Open Access
Add to BibSonomyAbstract
We report the first experimental demonstration of a compact on-chip optical pulse timing discriminator that is able to provide an output voltage proportional to the relative timing of two 60-ps input pulses on separate paths. The output voltage is intrinsically low-pass-filtered, so the discriminator forms an interface between high-speed optics and low-speed electronics. Potential applications include timing synchronization of multiple pulse trains as a precursor for optical time-division multiplexing, and compact rangefinders with millimeter dimensions.
© 2016 Optical Society of America
1. Introduction
Detecting the relative timing of two signals is a very common process in electronics; for example, in radar systems, metrology, delay-locked loops, and data stream synchronization. In optics, pulse timing also is used in LIDAR and in measurement systems. It is also a precursor for optically-time division multiplexing optical data streams [1], as these streams may originate from different locations, yet need to have a precise relative timing if they are to be interleaved.
Awad et al. first demonstrated using a semiconductor optical amplifier (SOA) to detect the relative timing between counter-propagating picosecond pulses. The first pulse to arrive at the SOA will saturate its gain, so that the second pulse receives less gain (cross-gain modulation); an external balanced photodetector then measured the relative energies of the exiting pulses, which depended on the relative timing of the pulses due to the saturation and gain recovery times [2]. Awad et al. then replaced the SOA with an electro-absorption modulator (EAM) to lock a 10-Gpulse/s mode-locked laser to a 160-Gbit/s pulse stream [3]. Both schemes required fiber directional couplers to separate the incoming and out-going pulses from each SOA/EAM end facet. Gopalakrishnapillai et al. used a similar method as a time-of flight rangefinder [4].
We have previously proposed and demonstrated using simulations [5], and later analyzed [6,7], an optical pulse discriminator based on a semiconductor optical amplifier with three contacts along its length. The advantage is that no fiber directional couplers are required, so the device can be very compact. The pulses to be compared are input the two ends of the SOA. The center contact supplies the middle portion of the SOA with current that gives it gain. This gain can be depleted by pulses coming from both ends of the SOA. If the pulses are non-overlapping, because the gain is rapidly depleted by strong optical pulses, the first pulse to arrive at a facet of the SOA will receive the most gain, and the second pulse will receive less gain. However, the center-SOA’s gain slowly recovers towards a steady-state between the pulses due to current injection; thus, simply by measuring the relative strengths of the output pulses (assuming that they have similar input amplitudes), their order and timing can be found. Conversely, if the pulses overlap, the difference in the output amplitudes of the pulses decreases as the overlap is increased, becoming zero for perfect synchronization.
Our contribution [5], over Awad’s work, was to propose adding voltage-sensing contacts at the ends of the SOA. Because the voltage is logarithmically proportional to the carrier density under the contact, and this is depleted by the strong pulses that are about to exit the facets of the laser, the voltages on these contacts will indicate the relative strengths of the counter-propagating pulses in the SOA. Thus a 3-contact SOA can replace a single-contact SOA with external couplers and photodiodes, potentially making a very compact device for range sensing [8] and other applications. Due to the lack of open foundries with SOA capabilities at that time, we were unable to demonstrate the 3-contact SOA idea experimentally.
In this paper, we present a Photonic Integrated Circuit (PIC) incorporating a 3-contact SOA device, and show that it is able to discriminate the relative delays between counter-propagating pulses using the contact voltages of the end SOA sections. A response of 160 μV/ps was obtained in the overlapping-pulse mode. Our PIC also has two couplers and photodiodes within it, which also enables us to demonstrate Awad’s photodiode design but in integrated form. This is the first experimental demonstration of an integrated optical pulse discriminator, and could have many applications in both telecommunications and instrumentation. The reported integrated device is a critical building block for enabling accurate and stable pulse timing manipulation on photonic signal processor chips that are of high interest for areas such as telecommunications, LIDAR, and spectroscopy.
2. Device description
The chip, shown in Fig. 1, was fabricated via JePPIX foundry service using SMART Photonics Indium Phosphide (InP) platform. The circuit is based on shallow-etched InP waveguides with a propagation loss of 3.5 dB/cm and coupling loss of about 4 dB from the facet to a lensed fiber. It comprises a serial cascade of 3 sections of SOAs (125 μm, 1000 μm and 125 μm) with electrical isolations in between. The SOAs are constructed in InGaAsP/InP quantum well with Q = 1.25 and exhibit a gain of 70 cm−1 at 1550 nm for a current density of 6 to 7 kA/cm2. Two 50/50 1 × 2 couplers are connected to either side of the SOA region to tap out optical power for detection using two integrated photodiodes.
Fig. 1 Photonic integrated circuit: (a) conceptual layout; (b) computer-aided design drawing; (c) photograph after fabrication.
