Vous êtes sur la page 1sur 8

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 28, NO.

1, JANUARY 1, 2010

123

Highly Sensitive Optical Multilevel Transmission of Arbitrary Quadrature-Amplitude Modulation (QAM) Signals With Direct Detection
Nobuhiko Kikuchi, Member, IEEE, and Shinya Sasaki, Senior Member, IEEE
AbstractA proposal for the realization of highly sensitive optical direct-detection (incoherent) multilevel transmission with arbitrary signal constellation without optical chromatic dispersion (CD) compensation is reviewed in this paper. We introduce transmitter-side signal processing, CD pre-equalization, and phase preintegration techniques to realize direct-detection CD-compensation free transmission of arbitrary optical quadrature-amplitude modulation (QAM) signals. We also adopt receiver-side signal processing, multi-symbol phase estimation (MSPE), and non-Euclidean metric-based symbol detection to improve the optical SNR (OSNR) sensitivity. We have performed up to 160-km standard single-mode ber transmission of a 40-Gbit/s 16QAM signal and achieved the OSNR sensitivities only 0.52.4 dB away from those of coherent detection. Index TermsAdaptive equalizers, amplitude shift keying, differential phase-shift keying, multilevel systems, optical ber communication, optical modulation, quadrature-amplitude modulation (QAM).

I. INTRODUCTION

ROWING Internet trafc continuously urges steady increase of the transmission capacity of wavelength-division multiplexing (WDM) optical ber networks. At the same time, the speed of individual information trafc is expected to achieve 100 Gbit/s or beyond with the appearance of next-generation ultrahigh-speed Ethernet like 100 GbE. In order to realize exible and efcient use of limited wavelength band in future WDM networks, single-wave transport solution of such high-speed trafcs is desirable. However, the modulation speed of high-end optical communication links already reaches several tenth of gigahertz, and further increase is now facing severe limit not only by the speed of electrical and optical devices, but also by the power consumption, cost, and transmission reach. Optical multilevel signaling is one of the most promising solution, since it enables the increase of information capacity of a single optical channel without increasing physical modulation speed (symbol rate) by cramming several information bits into a single symbol. Emerging digital coherent transmission technology [1], [2] is one of the enabling techniques of the
Manuscript received May 26, 2009; revised August 12, 2009. First published November 10, 2009; current version published December 23, 2009. The authors are with the Central Research Laboratory, Hitachi Ltd., Kokubunji, Tokyo 185-8601, Japan (e-mail: nobuhiko.kikuchi.ca@hitachi.com). Digital Object Identier 10.1109/JLT.2009.2035827

practical optical multilevel signaling. However, the coherent receiver needs an extra local laser source for optical eld reference, and has severe polarization dependency. The latter can be solved by the use of polarization diversity conguration, but the receiver size and cost will be doubled. The polarization multiplexing technique can be used to double the channel rate by using two polarization axes for independent signal transmission and to cancel out the adverse effect of the polarization diversity (for example, [3]). Nonetheless, such a dual-polarization coherent transceiver still tends to be bulky and complex, and the realization of single-wave simpler multilevel transceiver is required. A direct-detection (incoherent) multilevel signaling technology is the potential candidate for it, in which optical delay-differential detection and intensity detection are used as the tools for detecting signal (differential) phase and amplitude components, respectively. The advantage of using the direct detection lies in receiver simplicity; it does not require a receiver-side local laser nor a frequency tracking mechanism, and has higher tolerance to laser phase noise. Since the direct detection is immune to the polarization states of the received signal and phase noise, neither polarization multiplexing nor diversity congurations are needed. Therefore, smaller, lower cost, and less power-consuming multilevel receiver can be realized. The simplest direct-detection multilevel receiver can be constructed by combining various binary delay detectors and/or intensity detectors: For example, 8-level (8DPSK [4], 8APSK [5], [6]) and 16-level (16APSK [7], [8]) signaling experiments have been reported, where DPSK stands for differential phase-shift keying and APSK stands for amplitudeand phase-shift keying, which is the combination of ASK (amplitude shift keying) and DPSK. In order to further increase the resolution of differential phase and amplitude detection, we proposed the introduction of digital signal processing, which also leads to the improvement of the transmission performance, such as optical SNR (OSNR) sensitivities, by the use of digital equalizers. Up to a 100-km 32-level APSK transmission [9] and a 1040-km 8-level DPSK transmission [10], experiments have been performed. However, these previous direct-detection multilevel signaling schemes still have various drawbacks, mainly because of the use of optical delay detectors for signal phase detection. For example, applicable multilevel signal constellations are limited to DPSK and APSK ( represents the number of signal levels), which are not necessarily ideal in terms of minimum signal point distance. Also, it has only limited receiver-side digital chromatic dispersion (CD) compensation capability [11],

