Dynamic power and chirp measurements of a quantum dash semiconductor optical amplifier amplified picosecond pulses using a linear pulse characterization technique

Quantum Dash (QDash) SOAs are of interest as an alternative to quantum dot SOAs, since they have some dot-like properties and are easier to fabricate such as to operate in the 1550 nm region, albeit with longer gain recovery times in the range of hundreds of picoseconds. QDash-SOAs are promising core component is optical subsystems for applications in next generation coherent communications and optical signal processing, which often involve pulse amplification. It is of interest to measure both the dynamic power, phase and associated chirp and thereby the spectrum of typical QDash-SOA amplified pulses. Non-linear measurement techniques are appropriate for pulsewidths less than 5 ps but have low sensitivity for wider pulsewidths often encountered in practice. This paper describes the use of a modified linear pulse characterization technique to measure the dynamic power and phase of the identical pulses of a QDash-SOA amplified 20 GHz repetition rate 19 ps pulsewidth pulse train. The pulse chirp and power from which the pulse chirp and pulse train power spectrum is calculated. The amplified pulse structure is strongly dependent on the amplifier gain and the input pulse shape and energy.


Introduction
Semiconductor Optical Amplifiers (SOAs) have many attractive properties, including small size, high gain and wide optical bandwidth that make them suitable as basic amplifiers in optical communication networks, including those based on advanced data modulation 36 Page 2 of 8 formats (Connelly 2017(Connelly , 2018Bonk et al. 2012). They also exhibit various fast non-linearities; useful in implementing a wide range of optical signal processing functions such as wavelength conversion, optical switching and optical logic (Connelly 2017). Compared to quantum well and bulk SOAs, Quantum dot (QD) and Quantum Dash (QDash) SOAs have faster dynamic responses, which make them more attractive for use in ultra-fast optical transmission systems (Akiyama et al. 2007, Connelly 2016. QDash-SOAs are of interest as an alternative to QD-SOAs, since they have some dot-like properties and can more easily be made to operate in the 1550 nm telecommunications wavelength range, although they have been shown to also have longer gain recovery times in the range of 100 ps (Connelly 2016; Zilkie et al. 2007). The principal causes of the distortion experienced by SOA amplified picosecond pulsewidth pulses are self-gain modulation and self-phase modulation (Agrawal 1989). The latter process, which is due to the SOA active region refractive index experiencing dynamic changes induced by the pulse power induced material gain saturation, affects the amplified pulse dynamic phase and hence its chirp and spectrum. In modeling and evaluating the performance of SOA signal amplification and optical signal processing applications, it is useful to have knowledge of the dynamic properties of the amplified pulses. Information on the pulse dynamic phase is especially important in systems using spectrally efficient coherent modulation techniques that employ phase and amplitude modulation of the carrier lightwave (Connelly 2018).
Two of most commonly used techniques for optical pulse power and phase measurement are frequency resolved optical gating and chronocyclic tomography (Trebino et al. 1997;Dorrer and Inuk 2003). However, although these techniques are often applicable to pulses having pulsewidths in the sub-picosecond to few picosecond range, they are not suitable for measuring pulses having powers and pulsewidths (10 s picosecond range) encountered in typical SOA applications (Connelly et al. 2019). In this paper we use a modified linear pulse characterization technique, capable of measuring relatively low power wide pulsewidth pulse trains, to determine the dynamic power and phase of a QDash-SOA amplified 19 ps pulsewidth, 20 GHz repetition rate pulse stream. The technique is described in detail in our previous work on bulk traveling-wave SOA and reflective SOA pulse amplification and on various types of pulse generators-Mach-Zehnder Modulator (MZM), Electroabsorption Modulator (EAM) and integrated SOA-EAM (Connelly et al. 2019(Connelly et al. , 2020Moreno et al. 2016). First the QDash-SOA structure, experimental setup and pulse measurement technique are described. Secondly, experimental results of the SOA input and output pulse temporal power and phase profiles are presented for three values of SOA current corresponding to low, moderate and high levels of gain saturation and thereby amplified pulse distortion. The corresponding pulse dynamic chirp and pulse train normalized power spectrums are also presented.

