**3. Fiber interferometers for sensor applications**

In this section, we will present three selected prototypes of fiber interferometer sensors developed in our laboratory recently, which are intentionally employed in the electric power industry for partial discharge sensing as well as for power-frequency electric field measurements. Some preliminary experimental results for demonstrating the performances of these sensors and experimental systems also are presented. The prototypes to be presented below will include: a fiber Sagnac interferometer-based acoustic sensor, a fiber Fabry-Perot interferometer-based acoustic sensor, and a twin-grating fiber Fabry-Perot interferometer-based power-frequency electric field sensor.

### **3.1. A fiber Sagnac interferometer-based acoustic sensor**

In the previous work, we had proposed a novel fiber Sagnac interferometer-based acoustic sensor for sensing of high-frequency and weak ultrasonic waves produced by the partial discharges [18].

### *3.1.1. System configuration and operation principles*

When the interferometer is placed in a varying environment, the reflection spectrum will shift as a whole, as a function of measurand without noticeable changes in its envelop [26]. For a Bragg wavelength shift ΔλB, which is induced by the measurand, such as temperature or strain,

the reflected light power R, in turn, is altered. This feature has been well utilized in sensor applications for monitoring of the changes in the measurand through the detection of the

**Figure 3** displays a reflection spectrum of a twin-grating fiber Fabry-Perot interferometer with a 10-mm long intrinsic cavity, measured at 24°C. The Bragg wavelength of each grating is

In this section, we will present three selected prototypes of fiber interferometer sensors developed in our laboratory recently, which are intentionally employed in the electric power industry for partial discharge sensing as well as for power-frequency electric field measurements. Some preliminary experimental results for demonstrating the performances of these sensors and experimental systems also are presented. The prototypes to be presented below will include: a fiber Sagnac interferometer-based acoustic sensor, a fiber Fabry-Perot interferometer-based acoustic sensor, and a twin-grating fiber Fabry-Perot interferometer-based

′ <sup>λ</sup> = FP λ−ΔλB . When the measurand fluctuates,

a new reflectance ′FP is obtained as FP

150 Optical Interferometry

intensity changes of the interference signal [12].

1542.392 nm and the power reflectivity is about 15%.

**Figure 3.** Reflection spectrum of twin-grating fiber Fabry-Perot interferometer.

**3. Fiber interferometers for sensor applications**

power-frequency electric field sensor.

The scheme of the proposed sensor is presented in **Figure 4(a)**, which is based on a fiber Sagnac interferometer consisting of a 3-dB, 2 × 2 fiber coupler (OC) and two no-isolator, CW DFB lasers (LD1 and LD2) with two nearly identical lasing wavelengths at 1528.60 nm and 1529.38 nm, respectively. Two AC current signals <sup>1</sup> and <sup>2</sup> are obtained, respectively, from the corresponding photodiodes packaged in laser modules. The final detection signal <sup>3</sup> is constructed through <sup>a</sup> cross-correlation operation with <sup>1</sup> and <sup>2</sup>. <sup>A</sup> fiber polarization controller PC in the fiber loop is utilized to adjust the total birefringence of the fiber. The fiber loop made of a 1-km long single-mode fiber forms a sensor coil with a diameter of 60 mm. The outer layers of this coil, close to one end of the OC, are utilized as the sensing region, so that the relatively high sensitivity can be achieved.

**Figure 4.** (a) Scheme of proposed fiber acoustic sensor and (b) a group of output waveforms obtained in a static state.

The feature of this sensor is to employ two laser diodes without optical isolators as two individual light sources to illuminate the fiber Sagnac interferometer from two sides, and at the same time, also as two in-line optical amplifiers to enhance the intensities of signal beams from the Sagnac interferometer. So the sensitivity of this sensor, compared with other sensors with the normal configuration, has a significant enhancement.

In principle, this sensor is a balanced fiber Sagnac interferometer, which, in the structural analysis, can be regarded as two individual Sagnac interferometers with a mirror-image relation. According to the analysis methods proposed in Ref. [19], for a balanced fiber Sagnac interferometer, two fringe signals <sup>1</sup> and <sup>2</sup> including laser intensity noises can be expressed as

$$i\_1 \propto -I\_0 \sin \phi\_\mathbf{b} \sin \Delta \varphi + \eta\_{\text{LD}\_1} \tag{13}$$

$$i\_2 \propto +I\_0 \sin \phi\_\mathbf{b} \sin \Delta \phi + \eta\_{\text{LD}\_2} \tag{14}$$

where <sup>0</sup> is the average intensity of laser beams; b is the nonreciprocal phase difference between two beams propagating in the fiber coil, which is a constant and arises from the fiber birefringence; Δ is another phase difference between two beams, caused by external disturbances; LD1 and LD2 are the intensity noises in two laser diodes, which all can be regarded as a zero-mean, additive white Gaussian noise process.

After taking a cross-correlation operation by simply multiplying <sup>1</sup> and <sup>2</sup>, the final detection signal <sup>3</sup> is obtained, expressed as

$$\dot{i}\_3 = \left\langle \left| \dot{i}\_1 \cdot \dot{i}\_2 \right| \right\rangle \propto I\_0^2 \sin \phi\_\mathbf{b}^2 \sin \Lambda \phi^2 \tag{15}$$

where ⋅ represents an absolute value operation; ⋅ denotes a time average realized in electric circuit by a low-pass filter (LPF), and all intensity noise items included in <sup>1</sup> and <sup>2</sup> will be removed after this time-average operation. The item sinb above is defined as the fringe visibility or the scale factor [19], which can be maximized by adjusting the PC to achieve b <sup>=</sup> /2.

