-
依据图4所示FBG解调实验系统进行实验。解调仪使用线阵InGaAs图像传感器[16]测量FBG反射谱的变化,该图像传感器在1525~1570 nm波长的范围内有256个像素点,像素间隔约为0.176 nm。FBG的3 dB带宽为0.2~0.3 nm。
-
该实验分单光栅实验和多光栅实验,探究不同波峰数的光谱中光强峰值和曝光周期的关系。
-
将单光栅光纤连入解调仪的1通道,分别在不同的曝光周期下曝光解调,求得不同曝光周期下谱峰的光强峰值,并根据记录的峰值-曝光周期点绘制关系图,结果如图5所示,由图可知在曝光周期达到200左右时,谱峰达到过饱和,峰值基本不再变化。
由于该实验研究的是光谱光强饱和值之下的峰值变化规律,所以截取饱和值之下的数据进行分析,比较不同方式的拟合效果,结果如表1所示。
其中线性拟合和多项式拟合的决定系数最大,说明线性拟合和多项式拟合x对y的解释程度非常高,从计算效率考虑,线性拟合优于多项式拟合,因此在单峰光谱的情况下可以认为峰值和曝光周期呈一次线性关系。
Fitting mode Relational expression Determining coefficient R2 Exponential fitting y=1 904e0.021 6x 0.754 4 Linear fitting y=245.93x−669.76 0.999 99 Logarithmic fitting y=10 877ln(x)−24 004 0.822 6 Polynomial fitting y=0.004 3x2+245.19x−652.06 0.999 99 Power function fitting y=94.961x1.204 7 0.987 2 Table 1. Comparison of different fitting methods of single grating fiber peak light intensity-exposure cycle
-
该实验选取的多光栅光纤带有10个光栅,将光纤光栅连入解调仪的2通道,多次曝光解调,记录不同曝光周期下每个峰的峰值,并根据每个峰的峰值-曝光周期数据绘制关系图,结果如图6所示,由图可知,每个谱峰的峰值与曝光周期也近似线性拟合,故分别进行线性拟合,结果如表2所示。
Peak serial number 1# 2# 3# 4# 5# Linear relation y=422.3x+1.944 y=260.4x+259.7 y=337.4x+62.25 y=211.7x+175.2 y=235.6x+352.8 Determining coefficient R2 0.999 75 0.999 97 0.999 69 0.999 74 0.99978 Peak serial number 6# 7# 8# 9# 10# Linear relation y=289.6x+239.3 y=254.1x+367 y=181.3x+135.6 y=221.8x+425.3 y=195.7x+346.7 Determining coefficient R2 0.999 996 0.999 98 0.998 99 0.999 89 0.999 99 Table 2. Linear fitting results of multi-grating fiber peak light intensity-exposure cycle
由于表2中线性拟合的决定系数极接近1,因此对于多峰光谱峰值和曝光周期之间的关系能够用一次线性关系来表述。
-
影响寻峰稳定性的因素包括噪声类型、光纤布拉格光栅谱型、噪声大小和寻峰算法的种类。噪声大小反映的是信噪比的大小,噪声类型和光纤布拉格光栅谱型的影响比噪声大小的影响相对较小,也不影响寻峰算法本身的优劣比较。信噪比对于寻峰稳定性的影响最大,信噪比大小的改变对寻峰稳定性的影响超过相同信噪比时各种算法间稳定性的差异,且信噪比对所有算法的影响基本一致[17]。在FBG反射光谱中信噪比的计算公式为:
式中:P表示谱峰的凸起高度,谱峰的凸起高度与峰值有关;N表示当前曝光周期下的全像素暗电流标准差,不同曝光周期下的标准差基本一致,可以当成常数。因此,信噪比的大小与谱峰峰值直接相关。经过上述分析可知,影响寻峰稳定性的主要因素为谱峰峰值,所以该实验主要研究谱峰峰值对寻峰稳定性的影响。
光纤光栅反射光中心波长的变化反映了外界信号的变化,针对FBG中心波长的分析处理是光谱解调算法的关键,因此衡量寻峰的稳定性从寻峰得到的中心波长稳定性分析。该实验选取了三个谱峰进行分析,在不同峰值下各采集10000帧光谱的数据,分别计算在不同光强峰值下的中心波长标准差,结果如图7所示。
分析并总结,由以上关系图可知随着谱峰峰值的增加,中心波长的标准差逐渐减小,当峰值达到饱和时,中心波长标准差急剧增加。所以图7中最优峰值区间是35 000~45 000,在该区间中心波长标准差在0.5 pm以内,谱峰峰值在10000以下时,中心波长标准差在1 pm以上,谱峰峰值35 000~45 000时的中心波长标准差比峰值在10 000以下时减小了50%,稳定性提高了50%。图7中光谱饱和值大约是50000,因此可以选取在饱和值的70%~90%内为最优峰值区间,使寻峰具有更高的稳定性,稳定性在0.5 pm以内。
-
该实验选取具有22个光栅的光纤传感器,由于在光路中产生了损耗,所以导致光谱强度严重不一致,光谱如图8(a)所示。根据图7所示的谱峰峰值与中心波长标准差的关系可知,图8(a)所示光谱图中后11个谱峰的峰值在5 000以下,产生了光强衰减的现象,严重影响了寻峰的稳定性。利用文中提出的自适应解调算法对光谱多次调节,结果如图8所示。
Figure 8. Adaptive demodulation spectrogram. (a) Exposure cycle/100 ns:234.4; (b) Exposure cycle/100 ns:304; (c) Exposure cycle/100 ns:503.1; (d) Exposure cycle/100 ns:659.6; (e) Exposure cycle/100 ns:967.2, (f) Exposure cycle/100 ns:2832.6; (g) Exposure cycle/100 ns:3881.2; (h) Exposure cycle/100 ns:5562.8
