1. Introduction

High-speed electro-optic modulators, that convert electrical into optical signals with low energy consumption, are key elements in optical transceivers of today’s information network. In the next decades, such optical transceivers are required to support bit rates of up to 10 Tb/s [1]. One of the suggested solutions for realization of such interfaces is scaling through spectral and spatial parallelism [1]. This requires array integration of transmitters and receivers on smallest footprint, preferably on the micrometer scale to meet the lane pitch spacing of electronic transceiver circuitry [2–4]. Additionally, to avoid inverse multiplexing at the electro-optic interface the bandwidth of modulators needs to be improved significantly, far beyond 100 GHz [5]. Furthermore, power hungry amplifiers for driving electronics should be avoided to realize a dense integration and feasible power management [6,7]. Power-efficient operation can be achieved through optimized synergies between driving electronics and electro-optic modulators. Consequently, the separate parts of an optical transceiver should be optimized by concurrent co-designing. Such optical interfaces may find applications in optical interconnects with simple intensity modulation (IM) and direction detection (DD) and transmission reaching from a few meters up to a few kilometers [8].

Several modulator concepts have already demonstrated optical interconnects with on-off keying (OOK) modulation beyond 100 Gb/s. One of the concepts is the Mach-Zehnder modulator (MZM) which has been realized on different technology platforms like Lithium Niobate [9,10], silicon-organic hybrid (SOH) [11], InP [12,13], and polymer photonic [14]. Another concept is the electro absorption modulator (EAM) which has been demonstrated either on the InP [15–17], or on the SiGe [18] technology platform.

More recently, plasmonic-organic hybrid (POH) modulators [19] have emerged. They allow for MZM [20,21] enabled amplitude modulation with large extinction ratios of more than 25 dB [22], small foot prints on the micrometer scale [20,23], an electro-optic bandwidth beyond 170 GHz [24] and a low drive voltage of less than 1 V_pp [25]. The small foot print of the POH modulators make this technology also essential for dense array integration of electro-optic components [3,26].

Still, the POH technology has lacked so far a demonstration of its full potential since it relied on electronics designed for conventional modulators.

In this paper, we report on a plasmonic transmitter operated with a mean peak-to-peak drive voltage of 178 mV_pp at the maximum eye opening at a bit rate of 120 Gb/s NRZ-OOK in an IM/DD optical interconnect. We introduce a new plasmonic Mach-Zehnder modulator with a dual-electrode design, which enables a differential voltage drop at the phase shifters and allows therefore for reduced driving voltages. The modulator’s active area has a length of 15 μm, which is below the RF wavelength and relaxes the design requirements for the RF electrodes. We measured an electro-optic bandwidth of up to 108 GHz, which shows no indication of a frequency dependent magnitude roll-off. The large electro-optic bandwidth allows for C-band transmission of 120 Gb/s NRZ-OOK signals through up to 100 m single mode fiber without any equalization. Data transmission of up to 500 m requires only a T-spaced feed forward equalizer with 7 taps to achieve a BER performance below the KP4-FEC limit.

This paper is based on initial results presented at the conference OFC’2019 [27].

2. Plasmonic dual-drive transmitter concept

The plasmonic dual-drive transmitter comprises a power multiplexer (PMUX) for generation of differential electrical signals and a dual electrode plasmonic-organic hybrid Mach-Zehnder modulator (POH MZM) for electro-optic conversion. Both parts of the transmitter are co-designed for optimal signal generation. This transmitter concept is shown in Fig. 1 and offers three main advantages.

 

Fig. 1 Dual-drive plasmonic transmitter. A power multiplexer (PMUX) generates the electrical differential signal (U_Drive, U̅_Drive) which is connected via S̅SS̅ electrodes with the plasmonic-organic hybrid (POH) Mach-Zehnder modulator (MZM). Colorized scanning electron microscope (SEM) pictures show the two plasmonic phase shifters, which are incorporated in the MZM structure operated in push-pull mode. Light is coupled via grating couplers (GC) to the chip and silicon photonic (SiP) waveguides direct the light to the plasmonic phase shifters. An additional thermo-optic phase shifter (PS) allows for adaption of the MZM’s operation point. (1) Cross section of the plasmonic MZM, which is covered by the organic electro-optic (OEO) material.

