Individual loss distribution measurement in 32-branched PON using pulsed pump-probe Brillouin analysis

Hiroshi Takahashi; Fumihiko Ito; Chihiro Kito; Kunihiro Toge

doi:10.1364/OE.21.006739

1. Introduction

In recent years, the number of fiber to the home (FTTH) service users has been increasing rapidly. Most FTTH installations employ a passive optical network (PON) configuration, in which a passive splitter is inserted between the customer’s premises and the central office, to save the cost of cables and optical transceivers. One of the critical issues when a PON is damaged is how to test the fiber. With a single star configuration, widely used optical time domain reflectometry (OTDR) allows us to monitor the entire loss distribution of the fiber and thus provides us with the fault location. However, with a PON, when we use this simple OTDR approach at the central office we obtain only a superposition of the reflectometric traces of all the tributaries, but not their individual loss distribution. As a result, the operator cannot accurately confirm the health of the optical fiber [1–5].

Some approaches have already been developed for measuring the branched sections of a PON including multi-wavelength OTDR [6–8], which is a technique where a wavelength sensitive routing device is inserted at the branch, and Brillouin-OTDR where different Brillouin frequency shifts are allocated to each tributary [9]. However, with these techniques the PON system requires additional optical components or changes to the optical fiber itself.

We have proposed a novel loss distribution measurement technique that uses Brillouin gain analysis [10], which can monitor the individual loss distribution in branched optical fibers without the need to add optical components to the PON system. This approach uses the end reflection of the each tributary. ITU-T L.66 [11] recommends the insertion of a test light cut-off filter at the end of the network. When the cut-off filter is realized by a fiber Bragg grating (FBG), by nature, the FBG filter provides nearly 100% reflection, which is available for the measurement. This means that no additional optical equipment is needed for a standardized PON system with the proposed method. At this point, we distinguish our method from those mentioned above. With the proposed method, if end reflection is available, the loss distribution of each branch can be measured individually by employing Brillouin gain analysis between a pulsed pump signal and a pulsed probe signal reflected at the far end of each branch. By using the difference between the lengths of each branch and test equipment with a fiber length identification resolution that is higher than this length difference, the loss distributions of all the branches in a PON can be monitored individually.

In this paper, we describe the new method in detail, and test it on a 32-branched PON. An event location resolution of 10 m, and a fiber length identification resolution of 2 m are achieved, which means that the system can distinguish between branches that have a 2 m length difference. In addition, by undertaking a theoretical analysis of the signal-to-noise ratio (SNR), we show that this measurement method has the potential to achieve a dynamic range of 25 dB.

2. Principle

Figure 1 shows a typical PON configuration and our proposed loss distribution measurement system. Our aim here is to measure the individual loss distribution of each tributary by launching a test beam into the fiber from the trunk side. It is assumed that a total or partial reflection is available at the far end of the fiber. In a standard PON, a test light cut-off filter installed in front of an optical network unit (ONU) can provide this reflection [11]. With our measurement technique, the branched optical fibers must all have different lengths.

Fig. 1 Configuration of event location technique in PON using Brillouin gain analysis, ’ $\times$ ’ shows the collision points of the pump and probe beams, which are located at a distance of υΔt/2 from the end reflection of each tributary. (υ: velocity of light in fiber, Δt: difference between the launch times of the probe and pump pulses.)

Download Full Size | PDF

The test beam we employ is a probe pulse accompanied by a pump pulse with a temporal interval of Δt. The optical frequency of the probe pulse is set at f₀-f_B, where f₀ is the optical frequency of the pump pulse and f_B is the Brillouin frequency shift of the optical fiber. This means that these beams yield a Brillouin interaction inside the fiber when they encounter each other. As shown in Fig. 1, the probe pulse, which is launched prior to the pump pulse, first reaches the reflection point at the end of the branched optical fibers, and acts as a counter-propagating probe. A Brillouin interaction occurs when the two pulses collide at a distance of υΔt/2 from the reflection point at the end of the branched optical fibers, where υ is the velocity of a lightwave in the optical fiber. The induced Brillouin gain G_B is given by [12]

G_{B} (z) = \exp [\int_{z}^{z + Δ L} g_{B} P_{p} (z) d z] ≅ 1 + g_{B} Δ L P_{p} (z),

where ΔL is the event location resolution, g_B is the Brillouin gain coefficient, and P_p(z) is the pump power at the interaction point z. Here we used an approximation where g_BP_p(z)ΔL << 1 in Eq. (1).

