Mode-locked laser

The mode-locked laser was fully developed at CSEM. A photo of a laser from the same series is shown in Figure 3-1. The laser is in a ruggedized package. All optical and electrical interfaces are situated on the same face of the housing. The laser optical output is delivered by a PM fiber pigtail.

Figure 3-1:

CSEM mode-locked laser head

The laser parameters are summarized in the table below:

Table 3-1:

MLL output parameters

Erbium-doped fiber amplifier

To generate the f-3f interferometer in the silicon nitride waveguide sufficient peak power is needed. For this reason, it was decided to implement an EDFA for amplification. The EDFA’s baseline design is a so-called parabolic amplifier generating output pulses with parabolic temporal shape. Such temporal pulse shape is ideally suited for compression in a single mode fiber to obtain short pulses and thus high peak power necessary for the f₀ detection. The amplifier output is split into two arms. One arm goes to the silicon nitride waveguide for f_o detection and the second arm provides light for photodetection and RF generation.

The fiber length on the waveguide arm is adjusted to obtain pulses as short as possible. Pulse parameters at the waveguide input are summarized in the Table 3-2.

Table 3-2:

Parabolic EDFA output parameters

Considering the parameters displayed in Table 3-2, an optical peak power of more than 4 kW or pulse energy of more than 400 pJ is reached at the input of the PIC waveguide which is more than sufficient to generate the required supercontinuum spectral bandwidth.

Pulse interleaver

The part used for the optical RF generation is connected to a fiber interleaver described in the following.

The laser has a repetition rate close to 395 MHz. In order to get maximum RF power into the 24th repetition rate harmonic (9.5 GHz), the laser repetition rate is multiplied using a three-stage fiber based interleaver multiplying the repetition rate by a factor of 8 to 3.16 GHz (see Figure 3-2). The interleaver is made of PM fiber and, in order to maximize the power for photodetection, the last combiner stage (S3 in Figure 3-2) is a polarizing beam combiner leading to alternate and orthogonal polarization of the output pulses [6]. The 3.16-GHz optical pulse train then impinges onto the HPPD, which detects the fundamental of the 3.16 GHz repetition rate pulses, its harmonics, and the spurious harmonics of the 395-MHz repetition rate of the laser. The third harmonic of the 3.16 GHz is isolated by an RF bandpass filter. In addition and to generate the best phase noise purity of the 9.5 GHz signal, sub-ps optical pulses impinging the HPPD are important to keep the phase noise reduction thanks to shot noise correlation [7]. In order to achieve this, a patch cord of dispersion compensating fiber (DCF) allows to recompress the pulses at the output of the fiber interleaver (several ps pulse duration). A measured pulse duration of 143 fs was achieved at the output of the interleaver.

Figure 3-2:

Fiber interleaver multiplying the laser repetition rate by a factor of 8 to 3.16 GHz.

Silicon nitride waveguide

The design of the nonlinear waveguides is based on finite element numerical simulations that yield group velocity dispersion as well as nonlinear optical characteristics for specific waveguide geometries. While the height of the waveguide was fixed to 800 nm (corresponding to Ligentec’s AN800 fabrication process), the (top) width was varied from 200 nm to 3000 nm, a sidewall angle of 1.5° was assumed for all geometries. Based on the simulated waveguide geometries, the super-continuum generation process was simulated using split-step simulation of the nonlinear optical dynamics in the waveguide. As a result of the simulation, 44 different waveguides have been chosen for the fabrication runs in a 5x10mm² large chip. While the first fabrication run relied on standard fiber-to-chip coupling by means of lensed fibers in a straight waveguide, in a second fabrication run SMF-28 single mode fiber compatible couplers (ExSpot from Ligentec) with better alignment tolerance and power handling capability were implemented at the waveguides input and output.

A broad super continuum spanning over hundreds of nanometers is observed. It was difficult to judge from the output spectra whether an output spectrum satisfies the conditions for f₀ detection. For that reason, the output of the PIC waveguides was analyzed with a photodiode. Three waveguides allowed for detection of f₀.

High-power handling photodiode

The used HP-PD was developed by Albis Optoelectronics in the ESA ARTES project HOPP in collaboration with CSEM (see previously mentioned contribution to this conference). The HP-PD used in this work was optimized for 10 GHz microwave generation. The main optimization parameter is the active diameter while the epitaxial structure was not modified with respect to the HOPP project. Indeed, a large photodiode diameter results in better heat management of the photodiode which yields better power handling capabilities and increased optical damage threshold. Furthermore, a large diameter allows to spread the optical input power across a larger area which allows high optical power operation without saturation effects taking place. On the other hand, a larger device diameter also results in a larger device capacitance which reduces the device bandwidth. The bandwidth must of course not be reduced below the targeted carrier frequency. The design rule derived from this consideration is that the PD bandwidth should be a factor of 1.4-1.5 above the targeted RF frequency – in our case about 14 GHz. The device responsivity is around 0.5 A/W. The HP-PD is mounted in a hermetic package with an optical fiber feedthrough on one side and a coaxial K-connector on the other side. The electrical feedthrough (RF output) is terminated by an internal matching impedance of 50 Ω. The bias voltage is applied to the HP-PD via two pins on the side of the package through an integrated bias-T. The HP-PD is aligned to a single mode optical fiber through the optical feedthrough to get the maximum responsivity.

f_rep and f₀ stabilization

The part of the laser light used for the stabilization of f_rep was optically filtered by a narrow (1 nm) optical band-pass filter and combined with the light coming from an ultra-stable CW laser which was itself locked onto a high-finesse Fabry-Pérot cavity (Finesse 200’000 and 10 cm length with fused-silica mirror substrates and ULE compensation rings) with the Pound-Drever-Hall technique. The optical beat note generated by the mixing of the CW laser and one mode-locked laser comb line was detected by a photodiode and band-pass filtered. A synthesizer stabilized to an active hydrogen maser served as a local oscillator for the f_rep control loop. A homemade analog PLL electronics acting on a piezo-electric transducer mounted in the MLL cavity was used for the f_rep stabilization.

