All-fiber wavemeter and Fourier-transform spectrometer
Mark Froggatt* and Turan Erdogant
* NASA Langley Research Center, Hampton, VA 23681 t The Institute of Optics, University of Rochester, Rochester, NY 14627 (Tel: (716) 275-7227; Fax: (716) 244-4936; e-mail: [email protected])
Fiber phase gratings fabricated by ultraviolet iddiation are known to couple light out the side of the fiber [1,2]. Simply put, the angle of the out-coupled radiation, a,measured with respect to the Itorma1 of the fiber axis, is related to the wavelength of light, A, and the period of the grating, A, by sins = n e . - A/A, where ne# is the effective refractive index of the fiber mode. Since the angle is dependent on wavelength, a simple spectrometer can be made by directly measuring the out-coupled intensity vs. angle [3]. This can be accomplished by focusing the radiation onto a detector array. Wagener, et al, have demonstrated such a device with excellent performance using a chirped fiber grating to focus the light through an index-matched prism onto a linear array [4]. In this paper we propose and demonstrate an analogous but distinct all-fiber device in which the outcoupled radiation is measured directly on the cladding of the fiber, making for a highly compact, simple device. Rather than measuring the far-field (focused) radiation, here we measure the interference pattern formed between counter-propagating out-coupled beams. Monochromatic light produces a sinusoidal pattern, while polychromatic light produces the Fourier transform of the spectrum. Here we demonstrate this device, and describe how it can be used as a high-resolution wavemeter, a network spectrum monitor (analogous to [4]), and a unique tool for in situ analysis of the fabrication process or post-fabrication testing of very long fiber Bragg gratings with demanding specifications (such as those used for dispersion compensation). ,-CCDor Figure l(a) shows a diagram of the device in one possible configuration. Here it is assumed that the counterpropagating light is produced by an in-line reflector whose reflectivity is chosen to optimize the out-coupled signal at the expense of fiber grating reflector transmitted loss (typically a few % is sufficient). A n air-gap reflector or a very short, broad-band fiber grating may be used for the reflector. If more signal is desired with minimum added loss, a c slightly less compact version is shown in Figure Figure 1: Two configurations for the Fiber Grating Fourier Transform SDectrometer. l(b), wherein two gratings and a loop are used. The light on the detector array is the superposition of the radiation from both gratings. To demonstrate this device, a fiber Bragg grating with a period of 533 nm (Bragg wavelength of 1545 nm) and a peak UV-induced index change of 1.4~10-3was used in the configuration of Fig. l(a). The grating had a gaussian profile with a FWHM of about 5 mm. A cleaved end-face of the fiber produced the reflected wave. The input consisted a tunable 780-nm external-cavity semiconductor laser (New Focus), which coupled light out exactly normal to the fiber axis at a wavelength of 779.87 nm. The outcoupled light was measured using a standard, high-dynamic range (16-bit) CCD chip with a 9 pm pixel spacing. The chip was more than a centimeter from the fiber since no special effort was made to remove the protective window (which ended up being partly responsible for extraneous interference fringes).
PD21-1
This setup was intended to demonstrate simultaneously the ability of the device to function as a wavemeter or spectrometer, and how a standard fiber grating designed for 1550-nm reflection may be characterized using a 780-nm source. Figure 2(a) shows a section of each of four fringe patterns ob,tained by illuminating the device with four different wavelengths. The wavelengths are 776.72, 777.22, '777.72, and 778.22 nm. Note that the mere 4% reflection producing the counter-propagating wave yields a fringe visibility of about 38%. Figure 2(b) shows a plot of the average fringe intensity vs. position along the array for each of the four cases in Fig. 2(a).
Figure 2: (a) Fringes measured at four different wavelengths; (b) average fringe intensity for each of the four patterns in part (a). Although the fringes appear to be reasonably well defined, it is evident from the fringe intensity plots in Fig. 2(b) that the fidelity for this demonstration is not very good. This is due in part to the use of a standard, windowed CCD array positioned at some distance from the fiber, rather than contacting the array to the fiber as shown in Fig. l(a). In theory, the spectrum I(A) is given by the magnitude of the Fourier transform of the measured fringe intensity I(z) minus the DC contribution lo,or
I ,
where K, is the spatial frequency. The transform may be readily computed using a Fast-FourierTransform (FET) routine on a computer, essentially in real time. For a fringe intensity pattern Z(z) consisting of N points (typically N is chosen to be 2m, where m is an integer), only the first N / 2 points are meaningful. These correspond to spatial frequencies varying between 0 5 Kz I ~ / pwhere p is the , pitch of the detector array (p = 9 pm for the experiments here). From this theory it can be shown that the range of measurable wavelengths U about a nominal wavelength A is approximately given by
M G - .
A2
4neffP
(2)
For example, for p = 10 pm at 1550 nm, the range is AA 42 nm. The resolution of the measurable spectrum is determined by the length Np of the detector array (assuming the grating length exceeds this
PD2 1-2
length). The minimum resolvable spatial frequency is the minimum resolvable wavelength interval ail:
Kz = 2n/Np, which
enables us to approximate
6.
2 =-. ~ a %?ffNP N
a2
So,for example, an array with a pitch of 10 pm and 1024 pixels (1 cm long) has a resolution of 6il= 0.08 nm a t 1550 nm. For the 780-nm experiment demonstrated here, the theoretical range and resolution for p = 9 pm andN = 512 are AA 12 nm and SA 0.045 nm. Taking the Fourier transform of the data sets in Fig. 2(b), we show the spectra for the tunable laser at each of the four wavelengths shown in Fig. 3(a). These may be compared to the spectra measured on a standard Optical Spectrum Analyzer (Ando) shown in Fig. 3(b) (with a 0.05 nm resolution). The wavelengths separated by 0.5 nm are clearly resolvable, although the spectra are somewhat noisy. This is mainly caused by extraneous interference fringes from the imperfect measurement system.
1.2
o-o .+
775
Figure 3:
.I
776 777 778 779 Wavelength (nm)
780
775
776 777 778 779 Wavelength (nm)
780
(a) Spectra of tunable laser at four different wavelengths measured with the Fiber Grating Fourier Transform Spectrometer; (b) same spectra measured on a standard optical spectrum analyzer (linear scale).
For in situ or post-fabrication characterization of 1550-nm fiber Bragg gratings, the 780-nm source wavelength is fixed. Then the opticalperid (ne#) of a short section of the grating may be monitored directly as that section of the grating is being fabricated. We estimate that standard (unblazed) gratings with a UV-induced index change as small as 5x10-5 may be monitored in this way. In conclusion, we present a device concept that may be applied to an extremely compact wavemeter or spectrum analyzer for source stabilization or network monitoring, but which also forms the basis for a nearly ideal long fiber Bragg grating fabrication tool, in that it allows direct monitoring of the optical period of the grating as it is fabricated.
[l] G. Meltz, W. Morey, and W. Glenn, In4iber Bragg grating tap, Paper TuG1, Opt. Fiber. Comm. Conf. (1990). [2] T. Erdogan and J. Sipe, Tilted fiber phase gratings,J. Opt. Soc. Am. A, 13,296 (1996). [3] P. St. J. Russell and R. Ulrich, Grating-fiber coupler as a high-resolution spectrometer, Opt. Lett., 10, 291 (1985). [4] J.L. Wagener, T. A. Strasser, J. R. Pedrazzani, J. DeMarco, and D. J. DiGiovanni, Fiber grating optical spectrum analyzer tap, Paper 448, Euro. Conf. on Opt. Comm. (1997).
PD2 1-3
