1. Introduction

There has been much effort to develop a next-generation multimode optical fiber (MMF) for local area network transmission systems as a physical layer for 10-gigabit data transmission links. The differential mode delay (DMD) of an MMF is a fundamental parameter which determines the bandwidth of an MMF [1]. There have been many reports on the concept and the measurement methods of DMD [2]. The manufacturers of an MMF consider a DMD measurement method to be an important diagnostic tool to optimize the design and the fabrication process of an MMF. This is because DMD is considered to be a standard specification parameter for an MMF to guarantee robust operation over time and a certain bandwidth regardless of a launching condition or a laser property [3].

Several measurement methods have been developed to determine the DMD of an MMF [4–7]. In particular, the time-domain DMD measurement method is the industry standard. In the conventional time-domain DMD measurement method [4], a short optical pulse is launched at one end of a test fiber. As each mode has a different propagation speed in an MMF, all modes of the sample MMF are selectively excited with an offset launching method, and the differences in propagation speed are measured. Depending on the different mode propagation speeds, the optical pulse experiences pulse broadening, or spreading. This broadening is measured using a fast detection system. The conventional time-domain DMD measurement system requires complicated and expensive instruments, such as a pulse laser and a fast detection system with a sampling oscilloscope, to measure the broadening of the short optical pulse. Although the recently proposed optical frequency-domain DMD measurement technique based on optical frequency-domain reflectometry (OFDR) improves measurement resolution, this method also uses an expensive tunable laser source (TLS) and a complicated subsidiary interferometer to suppress noise associated with nonlinear frequency sweep in an optical source [5, 6].

In this paper, we present a new DMD measurement technique based on Fourier-domain low-coherence interferometry (fLCI). A broadband light source and an optical spectrum analyzer (OSA) are used as a light source and a detection system, respectively. The interference signals in the spectral domain generated by intermodal interference in an MMF are Fourier transformed to obtain delays in the time domain. The intermodal interference in a two-or few-mode optical fiber has been already used in devices to sense temperature, sound, and pressure [8–10]. Here, we have demonstrated that this concept can easily retrieve modal delay information if combined with Fourier-domain low-coherence interferometry (fLCI) [11–13] to measure DMD in a multimode optical fiber. This new technique is simple in principle, fast in measurement, and can be used in both fibers with very short and long length.

2. Experiment and results

Figure 1 shows a schematic diagram of our experimental arrangement used for the DMD measurements of an MMF. It is based on a common path intermodal interferometer. An amplified spontaneous emission (ASE) light source made of an Erbium doped fiber was used as an optical source, and an optical spectrum analyzer (OSA, Ando 6324B) was used for optical signal detection. The ASE had a center wavelength of 1530 nm. An OSA placed at the output of the test optical fiber measures a spectral interferogram, which contains the intermodal delay information of an MMF. A 15 m long MMF (InfiniCor SX+50/125, Corning Inc.) with a core diameter of 50 µm was prepared as a test sample. As we need to measure the temporal delay between the fastest and the slowest modes in a sample MMF, it is quite important to excite all available modes in an MMF. Conventional DMD measurement methods use the scanning offset launching method to selectively excite all available modes in an MMF [4]. However, this offset launching method takes a long time to selectively excite individual mode group in an MMF. To overcome this problem, we used two mode scramblers (Newport FM-1) at the beginning and the end of a sample fiber. The first mode scrambler is to couple light from a single mode fiber into all the available modes in an MMF. On the other hand, the second one at the end of the sample MMF is to make lights from all of the guided modes in the MMF which are to be coupled back into a single mode fiber connected to the OSA. This generates spectral intermodal interference signals on the OSA between all of the guided modes in the MMF. A short length of an MMF is sandwiched between two corrugated metal surfaces with a gentle pressure controlled with a precision translation stage in a mode scrambler. Microbending in an MMF generated by the device helps make sufficient mode coupling among guided modes in the fiber such that the distribution of power among the guided modes becomes independent of the launch conditions of light. The wavelength span in the OSA was only about 10 nm from 1545 nm to 1555 nm in order to reduce the chromatic dispersion effect in the measurement.

 

Fig. 1. Experimental setup of our DMD measurement method for a multimode optical fiber.

