Bandwidth enhancement of SOA-based switches using optical filtering: theory and experimental verification

doi:10.1364/OE.14.001260

« Show journal navigation

Bandwidth enhancement of SOA-based switches using optical filtering: theory and experimental verification

Mads Nielsen and Jesper Mørk »View Author Affiliations

Optics Express, Vol. 14, Issue 3, pp. 1260-1265 (2006)
http://dx.doi.org/10.1364/OE.14.001260

View Full Text Article

Acrobat PDF (207 KB)

Journal Search
Article Lookup

Article Tools

Citations

Alert me when this article is cited
Export Citation/Save

Article Navigation

▸ Article Outline
– Abstract
1. Introduction
2. Theoretical predictions
3. Experimental results
4. Summary
– References and links

Article Tools
Full Text
Article Info
References
Cited By
Figures
Metrics
Download PDF

Viewing Equations in the
HTML Article

Optimal Viewing Recommendations

Full Text
Article Info
References (7)
Cited By
Figures (4)
Metrics

Abstract

We present small-signal response measurements on all-optical switches based on a single SOA followed by an optical filter. Both asymmetric MZI and band-pass type filters are investigated and theoretical predictions of a large modulation bandwidth enhancing effect are verified. The small-signal measurements of the DISC switch (SOA and asymmetric MZI) are the first of their kind.

1. Introduction

All-optical switches based on semiconductor optical amplifiers (SOAs) have demonstrated great potential for wavelength conversion, regeneration, and logic operations [1-3

Y. Liu, E. Tangdiongga, Z. Li, S. Zhang, H. de Waardt, G. D. Khoe, and H. J. S. Dorren, “160 Gb/s SOA-based wavelength converter assisted by an optical bandpass filter,” Technical Digest of OFC 2005 , PDP 17, Anaheim, (2005) [CrossRef]

]. The advantages of SOA-based architectures include relatively high-speed operation, small footprint and low power consumption (high nonlinearity). However, the carrier density modulation (CDM), which provides the gain and/or phase modulation facilitating the switching operation, also leads to the problem of patterning effects at high bitrates.

In a recent theoretical study we have shown that the CDM-induced patterning effects may be divided into two categories, namely linear patterning and nonlinear patterning [4

M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]

]. Linear patterning is the contribution stemming from the slow recovery of the switched signal, and this is the dominant patterning effect for switches based on a) cross-gain modulation (XGM), and b) so-called standard-mode interferometric switch configurations, such as Mach Zehnder (SOA-MZI) or Michelson interferometers (SOA-MI), where the phase of only one of the interfering signals is modulated [4

M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]

]. The problem of linear patterning, which manifests itself as inter-symbol interference, has been addressed by operating interferometric switches in the so-called differential mode (DM) [2

Y. Ueno, S. Nakamura, and K. Tajima, “Nonlinear phase shifts induced by semiconductor optical amplifiers with control pulses at repetition frequencies in the 40-160 GHz range for use in ultrahigh-speed all-optical signal processing,” J. Opt. Soc. Am. B 19, 2573 (2002) [CrossRef]

], which effectively cancels the slow recovery of the carrier density induced phase changes. This approach creates a short switching window, and eliminates the patterning contribution from the slow recovery. In other words, it eliminates linear patterning. However, in the absence of linear patterning, the effect of nonlinear patterning dominates [4

M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]

] – an effect which can be attributed to carrier density saturation and is observed as a fluctuation of the switched pulse power. Nonlinear patterning is the biggest challenge in operating SOA-based switches at bitrates beyond 100 Gb/s, and potential remedies include a combination of holding beams and filtering [3–4

J. Leuthold, D. M. Maron, S. Cabot, J. J. Jaques, R. Ryf, and C. R. Giles, “All-optical wavelength conversion using a pulse reformatting optical filter,” J. Lightwave Technol. 22, 186 (2004) [CrossRef]

In this paper we experimentally demonstrate that linear patterning can be compensated by optical switches consisting of an SOA and an optical filter. Two types of optical filters are investigated, namely the asymmetric MZI (AMZI) filter (see Fig. 1), and a band-pass filter (BPF). The combination of an SOA and an AMZI is often referred to as the delayed interference signal converter (DISC) [2

, 3

]. The SOA is operated in the small-signal regime, which per definition eliminates nonlinear patterning [4

M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]

]; a small-signal analysis thus provides a best-case estimate for the modulation bandwidth of the system. We measure, for the first time to our knowledge, the small-signal frequency response (SSFR) of a DISC switch. The experimental results are compared to theoretical predictions in Ref. [4

M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]

] and show excellent agreement.

