1. Introduction

Since the first demonstration of the high sensitivity of cavity ring-down spectroscopy (CRDS) by O’Keefe and Deacon in 1988 [1], CRDS has attracted more and more interests and been widely used for weak gas detection [2,3]. Unlike conventional direct measurement schemes [4], in which the attenuation of transmitted light through the target gas is measured, CRDS measures the intensity decay rate of an optical pulse as it is ringing down within a high-finesse optical cavity while interacting with the target gas [5]. The absorption loss of the target gas can be deduced from the pulse 1/e ring-down time [6]. Within the past ten years, high-finesse optical cavities have been extensively used to increase the sensitivity of gas detection with several obvious advantages, such as a longer effective interaction length of light and gas sample [7], an ultra-sensitive detection capability, and a simple experimental setup [8,9].

High reflection of mirrors and the stability of precision mechanism are crucial to CRDS. Compared with the conventional cavity ring-down (CRD) structure [10–13], in-fiber ring-down cavity [14–16], can be made very robust and inexpensive, avoiding alignment of any free-space components as well as other complications such as the effects of input beam quality and interference [17]. Light can be coupled into the loop in a variety of different ways using commercial fiber–fiber couplers [18,19], or custom interfaces [20], etc.

In order to improve the measurement precision effectively, some works have been done on CRDS with active fiber loop. The active fiber, as the gain medium, is used to compensate the power loss of the laser in CRD and increase the cavity round trips [16,17,19,21]. Ring-down time of tens of microseconds can readily be achieved and extended to several hundred microseconds [21]. But gain fluctuations noise and amplified spontaneous emission (ASE) spectrum noise will limit the accuracy and repeatability to a certain extent. Some means have been employed to eliminate the noises. For example, G. Stewart et al. [16,21] applied gain clamp compensation to stabilize the gain of the EDFA. N Ni et al. [22] proposed a digital least-mean-square (LMS) to reduce the ASE noise.

The cavity loss can be deduced from the pulse 1/e ring-down time. Then the concentration of target gas is obtained from the variation of the cavity loss, which is a key factor to measurement precision. In general, it is supposed that the cavity loss is a constant except for the gas absorption loss [16,21,23]. In fact, it is inevitable that some external factors (draw, pressure, twist, etc.) will bring a certain loss into the system and cause serious measurement errors especially in remote sensing. It is necessary to select an appropriate signal as reference for the cavity loss in real time.

Differential optical absorption method [24–29] gives a good way to obtain reference signal (baseline) for system. Generally, it employs two light sources, one of which is tuned to the absorption line peak to measure the absorption, and another that measured the baseline. It has already been investigated for its merit, such as in remote sensing of atmosphere [24,25], especially in lidar system [26–29]. Y.Chen et al. [13] used a light at a reference wavelength to measure the baseline as reference signal of cavity loss. And a semiconductor optical amplifier (SOA) worked as an optical amplifier as well as a switch to control injection of the laser light into the cavity. But there was no gain in the cavity to compensate cavity loss. Hence, its ring down cavity was consisted of superhigh reflectivity mirrors (over 99.9993%), which was very complicated and needed the alignment of free-space components as well as other complications.

We designed a novel active fiber CRD system based on dual wavelengths differential absorption method. Two lasers of different wavelengths were modulated by taking pulse signal and injected into a same fiber loop alternately. By introducing dual wavelength reference technology, it can eliminate the influence of the cavity loss variety and photoelectric device drift in the system effectively. Two fiber Bragg gratings (FBGs) were employed to get rid of ASE spectrum noise as filters. And then, different concentration ethyne samples were tested in order to further evaluate the effect of reference.

2. Principle

The fiber gas measurement method of CRDS is based on absorption spectrum. When a narrow pulse light is injected into the fiber loop and as the pulse rings around the cavity, a decaying pulse train will be coupled out. The initial intensity of the light is I₀. The intensity of the light after one cavity round trip is I [30].

The round-trip loss for the pulse light contains the total loss and the gain in the cavity, which can be written as [21]:

When ${\delta }_{\text{gas}}$ is zero, the inherent loss $\delta \text{'}$ is equal to the round-trip loss with 0% target gas in gas cell.

In general, ${\delta }_{\text{gas}}$ can be obtained by:

The inherent loss $\delta \text{'}$ doesn’t change ideally, but it is inevitable that some external factors (draw, pressure, twist, etc.) will bring a certain loss into the cavity especially in long distance sensing. That is to say the inherent loss $\delta \text{'}$ is not constant in fact. So the expression of Eq. (6) is not identical always. It is necessary to calibrate the inherent loss $\delta \text{'}$ in real time in some practical applications.