0733-8724/$26.00 2009 IEEE

124

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 28, NO. 1, JANUARY 1, 2010

[12] because the absolute phase of the received optical eld cannot be detected. Each results in degraded OSNR sensitivity and limited ber transmission distance without optical CD compensators. In this paper, we introduce various new techniques to overcome these problems with the help of transmitter- and receiver-side digital signal processing. The former includes the signal preequalization for CD compensation [13] and our recently proposed phase preintegration [14] for the use of arbitrary signal constellation by canceling the phase-differential operation of optical delay-detection process. The latter is intended to improve the OSNR sensitivity and includes the multi-symbol phase estimation (MSPE) [15], [16] to reduces the extra phase noise induced by the delay detection, and the non-Euclidean metric for multilevel symbol detection to increase the tolerance to the extra phase noise. By using these techniques, we realize highly sensitive direct-detection multilevel signaling using the arbitrary signal constellation with the digital preequalization capability of ber CD. We rst demonstrate direct-detection 30-Gbit/s 8 quadrature-amplitude modulation (QAM), 35.8-Gbit/s 12QAM, and 40-Gbit/s 16QAM signaling experiments, each with 400-, 240- and 160-km standard single-mode ber (SSMF) transmission without using optical CD compensators. As far as we know, these are the rst experimental demonstration of more than four-level signaling using CD preequalization at the highest bit rate (up to 40 Gbit/s), due to the multilevel signaling that enables the drastic increase of the signal bit rate even with the limited sampling speed of digital-to-analog (D/A) converters (such as 20 GSa/s). Also, we achieve and conrm high OSNR sensitivities in the direct-detection multilevel signaling experiments with various signal constellations from differential QPSK (DQPSK) to 16QAM, which are within 0.5 to 2.4 dB away from those obtained in the recent coherent signaling experiments.

Fig. 1. Experimental setup of direct-detection multilevel signaling with the transmitter-side digital phase preintegration and CD pre-equalization. (a) Transmitter and ber transmission line. (b) Receiver. SYMGEN: complex multilevel symbol generator, ELD: external cavity laser diode, IQ-Mod: LN-IQ modulator, EQ: digital adaptive equalizer, and SYMDET: complex multilevel symbol detector.

II. TRANSMITTER-SIDE DIGITAL SIGNAL PROCESSING

A. Conguration of Direct-Detection Multilevel Transceiver Fig. 1 shows the experimental congurations of: 1) a multilevel transmitter and a ber transmission line and 2) a direct-detection multilevel receiver used throughout in this paper. at First, complex multilevel symbol sequences 10-Gsymbol/s (32 768 symbols in length) are generated at a symbol generator in an ofine preprocessing part. After and quadrasome digital signal processing, their in-phase components are fed to two arbitrary waveform ture-phase generators emulating high-speed D/A converters running at the sampling speed of 20 GSa/s. The frequency responses of electrical signal paths and nonlinearity of optical modulators are equalized by digital lters (not shown). At the transmitter, a continuous-wave (CW) laser wave from an external cavity laser diode (linewidth 100 kHz) is converted to a multilevel signal eld by using a lithium niobate (LN)-IQ modulator [17], and then transmitted over a ber transmission line made of several

80-km-long SSMF spans and Erbium-doped ber amplier (EDFA) repeaters. The direct-detection multilevel receiver shown in Fig. 1(b) has a function of detecting differential phase and eld amplitude from the received optical eld . Its conguration is almost the same as the one used in our previous 32-level signaling experiments [9] and its front end includes two delay-interferometric receivers ps, interferometric phase ( ) and a photo detector (PD) working as an intensity 0 and , and . Each rereceiver with the output signals ceiver is followed by a high-speed analog-to-digital (A/D) converter (sampling rate 10 GSa/s). No clock extraction is implemented and only the sample sequences (length 100 Ksamples) taken at the center of symbols are used for the following ofine post signal processing. The sampled output signals and ( is the sample number) can be expressed as (1)(3). In an ofine postdigital signal processing part, sampled difand eld amplitude are calculated ferential phase from the two argument inverse tangent of and , and , respectively. Then an adaptive digital the square root of equalizing lter (ten FFE and ten DFE taps; the rst 2500 bits of a received pattern are used for the tap adaptation with the least mean squares (LMS) algorithm by minimizing the error from the ideal signal constellation) is applied to reduce signal intersymbol interference (ISI), and a multilevel symbol detector

KIKUCHI AND SASAKI: ARBITRARY QAM SIGNALS

125

detects received symbols from signal bit error ratio (BER)

and

and calculates (1) (2) (3)

