Widely-tunable low-phase-noise coherent receiver using an optical Wadley loop

The Wadley Loop is a method of down-converting RF signals over a wide frequency range using a low-quality widely-tunable oscillator and a high-stability frequency comb reference. Together the widely tunable oscillator and high-stability comb source provide a widely-tunable highstability receiver. In this paper, we demonstrate an electro-optic version of the Wadley Loop that is able to provide a widely-tunable, high phase stability coherent receiver. This could have applications in Quadrature Amplitude Modulation (QAM) receivers with high constellation sizes, optical OFDM receivers with long symbol durations, and wide-range high spectral resolution optical spectrum analysers. ©2015 Optical Society of America OCIS codes: (060.2330) Fiber optics communications; (060.4080) Modulation; (060.4230) Multiplexing; (060.1660) Coherent communications. References and links 1. T. Kimura, “Coherent optical fiber transmission,” J. Lightwave Technol. 5(4), 414–428 (1987). 2. E. Ip, A. P. T. Lau, D. J. F. Barros, and J. M. Kahn, “Coherent detection in optical fiber systems,” Opt. Express 16(2), 753–791 (2008). 3. P. J. Winzer and R. J. Essiambre, “Receivers for advanced optical modulation formats,” in 16th Annual Mtg. of Lasers and Electro-Optics Society, LEOS (Baltimore, MD, 2003), pp. 759–760. 4. W. Shieh, H. Bao, and Y. Tang, “Coherent optical OFDM: theory and design,” Opt. Express 16(2), 841–859 (2008). 5. S. L. Jansen, I. Morita, N. Tadeka, and H. Tanaka, “20-Gb/s OFDM transmission over 4,160-km SSMF enabled by RF-pilot tone phase noise compensation,” in Conference on Optical Fiber Communication, OFC (Anaheim, CA., 2007), p. PDP15. 6. S. Beppu, K. Kasai, M. Yoshida, and M. Nakazawa, “2048 QAM (66 Gbit/s) single-carrier coherent optical transmission over 150 km with a potential SE of 15.3 bit/s/Hz,” Opt. Express 23(4), 4960–4969 (2015). 7. E. Ip and J. M. Kahn, “Feedforward carrier recovery for coherent optical communications,” J. Lightwave Technol. 25(9), 2675–2692 (2007). 8. T. Pfau, S. Hoffmann, and R. Noé, “Hardware-efficient coherent digital receiver concept with feedforward carrier recovery for M-QAM constellations,” J. Lightwave Technol. 27(8), 989–999 (2009). 9. S. Zhang, P. Y. Kam, C. Yu, and J. Chen, “Decision-aided carrier phase estimation for coherent optical communications,” J. Lightwave Technol. 28(11), 1597–1607 (2010). 10. “Wadley Loop,” http://en.wikipedia.org/wiki/Wadley_Loop. 11. D. L. Talbot, “Frequency Synthesis Applications,” in Frequency Aquisition Techniques for Phase Locked Loops (John Wiley and Sons, New Jersey, 2012), pp. 166–173. 12. Y. Ma, Y. Qi, T. Yan, C. Simin, and W. Shieh, “1-Tb/s Single-channel coherent optical OFDM transmission with orthogonal-band multiplexing and subwavelength bandwidth access,” J. Lightwave Technol. 28(4), 308– 315 (2010). 13. S. Savory and A. Hadjifotiou, “Laser linewidth requirements for optical DQPSK systems,” IEEE Photonics Technol. Lett. 16(3), 930–932 (2004). 14. B. Hilpert, “Racal RA-17 & RA-117 Technical Description,” (2008), http://www.cs.ubc.ca/~hilpert/e/radio/RacalRA117/technical.html#wadley2015. 15. M. Nazarathy and A. Tolmachev, “Subbanded DSP architectures based on underdecimated filter banks for coherent OFDM receivers: overview and recent advances,” IEEE Signal Process. Mag. 31(2), 70–81 (2014). 16. A. Albores-Mejia, T. Kaneko, E. Banno, K. Uesaka, H. Shoji, and H. Kuwastuka, “Optical-comb-line selection from a low-power/low-OSNR comb using a low-coherence semiconductor laser for flexible ultra-dense short range transceivers,” in Optical Fiber Commun. Conf.(OSA, 2015), pp. W2A.23. 17. S. J. Savory, “Digital filters for coherent optical receivers,” Opt. Express 16(2), 804–817 (2008). #242888 Received 12 Jun 2015; revised 14 Jul 2015; accepted 15 Jul 2015; published 22 Jul 2015 © 2015 OSA 27 Jul 2015 | Vol. 23, No. 15 | DOI:10.1364/OE.23.019891 | OPTICS EXPRESS 19891 18. C. Zhu, B. Corcoran, A. V. Tran, and A. J. Lowery, “Nyquist-WDM with low-complexity joint matched filtering and adaptive equalization,” IEEE Photonics Technol. Lett. 26(23), 2323–2326 (2014). 19. J. C. M. Diniz, J. C. R. F. de Oliveira, E. S. Rosa, V. B. Ribeiro, V. E. S. Parahyba, R. da Silva, E. P. da Silva, L. H. H. de Carvalho, A. F. Herbster, and A. C. Bordonalli, “Simple feed-forward wide-range frequency offset estimator for optical coherent receivers,” Opt. Express 19(26), B323–B328 (2011).


