One of the key functions required for optical receivers in future 160-Gbit/s optical time-division multiplexed (OTDM) fiber transmission systems is optical time-division demultiplexing, which acts as a bridge between low bit-rate electrical signals and high bit-rate OTDM signals. All of the demultiplexing schemes reported so far are based on the optical time gate switching, where an OTDM signal is switched at the base-clock frequency by using either all-optical switches or optoelectronic switches. For 160-Gbit/s systems, the temporal width of the optical time gate must be less than 3 ps. Such an ultra-fast switching window is demonstrated only by all-optical switches using a highly nonlinear fiber [1
1. T. Morioka, H. Takara, S. Kawanishi, T. Kitoh, and M. Saruwatari, “Error-free 500 Gbit/s all-optical demultiplexing using low-noise, low-jitter supercontinuum short pulses,” Electron. Lett. 32,833–834 (1996). [CrossRef]
2. S. Watanabe, R. Okabe, F. Futami, R. Hainberger, C. Schmidt-Langhorst, C. Schubert, and H. G. Weber, “Novel fiber Kerr-switch with parametric gain: demonstration of optical demultiplexing and sampling up to 640 Gbit/s,” ECOC2004, PDP, Th4-1-1 (2004).
] or optoelectronic switches using specially-designed ultra-fast electroabsorption modulators (EAMs) [3
3. M. Daikoku, T. Miyazaki, I. Morita, H. Tanaka, F. Kubota, and M. Suzuki, “160 Gbit/s-based field transmission experiments with single-polarization RZ-DPSK signals and simple PMD compensator,” ECOC2005, We2.2.1 (2005).
4. H. Murai, M. Kagawa, H. Tsuji, and K. Fujii, “EA modulator-based optical multiplexing/demultiplexing techniques for 160-Gbit/s OTDM signal transmission,” IEICE Trans. Electron. E88-C,309–318 (2005). [CrossRef]
]; however, these devices are still premature.
In this paper, we propose a simple optoelectronic demultiplexing scheme based on phase modulation and spectral filtering. The basic principle of operation is similar to that of the all-optical switch using cross-phase modulation in nonlinear fibers and subsequent optical filtering [5
5. B. E. Olsson, P. Öhlén, L. Rau, and D. J. Blumenthal, “A simple and robust 40-Gb/s wavelength converter using fiber cross-phase modulation and optical filtering,” IEEE Photon. Technol. Lett. 12,846–848 (2000). [CrossRef]
], but our scheme features more practical implementation of this principle: In our scheme, the maximum frequency of the signal processing is the quarter of the bit rate of the OTDM signal; therefore, we can demultiplex a 160-Gbit/s OTDM signal into 40-Gbit/s tributaries using only a commercially available 40-GHz phase modulator and optical bandpass filter (BPF) for telecom systems. After we experimentally confirm the principle of operation at 40 GHz and obtain the design guideline of our receiver, we demonstrate optical time-division demultiplexing of a 160-Gbit/s OTDM signal into 40-Gbit/s tributaries.
2. Principle of operation of the proposed scheme
explains the principle of operation of our proposed scheme, where an OTDM signal is demultiplexed into four tributaries. For simplicity, we assume a sinusoidal intensity modulation at a frequency of f
as the all-mark OTDM signal at the bit rate of f
, as shown in Figs. 1(a)
First, the OTDM signal is phase-modulated at f
/4, and the phase modulation index is m
, as shown in Fig. 1(c)
. The phase difference between the OTDM signal and the phase modulation is adjusted so that the maximum rising slope of the phase modulation coincides with the peak of one of the tributaries of the OTDM signal, as indicated by A in Fig. 1(a)
. The phase modulation induces a frequency chirp on the OTDM signal [Fig. 1(d)
], and then a part of the optical spectrum of the OTDM signal is shifted by ±f
/4 [Figs. 1(e)
]. Note that the spectral component shifted by f
corresponds to the tributary A, while that shifted by -f
corresponds to the tributary C. On the contrary, the non-shifted components correspond to tributaries B and D. Filtering out only the spectral component shifted by f
, the tributary A is extracted as shown in Figs. 1(g)
Note that the bit rate of the demultiplexed OTDM signal is quad the frequency of phase modulation. Since 40-GHz phase modulation can apply to demultiplex the 160-Gbit/s OTDM signal, we can easily construct a demultiplexer for 160-Gbit/s OTDM systems using a commertially available 40-Gbit/s LiNbO3 (LN) phase modulator.
