Optical single-sideband OFDM transmission based on a two-segment EAM

Hsuan-Lin Cheng; Wei-Hung Chen; Chia-Chien Wei; Yi-Jen Chiu

doi:10.1364/OE.23.000982

1. Introduction

Next-generation metro-access networks are fueling demand for bandwidth to accommodate services with dynamic reconfigurability and high scalability [1, 2]. Developing such communication systems are encumbered by the need to limit the cost, energy consumption and size of components. These requirements have led to the favor of intensity modulation (IM) and direct detection (DD) schemes. Thus, one of visible solutions is to use an electro-absorption modulated laser (EML) and a PIN detector for modulation and detection, respectively, combined with an advanced modulation format [3]. Orthogonal frequency-division multiplexing (OFDM) is widely applied in wired as well as wireless broadband communication systems, due to its high tolerance for inter-symbol interference and flexibility in bandwidth allocation. The applicablility of OFDM in optical communications, such as metro-access applications, has also been demonstrated [4,5]. Unfortunately, the application of OFDM to IM/DD-based metro-access networks tends to result in frequency-selective power fading and nonlinear distortion associated with fiber dispersion [6]. The fact that nonlinear distortion is as a form of subcarrierto-subcarrier intermixing interference (SSII) has led to the development of decision-aided SSII cancellation schemes [7, 8]. Nonetheless, unrecoverable dispersion-induced power fading can be attributed to the double-sideband (DSB) nature of an IM/DD-based signal, wherein the upper and lower sidebands can be rendered out-of-phase following dispersive transmission. Thus, the frequency selectivity of power fading depends on transmission distance. Even the application of bit- and power-loading to OFDM subcarriers as a means of optimizing bandwidth utilization can still leave open the possibility of reduced bandwidth due to power fading [3]. This problem is particularly evident in reconfigurable metro-access systems, in which the distance is unfixed to pursue the efficiency and flexibility of the network resulting from dynamic routing. Previous researchers have indicated the need for a connection length of more than 200 km in such systems [1]; i.e., distances that renders IM/DD signals extremely susceptible to power-fading [6].

A number of IM-based single-sideband (SSB) schemes have been proposed to resolve the problem of power fading [5, 9–11]. Unfortunately, none of these solutions have been able to produce SSB OFDM signals with high sideband power ratio (SBPR) across all subcarrier frequencies, such that residual power fading can occur following transmission. Residual power fading can be attributed to non-perfect optical filtering [5, 9] or the fact that matching conditions between frequency modulation and IM-driven dual-EML can be satisfied only at specific frequencies [10, 11].

This study developed a compact electro-absorption modulator (EAM) comprising a pair of independently cascaded segments for optical SSB modulation. Differences in the chirp characteristics of two segments of the EAM make it possible to design driving signals capable generating an optical SSB OFDM signal with 8 ∼ 15-dB SBPR, which greatly reduces dispersion-induced power fading. We also developed a modified SSII cancellation scheme to mitigate nonlinear distortion produced by fiber dispersion and cascaded modulation, resulting in dispersion-tolerant transmission. Our experiment results demonstrate optical OFDM transmissions at 13.5-Gbps over a 0 ∼ 200-km IM/DD system with bit error rate (BER) of less than 3.8 × 10⁻³ without the need for dispersion compensation or distance-dependent bit- or power-loading.

2. Concept of proposed SSB modulation scheme

Figure 1 presents a schematic illustration of the proposed SSB modulation scheme based on a two-segment EAM, modulated by OFDM driving signals, $X_{i} = \sum_{n = - N}^{N} x_{i} (n) exp (j n ω t)$ , where i = 1 or 2 denotes the signal for the first or the second segment, respectively; N is the total number of subcarriers; ω/(2π) is the subcarrier spacing; x_i(n) represents the modulation of the n-th subcarrier; and $x_{i}^{*} (n) = x_{i} (- n)$ is required for X to be real. Assuming that the EAM segments would result in linear IM and only the i-th segment of the EAM is modulated, the normalized output optical field could be approximated as [6]