在对光谱自适应解调之前,首先求出光谱的饱和值,即用曝光周期上限值对光谱解调,可以得到光谱的饱和值为48380;然后根据70%~90%的最优峰值区间范围划定峰值区间,认为峰值超过饱和值90%的谱峰为不稳定谱峰;再根据自适应解调算法求出剩余谱峰中最高峰达到最优峰值区间的曝光周期,结果如表3所示。
Adjustment order Exposure cycle/100 ns Spectral peak number for each demodulation Peak/saturation value 1 234.4 1# 85.90% 2 304.0 2# 89.05% 3 503.1 3#, 4#, 5#, 6# 72.74%, 89.70%, 73.22%, 76.68% 4 659.6 7#, 8#, 9# 88.47%, 83.36%, 79.59% 5 967.2 10#, 11# 89.61%, 88.13% 6 2 832.6 12#, 13#, 15#, 16#, 17# 88.52%, 73.87%, 72.71%, 74.07%, 79.82% 7 3 881.2 14#, 18#, 19#, 20# 88.75%, 85.88%, 76.35%, 76.28% 8 5 562.8 21#, 22# 87.52%, 75.02% Table 3. Experimental results of adaptive demodulation algorithm
以表3中的曝光周期分别曝光解调,图8中(a)~(h)为每次解调得到的光谱图。该实验共经过八次曝光解调,每次只取峰值位于最优峰值区间内的解调结果,避免了不稳定谱峰的解调结果。图8中超出最优峰值区间的谱峰,属于不稳定谱峰,会严重影响寻峰的稳定性。实验结果如表3所示,解调得到的谱峰峰值均位于饱和值的70%~90%的最优峰值区间,寻峰得到的中心波长标准差在0.5 pm以内,寻峰稳定性比单次曝光解调提高了50%。程序运行时间为74.422 ms,可以满足快速解调的要求。
Adaptive demodulation of reflection spectrum intensity of fiber grating link
doi: 10.3788/IRLA20200440
- Received Date: 2021-11-03
- Rev Recd Date: 2022-02-27
- Available Online: 2022-06-08
- Publish Date: 2022-06-08
-
Key words:
- optical fiber sensing /
- spectral intensity /
- exposure cycle /
- peak-seeking threshold /
- adaptive demodulation algorithm
Abstract: In view of the large difference in the intensity of single channel multi-grating reflection spectrum of demodulator, which leads to the failure of peak-seeking or the increase of peak-seeking error, a method of multiple exposure demodulation in different exposure cycles was proposed to adjust the peak value, and the peak-seeking threshold was determined according to the histogram of the spectral data. The influence of peak value of spectrum on the stability of peak-seeking was analyzed, the adaptive adjustment rules of exposure cycle and peak-seeking threshold were established, and the adaptive peak-seeking demodulation algorithm was realized by LabVIEW software. Through the actual fiber grating sensor test, the automatic exposure and peak-seeking demodulation of the reflection spectrum with large difference could be completed. On the premise of ensuring the peak-seeking stability, the number of spectral peak identification, the adaptive ability of the demodulation system and the working reliability were effectively improved. The experimental results show that the peak stability is the highest when the peak value is in the range of 70%-90% of the saturation intensity. The standard deviation of the center wavelength obtained by peak-seeking is within 0.5 pm, the stability is increased by 50% compared with single exposure demodulation, and the program running time is within 100 ms, which can realize fast demodulation.