First, we observe the returning probe pulse power, P_probe, at the detector as a reference. This can be accomplished by observing the probe pulse without launching the pump pulse. Next, the pump pulse is launched, and the amplified probe pulse is observed. If we observe the amplified probe powers for two different collision points at z = 0, and z = z₀, the powers of the probe pulses are given by

P_{a m p} (0) = P_{p r o b e} G_{B} (0) = P_{p r o b e} (1 + g_{B} P_{p} (0) Δ L),

and

P_{a m p} (z_{0}) = P_{p r o b e} G_{B} (z_{0}) = P_{p r o b e} (1 + g_{B} P_{p} (z_{0}) Δ L) .

From these equations, we obtain

P_{p} (0) = \frac{P_{a m p} (0) - P_{p r o b e}}{g_{B} Δ L},

and

P_{p} (z_{0}) = \frac{P_{a m p} (z_{0}) - P_{p r o b e}}{g_{B} Δ L} .

Therefore, the fiber loss L(z₀) between z = 0 and z₀ is obtained by

L (z_{0}) = \frac{P_{p} (z_{0})}{P_{p} (0)} = \frac{P_{a m p} (z_{0}) - P_{p r o b e}}{P_{a m p} (0) - P_{p r o b e}} .

Note that, to obtain the loss between the two points L(z₀), it is not necessary to know g_BΔL, which is the Brillouin gain factor.

The loss L(z₀) of the corresponding branch can be distinguished because the return times of the probes from different branches are different. To distinguish between different branches, the difference between the branch lengths must exceed half of the pulse width.

By sweeping the difference Δt between the launch times of the probe and pump pulses, we can sweep the interaction point z₀. Therefore, this method can measure the loss distribution by sweeping the time difference Δt from zero to the time difference for the maximum branched fiber length.

The pulse widths of the probe and pump pulses determine the resolution of the measurement system in different ways. For simplicity, we assume that the lifetime of the refractive index grating induced by the Brillouin interaction is shorter than the pulse widths. Hence, the Brillouin grating disappears faster than the pulse duration, so the pulse shapes are not significantly deformed. In such a case, the reflected probe pulses from different tributaries can be always distinguished if half of probe pulse width is smaller than the minimum length difference of the tributaries. Therefore, the probe pulse width determines the required fiber length difference, or “fiber length identification resolution”. On the other hand, the pump pulse width is related to the resolution of the location of the loss, or “event location resolution”. The event location resolution ΔL₁ and the fiber length identification resolution ΔL₂ are given by

Δ L_{1} = \frac{υ Δ T_{1}}{2}, Δ L_{2} = \frac{υ Δ T_{2}}{2},

respectively, where υ is the optical velocity in optical fiber, ΔT₁ is the pump pulse width, and ΔT₂ is the probe pulse width. The required difference between the branch fiber lengths is determined by the probe pulse width (on condition that the pulse width is greater than the lifetime of the refractive index grating).

3. Experiment

Figure 2 shows our experimental setup for distribution measurement in a PON using Brillouin gain analysis. Both the pump and probe beams were generated from a conventional distributed feedback (DFB) laser. The frequency of the probe beam was downshifted by 10.77 GHz, which corresponds to the Brillouin frequency shift of single mode fiber, by using a single sideband modulator (SSBM) as a frequency shifter. The pump beam was amplified by an erbium doped fiber amplifier (EDFA). Both beams were intensity-modulated by acousto-optic modulators (AOMs), which provided a 100-ns-long pump pulse and a 20-ns-long probe pulse. Both pulse widths are similar to or longer than the estimated lifetime of the grating formed by the Brillouin interaction. The AOMs also caused optical frequency shifts of f_AOM = 250 MHz in the pulses, and yielded a frequency difference f_AOM between the local and probe beams. The peak launch powers of the probe and pump pulses were 2.0 and 16.0 dBm (similar to that of commercial OTDR), respectively. The power of the reflected probe pulse was observed by heterodyne detection. The amplitude of the beat signal of the frequency f_AOM ( = 250 MHz) was observed.

