November 1, 2014 / Vol. 39, No. 21 / OPTICS LETTERS
6301
Electrical filter-based and low-complexity DPSK coherent optical receiver Marco Presi,* Massimo Artiglia, and Ernesto Ciaramella Scuola Superiore Sant’Anna University, TeCIP Institute, via Moruzzi, 1-56037 Pisa, Italy *Corresponding author: [email protected] Received August 5, 2014; revised October 3, 2014; accepted October 6, 2014; posted October 7, 2014 (Doc. ID 220466); published October 28, 2014 Direct-detection of differential phase shifted keying (DPSK) optical signals is implemented either by ad hoc optical filtering or one-bit delayed optical interferometers. We show a coherent receiver where the filtering is performed in the electrical domain after down-converting the incoming signal to the baseband or to an intermediate frequency. This can be particularly advantageous whenever optical filters cannot be used or when extremely narrow filtering (sub-GHz) would be required [for example ultra-dense wavelength division multiplexing (WDM) passive optical networks]. Electrical filters can be realized more accurately than their optical counterpart. In addition, in a coherent receiver, this operation is colorless and avoids digital signal processing. We show experimentally this reduced complexity receiver and compare it with the classical one based on the delay and multiply block. © 2014 Optical Society of America OCIS codes: (060.1660) Coherent communications; (070.1170) Analog optical signal processing; (060.4080) Modulation. http://dx.doi.org/10.1364/OL.39.006301
Coherent receivers are currently under investigation to implement ultra-dense WDM optical systems for access networks: by using the local oscillator as WDM channel selector, they can realize the wavelength to the user paradigm [1]. However, while digital coherent receivers offer high-end functionalities [2], it is still not clear if they match cost and consumption levels demanded by the mass deployment. In this scenario, DPSK signaling is interesting not only because it shows an improved sensitivity compared to the amplitude shift keying (ASK) format, but also because it can be generated by using directly modulated lasers, still achieving minimal spectrum occupation [3]. Detection of DPSK signals is usually performed by a one-bit delay and multiply block [Fig. 1(c)], which might be tricky to implement in an analog coherent receiver. Here we analyze an alternative approach where the coherent receiver mimics the DPSK direct-detection scheme, which is based on optical filtering [4–6], [see Figs. 1(a) and 1(b)]. To do this, we move the filtering stage in the electrical domain, right after the down-conversion [see Fig. 1(d)]. By doing this, the filtering becomes colorless and the analog receiver structure is greatly simplified. In this Letter we present the design of such a receiver. After showing the equivalence of the two detection schemes, we demonstrate it experimentally and discuss its benefit in the implementation of low-cost, DSP-free coherent receivers. DPSK direct-detection receivers are traditionally realized by a delay-line interferometer (one bit delay time) or, equivalently, by a Gaussian filter having a bandwidth of about 60% of the signal bit-rate [Figs. 1(a) and 1(b)]. The latter approach simplifies the receiver structure and has been used to enable WDM demultiplexing [7]. Similarly, the conventional DPSK coherent receiver is based on a delay and multiply scheme [Fig. 1(c)], which might be unpractical to realize in an analog processing block. On the contrary, emulating the all-optical approach an electrical filter followed by a square function (an 0146-9592/14/216301-03$15.00/0
envelope detector) would replace the delay and multiply block, greatly simplifying a fully analog receiver realization. The bandwidth of the Gaussian filter should match the one required in the all-optical approach, i.e., 60% of the bit-rate, while its central frequency corresponds to the intermediate frequency (IF) or 0 in case of heterodyne and homodyne detection, respectively. Notably, besides the receiver simplification, the proposed approach also reduces the 3 dB bandwidth of the electrical filter if an homodyne down-conversion is adopted. The equivalence of the approaches can be easily demonstrated for a phase-diversity coherent receiver. We consider a phase-modulated optical signal st oscillating at a frequency f s and with phase ψ s t:
Fig. 1. DSPK detection schemes. (a) and (b) show direct detection by means of optical delay interferometer and Gaussian filter, respectively. (c) and (d) show coherent detection achieved by one-bit delay and multiply block or Gaussian filter, respectively. (e) and (f) show the equivalence of the filtering operated in the optical or electrical domain after the downconversion. © 2014 Optical Society of America
6302
OPTICS LETTERS / Vol. 39, No. 21 / November 1, 2014
st e|ψ s t e|2πf s t cos ψ s t | sin ψ s te|2πf s t sI t |sQ te|2πf s t .