2. Theoretical predictions

In Ref. [4

M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]

] it was predicted that the small-signal carrier density response (CDR), also referred to as the XGM response, of an SOA can be compensated by passing the modulated probe signal through an equalizing optical filter. The normalized SSFR for the CDR, T ^CDR _N(Ω) = 1/(-jΩτ_e -1), where Ω is the modulation angular frequency, is a low-pass function with a cut-off identified by an effective carrier lifetime τ_e and a high-frequency roll-off of -10 dB/decade. For the AMZI filter, it may be shown that by properly adjusting the phase bias Φ₀ of the AMZI, the response of the SOA can be flattened, reaching a -2 dB bandwidth of Ω_2dB = π/τ , where τ is the differential delay of the AMZI [4

M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]

]. The corresponding value for the phase bias, Φ^f ₀ , is determined by [4

M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]

]

τ_{e}^{- 1} \tan (Φ_{0}^{f} / 2) = \frac{2}{ατ}

(1)

where α is the linewidth enhancement factor.

The normalized SSFR of the DISC can be described by [4

M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]

]

T_{N}^{AMZ} (Ω) = {0.5 T}_{N}^{CDR} (Ω) [(1 + e^{j Ωτ}) - γ (1 - e^{j Ωτ})]

(2)

where γ = αtan(Φ₀/2) is referred to as the bandwidth enhancement factor, since choosing Φ₀ according to Eq. (1) enhances the small-signal bandwidth by approximately a factor of γ [4

M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]

Moreover, a filter function providing a flat SSFR can be shown to have the following slope at the carrier wavelength λ_P [4

M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]

]

\frac{d}{dλ} {\log | H (λ) |^{2} |}_{λ = λ_{P}} = \pm \frac{40}{\ln (10) 10^{9}} \frac{πc τ_{e}}{λ_{P}^{2} α} (dB / nm)

(3)

where + and − correspond to enhancing the red and blue sideband, respectively. Eq. (3) is valid for any filter providing a flat SSFR, i.e. not just for an AMZI. In the following section, Eqs. (1–3) will be verified experimentally.

3. Experimental results

The experimental setup for small-signal measurements of a switch based on an SOA and a filter is illustrated in Fig. 1(a). It comprises a network analyzer with an optical interface, creating a small-signal modulation at λ_C = 1557 nm at frequencies up to 20 GHz. This signal is combined with a continuous wave (CW) beam from a tunable external cavity laser (ECL) and launched into a commercial SOA, biased at 150 mA. In the case of the DISC switch, the AMZI is realized “in-line”, where different paths for orthogonal polarization components constitute the arms of the AMZI [see Fig. 1(b)]. Different differential delays τ of 2, 5, and 10 ps are obtained using birefringent calcite crystals of different length. At the input of the calcite crystal a quarter (Q) and half-wave (H) plate, as well as a polarizer, align the probe polarization to a linear state 45 degrees between the principal states of the crystal, corresponding to a 50/50 splitting ratio. The polarizer at the crystal input acts as an in-band ASE filter in this in-line implementation of the AMZI. A discussion of this falls outside the scope of this paper, but details can be found in Ref. [6

M. L. Nielsen, J. MØrk, J. Sakaguchi, R. Suzuki, and Y. Ueno, “Reduction of nonlinear patterning effects in SOA-based all-optical switches using optical filtering,” Technical Digest of OFC (Optical Society of America 2005), paper OThE7, Anaheim, (2005).

]. The phase bias Φ₀ is adjusted by rotating Q and P at the calcite output, while monitoring the shift of the fringe pattern carved in the SOA’s ASE spectrum by the AMZI, using an optical spectrum analyzer (OSA). Details can be found in Ref. [5

M. L. Nielsen, J. MØrk, R. Suzuki, J. Sakaguchi, and Y. Ueno, “Theoretical and experimental study of fundamental differences in the noise suppression of high-speed SOA-based all-optical switches,” Opt. Express 13 , 5080–5086 (2005), http://www.opticsexpress.org/abstract.cfm?URI=OPEX-13-13-5080 [CrossRef]

] and Ref. [6

], and references therein. The output is sent back to the network analyzer for SSFR measurements via a 1.2 nm wide band-pass filter (BPF), also shown in Fig. 1(b), centered at the CW wavelength λ_P. The BPF is wide enough that is does not introduce any shaping of the probe signal.