E_{i} \approx 1 + \underset{S_{i}^{(1)}}{\underset{︸}{\frac{1 - j α_{i}}{2} X_{i}}} + \underset{S_{i}^{(2)}}{\underset{︸}{\frac{- 1 - α_{i}^{2}}{8} X_{i}^{2}}},

where α_i is the chirp parameter of the i-th segment;

S_{i}^{(k)}

represents the k-th order terms; and those terms, of which the orders are greater than 2, are omitted for simplicity. When two segments are modulated simultaneously, the cascaded output field becomes

E_{ssb} \approx (1 + S_{1}^{(1)} + S_{1}^{(2)}) \times (1 + S_{2}^{(1)} + S_{2}^{(2)}) \approx 1 + \underset{1 st-order signal}{\underset{︸}{S_{1}^{(1)} + S_{2}^{(1)}}} + \underset{2 nd-order term}{\underset{︸}{S_{1}^{(2)} + S_{2}^{(2)} + S_{1}^{(1)} S_{2}^{(1)}}} .

Similar to (1), (2) shows only the terms, of which the orders are less than or equal to 2, and the other terms are omitted. To realize an SSB signal in the desired form of

S = \sum_{n = - N}^{N} s (n) exp (j n ω t)

,

S_{1}^{(1)} + S_{1}^{(2)}

in (2) is required to be equal to S. In accordance with (1), it follows that

\frac{1}{2} (1 - j α_{1}) x_{1} (n) + \frac{1}{2} (1 - j α_{2}) x_{2} (n) = s (n)

or

\frac{1}{2} (1 + j α_{1}) x_{1}^{*} (n) + \frac{1}{2} (1 + j α_{2}) x_{2}^{*} (n) = s^{*} (n)

. Assuming that the lower sideband (n < 0) is suppressed; i.e., s(n < 0) = 0, then the driving signals must be as follows:

[\begin{matrix} s (n) \\ s^{*} (- n) \end{matrix}] = [\begin{matrix} s (n) \\ 0 \end{matrix}] = \frac{1}{2} [\begin{matrix} 1 - j α_{1} & 1 - j α_{2} \\ 1 + j α_{1} & 1 + j α_{2} \end{matrix}] [\begin{matrix} x_{1} (n) \\ x_{2} (n) \end{matrix}], for n > 0 .

Thus, the required driving signals X₁ and X₂ can be used to produce the desired SSB signal. In accordance with (3), however, α₁ ≠ α₂ is a prerequisite for the proposed scheme. The chirp characteristics of an EAM is bias-dependent [12]; therefore, the requirement could be satisfied simply by adjusting bias voltages. Furthermore, larger difference between α₁ and α₂ may reduce the required amplitudes of the driving signals. In principle, the proposed scheme is able to maintain SBPR at a high value for all subcarriers, such that power fading is minimized for all frequencies, unlike the other IM-based SSB schemes [5, 9, 10].

Fig. 1 The schematic plot of the proposed SSB modulation scheme based on a two-segment EAM.

Download Full Size | PDF

Nonetheless, the proposed SSB scheme requires two different driving signals. Thus, compared with other schemes [9–11], an extra signal generator would be necessary, implying that higher transmitter complexity and more power consumption would be the price of maintaining high SBPR for all frequencies by the proposed scheme. As a matter of fact that two driving signals are designed to generate one desired SSB signal, they are not independent; thus, appropriate designs of the transmitter may reduce the complexity and/or the power consumption. For instance, in accordance with (3), the relationship between two driving signals can be determined as follows:

\frac{x_{2} (n)}{x_{1} (n)} = - \frac{1 + j α_{2}}{1 + j α_{1}} \equiv H (n), for n > 0,

and H(n) = H^*(−n) for n < 0. As shown in Fig. 1, a digital filter with the frequency response H(n) can be used to save the second signal generator.