(1)
In Eq. (1), we defined sI t cos ψ s t and sQ t sin ψ s t as the baseband in-phase and quadrature modulation components. In a direct-detection receiver where the signal is preprocessed by an optical filter Gf − f s [see Figs. 1(a) and 1(b)] the detection current idd is given by: Z 2 |2πf t idd t R Gf S I f |S Q f e df ;
(2)
where R being the photodiode responsivity, S I f and S Q f the spectrum of sI t and sQ t, respectively. In order to perform demodulation of DPSK signals, the filter G· usually has either a cosine or Gaussian shape [4,5,8]. Similarly, in the phase-diversity coherent receiver illustrated in Fig. 2, N copies of the received signal st are first down-converted to the baseband by mixing them with N, evenly phase-shifted copies of a local oscillator, of frequency f LO and phase ψ n t: sLO t ELO e|2πf LO tψ n tφk , φk 2πk∕N, k 0; 1; 2; …. Assuming perfect homodyne beating (f s f LO ), this operation produces multiple electrical currents, each other differing only in the relative phase shift: E cosψ s t − ψ n t φk . ık R pLO N
Z 2 E 2LO Gf S I f e|2πf t df N Z 2 |2πf t . Gf S Q f e df
ied t R2
(3)
In analogy with (1), the kth current can be decomposed in its in-phase and quadrature components that are common to all the currents: E cosφk − ψ n tsI t − sinφk − ψ n tsQ t. ik R pLO N (4) In most cases of practical interest, the phase noise term ψ n t is slowly varying compared to the Gf bandwidth, i.e., it can be considered constant over one bit, and can be neglected. Following the down-conversion, the currents are first individually processed by the filters Gf then squared and eventually summed together. It can be easily shown that the resulting current ied provides the demodulated signal envelope and is given by:
Fig. 2. Experimental setup. TX: DPSK transmitter based on external modulation. PPG, pulse pattern generator; DFB, distributed feed back laser; RTO, real time oscilloscope.
(5)
By comparing this signal to the one obtained by direct detection (2), we see that the following relation holds: ied t ∝ RE2LO idd t.
(6)
Equation (6) shows that in a coherent receiver where the signal processing stage is implemented by an electrical filter followed by an envelope detection stage is equivalent to a direct-detection receiver based on optical filtering. The coherent receiver, however, provides a better sensitivity, and it is colorless and provides for an higher channel selectivity. The above equivalence can be extended to heterodyne detection, although in this case, the filter G· becomes a band-pass filter centered at the heterodyne frequency. We investigated the behavior of the proposed receivers for DPSK signal in a proof-of-concept experiment. We used an homodyne phase-diversity coherent receiver based on a 3 × 3 optical coupler that provides for 120° relative phase shift among the output ports (Fig. 2). The receiver could be realized also by using the conventional 90° hybrid coupler, which is however more expensive. The three photodiodes had 10 GHz bandwidth, with an average responsivity of 0.65 A∕W. A real-time oscilloscope (RTO, 4 channels, 20 GSa∕s and 13 GHz analog bandwidth) was used to implement pre-detection filtering, envelope detection, and Q-factor estimation with maximum flexibility, although a fully analog implementation of the receiver would be desirable. The use of a RTO allowed to emulate the analogue processing described above, and to test filters of varying bandwidth. The filter shape was fixed to a low-pass Gaussian response with flat phase response. Two independent, free-running lasers served as local oscillator and as signal source, respectively. The DPSK signal, carrying PRBS sequences is generated by using a LiNbO3 Mach–Zehnder external modulator biased at null. Bit-rate is fixed at 1.25 Gb∕s, which is of interest for UDWDM access systems [1]. The main result is reported in Fig. 3, where we show the measured values of the bit-error-ratio (BER) as a function of the Gaussian filter bandwidth. The BER measurements for the different filter bandwidths have been
Fig. 3. Experimental BER measurement as a function of the low-pass Gaussian filter bandwidth. Insets show the typical eye diagram (horizontal timescale: 0.2 ns∕div; vertical scale: 10 mV∕div).