2. Experiments
Figure 2 illustrates arrangement of the SOA and the test waveform generators. A gain-switched semiconductor laser is used as the pulse source. The laser is a standard DFB telecommunications laser from a Siemens TransExpress 10-Gbit/s externally modulated system, which conveniently has a direct-modulation input. The laser is driven by 50-ps electrical pulses at a 1-GHz rate generated by a Tektronix 70000-series Arbitrary Waveform Generator (AWG), and amplified to 21 dBm by an SHF-807 amplifier to give a 140-mA p-p current swing, and biased with approximately 17 mA (lasing threshold = 13 mA). The measured optical pulse width is 60 ps, using a measurement system with a 26-ps rise-time [9]. The gain-switched laser had a jitter of several tens of picoseconds, but this is irrelevant, as the difference between two delayed versions of the same pulse is measured by the discriminator.
The output of the laser is amplified with an EDFA to 21 dBm, passed through an optical isolator, and polarization controlled, then split into two paths using a polarization-maintaining beam splitter; each path has an adjustable 0-350 ps optical delay. One optical delay has been converted to be driven by a crank and connecting rod, running at approx. 50 r.p.m., rather like a reciprocating engine, using Meccano parts. The crank also has an optical sensor to allow triggering of a digital oscilloscope. This arrangement gives a delay sweep range of 230 ps. The peak input powers to the SOA facets are 8 mW per facet.
The three SOA sections are biased with DC currents, IL, IC and IR (left, center, right) of 15 mA, 60 mA and 15 mA, respectively, derived from voltage sources via 100-Ω resistors. The voltage of the end SOA contacts is tapped off via 450-Ω resistors and then sent to two channels of the digitizing oscilloscope (50-Ω input impedance). This is effectively a 10:1 voltage divider. The DC voltages on the end SOAs’ contacts are ≈1.079 V at the given bias currents. The SOAs have a low differential resistance of about 15 Ω, because they are forward-biased diodes. The difference between the SOAs’ contact voltages is calculated using the oscilloscope mathematical functions, and then passed through a 100-Hz low-pass filter, also within the oscilloscope. The first pulse to enter the SOA regions will saturate their gains, but these regions still amplify the pulse significantly. Thus the SOA section at the opposite end to which the first pulse was input will be heavily saturated by the strongly amplified pulse, reducing its contact voltages in proportion to the logarithm of the carrier density divided by the intrinsic carrier density. A full analysis is presented in references [6] and [7].
The outputs of the two photodiodes are not externally reverse biased and are not followed by transimpedance amplifiers; they are in photovoltaic mode (4th quadrant of the LI characteristic), so the voltage change (superimposed upon around 600 mV DC) is limited by the photodiodes themselves.
2.1. Calibration of the applied delay
The sweep in delay is not linear because a short connecting arm is used to drive the piston-like retroreflector in the optical delay. The z-position of a piston driven by a crank of radius r and a connecting arm of length l is:
For a delay implemented using a retro-reflector in an air path will provide a delay of 2z/c seconds, where c is the speed of light.Close to the center of the sweep (θ = 90°), the rate of change of the delay is approximately:
where Tcrank is the period of the crank motion.3. Experimental results
3.1 Outputs from SOA contact voltages
Figure 3 shows the oscilloscope screen grab when the SOA contacts were used for detection. The steep up/down changes in voltage of the inputs correspond to when the two delays are almost equal, giving overlapping pulses. One change corresponds to the relative delay increasing (the retroreflector mirror moving backwards through the midpoint), and the other to the relative delay decreasing (mirror moving forwards). There is approximately a 4-mV differential change in contact voltage while the pulses overlap (after the 10:1 attenuation). Note that the outputs from the individual contacts have a lot of high-frequency electrical interference. This was reduced by low-pass filtering the difference between them to around 100-Hz using the oscilloscope functions, but could be improved using a better electrical layout with preamplifiers close to the actual SOA.
Fig. 3 Electrical waveforms from the SOA contacts (yellow, second from bottom: green, bottom), and difference waveform (cyan, top). The blue waveform (second top) is the trigger.
Figure 4 shows the differential SOA-contact voltage (after attenuation) plotted against the actual delay as determined from the position of the mirror (Eq. (1). The multiple crossings are due to the forwards and backwards transitions of the mirror. Low-frequency noise has affected the traces. This could be reduced by better design of the electronics. The sensitivity to relative delay is approximately 160 μV/ps (noting the 10:1 attenuation) when the pulses overlap, and the differential range is approximately 40 ps.
Fig. 4 Re-scaled waveforms showing the differential output of the SOAs versus the actual delay time determined using Eq. (1).
3.2 Outputs from integrated photodiodes with zero bias
Figure 5 shows the oscilloscope screen grab when the integrated photodiodes were used. The waveforms are far less noisy in this case. Again, the steep changes in voltage of the inputs correspond to when the two delays are almost equal and the pulses substantially overlap.