Download Full Size | PDF

Firstly, balanced S̅SS̅ operation offers twice the voltage swing for a negligible increase of complexity compared to standard unbalanced GSG operation. Here, the potential difference between signal S and the inverted signal S̅ is applied at each phase shifter for push-pull operation, respectively (see the cross section in the inset of Fig. 1). The S̅ signal path is split on the chip, where the longest electrical path dimensions are around 100 µm. Consequently, the on-chip electrode design can be considered as lumped since dimensions are smaller by factor of more than 10 compared to the RF wavelength. Secondly, an additional factor of two is exploited, since the plasmonic MZM can be assumed as a lumped capacitance, which does not require an ohmic far-end termination. Lastly, the PMUX is a driver concept in which the final MUX selector stage directly drives the load. The PMUX concept offers lower power consumption and better signal quality than a MUX with an external amplifier [28]. The POH intensity modulator [22] comprises two compact plasmonic phase shifters of 15 µm length and a slot width of 100 nm. The plasmonic phase shifters are incorporated into a balanced interferometric Mach-Zehnder configuration in push-pull mode, which allows for amplitude, respectively, intensity modulation. The phase relation of the two arms and hereby the MZM’s operating point is tunable by a thermo-optic phase shifter. The MZM offers a voltage-length product of about 110 Vµm, exploiting the high nonlinearity of the organic electro-optic material HD-BB-OH/YLD124 [29]. The voltage-length product and the length of the device translate into a V_π of the MZM in push-pull configuration of around 7.3 V. As discussed before, the proposed balanced driving scheme (S̅SS̅) of the lumped device leads to a four times stronger voltage drop at the modulator compared to a conventional dual electrode GSGS̅G driving scheme with 50 Ω termination. On-chip losses of the plasmonic device are ~10 dB. Additional losses of ~8.5 dB are attributed to fiber-to-chip coupling losses [30].

3. Experimental setup

Figure 2 depicts the experimental setup to evaluate the dual-drive plasmonic transmitter in an optical interconnect scenario. The plasmonic transmitter is an intensity modulated (IM) transmitter and comprises the dual-drive POH MZM under test, a tunable laser source (TLS), and PMUX. The transmitter is operated at a wavelength of 1545 nm with an optical input power of 10 dBm and generates a 120 Gb/s NRZ-OOK signal. The PMUX [28] multiplexes eight independent 15 Gb/s NRZ signals from an arbitrary waveform generator (AWG) into a single 120 Gb/s NRZ signal. The eight 15 Gb/s NRZ signals are four pairs of differential random bit sequences of length 2¹⁶ whereby the differential pairs are additionally de-correlated by different coaxial cable lengths of 0.75 m. A radio frequency (RF) synthesizer provides a 60 GHz clock to the PMUX, which the PMUX divides by a factor four, and an off-chip frequency divider by another factor of two, before it is sent to the AWG as a reference clock source. The differential electrical signal is directly applied via two GSG microwave probes to the POH MZM without any additional electrical amplifiers. A digitally interpolated eye diagram of the single-ended signal is shown in the first inset of Fig. 2. It is measured with a real-time oscilloscope (63 GHz electrical bandwidth). The mean voltage swing is 450 mV_pp at the maximum eye opening. The drive voltage can be adjusted by the operation current of the PMUX’s output stage. The POH MZM is operated at the quadrature point, which is adjusted by the thermo-optic phase shifter to generate intensity modulated signals. The second inset in Fig. 2 shows the optical spectrum of the 120 Gb/s NRZ-OOK signal as normalized optical power versus wavelengths. The two clock tones are spaced apart by 240 GHz which are located at the first base band zero of 120 GHz of the 120 Gb/s NRZ signal’s spectral shape.

 

Fig. 2 Experimental Setup. The intensity modulated transmitter comprises an arbitrary waveform generator (AWG) connected to a power multiplexer (PMUX) with a differential output for electrical signal generation and a plasmonic-organic hybrid (POH) Mach-Zehnder modulator (MZM) fed by a tunable laser source (TLS). The optical signal is amplified by an erbium doped fiber amplifier (EDFA). The transmission channel comprises standard single mode fibers (SMFs) of up to 500 m in length. The direct detection receiver comprises a variable optical attenuator (VOA), a photodiode (PD), and a real-time oscilloscope (RTO). The offline digital signal processing (DSP) includes normalization, timing recovery, T-spaced feed forward equalization (FFE), and bit-to-error ratio (BER) calculation.