November 1, 2014 / Vol. 39, No. 21 / OPTICS LETTERS
Fig. 4. Experimental measurement of the power penalty at BER 10−3 as a function of the signal-LO detuning.
made on the same single acquisitions, in order to avoid other sources of error. In this case, the frequency detuning was kept below 10 MHz. The optimal filter bandwidth (the one that minimizes the BER) is between 500 and 550 MHz. In optical filter-based DPSK receivers, the optimal filter bandwidth corresponds to about 60% of the signal bit-rate. However, it should be noted that usually the bandwidth of optical filters is measured over the power spectrum, i.e., the pmeasured bandwidth appears decreased by a factor 1∕ 2 in case of a Gaussian-shaped filter. By applying this correction factor, for a 1.25 Gb∕s signal, the optimal Gaussian filter FWHM results to be 1.06 GHz. In case of an homodyne receiver, the electrical filter is low-pass (see Fig. 2), and the optimal filter bandwidth becomes 530 MHz: this value matches exactly the one obtained experimentally. We found that the recovered eye-diagrams show an inverse-RZ-like shape that is stronger than usual because of the spurious intensity modulation due to the use of the Mach–Zehnder biased at null. This required to use a post-detection Bessel filter slightly larger than usual: 1.2 GHz, rather the 933 MHz as expected for 1.25 Gb∕s systems. Tolerance to detuning is intrinsically low in DPSK receivers: this has been observed both in direct detection and coherent analog receivers [9,3]. Replacing the filtering in the electrical domain makes no exception, as reported in Fig. 4, where we report the power penalty as a function of the detuning. For comparison, we also report the power penalty observed for the classical delay-and-multiply demodulation scheme. As can be observed, in both cases a power penalty arises as the detuning exceeds 50 MHz. It is worth noting that the same processing can be performed in case of heterodyne
6303
detection: in this case, a single filter and a single envelope-detection stage are required, which further simplifies the receiver structure. In conclusion, we presented and experimentally demonstrated a simplified coherent detection scheme for 1.25 Gb∕s DPSK signals, which avoids electrical delay lines. It mimics the Gaussian filtering approach used in direct-detection optical receiver, but it is implemented in the electrical domain; by doing this, the receiver is based on analog signal processing, and the filtering becomes colorless and without additional optical insertion losses. The proposed solution could be implemented in simplified and low-cost receivers to be adopted in optical access network, [1], but also in other scenarios where it could be useful to implement coherent detection. Examples include chip/board level optical interconnects [10] and laser radar systems [11]. This work has partly been supported by the European FP7 project COCONUT (contract no. 318515) References 1. J. Prat, M. Angelou, C. Kazmierski, R. Pous, M. Presi, A. Rafel, G. Vall-llosera, I. Tomkos, and E. Ciaramella, ICTON 2013 (IEEE, 2013), paper Tu.C3.2. 2. H. Rohde, E. Gottwald, A. Teixeira, J. D. Reis, A. Shahpari, K. Pulverer, and J. S. Wey, J. Lightwave Technol. 32, 2041 (2014). 3. I. Cano, A. Lerin, V. Polo, and J. Prat, IEEE Photon. Technol. Lett. 26, 973 (2014). 4. E. A. Swanson, J. C. Livas, and R. S. Bondurant, IEEE Photon. Technol. Lett. 6, 263 (1994). 5. D. Penninckx, H. Bissessur, P. Brindel, E. Gohin, and F. Bakhti, ECOC 2001, Vol. 3 (IEEE, 2001), pp. 456–457. 6. I. Lyubomirsky and C.-C. Chien, IEEE Photon. Technol. Lett. 17, 492 (2005). 7. A. D’Errico, R. Proietti, L. Giorgi, G. Contestabile, and E. Ciaramella, Electron. Lett. 42, 112 (2006). 8. A. H. Gnauck and P. J. Winzer, J. Lightwave Technol. 23, 115 (2005). 9. F. Vacondio, A. Ghazisaeidi, A. Bononi, and L. A. Rusch, J. Lightwave Technol. 27, 5106 (2009). 10. X. Zhang, A. Hosseini, X. Lin, H. Subbaraman, and R. T. Chen, IEEE J. Sel. Top. Quantum Electron. 19, 3401115 (2013). 11. R. G. Frehlich and M. J. Kavaya, Appl. Opt. 30, 5325 (1991).