The wavelength λ₁ locates in an absorption line of target gas. When a narrow pulse light at wavelength λ₁ is injected into the fiber loop, a portion of the light will be absorbed by the target gas. So the fiber round-trip loss $\delta ({\lambda }_1)$will include the target gas absorption loss. To the opposite, the wavelength λ₂ is not covered with the target gas absorption line, so the light is not absorbed by the target gas. That is to say, the gas absorption loss is equal to zero at wavelength λ₂. According to Eq. (4), the round-trip loss $\delta ({\lambda }_2)$ is always equal to the inherent loss $\delta \text{'}$of the fiber loop. When the inherent loss $\delta \text{'}$ is changed by the external factors, the round-trip loss $\delta ({\lambda }_2)$will be changed synchronously. The variations of them are equal. So the round-trip loss $\delta ({\lambda }_2)$can be used as reference of the inherent loss in real time in the system.

By selecting the wavelength λ₁ and λ₂ to be very close to each other, the inherent losses under these two wavelengths will be the same. So the revised gas absorption loss at wavelength λ₁ is

The absorption relationship between the light power and concentration of target gas is governed by the Beer-Lambert law [30]:

The gas absorption loss can be obtained:

Then the gas concentration can be obtained:

Once the output pulse train is acquired, the 1/e ring-down time and the total loss can be reached. And the gas concentration can be obtained by the value of the gas absorption loss.

3. Experiment system

The configuration of the proposed dual wavelengths active fiber CRD gas detection system was shown in Fig. 1. In this system, the fiber loop is composed with coupler₂, coupler₃, delay line, gas cell, circulator, FBG₁, FBG₂ and an EDFA.

 

Fig. 1 Configuration of the proposed dual wavelengths active fiber ring-down gas detecting system (DFB LD: Distributed Feedback Laser Diode; FBG: fiber Bragg grating; PD: photo diode; EDFA: Erbium Doped Fiber Amplifier)

Download Full Size | PDF

Two DFB LDs (LD₁ and LD₂) are employed, whose wavelengths are λ₁ (1530.37 nm) and λ₂ (1530nm) respectively. And the output powers of LD₁ and LD₂ are 5.18mW and 5.23mW respectively under continuous-wave (CW) operation (driving current @50 mA; temperature @25°C). These two LDs are modulated by taking pulse signals and are controlled by a driver alternately. In another words, these two LDs work alternately to insure only one laser signal run in the loop. Temperature control is well done to keep the wavelengths of LDs stable.

The wavelength λ₁ locates in an absorption line of target gas (ethyne in this paper). The higher target gas concentration is, the higher round-trip loss is. To the opposite, the wavelength λ₂ is not covered with the target gas absorption line, and the light of LD₂ is used as reference of the round-trip loss in the system.

The gas cell is formed with a pair of collimators at a distance of 65 mm, whose structure is very simple and stable [31]. And its loss is 0.43dB. The split ratios of coupler₁, coupler₂ and coupler₃ are 0.499, 0.987 and 0.989 at 1530nm. The insertion loss of the circulator is 0.7dB. The reflectivity of the FBG₁ is 98% at 1530.39 nm and its full width at half maximum (FWHM) is 0.198nm, while the FBG₂’s is 98% at 1530.02 nm and its FWHM is 0.203nm. The length of FBG₁ and FBG₂ is 1cm, and the fiber length between them is about 1cm, so the difference of round-trip time between the sensing signal (λ₁) and reference signal (λ₂) can be neglected. Temperature control is needed for the FBG when it is used as a filter. A TEC temperature control system is employed to keep the FBGs’ center wavelengths constant. The FBGs are placed on a thermoelectric cooler in free condition with thermally conductive silicone to conduct heat, to make sure the FBGs aren’t affected by any forces. Target temperature is set to 27 °C in this paper. The effect of temperature control is monitored in real time by PC using a RS232 interface. The temperature will keep stable after 2 minutes and the temperature fluctuation is about 0.01 °C.

EDFA with Automatic Gain Control (AGC) is used to compensate the power loss of the laser in the ring-down cavity, which will increase the cavity round-trips and improve the precision of gas detection. Pump (980nm) and controlling circuit are integrated in it. It provides excellent gain flatness with average gain accuracy of ± 0.5dB. As some researches [16,21] mentioned, the gain fluctuations would be observed when EDFA operated, and it would be more serious when the gain was close to the threshold condition. The gain fluctuation can be decreased significantly by AGC, but small signal fluctuations are still observed. It is an efficient way to sacrifice a fraction of sensitivity to get stability. The gain of the fiber loop is kept less than the cavity inherent loss. In order to eliminate the instability of ring-down cavity caused by ASE noise, two FBGs are employed to get rid of ASE spectrum noise as filters. Besides gain noise from EDFA, the photoelectric device drift noise is also a source to worsen sensitivity of the system. In our system, the sensing signal (λ₁) and reference signal (λ₂) are detected by a uniform detection circuit. Hence, some circuit noise can be eliminated efficiently by comparing the sensing signal (λ₁) with reference signal (λ₂).