Introduction
Coherent optical communications requires a local-oscillator laser at the receiver to mix with the incoming signal [1,2].If the receiver is to work across, say, the C-band of optical telecommunications, the local oscillator must tune across this 35-nm band.Additionally, if the received signal is very sensitive to phase noise, such as in QAM modulation with large constellations [3], or Coherent Optical Orthogonal Frequency Division Multiplexing (CO-OFDM) [4] with long symbol lengths, then the local oscillators must have high phase stability (a narrow linewidth) in the order of 100 kHz.Furthermore, if advanced spectrum shaping techniques, such as Nyquist filters or CO-OFDM are employed, zero frequency offset is preferred to allow accurate operation of the matched filter or FFT.These requirements are often conflicting when designing a suitable local oscillator laser, and so techniques such as transmitting a pilot tone [5,6] must be used.Alternatively, digital techniques can be used to track and correct frequency [7] and phase errors [8,9], although these are generally limited to when the laser linewidth is much less than the inverse of the symbol duration.
The Wadley Loop was developed in the 1940's to provide a technique to accurately downconvert radio frequency signals covering a wide frequency range to a specific and narrower frequency range for reception by a narrower-band receiver circuit [10,11].The accuracy of down-conversion was required due to the narrowness of the radio channels (a few kHz) compared to the wide range of frequencies that they might be transmitted on (tens of MHz); that is, any phase noise or drift in the conversion process would mean that the final stages of the receiver could not select them.Thus, the problem being solved is similar to receiving a narrow-band OFDM subcarrier in a 'superchannel' [12], or a signal with low tolerance to phase noise [13].In both cases, a narrow-linewidth, highly tunable, high frequency stability, local oscillator is required.Although the Wadley Loop formed the basis of the RACAL RA-17 and RA-117 1-30 MHz (HF band) radio receivers in the 1950's [14], it fell out of use with the advent of phase-locked loops that enabled the local oscillator to be locked to a crystal oscillator by frequency synthesis.However, the architecture of the Wadley Loop is very suitable for implementation as an electro-optic circuit, and thus it provides a means of implementing widely-tunable highly-accurate optical receivers.This is achieved by combining a lower-quality widely tunable laser, as is available commercially, with a highly stable optical comb source, such as a mode locked laser.For economy, only one mode-locked laser is required for a bank of optical receivers or the whole exchange, and advantageously, all of the receivers will reference from this source so that post-processing techniques can be applied over the outputs of many receivers for nonlinearity compensation over many bands, for example as in [15].A widely tunable, frequency accurate receiver is also useful for highresolution spectral analysis.
In this paper we cover the principle of the Wadley Loop as implemented in a radio receiver, then show how it can be implemented in an optical receiver.We then present results from a laboratory demonstration using commercial off-the-shelf components.We show that the system leads to reduced rates of phase wander and frequency offset in the constellations of a received optical QPSK signal, and so improves signal quality over a wide range of local oscillator tunings.