Fig. 2 Experimental setup for distribution measurement in PON using Brillouin gain analysis.

Download Full Size | PDF

Reflected light from the fiber under test (FUT) contains the pump pulse together with the desired probe pulse, so this measurement must remove the pump pulse from the reflected light. This undesired pump pulse can be removed via the heterodyne detection, by using the frequency difference between the local and pump beams, which exceeds 10 GHz. However, since the local beam also has a frequency component (f₀) derived from the residual carrier frequency by SSBM (~25 dB down from −1st sideband), the beat frequency of the pump pulse (f₀) and the frequency component (f₀) in the local beam drop to the same frequency as the beat signal. When branched fiber #1 has an excess loss 5 dB, the power of the reflected pump pulse from other branches is about 25 dB stronger than that of the reflected probe pulse from branched fiber #1. Thus the beat noise amplitude derived from the reflected pump pulse and the local beam is detected at a similar level to the beat signal amplitude. Therefore, to suppress the beat noise, an optical filter (XTM-50, Yenista OPTICS: suppression ratio > 40 dB at 10 GHz deviation) was inserted in front of the detector and removed the reflected pump pulse. The loss of this optical filter L_of at the pass band was 5 dB.

The measured beat signal was converted to a baseband signal by envelope detection, which consists of a square function and a low pass filter. The envelope was extracted from the squared signal with a low pass filter (LPF). The envelope signal was converted to a digital signal by using a 12-bit 200 MS/s A/D converter.

We employed a pump pulse width of 100 ns and a probe pulse width of 20 ns. These pulse widths correspond to an event location resolution of 10 m, and a fiber length identification resolution of 2 m, respectively. The measured power was averaged 10000 times for each measurement. This measurement time was 90 seconds. The FUT consisted of a 32-branched optical splitter, 16 branched fibers and 16 end reflectors. The trunk fiber was 2000 m long. The branch lengths were 950 m (#1), 1000 m (#2), 1005 m (#3), 1010 m (#4), 1020 m (#5), 1030 m (#6), 1040 m (#7), 1045 m (#8), 1050 m (#9), 1058 m (#10), 1060 m (#11), 1070 m (#12), 1080 m (#13), 1090 m (#14), 1100 m (#15), and 1105 m (#16). The minimum branch length difference (between #10 and #11) was 2 m. An intensive bending loss was added to branched fibers #1 (5.5 dB), #6 (3.0 dB), #10 (4.0 dB), and #15 (3.0 dB) at distances of 500, 700, 550, 500 m from the splitter, respectively.

Figure 3 shows the results of the measurements. The vertical axis indicates the fiber loss from the input point of the 32-branched optical splitter. Figure 4 shows that an intensive loss was observed in branched fibers #1, #6, #10, and #15. Branched fibers #10 - #11 in Fig. 4 show that the loss distributions can be measured individually with a branched fiber length difference of 2 m.

Fig. 3 Measured Brillouin gain analysis results with 32-branched optical fiber. 16 tributaries were measured. No fiber was attached to the remaining 16 ports.

Download Full Size | PDF

Fig. 4 Accuracy of pulsed pump-probe Brillouin analysis, the same measurement was performed ten times. The bold points show the average, and the bars shows the maximum and minimum values of the results.

Download Full Size | PDF

Furthermore, to confirm the accuracy of the loss, the measured loss was compared with the added value. The average measured loss and the deviation for 10 results are shown in Fig. 4. The horizontal and vertical axes indicate the added and measured losses, respectively. It is seen that the measured losses well reflect the actual values up to 6 dB. The increase in the deviation of the measurement is due to the degradation of the SNR. This result shows that less than 1 dB loss event is detectable.