Following fiber transmission and square-law DD, the detected signal can be written as follows:

\begin{array}{l} {| Θ {E_{ssb}} |}^{2} & \approx 1 \\ + \underset{1 st-order signal}{\underset{︸}{2 ℜ {Θ {S_{1}^{(1)} + S_{2}^{(1)}}}}} + \underset{2 nd-order SSII}{\underset{︸}{{| Θ {S_{1}^{(1)} + S_{2}^{(1)}} |}^{2} + 2 ℜ {Θ {S_{1}^{(2)} + S_{2}^{(2)} + S_{1}^{(1)} S_{2}^{(1)}}}}} \end{array}

where Θ{·} represents the effect of fiber dispersion [6, 7]. For instance, Θ{X_i} represents

\sum_{n = - N}^{N} x_{i} (n) exp (j n ω t + j n^{2} ω^{2} β_{2} L / 2)

, where β₂ and L are the dispersion parameter and length of fiber, respectively. The sum of the last two terms in (5) represents 2nd-order nonlinear distortion, which can be mitigated using the decision-aided SSII cancellation scheme [7, 8], in which the SSII model requires modification in accordance with (5).

3. Two-segment EAM and experiment setup for SSB modulation

A top-view diagram of a two-segment EAM is presented in Fig. 2(a). Two segments are monolithically integrated in a material system used for the fabrication of an optical waveguide for EAM, and each segment in this study was set at 200 μm. The material includes an active region formed by InGaAsP multiple quantum wells (MQWs) as well as top p-InP, and bottom n-InP cladding layers. Electrical isolation is achieved between the two segments using H⁺ implantation, which makes it possible to bias and modulate the two segments independently. Due to the quantum confined Stark effect (QCSE) in MQWs, variations in optical absorption near the bandgap enhance sensitivity to electrical signal and bias, which changes the optical index resulting in variations in chirp. Thus, monolithic integration and QCSE make it possible to use the bias and driving signal to modulate chirp and optical intensity in the cascaded structure. Figures 2(b) and 2(c) present the measured transmission and chirp parameters of the segments with optical input power of 12 dBm at 1550nm. It should be noted that the characteristics of segment 2 are affected by the bias of segment 1 because the difference in absorption of segment 1 results in different launch power into segment 2. To obtain independent measurements of modulation characteristics in the two segments, the bias voltage of segment 1 or 2 was fixed at 0.2 or 0 V, respectively, while the other segment was being tested. Employing a network analyzer to individually measure the small-signal frequency response of each segment, the chirp parameter was characterized using the fiber response peak method [13]. Based on the QCSE in MQWs, the chirp parameter of EAM is a function of bias. Higher and lower chirp parameters can be obtained under lower and higher bias regimes, respectively, as shown in Figs. 2(b) and 2(c). Despite the fact that the chirp parameters of the two segments can be differentiated by setting different bias values, bias values that result in a high degree of modulation nonlinearity and/or high optical absorption should be avoided. Therefore, we set the range of electrical signal with bias at approximately −0.5 ∼ 0.7 V.

Fig. 2 (a) The top view of the two-segment EAM. The measured transmission and chirp parameters of (b) segment 1 and (c) segment 2.