Principle of a Wadley Loop
Figure 1 illustrates the Wadley Loop [14], and the evolution of the signals' frequencies along the paths, which are color keyed to the blocks in the signal flow diagram.Gain stages are omitted for clarity.An input signal with a test frequency fin between zero and f in,max , is mixed (mixer M1) with the output of a widely-tunable oscillator (WTO) operating in a range above the input frequency range.Mixer M1, in combination with an IF (Intermediate Frequency) filter, produces an IF signal with frequency: The WTO can be thought of as a 'band selector'.The WTO should be roughly tuned in frequency steps, Δf, equal to those of the frequency comb generator (CG), though its exact tuning (and subsequent drifts in phase and frequency) is unimportant, as will be shown.The bandwidth of the IF1 filter (illustrated as a colored box) should be somewhat larger than Δf to accommodate drifting, mistuning and frequency noise in the WTO.
Because the WTO drifts in frequency, the IF frequency of the test signal at the output of IF1 will also drift.The Wadley Loop compensates for this drift using a mixer M3 and IF filter IF2 to generate and select the difference frequency between IF1's output and a similarly drifting reference signal (R1).
The frequency of the reference signal, fR1, equals the frequency difference between the WTO and one comb line of a CG, obtained as follows.The CG is locked to a crystal oscillator that provides a precise frequency f XTAL ; thus the CG provides a band of frequencies n.f XTAL , with n = 1, 2, 3… N, where N represents the number of comb lines.Mixer M2 produces an inverted comb of frequencies, to give: The broadening of these lines in the lower part of Fig. 1 is used to indicate that there are frequency errors due to mixing with the high-linewidth WTO.A narrow band filter, BPF, selects one of these frequencies to become the reference frequency, where n sel stands for the index of the desired comb line: The BPF's center frequency, f BPF , is designed so that R1's frequency is an integer number of comb lines above the center frequency of IF1.The bandwidth of the BPF must be sufficient to accommodate the frequency fluctuations and mistuning of the WTO, but narrow enough to reject adjacent comb lines.As the crystal oscillator is relatively stable, the phase error of the reference signal comes mostly from the WTO, and so can cancel the phase error in IF1's output.
Mixer M3 produces the sum and difference frequencies between the IF1-filtered output M1 and the reference signal R1.The difference frequency is selected by filter IF2, to give the output frequency: Because n sel is an integer and assuming that f XTAL is precise, then the phase noise and frequency error introduced by the down conversion are minimal.It is also obvious that the frequency of the WTO has no impact on the output frequency, apart from it selecting the value of n to be n sel .Because n is an integer, the frequency conversion occurs in integer steps.
In reality, because of the smooth transitions of the band-pass filters' responses, setting the WTO away from the center frequency of each band will degrade the output of the system in terms of gain and crosstalk; however, small tuning offsets of the WTO will not matter.The beauty of this design is that only one crystal is required to provide the necessary frequencies for all of the desired bands in the input spectral range.Also, the input spectral range can cover many octaves without problems of images passing through the system.Moreover, all of the IF filters are fixed in frequency, so can be optimized for their spectral shape, which ultimately affects the fidelity of the received signal and the rejection of neighboring channels.The only tuning required is that of the widely tunable oscillator, which is a relatively simple circuit, particularly as its frequency stability is unimportant, and also because it only needs to operate over less than an octave range.