Fig. 1. (a) Experimental setup for small-signal measurements of switch based on an SOA and a filter. (b) Implementation of AMZI filter.

Figure 2 Fig. 2 shows a plot of ∣T ^AMZ _N (Ω)∣ from Eq. (2), for τ = 10 ps and three different values of Φ₀, corresponding to full compensation (flat SSFR) of the CDR, as well as over and under compensation. An α-parameter of 8.0 is used, which is estimated from results shown in Fig. 3 below. The experimental data is measured for exactly the same values of phase bias Φ₀, and is observed to match the theoretical results remarkably well, in the frequency range up to 20 GHz that could be measured. Notice the comparison with the low-pass CDR. It should be noticed that Fig. 2 shows the SSFR normalized to the response for Ω = 0. To obtain a flat response up to ultra-high frequencies, the attenuated high frequency contents of the probe signal at the SOA output is enhanced by the AMZI, at the expense of the low-frequency components, which contain the most power. This leads to an inevitable trade-off between modulation bandwidth and optical signal-to-noise ratio (OSNR) [4

M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]

–7

M. L. Nielsen, B. Lavigne, and B. Dagens, ”Polarity-preserving wavelength conversion at 40 Gb/s using band-pass filtering,” IEE Electron. Lett. 39, 1334 (2003). [CrossRef]

]. However, operation at bitrates up to 160 Gb/s has been successfully demonstrated [1–2

], which underlines the feasibility and importance of this scheme.

Fig. 2. Normalized carrier density response (CDR) and small-signal frequency response (SSFR) for DISC for different Φ₀ values. Markers are measurements and dashed curves are theoretical, corresponding to α=8.0.

Figure 3(a) shows plots of Eq. (1) for values of a ranging from 6 to 10 (dashed lines), as well as experimental data obtained by measuring the phase bias providing a flat SSFR, for all three different values of τ and two different λ_P of 1550 nm (circles) and 1567 nm (squares). The different wavelengths give rise to different CDR bandwidths, and thus different effective lifetimes τ_e, due to the wavelength dependence of the differential gain. Moreover, the linewidth enhancement factor α increases as the wavelength is increased towards the band-edge. The measurements clearly verify the relationship in Eq. (1), as well as providing an estimate for α at the two wavelengths.

Fig. 3. (a) Optimum phase bias vs. τ for different wavelengths. (b) Optimum AMZI transfer functions for τ=2,5,10 ps, with indication of identical slope at λ=λ_P=1550 nm

Equation (3) is tested in Fig. 3(b), where the measured sinusoidal power transfer functions of the AMZI, ∣H _AMZ(λ)∣² , corresponding to flat SSFRs, are shown for τ = 2, 5, and 10 ps. As predicted by Eq. (3), the slopes at λ = λ_P = 1550 nm are equal, independently of τ, and estimated to be 85.3 dB/nm, which using Eq. (3) translates into α = 6.4. This value corresponds well with the data in Fig. 3.

As mentioned, Eq. (3) is completely general and thus applies to any filter transfer function. Small-signal response measurements were carried out for a switch comprising an SOA and a 0.38 nm wide (FWHM) thin film BPF, using the setup in Fig. 1, with the AMZI and 1.2 nm wide BPF replaced by the 0.38 nm wide BPF. The power transfer function and the slope of the BPF are shown in Fig. 4(a). The SSFR measurements were done with a probe wavelength λ_P of 1567 nm, which corresponds to an α-parameter of 8.0, and an effective carrier lifetime τ_e of 90.1 ps. Using Eq. (3), we find a flat SSFR filter slope requirement of ±75 dB/nm, where + and − signs may be obtained by detuning the BPF to the red and blue side of the carrier wavelength, respectively. From Fig. 4(a) it is clear that this requirement can not be met by the BPF, since the slope does not exceed ±40 dB/nm. However, by increasing the input power to the SOA, and thereby increasing the level of saturation, τ_e can be reduced to 45.2 ps, as deduced from the 3 dB bandwidth of the CDR in Fig. 4(b). This reduces the flat SSFR filter slope requirement to ±37.7 dB/nm, which is within range of the BPF. Since blue-shifting the BPF shows superior performance under large-signal operation [1