Download Full Size | PDF

Figure 3 presents the setup of the experiment on SSB modulation and transmission. A tunable laser set to 1550 nm was used as the light source. It should be noted that this laser could be replaced by a DFB laser and integrated within the EAM to reduce overall cost and produce a device with a smaller footprint. The lack of a suitable digital filter made it necessary to generate both of the electrical driving signals in accordance with (3) using an arbitrary waveform generator (AWG, Tektronix AWG7122B), at a sampling rate of 12 GS/s and digital-to-analog conversion resolution of 8 bits. A fast-Fourier transform size of 512 resulted in subcarrier spacing of 23.44 MHz, and cyclic prefix (CP) was set to 1/64. The 1st∼211th subcarriers were used to encode data, resulting in a signal bandwidth of 5 GHz. A simple verification of transmission performance was performed at the beginning of the experiment by fixing the modulation format as QPSK. We then applied a bit-loading algorithm according to the measured performance of the QPSK-OFDM signals at optical back-to-back (BtB) with the aim of adjusting the modulation format and power of each subcarrier in order to achieve the maximum capacity at the FEC limit. In order to meet the necessary condition of α₁ ≠ α₂, the bias voltages of the two-segment EAM were set at 0.2 V and −0.5 V for segments 1 and 2, respectively. As shown in Figs. 2(b) and 2(c), the corresponding chirp parameters are α₁ = 1 and α₂ = 0.5. According to Fig. 2(c), the total insertion loss is around 30 dB, which includes a high fiber-EAM/EAM-fiber coupling loss of 15 dB due to the lack of anti-reflection coating. An Erbium-doped fiber amplifier (EDFA) was therefore inserted to compensate for the high insertion loss, resulting in the fiber launch power of 0 dBm. It should be noted that, although a higher value of α₁ − α₂ can be achieved by setting higher and lower bias voltages for segments 1 and 2, respectively, the price would include lower modulation efficiency, higher optical absorption, and higher modulation nonlinearity. The bias voltages and corresponding modulation slopes of the two segments differ; therefore, the amplitude of the first driving signal was adjusted using an electrical amplifier and an attenuator in order to identify the best modulation conditions. For comparison, we also performed conventional DSB modulation of only the second segment, leaving the bias voltages unchanged. The carrier-to-subcarrier power ratios (CSPRs) of DSB and SSB modulations were also maintained at the same level by controlling the amplitudes of the driving signals. The optical spectra of SSB and DSB QPSK-OFDM signals obtained using an optical spectra analyzer (OSA, APEX 2441B) at a resolution of 0.16 pm (i.e., 20 MHz) are presented in the inset of Fig. 3. This figure shows that the SBPR of the SSB signal is more than 15 dB for subcarriers operating at low frequencies. SBPR was reduced to approximately 8 dB for subcarriers operating at high frequencies; however, this value could be affected by the interference floor at approximately −50 dBm. After optical IM based on the two-segment EAM and transmission over 0 ∼ 200-km standard single-mode fiber (SSMF) without dispersion compensation, optical OFDM signals at the optical received power of −13 dBm were detected using a 10-GHz PIN detector, and the received electrical signals were recorded using a digital storage oscilloscope (DSO, Agilent DSO81204A) with a sampling rate of 40 GS/s, analog-to-digital conversion resolution of 8 bits and 3-dB bandwidth of 12 GHz. The OFDM signal was demodulated using an off-line DSP program that includes the proposed SSII cancellation for the suppression of 2nd-order nonlinear distortion caused by SSB/DSB modulation and dispersion. Following demodulation, signal performance was evaluated according to signal-to-noise ratio (SNR) and BER, which was determined by bit-by-bit comparison. It should be noted that, when the format was fixed as QPSK, BER was not measured and SSII cancellation was employed without taking into consideration any decision error [7, 8]. For the bit-loaded signals, BER was measured to verify the transmission performance, and decision errors were taken into account while applying SSII cancellation.

Fig. 3 The setup of the experiment (inset: optical spectra of the SSB and DSB signals.)

Download Full Size | PDF

4. Experimental results and discussion

To investigate frequency-selective power fading induced by dispersion, we measured the average electrical powers of the received QPSK subcarriers for SSB and DSB modulations, as shown in Figs. 4(a) and 4(b), respectively. The power values were normalized using those obtained at optical BtB. The behavior associated with severe power fading in Fig. 4(b) was determined by dispersion and chirp through the following formula $(1 + α_{2}^{2}) {cos}^{2} (n^{2} ω^{2} β_{2} L / 2 - {tan}^{- 1} α_{2})$ [6, 13]. In contrast, the SSB signal suffered a variation in power of less than 3 dB, which verifies that SSB modulation is able to tolerate dispersion-induced power fading. Figures 5(a) and 5(b) present the measured SNR of the SSB and DSB QPSK-OFDM signals, respectively. All SNR values in Fig. 5 were obtained with the assistance of the SSII cancellation scheme. The SNR at low frequencies using SSB modulation failed to equal that obtained using conventional DSB modulation due to the fact that SSB modulation suppressed one sideband and half of the electrical subcarrier power. In contrast, subcarriers at high frequencies using SSB and DSB modulations presented similar SNR values due to the fact that performance is dominated by the transmitter. Specifically, the frequency responses of AWG and EAM are limited such that the performance of those subcarriers was less affected by the received electrical subcarrier power; therefore, in order to evaluate the effect of dispersion on the SNR, we plotted in Figs. 5(c) and 5(d) the relative SNR values normalized by the respective SNR at optical BtB. The DSB OFDM signal was shown to suffer from a serious SNR penalty after dispersive transmission. A comparison of Figs. 4(b) and 5(d) reveals similar trends in the relative power and relative SNR, demonstrating that power fading is the key issue. Similar to a small variation in the relative power in Fig. 4(a), the relative SNR values in Fig. 5(c) reveal a variation of approximately 3 dB, which confirms the effectiveness of SSB modulation in an IM/DD system.