As a note, the EDFA can be placed between port 1 of the circulator and the gas cell, or between the gas cell and the delay line, which is useful to filter the ASE noise by FBGs. But we found that some high frequency noise was brought into the system without optical isolation before the EDFA. In this paper, the EDFA is placed after circulator as shown in Fig. 1. The high frequency noise isn’t observed in this connection sequence. Of course, it is possible to bring a little ASE noise into the loop, but it can be filtered by FBGs before the light is amplified again when the light rings in the loop. Besides, we make the ASE noise leak to the detector as small as possible by adjusting the suitable gain of EDFA. Unlike high frequency noise, the ASE noise only produce a small direct current (DC) signal (offset) which is easier to remove, so that the influence can be neglected.

The DFB LDs emit a 900 ns wide pulse at a 10 kHz repeat rate into the fiber loop, alternately. A delay line (252m) enough long is absolutely necessary to separate two adjacent laser pulses in time domain. The output pulse train (ring-down signals) from port 2 of the coupler₃ is received by a high-speed photo detector, and then the signal is averaged for 16 times by the oscilloscope (Agilent DSOX2014A) to reduce the noises.

4. Experiments and results

When a narrow pulse light is injected into the loop, it will ring around the cavity, and dozens of decaying pulses can be observed. The maximum values of all pulses are recorded, and the data are fitted to an exponential decay curve in the format of

All the calibrating gases used in this paper were provided by the certified standard gas sample supplier. The test gas samples were filled in the gas cell at the standard atmospheric pressure. When the pure nitrogen (0% ethyne) was filled in the gas cell, the output ring-down signals and the fitted curves were shown in Fig. 2. Signal₁ and signal₂ were measured ring-down signals at wavelength λ₁(1530.37nm) and λ₂(1530nm) respectively. The dashed curves of fitting₁ and fitting₂ gave the fitting exponential decay curves of the peaks of signal₁ and signal₂ respectively. The first pulse of the train is coupled out directly, that is to say it doesn’t pass through the gas cell and doesn’t give any information about the gas concentration. Thus it should be unconsidered in calculating.

 

Fig. 2 Output ring-down signals and the fitted curves of pure nitrogen

Download Full Size | PDF

The parameters of fitting exponential decay curves were shown in Table 1. From Table 1, the 1/e ring-down time with pure nitrogen at wavelength λ₁ is 10.27452μs. The correlation coefficient R² of the fitting is larger than 0.998, which shows the fitting exponential decay curve with better accuracy. The round-trip loss is calculated as 0.52837 dB by the Eq. (3). The target gas absorption loss is zero in the condition. The 1/e ring-down time with pure nitrogen at wavelength λ₂ is 10.26467μs. The round-trip loss is calculated as 0.52888 dB according to Eq. (3), which is consistent with the round-trip loss at wavelength λ₁, with the relative deviation of 0.096%.

Table 1. Parameters of fitting exponential decay curves

View Table

From Fig. 2 we can see that the measured round-trip time of the pulse light $\tau$ is 1.25 μs, which describes the interval between two adjacent output pulses. For SMF-28, the effective refractive index of the core n_eff is 1.468, so the total length of the loop $l$ is calculated about 255.4m by the equation$l=\tau \cdot c/n_{eff}$, which is the length of the delay line and the connected optical fiber of other components in the loop. The length is enough to separate two adjacent laser pulses in time domain.

In the same way, the output ring-down signals and the fitted curves of 10% calibrating ethyne in nitrogen were shown in Fig. 3. Signal₁ and signal₂ are measured ring-down signals at wavelength λ₁(1530.37nm) and λ₂(1530nm) respectively. Fitting₁ and fitting₂ give the fitting exponential decay curves of the peaks of signal₁ and signal₂ respectively. Parameters of them are also shown in Table 1. From Table 1, the 1/e ring-down time with 10% ethyne in the gas cell at wavelength λ₁ is 2.20208μs. Once more using Eq. (3), the round-trip loss is calculated as 2.46528dB, which contains the target gas absorption loss. The 1/e ring-down time with 10% ethyne at wavelength λ₂ is 10.26025 μs from Table 1. The round-trip loss is 0.52911 dB. The target gas absorption loss is zero in the condition, because the wavelength λ₂ is not covered with the absorption line of ethyne. According to Eq. (7), the absorption loss in 10% ethyne at 1530.37nm is about 1.93617dB.

