoid the excess attenuation.

 

Fig. 6 Measurement results for the reflectivity of backscattered/reflected light wave.

Download Full Size | PDF

We placed the FUT in two different laboratory environments to check the spatial resolution since it is influenced by the environment. The total sound pressure was 58.1 dB in a normal laboratory environment and 60.7 dB in a relatively noisy laboratory environment. The sound pressure densities of two different experimental environments are shown in Fig. 7(d) while the details of the reflection peaks at different environments are shown in Figs. 7(a)–(b). For the reflection peaks that occur around 40 km (N × 2.5 km), which is an optimum compensation position, a 3-dB spatial resolution of 8 mm (240 Hz in frequency domain) is obtained in a normal environment and it deteriorates to about 2.5 cm (750 Hz in frequency domain) in a relatively noisy environment. For the reflection peaks that occur around 41.25 km ((N + 0.5) × 2.5 km), which is the border of the 16th compensation section, the spatial resolution is still 8 mm in a normal environment, and it deteriorates to about 5 cm (1500 Hz in frequency domain) in a relatively noisy environment.

 

Fig. 7 Details of reflection peaks in different environments. (a) Reflection peaks occurred around 40 km; (b) Reflection peaks occurred around 41.25 km; (c) Reflection peak measured in the field environment; (d) Sound pressure density of two laboratory environments.

Download Full Size | PDF

We compared the results to those obtained without bandwidth-division process at a sweep rate of 1 THz/s, the reflection peaks almost remain the same in the frequency domain. Therefore, it is obvious that the environmental perturbation caused the added frequency components. By sweeping the optical source frequency faster, we decreased the ratio of these added frequency components to beat frequency, therefore making spatial resolution improved. In the normal laboratory environment, we obtained a spatial resolution of 8 mm at any position within the entire measurement range.

Since the results are dependent on the environment, it is thus necessary to test the setup in a field environment to check its performance. The field trial was performed in Japan on an optical cable that was installed underground to link two cities, and that is still in use. The fiber for the test was a 39.2-km-long dispersion-shifted fiber (DSF), and was temporarily out of use. The detailed reflection peak at the final connector is also shown in Fig. 7(c), revealing a spatial resolution of 5 cm in the field environment.

5. Summary

We have presented a bandwidth-division PNC-OFDR technique, capable of reducing the influence of environmental perturbation via a fast sweep of the optical source frequency, by decreasing the ratio of the acoustic noise band to beat frequency. The bandwidth-division scheme permits us to deal with high-frequency signals by adopting a fast sweep of the optical source frequency. Meanwhile, the reduction of the signal bandwidth enables us to sample and process more data, thus helping us to improve the spatial resolution. By using this technique, we realized a sub-cm spatial resolution over 40 km in a normal laboratory environment, and a 5 cm spatial resolution of 39.2 km in a field trial.