6301
Electrical filter-based and low-complexity DPSK coherent optical receiver Marco Presi,* Massimo Artiglia, and Ernesto Ciaramella Scuola Superiore Sant’Anna University, TeCIP Institute, via Moruzzi, 1-56037 Pisa, Italy *Corresponding author: [email protected] Received August 5, 2014; revised October 3, 2014; accepted October 6, 2014; posted October 7, 2014 (Doc. ID 220466); published October 28, 2014 Direct-detection of differential phase shifted keying (DPSK) optical signals is implemented either by ad hoc optical filtering or one-bit delayed optical interferometers. We show a coherent receiver where the filtering is performed in the electrical domain after down-converting the incoming signal to the baseband or to an intermediate frequency. This can be particularly advantageous whenever optical filters cannot be used or when extremely narrow filtering (sub-GHz) would be required [for example ultra-dense wavelength division multiplexing (WDM) passive optical networks]. Electrical filters can be realized more accurately than their optical counterpart. In addition, in a coherent receiver, this operation is colorless and avoids digital signal processing. We show experimentally this reduced complexity receiver and compare it with the classical one based on the delay and multiply block. © 2014 Optical Society of America OCIS codes: (060.1660) Coherent communications; (070.1170) Analog optical signal processing; (060.4080) Modulation. http://dx.doi.org/10.1364/OL.39.006301
Coherent receivers are currently under investigation to implement ultra-dense WDM optical systems for access networks: by using the local oscillator as WDM channel selector, they can realize the wavelength to the user paradigm [1]. However, while digital coherent receivers offer high-end functionalities [2], it is still not clear if they match cost and consumption levels demanded by the mass deployment. In this scenario, DPSK signaling is interesting not only because it shows an improved sensitivity compared to the amplitude shift keying (ASK) format, but also because it can be generated by using directly modulated lasers, still achieving minimal spectrum occupation [3]. Detection of DPSK signals is usually performed by a one-bit delay and multiply block [Fig. 1(c)], which might be tricky to implement in an analog coherent receiver. Here we analyze an alternative approach where the coherent receiver mimics the DPSK direct-detection scheme, which is based on optical filtering [4–6], [see Figs. 1(a) and 1(b)]. To do this, we move the filtering stage in the electrical domain, right after the down-conversion [see Fig. 1(d)]. By doing this, the filtering becomes colorless and the analog receiver structure is greatly simplified. In this Letter we present the design of such a receiver. After showing the equivalence of the two detection schemes, we demonstrate it experimentally and discuss its benefit in the implementation of low-cost, DSP-free coherent receivers. DPSK direct-detection receivers are traditionally realized by a delay-line interferometer (one bit delay time) or, equivalently, by a Gaussian filter having a bandwidth of about 60% of the signal bit-rate [Figs. 1(a) and 1(b)]. The latter approach simplifies the receiver structure and has been used to enable WDM demultiplexing [7]. Similarly, the conventional DPSK coherent receiver is based on a delay and multiply scheme [Fig. 1(c)], which might be unpractical to realize in an analog processing block. On the contrary, emulating the all-optical approach an electrical filter followed by a square function (an 0146-9592/14/216301-03$15.00/0
envelope detector) would replace the delay and multiply block, greatly simplifying a fully analog receiver realization. The bandwidth of the Gaussian filter should match the one required in the all-optical approach, i.e., 60% of the bit-rate, while its central frequency corresponds to the intermediate frequency (IF) or 0 in case of heterodyne and homodyne detection, respectively. Notably, besides the receiver simplification, the proposed approach also reduces the 3 dB bandwidth of the electrical filter if an homodyne down-conversion is adopted. The equivalence of the approaches can be easily demonstrated for a phase-diversity coherent receiver. We consider a phase-modulated optical signal st oscillating at a frequency f s and with phase ψ s t:
Fig. 1. DSPK detection schemes. (a) and (b) show direct detection by means of optical delay interferometer and Gaussian filter, respectively. (c) and (d) show coherent detection achieved by one-bit delay and multiply block or Gaussian filter, respectively. (e) and (f) show the equivalence of the filtering operated in the optical or electrical domain after the downconversion. © 2014 Optical Society of America
6302
OPTICS LETTERS / Vol. 39, No. 21 / November 1, 2014
st e|ψ s t e|2πf s t cos ψ s t | sin ψ s te|2πf s t sI t |sQ te|2πf s t .