 

Fig. 3 Output ring-down signals and the fitted curves of 10% ethyne

Download Full Size | PDF

Several different calibrating samples mixed by ethyne and nitrogen were tested. Measured 1/e ring-down time versus gas concentration was shown in Fig. 4(a). The black-square and red-triangle curve represent 1/e ring-down time t_r(λ₁) and t_r(λ₂)of the output pulses at wavelength λ₁ and λ₂ respectively. 1/e ring-down time t_r(λ₁) decreases with increasing of ethyne concentration, while 1/e ring-down time t_r(λ₂) fluctuates slightly. The round-trip losses can be obtained by Eq. (3). The round-trip loss δ(λ₁) at wavelength λ₁ increases linearly with increasing of the ethyne concentration as the black-square curve shown in Fig. 4(b). If external factors don’t bring any loss into the cavity, the inherent loss of the loop is almost invariant. So the round-trip loss δ(λ₂) at wavelength λ₂ changes very slightly as red-triangle curve shown in Fig. 4(b). The gas absorption losses can be obtained by Eq. (7). The relationship between gas concentration and absorption loss is shown in Fig. 4(c). Experimental data are shown by black-square curve which can be fitted to a linear curve in the format of$y=kx+b$.

 

Fig. 4 (a). the 1/e ring-down time versus ethyne concentration; (b). the round-trip loss versus ethyne concentration; (c). the ethyne concentration versus absorption loss.)

Download Full Size | PDF

The equation of the fitted line (red in Fig. 4(c)) can be described as

In order to evaluate the effect of the reference in this method, an attenuator was employed to give an additional loss Δδ to the loop cavity. And a sample of 0.1% ethyne in nitrogen was tested. Both the round-trip loss δ(λ₁) and δ(λ₂) will synchronously change with the additional loss as shown in Fig. 5(a). The black-square and red-triangle curve represent the round-trip loss at wavelength λ₁ and λ₂ respectively. And the varieties of the round-trip loss δ(λ₁) and δ(λ₂) are equal to the additional loss Δδ. The measured absorption loss δ_gas of the sample gas can be obtained from Eq. (7) as Fig. 5(b) shown, which is less than 0.02dB. That is why the curves representing the round-trip loss δ(λ₁) and δ(λ₂) are very close to each other in Fig. 5(a). Because the round-trip loss δ(λ₁) and δ(λ₂) change with the additional loss synchronously, the measured absorption loss of definite gas is invariable according to Eq. (7). The measured concentration C can be obtained from Eq. (12) as shown in Fig. 5(c). And the relative deviation of measured concentration is found to be less than ± 0.4% of 0.1% ethyne when a certain additional loss from 0 to 1.2dB is introduced to the cavity. The results show it can efficiently eliminate the influence of the inherent loss variation in the system and maintain the measurement with a high accuracy by introducing dual wavelength reference technology.

 

Fig. 5 (a). the round-trip loss versus additional loss; (b). the absorption loss versus additional loss; (c). the ethyne concentration versus additional loss.)

Download Full Size | PDF

In order to further evaluate the stability of the system, the 0.1% calibrating sample mixed by ethyne and nitrogen had been measured for 24 hours. The data was recorded per twenty minutes. And the concentration relative deviation is less ± 0.5% for 24 hours. The result in Fig. 6 shows that the system has a good performance in stability and sensitivity.

 

Fig. 6 Ethyne concentration variation for 24 hours

Download Full Size | PDF

5. Conclusion

A novel active fiber loop CRD system based on dual wavelengths differential absorption method is designed. One wavelength centered within and another centered outside of the target gas absorption line. By introducing dual wavelength reference technology, it eliminates the influence of the cavity loss variety and photoelectric device drift in the system effectively. An EDFA with AGC is used to compensate the power loss of the laser in the ring-down cavity, in order to increase the cavity round trips and improve the precision of gas detection. And two FBGs are employed to filter ASE spectrum noise. And then, some calibrating mixed samples of ethyne and nitrogen are tested with a 65 mm long gas cell to further evaluate the effect of reference wavelength. The results show the relative deviation is found to be less than ± 0.4% of 0.1% ethyne when a certain additional loss from 0 to 1.2dB is introduced to the cavity. And the relative deviation of measured concentration is less than ± 0.5% over 24 hours.