4. Signal to noise ratio analysis

This section discusses the SNR to evaluate the range performance of this measurement method. The signal power P_signal of the Brillouin analysis is given by

P_{s i g n a l} = P_{a m p} - P_{p r o b e} = g_{B} Δ L P_{p} (z_{0}) .

The following calculation uses a typical value for the Brillouin gain factor

g_{B} Δ L = 2.0 \times 10^{- 3}

in single mode fiber for a pump pulse width of 100 ns [13]. The signal power of the photocurrent induced by the probe beam is given by

{\bar{I}}_{_{s}}^{2} = {(\frac{η e}{h ν} \sqrt{P_{s i g n a l}})}^{2},

where e is the electron charge, h is Planck’s constant, η is the quantum efficiency of the photo detector and ν is the light wave frequency. The noise incorporated in the measurement includes

• spontaneous Brillouin scattering (SPBS) and
• detector noise.

The SPBS power generated with a loss of L in the fiber is given by [12], [13]

P_{s p} = r_{B} P_{p} (0) {L}^{2},

where r_B is the Brillouin backscattering factor. The following calculation employs

r_{B} = 1.58 \times 10^{- 8}

for a pump pulse width of 100 ns [13]. When the fiber has K branches,

P_{s p} = \sum_{i = 1}^{K} r_{B} P_{p} (0) {L_{i}}^{2},

where L_i is the loss between the SPBS generation point in an i-branched fiber from the pump pulse launch point. The photocurrent induced by the SPBS is given by

{\bar{I}}_{_{s p}}^{2} = {(\frac{η e}{h ν} \sqrt{P_{s p} (z)})}^{2} = {(\frac{η e}{h ν} \sqrt{\sum_{i}^{K} r_{B} P_{p} (0) {L_{i}}^{2}})}^{2} .

When I_sp is received with the amplified signal, the noise power of

σ_{s p}^{2} = \frac{2 {\bar{I}}_{s} {\bar{I}}_{s p} B}{Δ ν_{B}} + \frac{{\bar{I}}_{s p}^{2} B}{Δ ν_{B}}

is incorporated [12], [14], [15], where Δν_B is the Brillouin gain bandwidth and B is the receiver bandwidth. The first and second terms in (13) represent the beat noise between the probe light and the SPBS, and between the SPBS and itself, respectively. We assume that the detector noise is dominated by the shot noise since this measurement adopted coherent detection. This is represented by

σ_{s h o t}^{2} = 2 e {\bar{I}}_{L O} B,

and

{\bar{I}}_{_{L O}}^{2} = {(\frac{η e}{h ν})}^{2} P_{L O} .

The total noise σ_amp incorporated in the amplified probe pulse is given by

σ_{a m p}^{2} = σ_{s p}^{2} + σ_{s h o t}^{2} .

On the other hand, the reference probe pulse induces a noise of

σ_{r e f}^{2} = σ_{s h o t}^{2} .

Since the signal is obtained by subtraction P_amp – P_probe, the variance of P_amp – P_probe is given by the sum of each variance, the total noise is

σ_{t o t a l}^{2} = σ_{a m p}^{2} + σ_{r e f}^{2} .

Therefore, the SNR is given by

S N R = \frac{{\bar{I}}_{s}^{2}}{σ_{s p}^{2} + 2 σ_{s h o t}^{2}} .

When the average number N, the single way dynamic range SWDR is given by

S W D R = \frac{{\bar{I}}_{s}^{2} \sqrt{N}}{σ_{s p}^{2} + 2 σ_{s h o t}^{2}} .

Figure 5 shows an example analysis of the SNR. In the analysis, we used the PON loss model shown in Fig. 5(a). The PON includes a 32-branch splitter whose loss is assumed to be L_c = 17.5 dB (3.5 dB / 2 divisions). The loss of the trunk line is L_f1 = 0.8 dB. The loss of the branched fiber is assumed to be L_f2 = 0.4 dB. There is excess loss induced by bending in one branch.