Fig. 1. Temporal and spectral waveforms of OTDM pulses before and after phase modulation (PM) and after optical filtering. (a) and (b): temporal and spectral waveforms of the OTDM signal with the bit rate of f
sig before PM. (c): PM at the frequency of f
sig/4. (d): frequency chirp induced by PM. (e) and (f): temporal and spectral waveforms after PM. (g) and (h): temporal and spectral waveforms after optical filtering.
3. Confirmation of operation principle of our scheme
Before performing the experiment on 160-Gbit/s optical time-division demultiplexing, we confirm the operation principle of our scheme and optimize operation conditions for phase modulation and spectral filtering, using the configuration shown in Fig. 2
, where a 10-GHz electrical clock switches a 40-GHz optical pulse train.
After the 40-GHz 10-ps optical pulse train was generated by an LN intensity modulator driven by 40-GHz clock RF from a CW light at a wavelength of 1558 nm, it was injected on our demultiplexer consisting of an LN phase modulator (Sumitomo Osaka cement T.PMH1.5-10) with V
= 5 V and a BPF (JDSU TB9166) with a bandwidth of 0.4 nm. The modulator was driven by a 29-dBm, 10-GHz electrical clock, and the phase modulation index m
was calculated to be ∼2π from the injected RF power and V
. The offset wavelength λ
of BPF from the center wavelength of the 40-GHz optical pulse train was adjusted from 0.4 nm to 1.2 nm. The phase difference between the 40-GHz optical pulse train and 10-GHz clock RF was adjusted by tuning an electrical delay line. The delay time τ is defined such that τ = 0 when the maximum rising slope of the phase modulation coincides with one of the peaks of the pulse train as shown in Figs. 1(a)
Fig. 2. Configuration of our demultiplexer consisting of a LiNbO3 phase modulator (LNPM) and an optical bandpass filter (BPF).
Fig. 3. (a). Optical spectra measured before and after PM and after BPF. The scale is expanded in the right hand side. (b). Temporal waveforms measured after BPF.
First, we measured the spectral and temporal waveforms after BPF when λBPF
was fixed to 0.9 nm and τ was tuned from 0 ps to 25 ps. Figure 3(a)
shows optical spectra after phase modulation and bandpass filtering. We observe the clear dependence of spectral waveforms after phase modulation on τ, and the period of the spectral change is 25 ps. Figure 3(b)
shows the temporal waveforms after BPF together with that of the input 40-GHz pulse train, which are measured by a digital sampling oscilloscope with an 80-GHz bandwidth. We obtain the clear 10-GHz pulse train separated from the 40-GHz optical pulse train when τ is 0 ps or 25 ps,
while the extracted pulse is distorted for τ = 10 ∼ 15 ps.
Fig. 4. Optical spectra (a) and temporal waveforms (b) measured after BPF. λBPF is the offset wavelength of BPF from the center wavelength of the 40-GHz pulse train.
Fig. 5. Design guideline for the phase modulation index and optical bandpass filtering. Optical spectra before (a) and after (b) phase modulation, and transmission spectrum of BPF (c). f
sig is the bit rate of the signal pulse, and f
PM is the frequency shift induced by phase modulation (PM). f
BPF and Δf
BPF are the offset frequency and bandwidth of BPF, respectively.
Next, in order to find the optimal condition for spectral filtering, we measured the dependence of the spectral and temporal waveforms on λBPF
, as shown in Figs. 4(a)
, respectively. When λBPF
≤ 0.6 nm, we can see the large tailing on the extracted pulse. On the
other hand, when λBPF
≥ 0.8 nm, we observe the pulse train without distortion. These results suggest that we can obtain the clear 10-GHz pulse train demultiplexed from the 40-GHz pulse train if we filter out only the spectral-shifted component.
The above-mentioned results provide us with the two following conditions for optimum phase modulation and bandpass filtering in the case where an OTDM signal with the bit rate of f
is demultiplexed. (1) The spectral shift f
induced by phase modulation must be broader than the spectral bandwidth of the OTDM signal: f
, as indicated in Figs. 5(a)
. Consequently, the required phase modulation index m
should be larger than 4. (2) The overlap between the OTDM signal spectrum and the transmission spectrum of the BPF must be suppressed, as indicated in Fig. 5(c)
, so that f
/2 > f
, where f
is the offset frequency away from the center frequency of the OTDM signal, and λf
is the bandwidth of the BPF. Satisfying the above-mentioned conditions, we can demultiplex a desired tributary channel from the OTDM signal.