B. Principle of CD Preequalization and Phase Preintegration To overcome the problems of the limited digital CD compensation capability and the limited applicable signal constellation, we adopt two transmitter-side signal processing: the digital CD preequalization and the phase preintegration. The former is the technique to cancel the linear transmission impairment like ber CD by applying its inverse transfer function on the transmitting eld in advance, using digital signal processing and an arbitrary optical eld modulator. So far, up to a 5120-km SSMF transmission with 10-Gbit/s binary signal (DPSK) has been reported with it [13]. But its application to high-speed signals like 40 Gbit/s is quite difcult, since it requires very high-speed D/A converters running at twice the symbol rate and also the size of CD equalization circuit, typically a nite impulse response (FIR) lter or a lookup table (LUT), grows rapidly in proportional to the square of the symbol rate. Multilevel signaling has very high afnity to the pre-equalization technique, because the reduction of signal symbol rate leads to the drastic decrease of the speed of D/A converters and the size of the CD equalization and , respeccircuit with the ratio of tively, from the binary signaling (for example, 1/4 and 1/16 in 16QAM case). The polarization-mode dispersion (PMD) is the another degradation factor, which will also limit the signal transmission distance, but it cannot be compensated by preequalization technique. The PMD tolerance is enhanced by the use of higher-order multilevel signals, for example, 40-Gbit/s nonreturn to zero (NRZ)-16-QAM signal is estimated to have the rstorder PMD tolerance of 15 ps without considering the improvement given by digital equalizers (DFE/FFE). A 1000-km ber transmission over the ber with PMD value of 0.14 ps (average PMD: 4.74 ps) is attainable with the outage probability assuming Maxwellian distribution. of The latter, phase preintegration [14] is a newly introduced technique, which cancels the phase-differential operation of optical delay detection process and enables the use of almost arbitrary multilevel signals. For example, we use 8QAM, 12QAM, and 16QAM signals in Fig. 2(a)(c) in this paper with this technique. The phase preintegration is the symbol-by-symbol numerical accumulation of the phase component of multilevel signals at the transmitter side. Denoting complex multilevel symbols generated at the symbol generator in polar coordinate as , its phase is accumuis recombined with lated as in (4). The accumulated phase the original amplitude component to generate as shown in Fig. 1(a). The resultant eld becomes cocentric constellations with increased number of signal points as the examples of 8QAM and 16QAM signals, as shown in Fig. 2(d) and (e), respectively. When such phase preintegrated signals are received by the direct-detection multilevel receiver (with the

Fig. 2. Experimental signal constellation diagrams of original and phase preintegrated QAM signals. (a) 8QAM. (b) 12QAM. (c) 16QAM. (d) 8QAM after phase preintegration. (e) 16QAM after phase preintegration.

), as shown in Fig. 1(b), the sampled output difcan be calculated as in (5), and it turns ferential phase of the original multilevel symbol, into the absolute phase since the phase preintegration acts as differential precoding and cancels out the effect of optical delay detection. By combining and separately detected eld amplitude , original mulcan be reconstructed. The initilevel symbols tial phase is unknown but it is not at all important, since it vanishes when differentially detected. Meanwhile, the use of should be avoided since the differential the origin phase of the following symbol would be lost. It can be done by simply not assigning the origin as a signal point of multilevel signal. Even though a received signal point with noise can accidentally hit near the origin, it just cause two successive differential phase detection errors and further error propagation is avoided by the nature of differential phase precoding. It should be noted that a similar phase preintegration concept [18] was proposed for wireless signal transmission with earlier electrical delay demodulation. Its implementation is different from the one shown in this paper, since optical delay-differential detection causes signicant intersymbol interference in the amplitude part. Therefore, we propose and verifyits optical implementation for the direct-detection multilevel signaling with coupled optical delay-interferometric receivers and a separate amplitude detector, for the rst time (4) (5)

C. Experimental Results We verify the feasibility of the proposed concept with ber transmission experiments of 30-Gbit/s 8QAM and 35.8-Gbit/s 12QAM signals [19]. In the transmitter-side ofine preprocessing shown in Fig. 1(a), we rst apply the phase preintegration on the generated multilevel symbol sequence, twice upsample it, and then apply a CD preequalization with a 53-stage digital FIR lter emulating a transfer function of , where is the average ber ber CD

126

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 28, NO. 1, JANUARY 1, 2010

Fig. 3. Reconstruction of directly detected 8QAM signal. (a) Differential constellation diagram (dI (n); dQ(n)). (b) Reconstructed signal constellation diagram (I (n); Q(n)).