Fig. 5 Example SNR analysis, (a) model of loss of analyzed PON, (b) parameter of SNR in single shot for 32-branched PON, Fiber loss L_f = trunk fiber loss L_f1 + branched fiber loss L_f2 = 1.2 dB, splitter loss L_sp = 17.5 dB (3.5 dB / 2 branch), optical filter loss L_of = 5 dB.

Download Full Size | PDF

Figure 5(b) shows ${\bar{I}}_{s}^{2}$ , $σ_{t o t a l}^{2}$ , $2 σ_{s h o t}^{2}$ and $σ_{s p}^{2}$ , as a function of the excess loss. The parameters used here are as follows: input pump power P_p(0) = 16 dBm, input probe power P_probe(0) = 2 dBm, receiver bandwidth B = 50 MHz. The result revealed that when the fiber loss is small the total noise power is dominated by the beat noise between the probe light and the SPBS. As the fiber loss becomes larger, in other words as the reflected signal power becomes smaller, the total noise power is dominated by the shot noise.

Figure 6 compares calculated and experimental results for the SNR with 10000 times averaging. In Fig. 6, these results are in good agreement. In Fig. 6, it is seen that the SNR becomes less than unity when the excess loss exceeds 6.6 dB. If the loss is larger than the value, the entire section of the branch is undetectable. Figure 7 shows the relationship between fiber length identification resolution and dynamic range with 10000 times averaging. In Fig. 7, when the fiber length identification resolution or the event location resolution is set at a lower value, the dynamic range is larger. However, when the fiber length identification resolution is set higher than about 1 m, the dynamic range cannot improve even if the event location resolution is set lower. According to the calculated SNR in Fig. 6, and the calculated dynamic range in Fig. 7, this method has the potential to measure an excess loss event less than 6.6 dB, and the dynamic range is 25.3 dB with a 10 m event location resolution, a 2 m fiber length identification resolution and 10000 times averaging.

Fig. 6 SNR with 10000 times averaging. The solid line and dots represent calculated and experimental SNR, respectively.

Download Full Size | PDF

Fig. 7 Relationship between fiber length identification resolution and dynamic range. The red, green, and blue lines indicate event location resolutions of 50, 10, and 5 m, respectively.

Download Full Size | PDF

5. Summary and discussion

A new method was described for examining the loss distributions of a PON individually. The method is applicable to a standardized PON that includes end reflection without the need for any additional equipment to distinguish the individual branches. A preliminary experiment was demonstrated that employed a 32-branched PON constructed in the laboratory. The loss distribution of every tributary was successfully monitored with an accuracy better than 1 dB. The SNR was analyzed. It was shown that the method can realize a 25 dB dynamic range (or excess loss of 6.6 dB in a 32-branched PON) in 90 seconds (10000 times averaging) with an event location resolution of 10 m, and a fiber length identification resolution of 2 m. These features would allow us to use the technology for quality test of physical layer at the construction and preventive monitoring, in which, a minor failure should be detected before it reaches data link disconnection.

The need for the fiber length difference would constitute a measure restriction when deploying the system. The required length difference is about 1~2 m, which is determined by the probe pulse width (or more intrinsically, by the lifetime of the refractive index grating). If a pair of branches have the same length, we cannot distinguish that pair, however, all remaining branches are individually observable.

Further research is necessary before the method can be deployed in the field. One is the wavelength of the test beam. The 1650 nm maintenance band is mentioned in ITU-T L.41 for monitoring PONs. Although many of the devices used in our experiment at 1550 nm are also available for use at 1625 ~1650 nm, at least the light source and boosting optical amplifier should be redesigned. Second, we must consider the Brillouin frequency shift variation in installed cables. As is well known, the Brillouin frequency shift changes with temperature (and strain) at ~1 MHz/°C. Therefore, the Brillouin frequency shift change of outside cables could be as large as several tens of MHz. Although the variation of the Brillouin frequency shift may increase the measurement time, the expected variation is only a few times as large as the typical Brillouin linewidth in fibers (30 ~40 MHz), so we can expect any elongation to be within the acceptable range. Additionally, the use of optical fibers with different Brillouin frequency shifts is not exclusive. The issue is being investigated.