4. Application of our scheme to 160-Gbit/s system
In order to investigate applicability of our scheme to 160-Gbit/s systems, we perform optical time-division demultiplexing of 160-Gbit/s OTDM signals into 40-Gbit/s tributaries using the experimental setup shown in Fig. 6
. A 160-Gbit/s OTDM signal with the average power of 4 dBm was generated by intensity modulation of a 10-GHz, 2-ps optical pulse train with a pseudo-random data pattern at 10 Gbit/s followed by optical time-division multiplexing. It was injected into our demultiplexer composed of a commercially available 40-Gbit/s LN phase modulator (Sumitomo Osaka cement T.PMH1.5–40) and BPF (JDSU TB9126). The modulator had the 3-dB bandwidth of 34 GHz, and it was driven by a 30-dBm, 40-GHz electrical clock. The modulation index m
was calculated to be larger than 1.5π from the injected RF power and V
. The bandwidth of BPF was 0.7 nm, and its center wavelength was set to 2 nm away from the center wavelength of the 160-Gbit/s OTDM signal. Note that these modulation and filtering conditions satisfy the requirements given in Sec. 3.
Fig. 6. Experimental setup for 160-Gbit/s demultiplexing using our scheme. P
in is the power of the 40-Gbit/s demultiplexed tributary before preamplification.
Fig. 7. (a). Optical spectra measured before and after PM and after BPF. (b). Eye patterns of the 160-Gbit/s data signal and the 40-Gbit/s demultiplexed tributary.
Fig. 8. (a). BERs of the best and worst 10-Gbit/s tributaries. Dots: BERs of demultiplexed tributaries. Open circles: the back-to-back results. Dashed curve: the theoretical BER determined from the ASE-signal beat noise. (b): P
in of the 40 Gbit/s tributary before preamplification, at which BERs of the 10-Gbit/s tributaries are 10−9.
show optical spectra before and after phase modulation, and after BPF. We can see that the spectral components are broadened by ±3 nm after phase modulation, and only the spectral-shifted component is extracted after BPF. Figure 7(b)
shows the typical eye pattern of the demultiplexed 40-Gbit/s tributary together with that of the original 160-Gbit/s optical data signal. The clear eye opening is observed for the demultiplexed tributary. The intensity fluctuation of the 40-Gbit/s tributary is due to that of the original 160-Gbit/s OTDM signal, which is induced by the multiplexing process.
In order to evaluate the degradation of the signal quality accompanied with our demultiplexing process, we measured the bit error rates (BERs) of all 10-Gbit/s tributaries of the 160-Gbit/s OTDM signal. The demultiplexed optical 40-Gbit/s tributary was preamplified and detected. Then an electrical demultiplexer separated the 40-Gbit/s tributary into four electrical 10-Gbit/s tributaries. We evaluated BERs of all 10-Gbit/s tributaries as we varied the power of the demultiplexed optical 40-Gbit/s tributary P
measured before preamplification. Dots in Fig. 8(a)
shows BERs of the 10-Gbit/s tributaries in the best and worst cases. Open circles indicate results of the back-to-back measurement when a 40-Gbit/s optical signal is directly incident on the pre-amplifier, and the dashed curve is the theoretical BER determined from the ASE-signal beat noise. Figure 8(b)
for all tributaries when BER=10−9
, and the dashed line is the back-to-back result. Values of P
are ranging from −30 dBm to −28 dBm for all 10-Gbit/s tributaries, and power penalties from the dashed line are less than 2 dB although the 160-Gbit/s OTDM signal is somehow degraded with the multiplexing process. Therefore, degradation of the signal quality accompanied with the demultiplexing process itself seems remarkably low.
We have proposed a simple optoelectronic time-division demultiplexing scheme based on phase modulation and spectral filtering. Our scheme has the advantage that we can cope with 160-Gbit/s systems only by using commercially available optical components. Based on our scheme, we have demonstrated optoelectronic demultiplexing of the 160-Gbit/s OTDM signal into 40-Gbit/s tributaries, and have shown that power penalties are maintained less than 2 dB for sixteen 10-Gbit/s tributaries electrically demultiplexed from the 40-Gbit/s tributaries.