(1)
In Eq. (1), we defined sI t cos ψ s t and sQ t sin ψ s t as the baseband in-phase and quadrature modulation components. In a direct-detection receiver where the signal is preprocessed by an optical filter Gf − f s [see Figs. 1(a) and 1(b)] the detection current idd is given by: Z 2 |2πf t idd t R Gf S I f |S Q f e df ;
(2)
where R being the photodiode responsivity, S I f and S Q f the spectrum of sI t and sQ t, respectively. In order to perform demodulation of DPSK signals, the filter G· usually has either a cosine or Gaussian shape [4,5,8]. Similarly, in the phase-diversity coherent receiver illustrated in Fig. 2, N copies of the received signal st are first down-converted to the baseband by mixing them with N, evenly phase-shifted copies of a local oscillator, of frequency f LO and phase ψ n t: sLO t ELO e|2πf LO tψ n tφk , φk 2πk∕N, k 0; 1; 2; …. Assuming perfect homodyne beating (f s f LO ), this operation produces multiple electrical currents, each other differing only in the relative phase shift: E cosψ s t − ψ n t φk . ık R pLO N
Z 2 E 2LO Gf S I f e|2πf t df N Z 2 |2πf t . Gf S Q f e df
ied t R2
(3)
In analogy with (1), the kth current can be decomposed in its in-phase and quadrature components that are common to all the currents: E cosφk − ψ n tsI t − sinφk − ψ n tsQ t. ik R pLO N (4) In most cases of practical interest, the phase noise term ψ n t is slowly varying compared to the Gf bandwidth, i.e., it can be considered constant over one bit, and can be neglected. Following the down-conversion, the currents are first individually processed by the filters Gf then squared and eventually summed together. It can be easily shown that the resulting current ied provides the demodulated signal envelope and is given by:
Fig. 2. Experimental setup. TX: DPSK transmitter based on external modulation. PPG, pulse pattern generator; DFB, distributed feed back laser; RTO, real time oscilloscope.
(5)
By comparing this signal to the one obtained by direct detection (2), we see that the following relation holds: ied t ∝ RE2LO idd t.
(6)
Equation (6) shows that in a coherent receiver where the signal processing stage is implemented by an electrical filter followed by an envelope detection stage is equivalent to a direct-detection receiver based on optical filtering. The coherent receiver, however, provides a better sensitivity, and it is colorless and provides for an higher channel selectivity. The above equivalence can be extended to heterodyne detection, although in this case, the filter G· becomes a band-pass filter centered at the heterodyne frequency. We investigated the behavior of the proposed receivers for DPSK signal in a proof-of-concept experiment. We used an homodyne phase-diversity coherent receiver based on a 3 × 3 optical coupler that provides for 120° relative phase shift among the output ports (Fig. 2). The receiver could be realized also by using the conventional 90° hybrid coupler, which is however more expensive. The three photodiodes had 10 GHz bandwidth, with an average responsivity of 0.65 A∕W. A real-time oscilloscope (RTO, 4 channels, 20 GSa∕s and 13 GHz analog bandwidth) was used to implement pre-detection filtering, envelope detection, and Q-factor estimation with maximum flexibility, although a fully analog implementation of the receiver would be desirable. The use of a RTO allowed to emulate the analogue processing described above, and to test filters of varying bandwidth. The filter shape was fixed to a low-pass Gaussian response with flat phase response. Two independent, free-running lasers served as local oscillator and as signal source, respectively. The DPSK signal, carrying PRBS sequences is generated by using a LiNbO3 Mach–Zehnder external modulator biased at null. Bit-rate is fixed at 1.25 Gb∕s, which is of interest for UDWDM access systems [1]. The main result is reported in Fig. 3, where we show the measured values of the bit-error-ratio (BER) as a function of the Gaussian filter bandwidth. The BER measurements for the different filter bandwidths have been
Fig. 3. Experimental BER measurement as a function of the low-pass Gaussian filter bandwidth. Insets show the typical eye diagram (horizontal timescale: 0.2 ns∕div; vertical scale: 10 mV∕div).