References and links

1. L. Wuilmart, V. Moeyaert, D. Daniaux, P. Megret, and M. Blondel, “A PC-based method for the localisation and quantization of faults in passive tree-structured optical networks using the OTDR technique,” in Proc. of IEEE Conf. on Lasers and Electro-Optics Society, 2, 122–123, (1996). [CrossRef]

2. L. Laferriere, M. Saget, and A. Champavere, “Original method for analyzing multipaths networks by OTDR measurement,” in Proc. of Opt. Fiber Commun. Conf., 99–101, (1997). [CrossRef]

3. Y. Enomoto, H. Izumita, and M. Nakamura, “Highly developed fiber fault isolation technique for branched optical fibers of PONs using high spatial resolution OTDR and frequency domain analysis,” The Rev. of Laser Eng., 33, No. 9, (2005).

4. N. Zou, Y. Namihira, C. Ndiaye, and H. Ito, “Fault location for branched optical fiber networks based on OFDR technique using FSF laser as light source,” in Proc. of Opt. Fiber Commun. Conf., 1–3, (2006).

5. T. Hasegawa and A. Inoue, “Monitoring of drop optical fibers in 32-branched PON using 1.65 μm Pulse-OCDR,” in Tech. Digests of National Fiber Optic Eng. Conf., paper NWA5, (2009).

6. K. Tanaka, M. Tateda, and Y. Inoue, “Measuring the individual attenuation distribution of passive branched optical networks,” IEEE Photon. Technol. Lett. 8(7), 915–917 (1996). [CrossRef]

7. U. Hilbk, M. Burmeister, B. Hoen, T. Hermes, J. Saniter, and F. Westphal, “Selective OTDR measurements at the central office of individual fiber links in a PON,” in Tech. Digests of Opt.Fiber Commun. Conf., (1997). [CrossRef]

8. P. J. Urban, G. Vall-llosera, E. Medeiros, and S. Dahlfort, “Fiber plant manager: An OTDR- and OTM-based PON monitoring system,” IEEE Commun. Mag. 51(2), S9–S15 (2013). [CrossRef]

9. D. Iida, N. Honda, H. Izumita, and F. Ito, “Design of identification fibers with individually assigned Brillouin frequency shifts for monitoring passive optical networks,” J. Lightwave Technol. 25(5), 1290–1297 (2007). [CrossRef]

10. H. Takahashi, X. Fan, Y. Koshikiya, and F. Ito, “Individual fault location in PON using pulsed pump-probe Brillouin analysis,” Electron. Lett. 47(25), 1384–1385 (2011). [CrossRef]

11. “Optical fibre cable maintenance criteria for in-service fibre testing in access networks,” ITU-T L.66

12. T. Horiguchi and M. Tateda, “BOTDA-nondestructive measurement of single-mode optical fiber attenuation characteristics using Brillouin interaction: Theory,” J. Lightwave Technol. 7(8), 1170–1176 (1989). [CrossRef]

13. T. Horiguchi, K. Shimizu, T. Kurashima, M. Tateda, and Y. Koyamada, “Development of a distributed sensing technique using Brillouin scattering,” J. Lightwave Technol. 13(7), 1296–1302 (1995). [CrossRef]

14. S. D. Personick, “Applications for quantum amplifiers in simple digital communication systems,” Bell Syst. Tech. J. 52(6), 329–331 (1989).

15. Y. Aoki, “Properties of fiber Raman amplifiers and their applicability to digital optical communication systems,” J. Lightwave Technol. 6(7), 1225–1239 (1988). [CrossRef]

Individual loss distribution measurement in 32-branched PON using pulsed pump-probe Brillouin analysis

Abstract

1. Introduction

2. Principle

3. Experiment

4. Signal to noise ratio analysis

5. Summary and discussion

References and links

Cited By

Figures (7)

Equations (20)

Optics Express