Fig. 4 The subcarrier powers related to those at optical BtB using (a) SSB and (b) DSB modulation.

Download Full Size | PDF

Fig. 5 The SNR of (a) SSB and (b) DSB QPSK-OFDM signals, and the relative SNR normalized by the respective SNR at optical BtB of (c) SSB and (d) DSB signals.

Download Full Size | PDF

In accordance with the measured SNR at optical BtB in Figs. 5(a) and 5(b), we applied the bit-loading algorithm to determine the appropriate power adjustment and the bit number of quadrature amplitude modulation (QAM) format for each subcarrier in order to generate bit-loaded SSB and DSB signals. The modulation order of each subcarrier after bit-loading is presented in Fig. 6(a), in which the total data rates were 13.5 and 15.7 Gbps for SSB and DSB signals, respectively. Due to the higher SNR of the DSB signal at optical BtB, a number of subcarriers of the DSB signal were assigned using modulation orders higher than those of the SSB signal, which led to a higher data rate. Figures 6(b) and 6(c) respectively plot the measured SNR of bit-loaded SSB and DSB signals with SSII cancellation at optical BtB or after 50, 100, 150, and 200-km fiber transmission. Figs. 6(d) and 6(e) plot the improvement in SNR of the bit-loaded SSB and DSB signals provided by SSII cancellation, respectively, and the improvements were in the range of 0 ∼ 1.8 dB, depending on modulation scheme, transmission distance and subcarrier frequency. The aim of the bit-loading algorithm was to equalize the BER performance of each subcarrier; therefore, the SNR values of the bit-loaded signals form step ladder profiles at optical BtB. However, after dispersive transmission, power fading would destroy these profiles, leading to worse BER. We measured the BER performance of the bit-loaded signals as a function of transmission distance, as shown in Fig. 7. Bit-loading was not applied adaptively in accordance with the SNR after transmission; therefore, the BER performance of the DSB signal deteriorated after longer transmissions due mainly to the more pronounced effects of power fading. Conversely, the proposed SSB signal is able to tolerate dispersion; therefore, BER performance can be maintained below 3.8 × 10⁻³ without the need for distance-dependent bit-loading. The selective constellations shown in Fig. 7 illustrate whether the signals are tolerant to fiber dispersion. The constellations of SSB signals show little variation after transmission; however, those of DSB signals clearly show the effects of the penalty induced by transmission. In particular, the constellations of the QPSK DSB subcarriers suffer much more from transmission, compared with those of the 8-QAM DSB subcarriers. This difference in the degree of dispersion-induced penalty is mainly caused by the difference in the degree of power fading. Figure 6(a) presents that the 8-QAM and QPSK formats were mostly assigned to the DSB subcarriers at frequencies of ≳ 3 GHz and ≲ 3 GHz, respectively; thus, in accordance with Fig. 4(b) or 5(d), the QPSK signals experienced more severe fading.

Fig. 6 (a) The bit number of each subcarrier of the bit-loaded signals. The measured SNR of (b) SSB and (c) DSB bit-loaded signals, and the improvement in SNR provided by SSII cancellation of (d) SSB and (e) DSB bit-loaded signals.

Download Full Size | PDF

Fig. 7 BER and selected constellations of the bit-loaded OFDM signals.