Electro-optic implementation
Figure 2 shows our proposal for implementing Wadley Loop as a set of electronic and photonic components.The optical input signal is mixed with the output of a widely tunable laser, serving as the WTO.The mixing M1 is achieved using a coherent receiver (C1) drawn here as a balanced-photodiode single-polarization homodyne receiver, with inphase (I) and quadrature (Q) outputs.The use of I and Q outputs allows the filters to operate at basebandthat is, the IF and bandpass filters in the radio system are replaced by pairs of low-pass filters operating on the I, Q signal pairs.An initial pair of low-pass filters (IF1) selects a band of frequencies from C1.The WTO also feeds a second coherent receiver (C2), along with the output of an optical comb source (mode-locked laser, MLL), and a second pair of low-pass filters (LPF) selects an appropriate frequency.The MLL serves the purpose of the CG.The outputs of IF1 and LPF are mixed together with an image-reject mixer (M3) which could be implemented in DSP.The output(s) of the mixer can be low-pass filtered (IF2).With the proposed design, for a coherent optical communication system, the frequency offset and phase noise caused by the low-quality WTO is eliminated (see Eq. ( 4).If the transmit laser carrier frequency roughly aligns to one of the MLL comb lines, the frequency offset of the signals after Mixer M3 is then defined only by the mismatch between the transmit laser carrier and the chosen comb line.This small residual offset can be compensated efficiently using standard frequency offset algorithms.It is also clear that the remaining phase noise of the output signals only depends on the transmit laser.Therefore, only a frequencystable low-linewidth transmit laser can lead to a system with negligible frequency offset and very small phase noise, which largely simplifies the digital signal processing and improves receiver sensitivity.The proposed system provides a similar phase stable reference to optically injection locking (OIL) a high-linewidth laser with one comb line of the MLL [16]; however, the Wadley Loop configuration is not limited by the locking bandwidth of the OIL laser, and may be able to compensate for larger frequency offsets.This limit to the allowable signal/comb offset is defined by the passband of the LPF used after coherent receiver (Coh.Rx.) C2 (Fig. 2), which can be several GHz.

Experimental demonstration
We experimentally demonstrated our optical Wadley Loop using off the shelf components.The experimental setup is shown in Fig. 3. To generate a test signal, an external cavity laser (ECL) with ~100 kHz linewidth (Alnair Labs TLG-300-C4) was modulated by an optical I/Q modulator, with its electrical drive signals generated from a 65 GSample/s arbitrary waveform generator (AWG).The optical test signal was designed to occupy a 10-GHz bandwidth, with QPSK modulation and root-raised-cosine pulse shaping with a 0.01 roll-off factor, which is also shown in Inset (a) of Fig. 3.The transmitter laser's frequency was set to 193.1 THz, which is the frequency of one tone of the MLL comb source at the receiver.
At the receiver, a JDSU Erbium-doped Glass Laser Oscillator (ERGO) passively modelocked laser (MLL) was used as a comb source, with the comb spacing locked to a 10.001 GHz sinewave from an RF synthesizer.As shown as Inset (b) of Fig. 3, only three tones of the MLL comb were selected by a Finisar WaveShaper for amplification, because our polarization-maintaining EDFAs did not have sufficient power to amplify more tones to a sufficient level for the coherent receiver.A second polarization-maintaining EDFA amplified the received test signal.The signal and comb were amplified to have the same average power.The received signal and comb powers were 0 dBm.A dual-polarization coherent receiver (U 2 T CPRV1220A) served as both coherent receivers (C1, C2) in Fig. 2.This was achieved by combining the amplified comb lines and amplified received test signal into the signal port of the receiver, using a polarization beam combiner (PBC).The local oscillator port of the receiver was fed from the WTO, which was a Fujitsu FLD5F6CX-E35 distributed feedback laser with a linewidth set to be 19.8MHz by adjusting the drive current, as measured with a high resolution spectrum analyzer (Inset (c)).The 'WTO' laser module's wavelength can be tuned using the internal Peltier cooler within the laser to change the laser-chip temperature.Inside the coherent receiver, the WTO is split into two paths and then combined with the two polarizations that are input to the signal port.This means that the X-polarization output of the receiver contains the test signal from the transmitter mixed with the WTO, the Y-polarization output contains the reference comb from the MLL mixed with the WTO.
We directly digitized the two in-phase and two quadrature outputs of the receiver (the X and Y polarizations) using an Agilent 40 GSa/s real-time oscilloscope and implemented the 3 pairs of filters and final mixer of the Wadley Loop concept with offline MATLAB processing.The final mixer is an image-reject mixer (a 'triple balanced mixer') using Inphase and Quadrature signals at both ports.