Quantum dash SOA
The 1 mm long SOA structure consists of six dash stack layers as shown in Fig. 1. The dashes are grown in a dash-in-a-barrier structure where the quantum dash emitting at 1.55 µm is buried in a quaternary barrier of λ g = 1.17 μm (Lelarge et al. 2007). The structure is processed using buried ridge stripe technology with a ridge width of 1.5 µm. The waveguides tilted at 7° include a tilted mode shape convertor and the end facets are anti-reflection coated. The packaged SOA shown in Fig. 1 includes a thermistor and thermoelectric cooler, which are used with a temperature controller to maintain the SOA temperature at 20 °C. The input and output coupling losses are approximately 2 dB. Typical Transverse Electric (TE) and Transverse Magnetic (TM) polarization resolved SOA Amplified Spontaneous Emission (ASE) spectra are shown in Fig. 1. The packaged SOA polarization dependent small-signal gain versus current and input power characteristics at 1530 nm are shown in Fig. 2. The TE mode amplification is significantly larger than that for the TM mode. For TE amplification, at a current of 500 mA, the saturation output power is 5 dBm. All of the pulse measurements in this paper are for the packaged SOA with TE polarized input pulses.

Experimental setup and pulse characterization technique
The pulse measurement experimental setup is shown in Fig. 3. A 1530 nm lightwave from a tunable external cavity laser is input to a MZM (OptiLab IM-1550-20 25 GHz bandwidth, low chirp intensity modulator) driven by a 20 GHz sinusoidal clock to generate a 20 GHz repetition rate pulse train. A polarization controller is used to align the input signal polarization to the SOA TE mode. A 1 nm bandwidth optical bandpass filter (not shown) is used at the SOA output to reduce the SOA ASE. The pulse power and phase profiles of the SOA input and output pulse trains are measured by the pulse characterization section, the operation of which and the required signal post-processing is described in detail in (Connelly 2019) as are details on the measurement sensitivity. Here, only a brief description is given. The pulse train spectrum is a line spectrum, whose Spectral Lines (SLs) are separated by integer multiples of the clock frequency (20 GHz) centered around the laser centre frequency (195.9 THz). The powers of each SL are first measured by directly connecting the pulse train to a 0.06 nm (7.7 GHz at 1530 nm) resolution bandwidth optical spectrum analyzer. MZMc (Avanex SD10 12.5 GHz bandwidth, chirp free intensity modulator) is biased at its minimum transmission point and driven by a 10 GHz (half the clock frequency) low-amplitude sinusoidal signal obtained by frequency dividing the pulse generator clock. This results in carrier-suppressed double-sideband modulation of the pulse train. The DC-18 GHz phase shifter (Fairview Microwave SMP1801) provides a controllable time delay over an equivalent phase range of 0 to 360° of MZMc's modulating signal with respect to the pulse train clock. The up-converted sideband of each SL of the input pulse train interferes with the down-converted sideband of the next SL. The resulting interference signal spectrum is a series of Interference Spectral Lines (ISLs) interlaced between the SLs of the original pulse spectrum. The pulse time-varying power p(t) and phase (t) profiles are reconstructed by post processing (in Matlab) a set of the output signal spectrums from MZMc corresponding to twelve values of the delay and the input pulse train SL powers. The pulse dynamic chirp is calculated as Δ (t) = 1∕(2 )d ∕dt.