Download Full Size | PDF

The measured spectral interferograms of the test fiber for several different corrugation pitch sizes of the mode scrambler are shown in Fig. 2(a). There are various frequency components in the measured spectral interferograms, and each frequency component corresponds to a relative group delay between two modes in the test fiber. The corrugation pitch of the mode scrambler is changed from 12.5 µm to 75.0 µm. The pressures of the first and the second mode scramblers were set equal and kept constant for the entire measurements.

 

Fig. 2. DMD measurement results for several different corrugation pitch sizes of a mode scrambler: (a) measured spectral interferograms by an OSA, (b) calculated relative time delay graphs obtained from spectral interferograms in Fig. 2(a) by Fourier transformation, and (c) calculated relative time delay graphs in log scale (10dB/div.)

Download Full Size | PDF

The Fourier transformation of a spectral interferogram generates another data in the time domain. Each colored interferogram in Fig. 2(a) is Fourier transformed and plotted in Fig. 2(b) with the same color. Peaks in the Fourier-transformed time-domain data correspond to the temporal delays between excited modes in a test multimode fiber. In order to do accurate numerical Fourier transformation, we need to have interferogram data in the frequency domain instead of the wavelength domain. The conversion of data in the wavelength domain into the frequency domain is done with a cubic spline fitting algorithm [11, 14]. After regenerating equally spaced interferogram data in the frequency domain, we did numerical Fourier transformation to obtain graphs in Fig. 2(b). Since there is a large peak at t=0 ps/m due to the DC signal in the interferogram, we have eliminated this large peak by using a high-pass filter in Fig. 2(b) and 2(c). The upper horizontal axis is the actual temporal delay, and the lower horizontal axis is normalized by the length of the sample fiber. It shows that the efficiency of the mode scrambler is increased, and the number of frequency components in the interferogram is increased as the corrugation pitch of the mode scrambler is increased from 12.5 to 75.0 µm. When the corrugation pitch was further increased, there was no further number of peaks in measured interferograms, and only noise in an interferogram is increased. The last peak is at 1.52 ps/m in the uppermost curve of Fig. 2(b), which is the DMD of the sample fiber. The temporal position of the last peak represents the relative group delay between the fastest mode and the slowest mode. It is when the corrugation pitch of the mode scrambler was 75.0 µm. Fig. 2(c) shows the log scale view of temporal group delay. Grid lines in the y-axis have a scale of 10dB/div. We could easily obtain a 30 dB signal-to-noise ratio in the calculated modal delay curves in Fig. 2(c).

We also measured the DMD of the same optical fiber by using a time-domain impulse response method to confirm our measured results [4]. A gain-switched semiconductor laser (OPG-1500, Optune Inc.) was used as the input pulse source. It has a 28 ps pulse width at 1550 nm wavelength, and its repetition rate is 50 MHz. A sampling oscilloscope was used for short optical pulse width detection (86100A, Agilent Inc.). The length of the test fiber was increased to 450 m for this measurement due to the measurement resolution of the time-domain technique. The measured results are shown in Fig. 3. Meanwhile, Fig. 3(a) shows the measured DMD profile using the conventional scanning offset launching method. In this method, a single mode probe fiber was scanned from one end of the core to the other end with an offset value of 1 µm in order to excite all the modes from the lowest mode group to the highest order mode group. The total mode group was divided into eight mode groups according to the mode excitation condition. It shows that the temporal delay between the leading peak and the trailing peak or the DMD of the fiber is about 1.45 ps/m.

 

Fig. 3. Modal delay measurements of an MMF from the time-domain method using a scanning offset launching method (a) and a mode scrambler method (b).

Download Full Size | PDF

Our measurement results agree well with the time-domain method within 4.6% error. Fig. 3(b) is the measured DMD profile of the sample fiber with a mode scrambler. One mode scrambler to excite the mode was placed at the beginning of the multimode fiber. It shows that the mode scrambler could excite all mode groups in the MMF. The corrugation pitch of the mode scrambler is approximately set to be 75 µm.