Download Full Size | PDF

The optical transmission line enables a back-to-back scenario as well as transmission over 100 m and 500 m standard single mode fiber (SMF). An erbium-doped fiber amplifier (EDFA) amplifies the modulated optical signal before the transmission line.

At the receiver, we detect the transmitted optical signal with direct detection (DD). The DD receiver includes a variable optical attenuator (VOA) to vary the received optical power, a 70 GHz photodiode (PD), and a 63 GHz real-time oscilloscope (RTO) with a sampling rate of 160 GSa/s. In the case of a symbol rate of 120 GBd, a 0-1 sequence has a fundamental frequency of 60 GHz, which is still covered by the RTO’s 63 GHz bandwidth. The digitized waveform comprises 5 million samples and is analyzed by offline digital signal processing (DSP). In the offline DSP we apply normalization, timing recovery [31], down-sampling to one sample per symbol, T-spaced feed forward equalization (FFE), symbol decision, and bit-to-error ratio (BER) calculation. The FFE has 51 filter taps and is trained in data aided operation with the first 10% of the received symbols and then applied in static mode to the whole signal.

4. Experimental results

4.1. Electro-optic bandwidth

In a first step, we evaluated the electro-optic bandwidth of the dual-drive POH MZM. The evaluation comprises two configurations. On the one side, the assessment with a broadband data signal for frequencies of up to 60 GHz and on the other side, the analysis with single harmonic tones between 10 GHz and 108 GHz.

In the first configuration, we used a broadband data signal (120 Gb/s NRZ) and analyzed the electro-optic-electric (EOE) bandwidth by offline DSP with help of the experimental setup depicted in Fig. 2. The frequency response was determined by a feed-forward equalizer (FFE) with 101 T/2-spaced filter taps which was trained data-aided by means of the least mean square (LMS) updating procedure. This allows for a frequency response analysis of up to 60 GHz. The measured frequency responses across the full link, i.e. the electrical circuits as well as the electro-optic modulator and the photodiode is depicted in Fig. 3(a) (square, black). The electrical back-to-back (Elect. BTB), i.e. the driver electronics, the electrical path, and the PMUX’s zero-order hold sinc roll-off, is shown by red circles in Fig. 3(a). The electrical back-to-back measurement corresponds to the 120 Gb/s data signal applied to the POH MZM, neglecting the frequency response of the microwave probes. The green and triangular markers in Fig. 3(a) present the EOE frequency response of the POH MZM which is retrieved by subtracting the frequency response of the electrical BTB measurement, the microwave probe, and the receiver photodiode from full link measurements. The frequency response of the photodiode has been measured in a separate calibration measurement and the frequency response of the microwave probe has been taken from its calibration file. It can be seen that the frequency response of the full link is not limited by the frequency response of the electro-optic modulator.

 

Fig. 3 (a) Experimental setup to measure the electro-optic (EO) frequency response of the modulator. Heterodyne beating of two lasers (f₁ and f₂) in a photo diode (PD) generates the electrical drive signal. (b) Normalized electro-optic-electric (EOE) and electro-optic (EO) frequency responses The EOE response of the full link comprises all electrical and optical components (black, square). The electrical back-to-back (Elect. BTB) frequency response comprises the electrical driver (red circles). The EOE frequency response of the plasmonic modulator is given by triangular green markers. The EOE frequency responses have been estimated by a feed-forward equalizer (FFE). The EO bandwidth has been measured with an optical spectrum analyzer (OSA) and the results are depicted as green crosses.