November 1, 2014 / Vol. 39, No. 21 / OPTICS LETTERS
Fig. 4. Experimental measurement of the power penalty at BER 10−3 as a function of the signal-LO detuning.
made on the same single acquisitions, in order to avoid other sources of error. In this case, the frequency detuning was kept below 10 MHz. The optimal filter bandwidth (the one that minimizes the BER) is between 500 and 550 MHz. In optical filter-based DPSK receivers, the optimal filter bandwidth corresponds to about 60% of the signal bit-rate. However, it should be noted that usually the bandwidth of optical filters is measured over the power spectrum, i.e., the pmeasured bandwidth appears decreased by a factor 1∕ 2 in case of a Gaussian-shaped filter. By applying this correction factor, for a 1.25 Gb∕s signal, the optimal Gaussian filter FWHM results to be 1.06 GHz. In case of an homodyne receiver, the electrical filter is low-pass (see Fig. 2), and the optimal filter bandwidth becomes 530 MHz: this value matches exactly the one obtained experimentally. We found that the recovered eye-diagrams show an inverse-RZ-like shape that is stronger than usual because of the spurious intensity modulation due to the use of the Mach–Zehnder biased at null. This required to use a post-detection Bessel filter slightly larger than usual: 1.2 GHz, rather the 933 MHz as expected for 1.25 Gb∕s systems. Tolerance to detuning is intrinsically low in DPSK receivers: this has been observed both in direct detection and coherent analog receivers [9,3]. Replacing the filtering in the electrical domain makes no exception, as reported in Fig. 4, where we report the power penalty as a function of the detuning. For comparison, we also report the power penalty observed for the classical delay-and-multiply demodulation scheme. As can be observed, in both cases a power penalty arises as the detuning exceeds 50 MHz. It is worth noting that the same processing can be performed in case of heterodyne
6303
detection: in this case, a single filter and a single envelope-detection stage are required, which further simplifies the receiver structure. In conclusion, we presented and experimentally demonstrated a simplified coherent detection scheme for 1.25 Gb∕s DPSK signals, which avoids electrical delay lines. It mimics the Gaussian filtering approach used in direct-detection optical receiver, but it is implemented in the electrical domain; by doing this, the receiver is based on analog signal processing, and the filtering becomes colorless and without additional optical insertion losses. The proposed solution could be implemented in simplified and low-cost receivers to be adopted in optical access network, [1], but also in other scenarios where it could be useful to implement coherent detection. Examples include chip/board level optical interconnects [10] and laser radar systems [11]. This work has partly been supported by the European FP7 project COCONUT (contract no. 318515) References 1. J. Prat, M. Angelou, C. Kazmierski, R. Pous, M. Presi, A. Rafel, G. Vall-llosera, I. Tomkos, and E. Ciaramella, ICTON 2013 (IEEE, 2013), paper Tu.C3.2. 2. H. Rohde, E. Gottwald, A. Teixeira, J. D. Reis, A. Shahpari, K. Pulverer, and J. S. Wey, J. Lightwave Technol. 32, 2041 (2014). 3. I. Cano, A. Lerin, V. Polo, and J. Prat, IEEE Photon. Technol. Lett. 26, 973 (2014). 4. E. A. Swanson, J. C. Livas, and R. S. Bondurant, IEEE Photon. Technol. Lett. 6, 263 (1994). 5. D. Penninckx, H. Bissessur, P. Brindel, E. Gohin, and F. Bakhti, ECOC 2001, Vol. 3 (IEEE, 2001), pp. 456–457. 6. I. Lyubomirsky and C.-C. Chien, IEEE Photon. Technol. Lett. 17, 492 (2005). 7. A. D’Errico, R. Proietti, L. Giorgi, G. Contestabile, and E. Ciaramella, Electron. Lett. 42, 112 (2006). 8. A. H. Gnauck and P. J. Winzer, J. Lightwave Technol. 23, 115 (2005). 9. F. Vacondio, A. Ghazisaeidi, A. Bononi, and L. A. Rusch, J. Lightwave Technol. 27, 5106 (2009). 10. X. Zhang, A. Hosseini, X. Lin, H. Subbaraman, and R. T. Chen, IEEE J. Sel. Top. Quantum Electron. 19, 3401115 (2013). 11. R. G. Frehlich and M. J. Kavaya, Appl. Opt. 30, 5325 (1991).