dispersion parameter and is the total ber length. Resultant complex samples are passed to D/A converters and converted to a transmitted optical eld. The ber transmission line consists of EDFA repeaters and three to ve 80-km SSMF spans (average loss and CD of 17.2 dB/span and 16.7 ps/nm/km) dBm. The totals of digital CD with the ber input power of and ps/nm for preequalization are equivalent to 240- and 400-km-long transmissions, respectively. An example of the reconstruction process of directly-detected 8QAM signals (back-to-back) is shown Fig. 3. Fig. 3(a) shows the differential constellation diagram, i.e., 2-D plot of obtained from the two optical delay interferometric receivers shown in Fig. 1(b). By combining the calculated from them with the sepadifferential phase , the signal constellation of rately detected 8QAM signal shown in Fig. 3(b) is reconstructed, which exactly coincides with that of original 8QAM signal shown in Fig. 2(a). It should be noted that the signal points in Fig. 2(a)(c) are slightly modied from the ideal grid-like patterns to improve experimental BER performances so that signal points have a bit wider radial separation. At rst, the 30-Gbit/s 8QAM signal is used for an optical-dispersion compensation free 400-km SSMF transmission. Without CD preequalization, the received differential signal is totally distorted, as shown constellation diagram in Fig. 4(a). With the CD preequalization, the original 8QAM signal constellation is reconstructed as in Fig. 4(b) and the BER is obtained. Next, a 35.8-Gbit/s 12QAM signal of is generated and transmitted over a 240-km SSMF with CD preequalization. The reconstructed signal constellation diagram in Fig. 4(c) shows good separation among 12 signal points and , which is well below the the BER is measured to be threshold of a super forward error correction (FEC) even considering its 7% overhead penalty not included in our experiment. III. SENSITIVITY IMPROVEMENT OF DIRECT-DETECTION MULTILEVEL SIGNALING A. Noise Distribution of Direct-Detection Multilevel Receiver Using Optical Delay Detection One of the characteristics of our direct-detection multilevel receiver lies in its noise distribution. Fig. 5 shows 2-D noise distribution on a normalized output signal point of various receiver congurations under amplied spontaneous emission (ASE)

Fig. 4. Reconstruction of directly detected 8QAM and 12QAM signals with transmitter-side phase preintegration and CD preequalization. (a) Differential constellation diagram of 8QAM signal after 400-km SSMF transmission (without CD preequalization and phase preintegration). (b) Reconstructed 8QAM signal after 400-km SSMF transmission with 6400-ps/nm CD preequalization. (c) Reconstructed 12QAM signal after 240-km SSMF transmission with 4080-ps/nm CD preequalization.

noise-limited condition, which is very typical in high-speed long-distance ber transmissions using optical preampliers and/or repeater ampliers. The distribution of coherently detected signal point with ASE is shown to be isotropic (noise ), as in Fig. 5(a). The one received with an variance optical orthogonally coupled delay-interferometric receivers is also almost isotropic, as shown in Fig. 5(b), when received eld , butits normalized amplitude noise and amplitude ) become phase noise variances (noise power when , ) than the coherent one due to twice ( the differential penalty [16]. In the coherent detection, the reference local light is not smeared by the ASE noise; however, in the delay demodulation process, both the detected symbol and the previous (reference) symbol eld have independent ASE-induced amplitude and phase noises. In the case of our multilevel receiver shown in Fig. 1(b), which has separate intensity detector, its noise distribution becomes nonisotropic and the noise variance in the angular direction becomes twice than that in the radial calculated from direction, since its amplitude component has almost the same amount of noise the square root of . Therefore, as that of the coherent detection when it can be expected that the OSNR sensitivity of our multilevel receiver lies between those of the coherent receiver and the pure orthogonally coupled delay interferometric receiver when QAM signals with amplitude modulation are detected. B. Symbol Detection With Non-Euclidean Metric and MSPE Under the aforementioned nonisotropic noise distribution with larger phase noise, the improvement of the OSNR sensitivity can be attained by the two techniques, i.e., the optimal symbol decision strategy and the further reduction of extra

KIKUCHI AND SASAKI: ARBITRARY QAM SIGNALS

127

Fig. 5. Schematic illustration of noise distribution on received signal constellation with the same average power. (a) Coherent detection r(n) exp(j(n)), (b) delay detection r (n)r (n 1) exp(j(n)), and (c) delay detection with separate intensity detection r (n) exp(j1(n)).  and  represent standard deviation of radial and angular noises.

of the current eld in reference to the th past received eld are dened and calculated as in (8) or Fig. 6(b) by removing unnecessary data phases ( decision results) using digital signal processing. They are averaged with the forgetting factor to reduce independent phase noise, and more reliable averaged as in (9) is obtained and used for differential phase . In the ideal the following symbol detection instead of case, the BER performances of differential phase detection with MSPE approach to those of the coherent detection by increasing the number of past phase references , but actual performance is limited by error propagation, signal intersymbol interference (ISI), etc.,

(8)
Fig. 6. Concept of MSPE. Black circle E (n): current received symbol eld (optical eld), open circles: past received symbol points, and black circle E (n): averaged phase reference.