Fig. 5 Electrical waveforms from the integrated photodiodes (yellow, top: green, bottom), and the calculated difference waveform (cyan, middle). Blue (second top) is the trigger.
Figure 6 shows the rescaled photodiode voltages. The response is 870 μV per picosecond. This is very similar in shape to the SOA measurements; however, there is far less noise than for the SOA measurements, which we suspect is due to the different electrical arrangement. For example, the SOAs are biased using power supplies that could introduce noise, whereas the photodiodes are directly connected to the oscilloscope. The SOAs also have voltmeters attached to their contacts, with reasonably long and unshielded leads that will also pick up electrical interference. Ideally, a preamplifier designed to for low-noise operation when fed from a low-impedance source would be designed and fabricated close to the SOA contacts. This would help to achieve a noise performance close to that predicted in our theoretical work [7].
Fig. 6 Re-scaled waveforms showing the differential output of the photodiodes versus the actual delay time determined using Eq. (1).
4. Discussions
There are several factors that affect the calibration of the device, as have been discussed in our theoretical papers. The output voltage and accuracy depends on the absolute input powers from both ends of the device; this effect could be calibrated out using photodiodes to monitor the input signal powers, and a microcontroller to post-process the results. If the pulse repetition rate is too high (around a GHz), then the gains will not fully recover between the pulses. The polarizations of the inputs may matter because of polarization dependence of the gain and hence gain saturation, though the PIC technology here is intended to be polarization independent. The photodetectors were used in photovoltaic mode, which is nonlinear and slow; a reverse bias and a transimpedance amplifier would improve nonlinearity and bandwidth. The end SOAs were also forward biased, giving them a very low impedance and a limited voltage change; this was necessary to give these sections gain, rather than making them absorptive to the input waves.
5. Conclusions
We have experimentally demonstrated that a compact 3-contact SOA is able to determine the relative timing of two 60-ps optical pulses, and provides a low-frequency electrical output proportional to the overlap. We have also shown that photodiodes on the same PIC can be used to monitor the relative output powers of the counter-propagating waves, to also provide an estimate of relative pulse delay.
Acknowledgments
This work was supported by Lowery’s Australian Research Council (ARC) Laureate Fellowship on Electro-Photonic Interchanges (FL130100041).
References and links
1. R. S. Tucker, G. Eisenstein, S. K. Korotky, U. Koren, G. Raybon, J. J. Veselka, L. L. Buhl, B. L. Kasper, and R. C. Alferness, “Optical time-division multiplexing and demultiplexing in a multigigabit/second fibre transmission system,” Electron. Lett. 23(5), 208–209 (1987). [CrossRef]
2. E. S. Awad, C. K. J. Richardson, P. S. Cho, N. Moulton, and J. Goldhar, “Optical clock recovery using SOA for relative timing extraction between counterpropagating short picosecond pulses,” IEEE Photonics Technol. Lett. 14(3), 396–398 (2002). [CrossRef]
3. E. S. Awad, P. S. Cho, N. Moulton, and J. Goldhar, “Subharmonic optical clock recovery from 160 Gb/s using time-dependent loss saturation inside a single electroabsorption modulator,” IEEE Photonics Technol. Lett. 15(12), 1764–1766 (2003). [CrossRef]
4. B. S. Gopalakrishnapillai, K. L. Lee, A. J. Lowery, M. H. Premaratne, S. Shinada, N. Wada, T. Miyazaki, A. Nirmalathas, and C. Lim, “Experimental demonstration of a simple time-of-flight rangefinder based on semiconductor optical amplifier,” in Optical Fiber Communications (OFC),(IEEE, Anaheim, CA, 2006), p. OTuL5.
5. A. Lowery and M. Premaratne, “Reduced component count optical delay discriminator using a semiconductor optical amplifier-detector,” Opt. Express 13(1), 290–295 (2005). [CrossRef] [PubMed]
6. M. H. Premaratne and A. J. Lowery, “Analytical characterizaton of SOA-based optical pulse delay discriminator,” J. Lightwave Technol. 23(9), 2778 (2005). [CrossRef]
7. M. Premaratne and A. J. Lowery, “Semiclassical analysis of the impact of noise in SOA-based optical pulse delay discriminator,” IEEE J. Sel. Top. Quantum Electron. 12(4), 708–716 (2006). [CrossRef]
8. A. Lowery and M. Premaratne, “Design and simulation of a simple laser rangefinder using a semiconductor optical amplifier-detector,” Opt. Express 13(10), 3647–3652 (2005). [CrossRef] [PubMed]
9. A. J. Lowery, “All-optical DAC using counter-propagating optical and electrical pulses in a Mach-Zehnder modulator,” Opt. Express 22(21), 26429–26437 (2014). [CrossRef] [PubMed]