Download Full Size | PDF

In the second configuration, we used single harmonic tones for frequencies of up to 108 GHz to measure the electro-optic (EO) bandwidth of the POH MZM. Figure 3(a) depicts the second configuration for the frequency response measurement. The single frequency tones were generated by heterodyne mixing of two lasers (f₁ and f₂) in a photodiode (PD) which is directly attached to the electrical input of the POH MZM. The PD’s electrical output has been determined with help of a 110 GHz electrical spectrum analyzer and was subtracted after the measurement. The EO frequency response is evaluated by measuring the ratio between the optical carrier and the modulated signal with the help of an optical spectrum analyzer (OSA) for frequencies between 10 GHz and 108 GHz. The modulation efficiency is then normalized and plotted in Fig. 3(a) (cross, green). The oscillations are related to the reflections between the POH MZM, which is a lumped capacitance, and the electrical measurement setup. There is no indication of a frequency dependent roll-off up to 108 GHz. Consequently, the dual-drive design of the POH MZM shows no destructive cross talk between positive and negative signal path, which would result in a strong frequency dependent roll-off. Besides, it underscores the observed flat frequency response of the POH MZM from earlier measurements [24].

4.2. Data transmission and eye diagrams

Figure 4 presents the bit error ratio (BER) performance at different received optical power levels in a back-to-back scenario and for data transmission over 100 m and 500 m SMF at a wavelength of 1545 nm. As a reference, we consider two different input BER thresholds of hard decision forward error correction (FEC) codes. The first code is known as KP4-FEC that is a Reed-Solomon RS(544,514) FEC code with an input BER threshold of 2.4 × 10⁻⁴ [32]. The second code is known as HD-FEC. The HD-FEC is a FEC code with two interleaved Bose–Chaudhuri–Hocquenghem BCH(1020,988) codes with an input BER threshold of 3.8 × 10⁻³, an overhead of 7%,and a net coding gain of 9.19 dB. It is defined by the OUT4 standard G.975.1 [33,34]. A BER performance below the KP4-FEC limit is achieved for transmission distance of up to 500 m. To realize a BER below the HD-FEC limit, optical power levels at the receiver of 0.3 dBm, 0.5 dBm, and 2 dBm are necessary for BTB, 100 m, and 500 m data transmission.

 

Fig. 4 Experimental results for 120 Gb/s NRZ-OOK signals. Bit error ratio (BER) as a function of received optical power for back-to-back (BTB) (square, black), for 100 m transmission (circle, red), and for 500 m transmission (triangle, green). As a reference the FEC limits at 3.8 × 10⁻³ and 2.4 × 10⁻⁴ (dashed grey) are shown.

Download Full Size | PDF

The corresponding eye diagrams are summarized in Table 1. They are digitally interpolated with a frequency domain raised cosine pulse-shape with a roll-off factor of 1, neglecting the brick wall characteristic of the real-time oscilloscope. We show the best eye diagrams for BTB, 100 m, and 500 m transmission without and with a T-spaced FFE with 51 filter taps. With equalization, a BER as low as 5.7 × 10⁻⁷ was achieved for optical BTB. Without equalization, we achieve a performance below the HD-FEC limit of 3.8 × 10⁻³ for both optical BTB and 100 m transmission.

Table 1. Analysis of signal integrity for different DSP complexities and transmission distances with a mean peak-to-peak drive voltage of 450 mV_pp.

View Table

4.3. Equalization requirements

We further analyzed the required complexity of the receiver DSP. Therefore, we reduced the amount of filter taps of the T-spaced FFE. Figure 5(a) shows the results of the BER as a function of filter taps for BTB (square, black), 100 m (circle, red), and 500 m transmission (triangle, green). By increasing the transmission distance, the effect of chromatic dispersion induced power fading after direct detection is increasing. For the 500 m transmission, the power fading accounts for a 3 dB roll-off at 60 GHz. The equalizer needs also to compensate for this effect in addition to the frequency responses of the different components. To achieve a BER performance below the HD-FEC limit no equalizer at all is required for transmission of up to 100 m and a 3 taps equalizer for transmission over 500 m. A performance below the KP4-FEC limit can be realized with a 3 taps FFE for transmissions of up to 100 m and a 7 taps FFE for 500 m data transmission.