(9)

phase noise. First, the best or nearly best BER performances can be obtained by the use of maximum a posterior (MAP) probability or minimum mean square error (MMSE) detector [20], but they require considerable amount of computation. As an approximation of them, we adopt non-Euclidean metric for symbol detection with higher weight on radial direction than angular direction [21], which can be implemented by simple xed LUTs. The proposed non-Euclidean metric is given by (7), where (6) represents conventional Euclidean metric that is the geometric distance of two point and . A compensation term is introduced to put higher weight on with constant radial direction than in angular direction. With it, the decision boundaries, equidistance lines between adjacent signal points, are distorted, so that each symbol decision area is widen in angular direction (6) (7) Secondary, we implement an MSPE ceiver equalizer part in order to reduce caused due to optical delay detection lizes additional past-detected function in the reexcess phase noise [16], [15]. It utidifferential phases and corresponding

C. Experimental Results We have implemented proposed functions into the ofine signal processing part of experimental multilevel receiver shown in Fig. 1(b); the symbol detector is modied to use , if not specied) for symbol non-Euclidean metric ( discrimination and the digital adaptive lter is changed to have an MSPE function using two additional past symbols with the forgetting factor . Fig. 7 shows various decision boundaries for direct-detection 8QAM signals (black bold lines) and its experimentally received and reconstructed signal point distributions (dots) under heavily ASE-loaded condition. When using the conventional in Fig. 7(a), received signal points Euclidean metric are shown to override the decision boundaries into angular direction and cause BER degradation due to the nonisotropic noise distribution described before. Newly dened non-Euclidean deand 3 are shown in cision boundaries with the weight Fig. 7(b)(d). The decision area is shown to be made wider in angular direction as the weight is higher, and the best BER is , where it best ts experimental signal point disgiven at tribution. Next, direct detection of a 16QAM signal has been veried. Fig. 8(a) shows a differential constellation diagram of a 16QAM signal at back to back, which shows tight signal point distribution that is the proof of very accurate signal modulation and demodulation processes. After replacing its amplitude component with separately detected , clear four-by-four constellation diagram of 16QAM signal is reconstructed as in Fig. 8(b). We have performed a 160-km SMF transmission (two 80-km ber spans, average loss: 16.1 dB/span, average dispersion: 16.7 ps/nm/km, ber dBm) of it with -ps/nm CD preequalinput power: ization by a 63-stage FIR lter. Using the digital adaptive

decision results to calculate a better phase reference by averaging out random phase noise on them. Its concept is schematically illustrated is nth received signal eld and is in Fig. 6, where its absolute phase. In the conventional delay demodulation, is measured in reference solely the differential phase or its phase . to the previous received eld On the other hand, MSPE utilizes further past received elds as additional phase references. The differential phases

128

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 28, NO. 1, JANUARY 1, 2010

Fig. 9. Experimental BER performances of directly detected 20-Gbit/s NRZDQPSK and 30-Gbit/s NRZ-8DPSK signals. Open symbols: with MSPE. Filled symbols: without MSPE.

Fig. 7. Symbol decision boundaries for direct-detection 8QAM signals (black bold lines) and experimental signal point distribution (dots). (a) Euclidean : , non-Euclidean metric with: (b) metric a , (c) a : , and (d) a : . a :

= 0 (BER = 8 4 2 10 ) = 2(BER = 2 7 2 10 ) = 1(BER = 4 5 2 10 ) = 3(BER = 3 12 2 10 )

Fig. 10. Experimental BER performances of directly detected 30-Gbit/s NRZ-8QAM, 35.8-Gbit/s NRZ-12QAM, and 40-Gbit/s NRZ-16QAM signals using non-Euclidean metric. Open symbols: with MSPE, lled symbols: without MSPE. Crossed symbol: with Euclidean metric without MSPE.

Fig. 8. Reconstruction of directly detected 16QAM signal. (a) Differential constellation diagram dI n ; dQ n . (b) Reconstructed Optical eld I n ;Q n .

( ( ) ( ))

( ()

( ))

lter with MSPE, the raw BER of is improved to and a successful transmission has been achieved. Next, we evaluate the experimental back-to-back OSNR sensitivities of various DPSK and QAM signals at 10 Gsymbols/s with the use of MSPE and/or non-Euclidean metric. Fig. 9 shows back-to-back BER characteristics of two DPSK signals; the OSNR sensitivities of 9.5 and 15.5 dB have been achieved for 20-Gbit/s NRZ-DQPSK and 30-Gbit/s NRZ-8DPSK signals, respectively, with MSPE, which results in 0.5- and 1.0-dB OSNR sensitivity improvement. It should be noted that the use of nonEuclidean metric does not have any impact on the DPSK signals. The experimental BER performances of three different QAM signals, 30-Gbit/s NRZ-8QAM, 35.8-Gbit/s NRZ-12QAM, and 40-Gbit/s NRZ-16QAM signals, are shown in Fig. 10, in which the use of not only MSPE, but also non-Euclidean decision metric contributes to the BER performance improvement. In the example of the 8QAM signal, the original OSNR sensitivity of (crossed symbols) with Euclidean 15.6 dB at

