A new photonic signal processor structure that can realise multiple-taps, with a general response capability, low-noise and widely tunable processor operation, is presented. It is based on a novel concept of employing positive and negative group delay slopes simultaneously by means of a dual-fed chirped fiber Bragg grating, and a new wavelength mapping scheme that enables wavelength re-use. The technique offers scalability, arbitrary responses with both positive and negative taps, tunability, and high frequency operation. Experimental results demonstrate widely tunable filters, with arbitrary bipolar tap generation, no PIIN noise, and with high FSR.
©2009 Optical Society of America
Photonic signal processing using optical delay line structures is a powerful technique for processing high speed microwave signals because it can overcome inherent electronic bottlenecks, and it also offers attractive features such as low loss, large time-bandwidth product and immunity to electromagnetic interference (EMI) .
Though several photonic signal processor structures have been previously reported, none of the previously reported photonic processor structures have solved the inherently difficult problem of simultaneously realising high-resolution, general response, low-noise and widely tunable processors. For instance, high-resolution or high-Q response requires a large number of taps, however previous photonic processors [2–4], have been limited by the excessive phase induced intensity noise (PIIN) generated by the optical interference in summing the multiple delayed optical signals [5,6]. Moreover, previous efforts to overcome the PIIN limitation [7–10], do not permit high-resolution, tunable processor operation. High-resolution, general response processors generically require a large number of taps and finite impulse response (FIR) to synthesise arbitrary taps. However, this invokes the critical problems of intractable hardware implementation since each tap basically requires its own delay structure, and of difficulty in tuning the multi-tap processor. These are hard though key problems, which have not been solved before. The issue of realising high-resolution, general response, low-noise, tunable photonic signal processors is a significant problem.
The object of this paper is to present a new photonic signal processor structure that can solve for the first time, the problem of realising high-resolution, general response, low-noise and widely tunable processors. It is based on the key innovations of employing positive and negative group delay slopes simultaneously by means of a dual-fed chirped fiber Bragg grating (CFBG) and a new wavelength mapping scheme that enables wavelength re-use. A feature of this topology is that it is scalable to a large number of taps, and that it can generate arbitrary responses with both positive and negative taps. Results are presented which demonstrate multi-tap processors with no PIIN noise, and with widely tunable performance.
2. Filter topology
Our concept is inspired by the fact that the FIR response has a symmetrical nature, as shown by the impulse response of Type I and Type II linear-phase FIR filters  illustrated in Fig. 1. Based on this feature, we introduce the idea of using both positive and negative group delay slopes versus wavelength simultaneously, by feeding a linearly CFBG with unity reflectivity from both its ends.
The topology of the new photonic signal processor is shown in Fig. 2. The WDM optical source, which can be implemented by either arrays of lasers or spectrum slices of a broadband source, is intensity modulated by an RF signal via a dual-input electrooptic modulator (EOM). As an example, spectrum sliced sources based on digital micro-mirror devices are available commercially , which can be used to generate a large number of wavelengths.
Bipolar taps can be obtained because the dual-input electro-optic modulator generates 180° phase difference between the modulation of its two arms. The output modulated signals from the EOM are then split into two branches and are launched into the two ends of a linearly CFBG with unity reflectivity. Hence the signals in Branch 1 and Branch 2 see opposite group delay slopes with positive dispersive group delay (TG (λi)) and negative dispersive group delay (T′G (λi)). The reflected modulated optical signals from the two ends of the chirped grating are detected by two photodetectors.
Our new wavelength mapping approach is explained with reference to an illustrative example for the impulse response of a typical filter to be synthesised, shown in Fig. 3(a). The wavelength mapping for the individual taps commences by synthesising the minimum delay τ that sets the fundamental time delay of the filter and its free spectral range (FSR). This is done by assigning λ1 to the nearest location from the centre wavelength λc (shown to the right of λc in Fig. 3(b)) so that the time delay between the reflection of the signal fed to the positive group delay slope end of the grating and the reflection from the other end corresponding to the negative group delay slope, is τ, (see Fig. 3(b)). We then assign λ2 to the left of λc to obtain a larger time delay such that |t1’ − t2| = τ = |t2’ − t1|, as required in Fig. 3(b). This wavelength mapping process continues recursively, each time increasing the time delay between successive taps by τ by alternating the wavelength assignment between opposite sides of the central wavelength λc, as shown in Figs. 3(b) and (a), until all the taps are assigned. The wavelength mapping procedure for synthesising an odd number of taps is similar, except that the first wavelength is allocated to the centre wavelength λc and the weight contribution for this centre tap comes from both branches. Note an important advantage of this new wavelength-mapping concept is that it is scalable to a large number of taps, and that it is efficient because it enables wavelength re-use to be used, so that for a given number of taps only half that number of wavelengths is required. As an example, it can be noted that CFBGs are currently commercially available with a bandwidth covering the entire C-band , and multiple-wavelength spectrum sliced sources are also available, which enables the potential of obtaining several hundred taps.