, 5

, 7

M. L. Nielsen, B. Lavigne, and B. Dagens, ”Polarity-preserving wavelength conversion at 40 Gb/s using band-pass filtering,” IEE Electron. Lett. 39, 1334 (2003). [CrossRef]

], we only consider this case in the following. Notice that for a blue-shift of Δλ_det, the filter transfer function at the carrier wavelength λ_P will be H(λ_P-Δλ_det). Figure 4(b) shows the SSFR for BPF detunings of -0.1 nm, -0.132 nm, -0.27 nm, and -0.798 nm. As the detuning is increased the response gradually improves, and becomes flat (up to at least 20 GHz) at Δλ_det = -0.27 nm. From Fig. 4 (a) it is observed that the slope at this detuning exactly matches the theoretically predicted requirement of -37.7 dB/nm. Further increasing the BPF detuning does not change the filter slope significantly, and this is the reason why the SSFR is still relatively flat at Δλ_det = -0.798 nm. The ripples observed in Fig. 4 (b) are primarily due to a low OSNR, caused by the large probe power reduction, which is an unavoidable result of the large filter detuning.

Fig. 4. (a) BPF power transfer function (left axis) and corresponding slope (right axis). (b) Normalized CDR and SSFR for switch consisting of SOA and BPF in (a) for different BPF detunings. (c) 10 GHz large-signal response using same parameters as in (b) for different BPF detunings.

Figure 4(c) shows large-signal equivalents of the SSFR traces in Fig. 4(b), in terms of switched output pulse traces. The SOA bias current, signal wavelengths, and the optical input powers are identical to the small-signal measurements, but the small-signal modulation has been replaced by a 10 GHz train of 2 ps wide pulses from a semiconductor mode-locked laser. The impact of blue-shifting the BPF in the large-signal regime is observed by following the waveforms from the bottom and up: the recovery time is clearly reduced, and for Δλ_det = -0.27 nm the inverted pulse is completely symmetric, indicating that the recovery time is as short as the fall-time, determined either by the stimulated recombination time or the response of the sampling scope (40 GHz bandwidth). Increasing the blue-shift to Δλdet = -0.366 nm and beyond, the polarity of the switched signal starts re-inverting, because the power in the carrier-peak, which determines the power in the mark-level of the inverted signal, is increasingly suppressed compared to the power in the blue sideband. This effect is also shown and explained in Ref. [7

M. L. Nielsen, B. Lavigne, and B. Dagens, ”Polarity-preserving wavelength conversion at 40 Gb/s using band-pass filtering,” IEE Electron. Lett. 39, 1334 (2003). [CrossRef]

]. For Δλ_det = -0.798 nm the carrier peak is suppressed by roughly 25 dB, which is sufficient to generate a switched signal with the same polarity as the control signal, and an extinction ratio of 8 dB.

4. Summary

A small-signal analysis is interesting as it provides a best-case limit for the modulation bandwidth. In this paper we have shown experimentally that the small-signal bandwidth of an SOA can be dramatically enhanced using an optical filter. We have investigated both asymmetric MZI and band-pass type filters, and find excellent agreement with previously published theoretical predictions of filter characteristics.