Electrical signal spectra at the output of the coherent receiver
Figure 4(a) is the spectrum of the X-polarization output of the coherent receiver.Figure 4(b) is the spectrum of the Y-polarization output of the coherent receiver (the WTO mixed with the MLL comb lines).The X-polarization output shows the result of mixing the transmitted test signal with the WTO.This is offset from zero frequency due to the drift of the WTO (1.25 GHz in Fig. 4(a)), which is to be expected in a laser without active frequency locking.Ideally, the transmitted spectrum would align with the matched filter (10 GHz 0.01 roll-off root-raised-cosine (RRC) filter) response shown as the blue dash curve.The Y-polarization has 3 lines corresponding to the 3 lines of the MLL comb that have been down-converted to baseband using the WTO.The spectral width of these lines is dominated by the spectral width of the WTO.One line will be selected using a digital low-pass filter (LPF) with a passband indicated by red dashed curve, which is 9-GHz wide in this case.The desired comb line can be selected if the actual offset between the transmit laser and the WTO is within ± half of the passband of the LPF, i.e., ± 4.5 GHz in this case.Note that the image band rejection filter (IF1 in Fig. 2) was not employed in the experiment, because we use a large oversampling rate (4 Sa/symbol) so that the image falls far enough away from the signal to be rejected by the matched filter at the receiver.

Frequency offset correction and phase noise reduction with Wadley Loop processing
Figure 5(a) shows the spectrum after Mixer M3.This shows that the frequency offsets at the outputs of the coherent receiver have cancelled out: the received signal now aligns with the matched filter.To show the effect of the Wadley Loop on the phase noise, the QPSK test signal was processed in the following manner.After matched filtering, the I and Q signals were resampled to 2 Sa/symbol.Adaptive equalization, based on constant modulus algorithm (CMA) and Viterbi-Viterbi phase estimation, was then used to recover the signals [17].
Figure 5(b) shows the phase evolution with and without Wadley Loop processing over 20,000 symbols.Each symbol is 100-ps long.Clearly, the Wadley Loop eliminates most of the phase error due to the WTO; the residual drift is due to the transmitter laser.If Wadley Loop processing is not used (that is, only the X-polarization output of the receiver is used), the DSP flow follows the approach in [18]: resampling, large tap frequency domain CMA to allow the adaptive filter converges to matched filter solution, then the algorithm proposed in [19] copes with the relatively large frequency offset for the DFB local oscillator, and finally Viterbi-Viterbi phase noise correction to recover the signals.Obviously without Wadley Loop processing, extra adaptive filter taps (81 taps vs. 21) and frequency offset estimation effort are required.Also, the signals still suffer a relatively high phase noise compared with when the Wadley Loop is employed, due to the broad linewidth of the WTO.Furthermore, in our experiment, when the Wadley Loop processing was applied, cycle slips were avoided, which may allow us to improve the total system performance by avoiding the penalty introduced by differential encoding [7].Again, these cycle slips are due to the broad laser linewidth of the WTO, which is compensated for by the Wadley Loop.
Figures 5(c) and (d) are the constellations of the received signals with and without the Wadley Loop.With Wadley Loop processing (Fig. 5(c)) there is about 1-dB improvement in signal quality, Q, compared with the pure DSP approach in [18] in back-to-back transmission.Although there is only a minor Q-factor improvement in this case, the digital processing power can be largely saved and the DSP design can be greatly simplified due to the elimination of most of the frequency offset and phase noise by the Wadley Loop.Furthermore, an additional performance enhancement is expected when multi-channel and high-order modulation formats are applied.