3. Discussion

We have shown that our proposed measurement method based on Fourier-domain low coherence interferometry can determine the DMD of an MMF very effectively. We used a mode scrambler to excite all available modes in a sample MMF at the beginning of the MMF. After all the mode groups have traveled through the MMF, another mode scrambler is used at the end of the test fiber to mix all of the mode groups to generate interference signals associated with path length differences between modes. Our proposed configuration is a form of common path interferometer, measuring relative intermodal delays within the test fiber without the need for a reference arm. This common path optical interferometer enhances the stability of an interferometric signal by decreasing the effect of external perturbations, such as air fluctuation and vibration, as compared to a conventional two-arm interferometer.

In our method, it is important to determine the appropriate condition of the mode scrambler to excite all available modes at the beginning of a sample fiber and to mix the modes at the end of the fiber. We have used a commercial mode scrambler where a periodic lateral pressure is applied to a sample fiber by using a pair of corrugated metal plates. Because the mode scrambler perturbs the optical fiber laterally, considerable microbending on a fiber above an appropriate level causes higher-order modes to couple to radiation modes, which in turn causes excessive loss in a fiber. When the pressure in the mode scrambler is too low below the appropriate level, mode mixing efficiency is not enough to excite all available modes in a fiber. Therefore, inappropriate conditions for the mode scrambler decrease the accuracy of DMD measurement. In our case, all modes were fully excited when the pitch of the corrugated metal plate was between 75 µm and 87.5 µm. We stripped the coating off the MMF near the position of the mode scrambler and then immersed it in index-matching gel to eliminate the cladding modes of the MMF and prevent unwanted coupling with the core mode. Although in principle, spectral fringe visibility can be degraded by polarization state mismatch between mode groups, we have experimentally observed that it does not influence measurement results; no changes in peaks in Fig. 2(b) were observed when the input polarization state to the MMF was changed.

Our proposed measurement method has advantages over a conventional time-domain measurement method or an optical frequency-domain OFDR method [4, 5]. First, our proposed technique is very simple in principle and cost effective because there is no special optical element or auxiliary instrument except for a mode scrambler. The multimode optical fiber plays an important role as an interferometer by itself. Although we used an OSA for detection in this experiment, it is easily exchanged with a spectrometer. In the conventional time-domain DMD method, an expensive optical short pulse source and a fast detection system, such as a sampling oscilloscope, are needed. In addition, it requires a complicated offset launching system to excite all mode groups. As it detects the broadening of an optical pulse in the time domain, the sensitivity or the signal-to-noise ratio of measurement is quite low; if the DMD of a sample fiber is too large, or if the sample length is too long, then the pulse may spread too much in the time domain to be detected as a pulse. The optical frequency-domain DMD measurement method based on the OFDR technique shows a good sensitivity and resolution as it is basically a self-heterodyne method. However, it also requires an expensive optical source, such as a TLS, and requires a complicated auxiliary interferometer to reduce the nonlinearity of the sweeping laser source [5, 6].

The sensitivity and the resolution of our proposed DMD measurement method based on the fLCI technique is as good as those of the frequency-domain method. Our technique can measure DMD for the lengths of a multimode optical fiber from several meters to several hundred meters. Therefore, we anticipate that timely quality testing during the manufacturing process of an MMF as well as practical testing of an already installed fiber can be readily achieved with our method. Although this method does not determine the exact time delay information for each mode group excited as does the time-domain method, our proposed method can measure differential modal delay information effectively. The proposed method can be applied to different wavelengths such as 850 nm or 1300 nm by replacing components such as an optical source, optical spectrum analyzer, and fiber couplers suitable for those wavelengths.

4. Conclusions

A novel DMD measurement technique for a multimode optical fiber using intermodal interference with Fourier-domain low-coherence interferometry was proposed. We demonstrated the validity of our proposed method by determining the DMD of a conventional multimode fiber. We also examined the optimum condition of a fiber mode scrambler for effective mode mixing in our proposed DMD measurement technique. We could obtain a high signal-to-noise ratio better than 30dB. Furthermore, comparison with the time-domain method showed that the measurement results are in good agreement with those of the time-domain method. The main advantage of our proposed method is that it is simple, compact, low in cost, and can measure conventional optical multimode fiber from a few meters to a few hundred meters. We expect our technique to be an alternative practical method for determining the modal dispersion of a conventional multimode optical fiber.