metric is improved to 1.4 dB by the application of non-Euclidean metric (lled boxes), and further use of MSPE results in the additional improvement of 0.7 dB and the OSNR sensitivity of 13.5 dB is achieved (open boxes). The best back-to-back sensitivities of 12QAM and 16QAM signals are 16.0 and 18.1 dB, respectively. Table I lists their experimental sensitivities and the improvement of it by the application of MSPE, non-Euclidean metric, and digital adaptive equalizer. Their contributions are different, depending on each modulation format; but in principle, non-Euclidean metric is more effective for QAM-like signal constellation in which nearest signal points have both phase and amplitude separations (modulations) at the same time, since it alleviates signal BER by actively using radial separation of them as in 8QAM and 16QAM cases. On the other hand, MSPE is more effective when the signal BER is limited by DPSK-like outer eight signal points without amplitude modulation, since it alleviates the contribution of phase noise. For example, the BER of 12QAM is improved mainly by MSPE, since its experimental BER is limited mainly by 8DPSK-like signal points, as shown in Fig. 2(b), not by inner four points. Finally. the digital adaptive equalizer alleviates the shift of signal points and ISI; therefore, it has larger contribution when the number of signal points is increased. These OSNR sensitivities are close to those obtained in the recent coherent experiments within 0.52.4 dB range. For example, our NRZ-8DPSK and NRZ-8QAM sensitivities are

KIKUCHI AND SASAKI: ARBITRARY QAM SIGNALS

129

TABLE I EXPERIMENTAL OSNR SENSITIVITIES OF DIRECTLY-DETECTED 10-GSYMBOL/S NRZ MULTILEVEL SIGNALS AT BER = 10 (OSNR RESOLUTION IS 0.1 NM, NE: NON-EUCLIDEAN METRIC, AND EQ: DIGITAL ADAPTIVE EQUALIZER)

TABLE II THEORETICAL OSNR SENSITIVITIES OF 10-GSYMBOL/S NRZ MULTILEVEL IS THE SIGNALS AT BER = 10 (OSNR RESOLUTION IS 0.1 NM AND NUMBER OF SIGNAL LEVELS)

2.5- and 0.5-dB behind from the coherent RZ-8PSK OSNR sensitivity in [22] (13.0 dB), respectively, despite 1- to 2-dB RZ-to-NRZ penalties. And our 16QAM sensitivity is 1.5- and 2.4-dB from those of the coherent experiments in [23] and [24] when converted to the same symbol rate. We compile the theoretical OSNR sensitivities of coherent PSK and QAM signals and DPSK signals presented in Table II based on some reference papers [20], [25][27]. The experimental sensitivity of our NRZ-DQPSK (9.5 dB) is 3.7 dB away from the coherent detection limit. Also, the OSNR sensitivities of the experimental direct-detected NRZ-8QAM and NRZ16QAM signals are 3.9 and 5.6 dB away from those of theoretical coherently detection of them, which show very high sensitivities of our experimental results. IV. CONCLUSION In this paper, we reviewed our recent proposal of the use of advanced signal processing for high-performance incoherent multilevel signaling. With the use of digital phase preintegration and CD preequalization, incoherent 30-Gbit/s 8QAM signal over 400-km SSMF, 35.8-Gbit/s 12QAM signal over 240-km SSMF, and 40-Gbit/s 16QAM signal over 160-km SSMF are successfully demonstrated without using optical CD compensators. We also improve the OSNR sensitivity of incoherent multilevel signaling with the use of non-Euclidean metric and MSPE, and achieve the sensitivities almost comparable to those of ever reported experimental coherent receivers. It should be noted that the reported experiments rely on ofine signal processing and not on real-time experiments, but we believe that it proves the feasibility of future multilevel signaling. REFERENCES
[1] M. G. Taylor, Coherent detection method using DSP for demodulation of signal and subsequent equalization of propagation impairments, IEEE Photon. Technol. Lett., vol. 16, no. 2, pp. 674676, Feb. 2004.