The RF spectrum of the beat note frequency has servo bumps at 100 kHz away from the coherent peak. This high locking bandwidth for a piezo actuator is possible thanks to the careful mechanical design of the CSEM MLL. The phase noise L(f) measurements of the f_rep lock alone and when the f₀ lock is engaged as well are shown in the left panel of Figure 5-1. The servo bump at around 100 kHz is present as well as other spurious noise features at f < 100 kHz. The integrated phase noise of the beat note frequency integrated between 1 Hz and 1 MHz amounts to 224 mrad in case of f_rep stabilization only and 220 mrad when f₀ is stabilized at the same time.

Figure 5-1:

Left: in loop phase noise of f_rep signal; Right: in loop phase noise of f₀ signal.

The part of the laser light used for the stabilization of f₀ was coupled to a Ligentec SiN waveguide. The output of the waveguide was coupled to a fibre and the fibre output was focused onto a photodetector. The generated RF signal was band-pass filtered and a very similar PLL as compared to the f_rep lock was implemented. It acts on the current of the MLL pump laser.

The f₀ RF spectrum exhibits servo bumps at 30 kHz from the coherent peak. The RF spectrum is clean and does not exhibit spurs. The phase noise double-sideband spectral density L(f) of the f₀ lock alone and simultaneously with the f_rep lock is shown in Figure 5-1, on the right panel. The servo bump at around 30 kHz is present as well and the spectrum shows only a few weak spurs. As the loop gain for the f₀ lock alone is slightly higher the phase noise is here about 10 dBc/Hz lower at intermediate frequencies than in the case for the two combined f₀ and f_rep locks. The servo bump of the f_rep lock at around 100 kHz can also be observed in the case where the two locking loops are active. f₀ integrated phase noise amounts to 912 mrad in case of f₀ stabilization only and 960 mrad when f_rep is stabilized at the same time.

High frequency phase noise measurement

In a first experiment, the HP-PD was illuminated with 15 mW of optical power. The photodiode bias voltage was varied from 8 to 15 volts and the phase noise was measured from 1 kHz to 30 MHz offset frequency (see Figure 5-2 below). The measurement is depicted in Figure 5-2. At low offset frequencies the phase noise is almost the same for all bias voltages. At offset frequencies above 10 kHz it can be observed that the higher the bias voltage the lower the phase noise. For the lowest bias voltage, the phase noise floor amounts to -169 dBc/Hz whereas at a bias voltage of 15 V noise floor is significantly improved to -175 dBc/Hz.

Figure 5-2:

Phase noise measurements versus different photodiode bias voltages

Phase noise for different locking configurations

In a last set of measurements, the influence of the f₀ and _frep stabilization on the microwave signal phase noise was tested. Those measurements were all carried out at a bias voltage of 15 V and a photocurrent of 8.2 mA. The RF power in the 9.5 GHz tone was about -0.5 dBm. The measurements are shown in Figure 5-3.

Figure 5-3:

Phase noise of 9.491 GHz tone with and without all combinations of f₀ and f_rep stabilization / P_RF≈-0.5dBm

In the case of a free running laser, the phase noise is on the order of 0 dBc/Hz at 1 Hz offset frequency and about -140 dBc/Hz at 10 kHz offset frequency. It continues to decay up to about 1 MHz where the phase noise floor of about -175 dBc/Hz is reached. When f₀ stabilization is activated the phase noise curve is essentially the same as with the free running laser. When the optical f_rep lock is activated without f₀ stabilization, the phase noise significantly drops up to a frequency of about 100 kHz which corresponds to the lock bandwidth of the f_rep lock. This indicates that low offset frequency phase noise of the free running laser is dominated, as expected, by fluctuations in f_rep. The servo bump of the f_rep lock (at 100kHz) can be noticed as well. In the low offset frequency range (especially below 10 kHz) the phase noise limit is attributed to acoustic noise and vibrations causing laser cavity length fluctuations. At offset frequencies above 100 kHz the phase noise is the same as for the case without repetition rate stabilization. When the f₀ stabilization is activated together with the f_rep stabilization, the phase noise further decreases by up to 30 dB within the lock bandwidth of the f₀ lock of 30 kHz. Both the servo bump of the f_rep lock and the f₀ lock can be observed on the phase noise curve. The phase noise floor remains unaffected by the servos.

These phase noise measurements were performed with a Rohde-Schwartz FSWP using cross-correlation. While this instrument provides state-of-the-art capabilities in a commercial instrument, it is not capable to measure the ultra-low phase noise of the photonics oscillator up to frequencies of about 10 kHz since a prohibitively high number of cross-correlations would be needed. Other setups are required to measure such low phase noise which is expected to be in the range of -100 dBc/Hz at 1 Hz offset corresponding to the 1 s relative instability of the reference cavity of about 10^-15 [8].