 

Fig. 5 (a) Complexity analysis of the feedforward equalizer (FFE) for 120 Gb/s NRZ-OOK signals. Bit-to-error ratio (BER) as a function of filter taps of a T-spaced FFE for back-to-back (BTB) (square, black), 100 m transmission (circle, red), and 500 m transmission (triangle, green). As a reference the FEC limits at 3.8 × 10⁻³ and 2.4 × 10⁻⁴ (dashed grey) are shown. (b) Bit error ratio (BER) as a function of the peak-to-peak drive voltage for optical back-to-back (BTB) measurements of 120 Gb/s NRZ-OOK signals. The given peak-to-peak drive voltage is measured single ended at a 50 Ω termination at the maximum eye opening. As reference the FEC limits at 3.8 × 10^{- 3} and 2.4 × 10⁻⁴ (dashed grey) are shown.

Download Full Size | PDF

4.4. Electrical drive voltage requirements

We investigated the requirements for the driver electronics by varying the amplitude of the drive voltage. Figure 5(b) shows the BER for different drive voltages in the back-to-back configuration. The drive voltages are given single-ended and are measured across a 50 Ω load. During the experiments the signal was applied to the microwave probe to contact the chip based POH MZM. The peak voltage swing is measured as mean value at the maximum eye opening as it can be seen in the first inset of Fig. 2. A peak-to-peak drive voltage of 330 mV_pp is sufficient to operate the plasmonic transmitter at 120 Gb/s NRZ-OOK with a BER performance below the KP4-FEC limit. With a peak-to-peak drive voltage of 178 mV_pp, we achieve a BER performance of 2.3 × 10^{- 3}, which is below the HD-FEC.

This shows that low electrical drive voltage operation is possible up to highest speed for a POH MZM if optimized electronics are used.

5. Outlook

One of the next steps is to remove 50 Ω transmission lines between the driver electronics and the plasmonic modulator by co-integration or by monolithically integration [28,35]. For such a co-design of the electronics and plasmonics, the modulator’s impedance and size characteristics are of particular interest. The required drive voltage and the device capacitance determine the energy consumption per bit, which has emerged as an EO modulator figure of merit. The charging or discharging of the capacitance occurs on average every second bit and the electrical energy consumption per bit can therefore be calculated by $W_{\text{bit}}=U_{\text{D}}^2{C}_{\text{MZM}}/4$ [36], where $U_D$ is the peak-to-peak voltage change at the device. As the purely capacitive POH MZM is operated in open-circuit, $U_D$ is twice the peak to peak voltage measured at a 50Ω termination. This means that $U_{\text{pp}}=2 U_{\text{pp},50\text{Ω}}$. Additionally, we have a differential voltage drop across the modulator which means the voltage swing $U_D$ extends from $U_{\text{pp}}$ to $-U_{\text{pp}}$ ($U_{\text{D}}=U_{\text{pp}}-(-U_{\text{pp}})=4\cdot U_{\text{pp},50\text{Ω}}$). Further, a MZM comprises two phase-modulators. If each has a capacitance of $C_{\text{PM}}$ then the capacitance of a MZM is $C_{\text{MZM}}=2\cdot C_{\text{PM}}$. The capacitance $C_{\text{PM}}$ in one arm of the MZM without RF contact pads is 3.4 fF which has been determined with frequency domain simulations in CST microwave studios [37]. This translates into a potential electrical drive power consumption per bit of 862 aJ/bit for BERs below the HD-FEC limit and 2.96 fJ/bit for BERs below the KP4-FEC limit with data rates of 120 Gb/s NRZ.

6. Conclusion

We demonstrated a plasmonic transmitter in an IM/DD optical interconnect link with a peak-to-peak drive voltage of 178 mV_pp operated at 120 Gb/s NRZ-OOK. This corresponds to a potential 862 aJ/bit operation. The low electrical drive power consumption demonstration was enabled by the plasmonic Mach-Zehnder modulator’s dual electrode design, which allowed for a differential voltage drop at each phase shifter. The RF electrode design showed no bandwidth limiting behavior and we measured a flat electro-optic bandwidth beyond 108 GHz with no indication of a magnitude roll-off. The 108 GHz bandwidth design of the plasmonic modulator allowed for 120 Gb/s NRZ-OOK data transmission in the C-band through single mode fiber without any equalization techniques. A transmission distance of 500 m requires only a T-spaced FFE with 3 taps to achieve a BER performance below the HD-FEC limit.