Hence the tap time delays obtained from the two ends of the linearly CFBG are combined in an interleaved manner with a common time delay difference τ between consecutive taps to form the FIR filter transfer function, by choosing the source wavelength allocation according to the following wavelength reuse scheme. Sources to satisfy this wavelength reuse scheme are calculated from the algorithm given below, for the cases of notch filters (1), multi-tap filters with even number of taps (2) and with odd number of taps (3),
where τ is the basic system time delay, τc is the group delay at λc and i = 1, 2, 3, … , N-1. Wavelength allocation corresponding to the even taps case and the odd taps case is solved by means of the recursive relations, Eq. (2) and Eq. (3) with its initial conditions respectively.
The amplitude response of the transversal filter with weights (Wi), corresponding to the output power of ith source can be expressed as,
Advantages of this topology, in addition to the wavelength reusability [14,15], include scalability, reconfigurability and tunability, and no PIIN noise. Reconfiguration of the bandwidth of the filter can be achieved by controlling the output power of the WDM sources, and scalability can be implemented by increasing the number of optical carriers. Both need only adjustment of the source part without modifying the rest of the structure. Tuning the centre frequency and the FSR of the multi-tap filter can be achieved by either reallocating a new set of wavelength or tuning the group delay slope of linearly CFBG without altering the centre wavelength . Additionally, very high frequencies of operation can be attained because the minimum delay, which is inversely proportional to the FSR, can be made very small, simply by choosing λ1 to be closer to the centre wavelength λc. Coherent interference effects and PIIN noise are also completely eliminated because of the use of two photodetectors and also because for the WDM signals the beat frequencies of the products fall outside the photodetector bandwidths.
3. Experimental results
To verify the proof of principle for the new structure, experiments were conducted. The setup was as shown in Fig. 2, except that due to the lack of a 100% reflectivity linearly CFBG, it was replaced by two identical 70% reflectivity linearly CFBGs having opposite group delay slopes, which were coupled individually to Branch 1 and Branch 2 with separate circulators. Note that for the dual-fed CFBG grating as shown in Fig. 2, the reflectivity of the grating needs to be close to 100% so that there is no feed-through leakage transmission which would degrade the performance. CFBG gratings having reflectivity of 99.9%, with the upper limit dependent on the bandwidth, are commercially available . As an alternative, we employed two separate but nominally identical 70% reflectivity CFBGs having opposite group delay slopes, which were coupled individually to Branch 1 and Branch 2 with separate circulators. Each of these gratings had a centre wavelength 1545.30nm, a bandwidth of 36nm, and one had a group delay slope of +330ps/nm and the other -330ps/nm. Errors in the CFBG time response between the two gratings can cause non-ideal tap delay response leading to RF response mismatch from the ideal case. However, the CFBGs used in the experiments had very low group delay ripple of about +-5ps and the results were in very good agreement with theory. The WDM sources comprised an array of external tunable lasers, followed by a 16 GHz dual input EOM, biased at the quadrature point. The wavelengths chosen were calculated from Eq. (1)–(3).
Experiments were set up on the new structure to investigate the realisation of several general response processors with different numbers of taps and FSR values. For the proof of concept demonstration and for ease of experimental setup, we demonstrated measurements for up to 11 taps, using only four wavelengths. A larger number of taps can be obtained by scaling up the number of taps according to the proposed wavelength reuse mapping by increasing the number of wavelengths e.g. using a spectrum sliced source, and it can be noted that this scalability does not require any modification in the rest of structure.
First, an 8-tap microwave photonic filter with positive unity weights was implemented. Figure 4 shows the frequency response and tunability of the filter. Starting from a fundamental filter centre frequency of 3.2GHz, which required laser wavelengths at [1541.986nm, 1543.880nm, 1545.773nm, 1547.677nm] calculated from Eq. (2), the filter frequency was tuned to 4.2 GHz by assigning a new set of source wavelengths [1542.775nm, 1544.218nm, 1545.661nm, 1547.104nm], and to a filter centre frequency of 4.8 GHz by assigning the source wavelengths to [1543.090nm, 1544.353nm, 1545.616nm, 1546.878nm]. This demonstrates a wide fractional frequency tuning range of 40%.
Next, the windowing characteristics of the filter were demonstrated for the 8-tap unity weighted filter at the centre frequency of 3.2GHz. Figure 5 shows that apodization using weightings [0.475, 0.608, 0.78, 1, 1, 0.78, 0.608, 0.475], implemented by modifying the laser power outputs, increases the sidelobe suppression from 13 dB in the original filter to 20 dB in the apodized filter, while slightly increasing in bandwidth.