Download Full Size | PDF

5. Conclusion

This work presents a novel SSB modulation scheme in an IM/DD system using a two-segment EAM. Each segment was modulated by driving signals designed in accordance with a desired OFDM signal. Although the total length of the EAM was 400 μm The generated SSB OFDM signal presented SBPR values of 8 ∼ 15-dB, indicating that the signal does not suffer from dispersion-induced power fading. This makes it possible for the proposed scheme to provide distance-insensitive transmission performance without the need for dispersion compensation. Combined with the modified SSII cancellation scheme, we succeeded in transmitting a 13.5-Gbps SSB OFDM signal over 0 ∼ 200-km fiber without the need for adaptive bit- and power-loading.

References and links

1. N. Madamopoulos, D. C. Friedman, I. Tomkos, and A. Boskovic, “Study of the performance of a transparent and reconfigurable metropolitan area network,” J. Lightwave Technol. 20, 937–945 (2002). [CrossRef]

2. K.-I Sato and H. Hasegawa, “Optical networking technologies that will create future bandwidth-abundant networks,” J. Opt. Commun. Netw. 1, A81–A93 (2009). [CrossRef]

3. D.-Z. Hsu, C.-C. Wei, H.-Y. Chen, W.-Y. Li, and J. Chen, “Cost-effective 33-Gbps intensity modulation direct detection multi-band OFDM LR-PON system employing a 10-GHz-based transceiver,” Opt. Express 19, 17546–17556 (2011). [CrossRef] [PubMed]

4. N. Cvijetic, “OFDM for next-generation optical access networks,” J. Lightwave Technol. 30, 384–398 (2012). [CrossRef]

5. T. M. F. Alves, A. A. P. Alberto, and A. V. T. Cartaxo, “Direct-detection multi-band OFDM metro-networks employing virtual carriers and low receiver bandwidth,” in Optical Fiber Communication Conference2014, paper Tu3G.5. [CrossRef]

6. C.-C. Wei, “Small-signal analysis of OOFDM signal transmission with directly modulated laser and direct detection,” Opt. Lett. 36, 151–153 (2011). [CrossRef] [PubMed]

7. C.-C. Wei, “Analysis and iterative equalization of transient and adiabatic chirp effects in DML-based OFDM transmission systems,” Opt. Express 20, 25774–25789 (2012). [CrossRef] [PubMed]

8. D.-Z. Hsu, C.-C. Wei, H.-Y. Chen, C.-Y. Song, I.-C. Lu, and J. Chen, “74.4% SSII cancellation in an EAM-based OFDM-IMDD transmission system,” in Optical Fiber Communication Conference2013, paper OM2C.7.

9. J. L. Wei, C. Snchez, R. P. Giddings, E. Hugues-Salas, and J. M. Tang, “Significant improvements in optical power budgets of real-time optical OFDM PON systems,” Opt. Express 18, 20732–20745 (2010). [CrossRef] [PubMed]

10. M. E. Chaibi, T. Anfray, K. Kechaou, C. Gosset, L. A. Neto, G. Aubin, C. Kazmierski, P. Chanclou, C. Aupetit-Berthelemot, and D. Erasme, “Dispersion compensation-free IM/DD SSB-OFDM transmission at 11.11 Gb/s over 200 km SSMF using dual EML,” IEEE Photon. Technol. Lett. 25, 2271–2273 (2013). [CrossRef]

11. H. Kim, “EML-based optical single sideband transmitter,” IEEE Photon. Technol. Lett. 20, 243–245 (2008). [CrossRef]

12. K.-C. Chang, S.-W. Shen, M.-C. Hsu, Y.-J. Chiu, C.-C. Wei, and C.-K. Lee, “Negative-chirped EAM-SOA for distance-insensitive optical OFDM transmission in long-reach OFDMA PONs,” in Optical Fiber Communication Conference2014, paper Tu3H.4. [CrossRef]

13. F. Devaux, Y. Sorel, and J. F. Kerdiles, “Simple measurement of fiber dispersion and of chirp parameter of intensity modulated light emitter,” J. Lightwave Technol. 11, 1937–1940 (1993). [CrossRef]

Optical single-sideband OFDM transmission based on a two-segment EAM

Abstract

1. Introduction

2. Concept of proposed SSB modulation scheme

3. Two-segment EAM and experiment setup for SSB modulation

4. Experimental results and discussion

5. Conclusion

References and links

Cited By

Figures (7)

Equations (5)

Optics Express