References and links

1.	Y. Liu, E. Tangdiongga, Z. Li, S. Zhang, H. de Waardt, G. D. Khoe, and H. J. S. Dorren, “160 Gb/s SOA-based wavelength converter assisted by an optical bandpass filter,” Technical Digest of OFC 2005 , PDP 17, Anaheim, (2005) [CrossRef]
2.	Y. Ueno, S. Nakamura, and K. Tajima, “Nonlinear phase shifts induced by semiconductor optical amplifiers with control pulses at repetition frequencies in the 40-160 GHz range for use in ultrahigh-speed all-optical signal processing,” J. Opt. Soc. Am. B 19, 2573 (2002) [CrossRef]
3.	J. Leuthold, D. M. Maron, S. Cabot, J. J. Jaques, R. Ryf, and C. R. Giles, “All-optical wavelength conversion using a pulse reformatting optical filter,” J. Lightwave Technol. 22, 186 (2004) [CrossRef]
4.	M. L. Nielsen and J. MØrk, “Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering,” J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]
5.	M. L. Nielsen, J. MØrk, R. Suzuki, J. Sakaguchi, and Y. Ueno, “Theoretical and experimental study of fundamental differences in the noise suppression of high-speed SOA-based all-optical switches,” Opt. Express 13 , 5080–5086 (2005), http://www.opticsexpress.org/abstract.cfm?URI=OPEX-13-13-5080 [CrossRef]
6.	M. L. Nielsen, J. MØrk, J. Sakaguchi, R. Suzuki, and Y. Ueno, “Reduction of nonlinear patterning effects in SOA-based all-optical switches using optical filtering,” Technical Digest of OFC (Optical Society of America 2005), paper OThE7, Anaheim, (2005).
7.	M. L. Nielsen, B. Lavigne, and B. Dagens, ”Polarity-preserving wavelength conversion at 40 Gb/s using band-pass filtering,” IEE Electron. Lett. 39, 1334 (2003). [CrossRef]

OCIS Codes
(060.2330) Fiber optics and optical communications : Fiber optics communications
(250.5980) Optoelectronics : Semiconductor optical amplifiers

ToC Category:
Optoelectronics

History
Original Manuscript: October 10, 2005
Revised Manuscript: January 9, 2006
Manuscript Accepted: January 22, 2006
Published: February 6, 2006

Citation
Mads Nielsen and Jesper Mørk, "Bandwidth enhancement of SOA-based switches using optical filtering: theory and experimental verification," Opt. Express 14, 1260-1265 (2006)
http://www.opticsinfobase.org/oe/abstract.cfm?URI=oe-14-3-1260

Sort: Author | Year | Journal | Reset

References

Y. Liu, E. Tangdiongga, Z. Li, S. Zhang, H. de Waardt, G. D. Khoe, and H. J. S. Dorren, "160 Gb/s SOA-based wavelength converter assisted by an optical bandpass filter," Technical Digest of OFC 2005, PDP 17, Anaheim, (2005) [CrossRef]
Y. Ueno, S. Nakamura, and K. Tajima, "Nonlinear phase shifts induced by semiconductor optical amplifiers with control pulses at repetition frequencies in the 40-160 GHz range for use in ultrahigh-speed all-optical signal processing," J. Opt. Soc. Am. B 19, 2573 (2002) [CrossRef]
J. Leuthold, D. M. Maron, S. Cabot, J. J. Jaques, R. Ryf, and C. R. Giles, "All-optical wavelength conversion using a pulse reformatting optical filter," J. Lightwave Technol. 22, 186 (2004) [CrossRef]
M. L. Nielsen, and J. Mørk, "Increasing the modulation bandwidth of semiconductor optical amplifier based switches using optical filtering," J. Opt. Soc. Am. B. 21, 1606 (2004) [CrossRef]
M. L. Nielsen, J. Mørk, R. Suzuki, J. Sakaguchi, and Y. Ueno, "Theoretical and experimental study of fundamental differences in the noise suppression of high-speed SOA-based all-optical switches," Opt. Express 13, 5080-5086 (2005), http://www.opticsexpress.org/abstract.cfm?URI=OPEX-13-13-5080 [CrossRef]
M. L. Nielsen, J. Mørk, J. Sakaguchi, R. Suzuki, and Y. Ueno, "Reduction of nonlinear patterning effects in SOA-based all-optical switches using optical filtering," Technical Digest of OFC (Optical Society of America 2005), paper OThE7, Anaheim, (2005).
M. L. Nielsen, B. Lavigne, and B. Dagens, "Polarity-preserving wavelength conversion at 40 Gb/s using band-pass filtering," IEE Electron. Lett. 39, 1334 (2003). [CrossRef]

Cited By

Alert me when this paper is cited

OSA is able to provide readers links to articles that cite this paper by participating in CrossRef's Cited-By Linking service. CrossRef includes content from more than 3000 publishers and societies. In addition to listing OSA journal articles that cite this paper, citing articles from other participating publishers will also be listed.

Click here to see a list of articles that cite this paper