Finally, a square-top microwave photonic filter with bipolar coefficients was demonstrated. This was designed to operate with an FSR of 3.1GHz. The 11-tap impulse response to be synthesized is shown in Fig. 6, and the coefficient values are [-0.1, 0, 0.2, 0, -0.6, 1, -0.6, 0, 0.2, 0, -0.1]. Because this contains zero weighted coefficients, 11 taps can be obtained from only four laser sources. The wavelength allocation to achieve the required filter response was calculated from Eq. (3), and the wavelengths are [1540.412nm, 1542.367nm, 1544.322nm, 1545.300nm, 1547.255nm, 1549.210nm]. The positive tap wavelengths 1542.367nm and 1545.300nm enter input Port 1 of the EOM, and the negative tap wavelengths 1540.412nm and 1544.322nm are coupled to input Port 2 of the EOM. The output power of each wavelength was adjusted according to the weighting profile except the one corresponding to the center wavelength in which case half of the value is required, since the weight contribution for this centre tap comes from both branches. Figure 7 shows the measured frequency response together with the theoretical frequency response calculated from Eq. (4). This shows very good agreement between measurement and calculated results.
A new photonic signal processor structure that can realise multiple-taps, with a general response capability, low-noise and widely tunable processor operation, has been presented. It is based on a novel concept of employing positive and negative group delay slopes simultaneously by means of a dual-fed CFBG, and a new wavelength mapping scheme that enables wavelength re-use. The technique offers scalability to a large number of taps, generation of arbitrary responses with both positive and negative taps, tunability, and high frequency of operation. Experimental results have demonstrated the realisation of filters with widely tunable performance, together with arbitrary bipolar tap generation, no PIIN noise, and operation with high FSR.
This work was supported by the Australian Research Council. The authors gratefully acknowledge Xudong Wang for his assistance in the experiments.
References and links
1. R. A. Minasian, “Photonic signal processing of microwave signals,” IEEE Trans. Microwave Theory Tech. 54, 832–846 (2006). [CrossRef]
2. B. Moslehi and J. W. Goodman, “Novel amplified fiber-optic recirculating delay line processor,” J. Lightwave Technol. 10, 1142–1147 (1992). [CrossRef]
3. N. You and R. Minasian, “A novel high-Q optical microwave processor using hybrid delay-line filters,” IEEE Trans. Microwave Theory Tech. 47, 1304–1308 (1999). [CrossRef]
4. E.H.W. Chan and R. A. Minasian, “High-resolution photonics-based interference suppression filter with wide passband,” J. Lightwave Technol. 21, 3144–3149 (2003). [CrossRef]
5. B. Moslehi, “Analysis of optical phase noise in fiber-optic systems employing a laser source with arbitrary coherence time,” J. Lightwave Technol. LT-4, 1334–1351 (1986). [CrossRef]
6. M. Tur, B. Moslehi, and J. W. Goodman, “Theory of Laser Phase Noise in Recirculating Fiber-Optic Delay-Lines,” J. Lightwave Technol. 3, 20“31 (1985). [CrossRef]
7. J. Capmany, D. Pastor, and B. Ortega, “New and flexible fiber-optic delay-line filters using chirped Bragg gratings and laser arrays,” IEEE Trans. Microwave Theory Tech. 47, 1321–1326 (1999). [CrossRef]
8. B. Vidal, V. Polo, J. L. Corral, and J. Marti, “Photonic microwave filter with tuning and reconfiguration capabilities using optical switches and dispersive media,” Electron. Lett. 39, 547–549 (2003). [CrossRef]
9. V. Polo, B. Vidal, J. L. Corral, and J. Marti, “Novel tunable photonic microwave filter based on laser arrays and N x N AWG-based delay lines,” IEEE Photon. Technol. Lett. 15, 584–586 (2003). [CrossRef]
10. F. Zeng and J. Yao, “Investigation of phase-modulator-based all-optical bandpass microwave filter,” J. Lightwave Technol. 23, 1721–1728 (2005). [CrossRef]
11. Tapio Saramaki, “Finite Impulse Response Filter Design,” in Handbook for digital signal processing, S.K. Mitra and J.F. Kaiser, eds. (Wiley, New York, 1993), pp. 155–189.
12. Newport Corporation, “Programmable spectral source user manual,” http://www.newport.com.
13. Proximion Fiber Systems, “Dispersion compensation module data sheet,” http://www.proximion.com.
14. B. Vidal, V. Polo, J. L. Corral, and J. Marti, “Efficient architecture for WDM photonic microwave filters,” IEEE Photon. Technol. Lett. 16, 257–259 (2004). [CrossRef]
15. Choong K. Oh, Tae-Young Kim, and Chang-Soo Park, “All-optical transversal filter with tap doubling and negative coefficients based on polarization modulation,” Opt. Express 15, 5898–5904 (2007) http://www.opticsinfobase.org/oe/abstract.cfm?URI=oe-15-10-5898. [CrossRef] [PubMed]
16. X. Y. Dong, P. Shum, N. Q. Ngo, C. L. Zhao, J. L. Yang, and C. C. Chan, “A bandwidth-tunable FBG filter with fixed center wavelength,” Microwave Opt. Technol. Lett. 41, 22–24 (2004). [CrossRef]
17. Redfern Optical Components, “Custom made FBG data sheet,” http://www.redferncomponents.com.