Resilience to frequency drift
As shown in Fig. 6(a), we forced the DFB laser frequency to drift from 193.11005 THz to 193.09064 THz by applying several step changes in current to its Peltier cooling element.During the subsequent drifts in frequency, we took 90 continuous measurements (each with 2 × 10 5 samples at 40 GSa/s) over approximately 54 seconds.Figure 6(b) depicts the estimated frequency offset with and without Wadley Loop processing.The Wadley Loop processing (red curve) eliminates most of the frequency offset if the original frequency offset (blue curve) is within ± 4.5 GHz.This stability allows the 9-GHz digital filter to select the desired comb line, and then allows the matched filtering to be implemented precisely and so simplifies the following digital signal processing.Figure 6(c) shows the measured Q-factor for the captured data sets, where Q-factor drops to zero indicating that none of the comb lines is within the 9-GHz digital filter passband (frequency offset is between ± 4.5 GHz to ± 5 GHz); therefore the signal failed to be detected after beating with the output of the digital LPF.When the frequency offset is greater than ± 5 GHz, another comb line is selected rather than the desired one, in this case phase noise is also mitigated effectively, but the frequency offset remains large, which requires turning the frequency offset compensation circuit on for signal recovery.If the LPF selects the non-desired comb line, the detection performance is slightly degraded.This degradation is because the non-desired comb line has a lower power, therefore a poorer signal to noise ratio, than the desired comb line (refer to Fig. 4(b)).

Conclusions
We have demonstrated that a new optical implementation of the radio-frequency Wadley Loop allows a low-performance widely tunable local-oscillator laser to be used to select a transmitted signal band without introducing additional phase noise or frequency drift.This is because a precise optical comb source is used to create a correction signal, by mixing with the widely tunable laser to generate an error signal, which can then be cancelled out.
An alternative architecture would be to use an optical filter to select one line from the comb source, to become a high-quality local oscillator at the receiver.The advantage of using a laser as the WTO and a Wadley Loop is that the laser can be made very small compared to commercial tunable optical filters, and could, therefore be integrated with the coherent receiver itself on a photonic chip.In both architectures one comb source can serve many receivers, and indeed transmitters; for example, one MLL could serve an entire telecommunications exchange.
The idea of using a Wadley Loop in optical systems could also be applied to optical wideband high-precision spectrum analysis.This could be achieved by selecting one portion of the spectrum at a time, using the WTO, and analyzing each portion using Fourier transform techniques.The results could then be stitched together.It is a tribute to Trevor Wadley that his ideas have such longevity.

Fig. 1 .
Fig. 1.Radio frequency implementation of the Wadley Loop (top) and a map of the frequencies along the system (bottom).The acronyms are explained in the text.

Fig. 2 .
Fig. 2. Electro-optic implementation of the Wadley Loop using coherent optical receivers.The acronyms are explained in the text.The blue shading indicates baseband electrical signals.

Fig. 3 .
Fig. 3. Experimental demonstration of the optical Wadley Loop.Inset (a) is the spectrum of the QPSK test signal.Inset (b) is the spectrum of the filtered MLL.Inset (c) is the spectrum of the WTO (blue) with a Lorentzian curve fit (red) indicating a linewidth of 19.8 MHz.All optical spectra were measured with an Agilent High-Resolution Spectrophotometer (HRS).

Fig. 5 .
Fig. 5. (a) Signal spectrum after Mixer M3; (b) comparison of phase noise evolution; (c) and (d) recovered signal constellations with and without Wadley Loop processing.