### *3.1.2. Experimental results*

We carried out several experiments based on this proposed sensor system. **Figure 4(b)** shows a group of signal waveforms obtained in a static state. The amplitude fluctuations in <sup>1</sup> and 2 traces reflect the intensity noises in laser beams. However, in <sup>3</sup> trace, these intensity noises have been obviously suppressed. As a direct benefit, the signal-to-noise ratio (SNR) of this sensor system is significantly improved.

We attempted to detect the acoustic emissions generated by high-voltage discharges between two electrodes. An experimental setup is illustrated in **Figure 5**, in which a pair of pin-type electrodes, connected with an AC high-voltage generator, was placed in an oil tank which was filled with transformer oil. The electrode gap was set at 8 mm. The fiber coil was attached to the shell of the oil tank, facing the electrodes. **Figure 6(a)** shows two groups of acoustic wave signals recorded at two time periods separated by about 10 seconds, when the imposed AC voltage was increased close to 5 kV and the discharges just started. It is obvious that, as shown in two signal traces, in this stage, the discharges in oil present a periodical change in intensity with a 50-Hz repetition frequency and the average intensity increases with time. **Figure 6(b)** displays a signal trace recorded a minute later. In this stage, the oil had become hot and dirty due to the discharge arcs burning the oil. Clearly, in this result, the signal amplitudes increase greatly and the repetition frequency of discharges then changes to 100 Hz, which indicates that the oil will be in a complete breakdown state and an oil-burning will come soon.

**Figure 5.** Schematic of an experimental setup for sensing of partial discharges in oil tank.

<sup>+</sup> <sup>1</sup> <sup>0</sup> <sup>b</sup> L1 <sup>D</sup> µ- D *Ii* sin sin f

µ+ D f

where

152 Optical Interferometry

signal

b <sup>=</sup> /2.

*3.1.2. Experimental results*

sensor system is significantly improved.

bances; LD1

and LD2

<sup>3</sup> is obtained, expressed as

as a zero-mean, additive white Gaussian noise process.

After taking a cross-correlation operation by simply multiplying

03

circuit by a low-pass filter (LPF), and all intensity noise items included in

2 b2 =× µ D *i i Ii* sin sin

f

where ⋅ represents an absolute value operation; ⋅ denotes a time average realized in electric

removed after this time-average operation. The item sinb above is defined as the fringe visibility or the scale factor [19], which can be maximized by adjusting the PC to achieve

We carried out several experiments based on this proposed sensor system. **Figure 4(b)** shows

have been obviously suppressed. As a direct benefit, the signal-to-noise ratio (SNR) of this

We attempted to detect the acoustic emissions generated by high-voltage discharges between two electrodes. An experimental setup is illustrated in **Figure 5**, in which a pair of pin-type electrodes, connected with an AC high-voltage generator, was placed in an oil tank which was filled with transformer oil. The electrode gap was set at 8 mm. The fiber coil was attached to the shell of the oil tank, facing the electrodes. **Figure 6(a)** shows two groups of acoustic wave signals recorded at two time periods separated by about 10 seconds, when the imposed AC voltage was increased close to 5 kV and the discharges just started. It is obvious that, as shown in two signal traces, in this stage, the discharges in oil present a periodical change in intensity with a 50-Hz repetition frequency and the average intensity increases with time. **Figure 6(b)** displays a signal trace recorded a minute later. In this stage, the oil had become hot and dirty

a group of signal waveforms obtained in a static state. The amplitude fluctuations in

1

traces reflect the intensity noises in laser beams. However, in

 j

> j

between two beams propagating in the fiber coil, which is a constant and arises from the fiber birefringence; Δ is another phase difference between two beams, caused by external distur-

<sup>0</sup> is the average intensity of laser beams; b is the nonreciprocal phase difference

2 2

 j

*n* (13)

<sup>1</sup> and

(15)

<sup>2</sup>, the final detection

<sup>1</sup> and

<sup>3</sup> trace, these intensity noises

<sup>2</sup> will be

<sup>1</sup> and 2

<sup>+</sup>*<sup>n</sup>* <sup>2</sup> <sup>0</sup> <sup>b</sup> L2 <sup>D</sup> *i I* sin sin (14)

are the intensity noises in two laser diodes, which all can be regarded

**Figure 6.** Measured partial discharges occurred in transformer oil. (a) A group of signal waveforms measured when partial discharges just started, and (b) signal waveforms measured a minute later.

**Figure 7.** Schematic of an experimental setup for electric field strength measurements.

We also tried to employ this sensor system to measure the strength of power-frequency electric field by detecting of the intensities of ultrasonic waves propagated in air, which were produced between two electrodes during ionization of air by the AC high-voltage electric field. The experimental setup is illustrated in **Figure 7**, in which the fiber coil was placed between two parallel, copper-plate electrodes separated by 10 cm.

In the experiment, the AC voltage imposed on the electrodes was increased gradually and the detection signals were recorded at several specified voltages. **Figure 8(a)** shows three groups of pulse train waveforms, recorded at 1800 V, 2800 V, and 3800 V, respectively. By observing these data, it is obvious that the number of pulses or pulse density within a cycle of AC voltage increases with the imposed AC voltage.

**Figure 8.** (a) Measured ultrasonic waves generated by air ionization, and (b) relationship between RMS output voltage and applied electric field strength.

**Figure 8(b)** shows a set of RMS (root-mean-square) voltages measured in a cycle of AC voltage under different electric field strengths from 175 to 475 V/cm. It is obvious that the RMS voltage increases proportionally with the applied electric field strength.