[2] S. Tsukamoto, K. Katoh, and K. Kikuchi, Unrepeated 20-Gbit/s QPSK transmission over 200-km standard single-mode ber using homodyne detection and digital signal processing for dispersion compensation, presented at the Opt. Fiber Commun. Conf. (OFC/NFOEC), Anaheim, CA, Mar. 2006, Paper OWB4. [3] C. R. S. Fludger, T. Duthel, D. Van Den Borne, C. Schulien, E.-D. Schmidt, T. Wuth, E. De Man, G. D. Khoe, and H. de Waardt, 10 Gbit/s, 50 GHz spaced, POLMUX-RZ-DQPSK transmission over 2375 km employing coherent equalisation, presented at the Opt Fiber Commun. Conf. (OFC/NFOFEC), San Diego, CA, Feb. 2008, Postdeadline paper PDP22. [4] M. Serbay, C. Wree, and W. Rosenkranz, Experimental investigation of RZ-8DPSK at 3 10.7Gb/s, presented at the 18th Annu. Meeting IEEE Lasers Electro-Opt. Soc., Sydney, Australia, Oct. 2005, Paper WE3. [5] S. Hayase, N. Kikuchi, K. Sekine, and S. Sasaki, Proposal of 8-state per symbol (binary ASK and QPSK) 30-Gbit/s optical modulation/demodulation scheme, presented at the 29th Eur. Conf. Opt. Commun. (ECOC/IOOC 2003), Rimini, Italy, Sep. , Paper Th2.6.4. [6] T. Miyazaki and F. Kubota, Superposition of DQPSK over inverse-RZ for 3-bit/symbol modulationdemodulation, IEEE Photon. Technol. Lett., vol. 16, no. 12, pp. 26432645, Dec. 2004. [7] K. Sekine, N. Kikuchi, S. Sasaki, S. Hayase, and C. Hasegawa, Proposal and demonstration of 10-Gsymbol/sec 16-ary (40 Gbit/s) optical modulation/demodulation scheme, presented at the 30th Eur. Conf. Opt. Commun. (ECOC 2004), Stockholm, Sweden, Sep. , Paper We3.4.5. [8] M. Serbay, T. Tokle, P. Jeppesen, and W. Rosenkranz, 42.8 Gbit/s, 4 bits per symbol 16-ary inverse-RZ-QASK-DQPSK transmission experiment without polmux, presented at the Opt. Fiber Commun. Conf. (OFC/NFOFEC 2007), Anaheim, CA, Mar. 2007, Paper OThL2. [9] N. Kikuchi, K. Mandai, K. Sekine, and S. Sasaki, First experimental demonstration of single-polarization 50-Gbit/s 32-level (QASK and 8-DPSK) incoherent optical multilevel transmission, presented at the Opt. Fiber Commun. Conf. (OFC/NFOEC), Anaheim, CA, Mar. 2007, Paper PDP21. [10] K. Mandai, N. Kikuchi, and S. Sasaki, First long distance (1040-km) WDM transmission of 16-channels 30-Gbit/s (480-Gbit/s) 8-ary differential phase-shift-keying (8-DPSK) signals, presented at the 12th OptoElectron. Commun. Conf. 16th Int. Conf. Integr. Opt. Opt. Fiber Commun. (OECC/IOCC 2007), Yokohama, Japan, Jul. , Postdeadline Paper P1.2. [11] N. Kikuchi, K. Mandai, S. Sasaki, and K. Sekine, Proposal and rst experimental demonstration of digital incoherent optical eld detector for chromatic dispersion compensation, presented at the 32nd Eur. Conf. Opt. Commun. (ECOC 2006), Cannes, France, Sep. , Postdeadline Paper Th.4.4.4. [12] X. Liu and X. Wei, Electronic dispersion compensation based on optical eld reconstruction with orthogonal differential direct-detection and digital signal processing, presented at the Opt. Fiber Commun. Conf. (OFC/NFOEC), Anaheim, CA, Mar. 2007, Paper OTuA6. [13] D. McGhan, C. Laperle, A. Savchenko, C. Li, G. Mak, and M. OSullivan, 5120 km RZ-DPSK transmission over G652 ber at 10 Gb/s with no optical dispersion compensation, presented at the Opt. Fiber Commun. Conf. (OFC/NFOEC), Anaheim, CA, Mar. 2005, Postdeadline paper PDP27. [14] N. Kikuchi and S. Sasaki, Incoherent 40-Gbit/s 16QAM and 30-Gbit/s staggered 8APSK (amplitude- and phase-shift keying) signaling with digital phase pre-integration technique, in Proc. IEEE LEOS Summer Top. Meeting, Acapulco, Mexico, Jul. 2008, pp. 251252, Paper WD3.2. [15] X. Liu, S. Chandrasekhar, A. H. Gnauck, C. R. Doerr, I. Kang, D. Kilper, L. L. Buhl, and J. Centanni, DSP-enabled compensation of demodulator phase error and sensitivity improvement in direct-detection 40-Gb/s DQPSK, presented at the 32nd Eur. Conf. Opt. Commun. (ECOC 2006), Cannes, France, Sep. , Postdeadline paper Th4.4.5. [16] X. Liu, Generalized data-aided multi-symbol phase estimation for improving receiver sensitivity in direct-detection optical M-ary DPSK, Opt. Exp., vol. 15, no. 6, pp. 29272939, Mar. , 2007. [17] S. Shimotsu, S. Oikawa, T. Saitou, N. Mitsugi, K. Kubodera, T. Kawanishi, and M. Izutsu, LiNbO3 optical single-sideband modulator, presented at the Opt. Fiber Commun. Conf. (OFC), Baltimore, MD, Mar. 2000, Postdeadline paper PD16-1.