Experimental results
The reconstructed pulse generator output pulse normalized power and chirp profiles are shown in Fig. 4a. A low power pedestal is present on the pulse power profile. Pedestals can have a significant influence on the amplified pulse shape. The relatively high power central part of the pulse has an almost symmetrical Gaussian shape with 10-90% rise and fall times of 13.5 and 13.9 ps respectively and a Full-Width at Half-Maximum (FWHM) pulsewidth of 19 ps. As expected for a low chirp MZM, the chirp across the center of the pulse is very low and also linear. Due to the different signs of the slope in the transfer function of the modulator, there is a transition in the phase between the high power central pulse region and the pedestal (Ji et al. 2013;Connelly et al. 2020). The sharp peaks in the chirp indicate fast changes in phase when going from the central region of the pulse to the pedestal as expected. The pulse power in the transition region is very low and as such the associated chirp does not significantly affect the pulse spectrum. The energy spectral density of an individual pulse is the square of the magnitude of its Fourier transform, calculated using the fast Fourier transform of the complex pulse envelope √ p(t) exp[j (t)] . Because the pulse train consists of a series of equally spaced identical pulses, the pulse train power spectrum has the same shape as the energy spectrum of an individual pulse; hence the normalized pulse energy and pulse train power spectrums are identical. The central part of the pulse is essentially a Gaussian pulse having negligible chirp, so it is expected that its energy spectrum will also be Gaussian. This is confirmed by the calculated pulse train normalized power spectrum shown in Fig. 4b. The FWHM spectral bandwidth is 21.4 GHz and the Time Bandwidth Product (TBP) is 0.41 (unchirped Gaussian pulses have a TBP of 0.44). The input pulse energy and peak power to the packaged SOA are 6.3 fJ and 0.32 mW respectively. The pulse power is such that the amplified pulses experience measurable distortion at three indicative values of the SOA current; 150, 300 and 450 mA at which the unsaturated amplifier gain is 12.9, 16.7 and 18.4 dB respectively. The amplified pulse dynamic normalized power and chirp are shown in Fig. 5a-c. The amplified pulse power profile becomes significantly more distorted at higher gains. As the saturation level increases the pedestal power tends to increase at the expense of the central part of the pulse as is the case for bulk traveling wave and reflective SOAs (Connelly 2020). The leading edge of the central part of the pulse is sharper than its trailing edge, which is also the case for both bulk travelling wave and reflective SOAs (Connelly 2020). This is mainly due to the QDash-SOA material gain recovery process, which after saturation by a high-power pulse consists of an ultrafast response followed by a slow response having typical characteristics times of a few picoseconds and 100 ps respectively (Ngo et al. 2011). The effects of these processes on the individual pulses of a pulse train depend on many factors in particular the pulse energy, shape, pulsewidth and repetition rate. The rise times of the central part of the pulse are 10.3, 8.1 and 5.7 ps and the fall times are 12.7, 19.9 and 19.2 ps in ascending order of the SOA current and thereby increasing saturation level. This is consistent with the rising edge of the pulse experiencing a larger gain than the trailing edge. The amplified pulse energies (in ascending order of SOA current) were 23.9, 57.8 and 80.7 fJ, hence the amplifier energy gains were 5.8, 9.6 and 11 dB. The corresponding pulse peak powers were 1.3, 2.9 and 3.8 mW and the FWHM pulsewidths were 20.3, 19.4, 22.2 ps. Although these FWHM values are similar to each other, such a simple metric is not particularly useful for highly asymmetric pulses.
The increasing dynamic gain saturation with increasing SOA current leads to more SPM and thereby more complex amplified pulse chirp profiles. In particular, the chirp in the central part of the pulse becomes more non-linear. The chirp profiles, especially in the high Fig. 4 Pulse generator output pulse a normalized power and chirp and b pulse train normalized power spectrum (the frequency scale is relative to the unmodulated laser frequency) power central part of the pulse, are similar in shape and magnitude to those measured for a traveling-wave bulk SOA with essentially the same input pulse stream (Connelly 2020). The pulse train normalized power spectrums, shown in Fig. 5d, are red-shifted relative to that of the input. They also become asymmetric and develop a high-frequency sidelobe, the magnitude of which increases with increasing SOA current. The TBPs of the output pulse energy spectrums (in ascending order of SOA current) are 0.61, 0.53 and 0.57. As the amplified pulses have significant asymmetry these similar TBFs are not necessarily a good measure of pulse quality, although they are significantly larger than that of the input pulse.

Conclusions
This work investigated the application of a linear pulse characterization technique to measure the dynamic power and chirp of a QDash-SOA amplified 20 GHz repetition rate, 19 ps pulsewidth Gaussian type pulse stream. The amplified pulse power profile becomes increasingly asymmetric as the amplifier gain increases. Although the input pulses are essentially chirp free, the amplified pulses acquire a significant non-linear chirp. The combination of the asymmetric pulse power profile and complex chirp profile leads to a pulse energy spectrum that is also asymmetric. The experimental results show that QDash-SOA amplified pulses can have a complex structure and as such gives insight into QDash-SOA pulse amplification applications in optical signal processing and communication systems.