130

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 28, NO. 1, JANUARY 1, 2010

[18] K. M. Aleong, H. Leib, and P. Kabal, A technique for combining equalization with generalized differential detection, in Proc. IEEE Int. Phoenix Conf. Comput. Commun., Scotsdale, AZ, Mar. 1993, pp. 416422. [19] N. Kikuchi and S. Sasaki, Optical dispersion-compensation free incoherent multilevel signal transmission over standard single-mode ber with digital pre-distortion and phase pre-integration techniques, presented at the 34th Eur. Conf. Opt. Commun. (ECOC 2008), Brussels, Belgium, Sep. , Paper Tu.1.E.2. [20] K.-P. Ho, Phase-Modulated Optical Communication Systems. New York: Springer-Verlag, 2005. [21] N. Kikuchi and S. Sasaki, Sensitivity improvement of incoherent multilevel (30-Gbit/s 8QAM and 40-Gbit/s 16QAM) signaling with nonEuclidean metric and MSPE (multi symbol phase estimation), presented at the Opt. Fiber Commun. Conf. (OFC/NFOFEC 2009), San Diego, CA, Mar. , Paper OWG1. [22] R. Freund, D.-D. Gros, M. Seimetz, L. Molle, and C. Caspar, 30 Gbit/s RZ-8-PSK transmission over 2800 km standard single mode bre without inline dispersion compensation, presented at the Opt. Fiber Commun. Conf. (OFC), San Jose, CA, Feb. 2008, Paper OMI5. [23] P. J. Winzer and A. H. Gnauck, 112-Gb/s polarization-multiplexed 16-QAM on a 25-GHz WDM grid, presented at the 34th Eur. Conf. Opt. Commun. (ECOC 2008), Brussels, Belgium, Sep. 2008, Postdeadline paper Th.3.E.5. [24] A. H. Gnauck, P. J. Winzer, C. R. Doerr, and L. L. Buhl, 10 ; 112-Gb/s PDM 16-QAM transmission over 630 km of ber with 6.2-b/s/Hz spectral efciency, presented at the Opt. Fiber Commun. Conf. (OFC), San Jose, CA, Feb. 2009, Postdeadline paper PDPB8. [25] R. Freund and M. Seimetz, Higher order modulation formats using coherent detection and electronic distortion equalisation for application in future backbone networks, in Proc. 2008 IEEE LEOS Summer Top., Acapulco, Mexico, Jul. 2008, pp. 253254, Paper WD3.3. [26] M. Nazarathy, E. Simony, and Y. Yadin, Analytical evaluation of bit error rates for hard detection of optical differential phase amplitude shift keying (DPASK), J. Lightw. Technol., vol. 24, no. 5, pp. 22482260, May 2006.

[27] E. Ip and J. M. Kahn, Carrier synchronization for 3- and 4-bit-persymbol optical transmission, J. Lightw. Technol., vol. 23, no. 12, pp. 41104124, Dec. 2005.

Nobuhiko Kikuchi (M02) was born in Tokyo, Japan, in 1965. He received the B.E. and M.E. degrees in precision mechanics from Tokyo University, Tokyo, in 1988 and 1990. Since 1990, he has been with the Central Research Laboratory, Hitachi Ltd., Tokyo, where he is involved in high-speed optical ber communication, covering advanced modulation schemes including multilevel modulations, ber nonlinear impairments, polarization-mode dispersion, etc. From 2001 to 2002, he was a visiting researcher at Optical Communication Research Laboratory, Stanford university. CA. Mr. Kikuchi is a member of the Institute of Electronics, Information, and Communication Engineers of Japan.

Shinya Sasaki (M87SM09) was born in Hokkaido, Japan. He received the B.Eng., M.Eng., and D.Eng. degrees in electronic engineering from Hokkaido University, Hokkaido, in 1975, 1977, and 1980, respectively. Since 1980, he has been with the Central Research Laboratory, Hitachi Ltd., Tokyo, Japan, where he is involved in high-speed lightwave transmission systems. From 1986 to 1987, he was a visiting researcher with Bellcore, Red Bank, NJ. Dr. Sasaki is a member of the Institute of Electronics, Information, and Communication Engineers of Japan.

Vous aimerez peut-être aussi