Optical channel aggregation based on modulation format conversion by coherent spectral superposition with electro-optic modulators

Advanced modulation techniques such as quadrature amplitude modulation (QAM) have attracted much interest for high spectral efficiency transmission to satisfy the sustained growth in demand for communication bandwidth. A conventional method for generating an optical M-QAM signal is to use the output from two electronic digital-to-analog converters (DACs) to drive an in-phase and quadrature (IQ) modulator. To generate an M-QAM optical signal with a symbol rate of R GBd, each of the two DACs needs a resolution and sampling rate of at least $log2M$ and R GS/s, respectively.¹ Therefore, electronic DACs significantly define the capacity of present-day optical networks. There is an upcoming limit to the future scalability of the CMOS or BiCMOS technology platforms, which can obstruct withstanding the future capacity crunch.^2,3 Additionally, these electronic DACs and IQ-modulator driver amplifiers have a finite linearity range at high baud rates, leading to inconsistencies in the data constellation points and considerably increasing the complexity and power consumption of the network.

Instead, all-optical signal processing techniques enable high data rate operation with flexible and efficient bandwidth utilization at low power consumption.⁴ Specifically, aggregating lower bit rate optical signals to a higher bit rate can be performed through coherent vector addition. Hence, several low-quality DACs could be implemented to generate low spectral efficiency optical signals that can be later optically aggregated to a channel with high spectral efficiency.^5–8 This operation can be performed on a transmitter to generate spectrally efficient modulation formats. Several system architectures have been proposed for a single wavelength transmitter based on segmented or nested modulators.^9–13 In these schemes, DACs producing lower-order modulation formats are used to drive different segments or stages of phase-shifters, which are then coherently combined to create higher-order modulation formats. However, this linear signal processing scheme requires a complex electronic and photonic design. It cannot be used to aggregate different channels in a network node, and it has yet to be shown that different wavelength channels can be aggregated.

Another way is based on different nonlinear optical effects like four-wave mixing (FWM), cross-phase modulation (XPM), self-phase modulation (SPM), parametric amplification (PA), and cross-gain modulation (XGM) arising from the second or third-order susceptibility of highly nonlinear fiber (HNLF), periodically poled lithium niobate (PPLN) waveguides, or semiconductor optical amplifiers (SOAs).^{5–8,14–20} For these methods, the parent channels need to be coherent, which is ensured by the use of an optical frequency comb as the source. In Refs. 14, 15, and 20, the aggregation of on–off keying (OOK) and M-PSK (M-ary phase shift keying) signals to a higher bit rate quadrature phase shift keying (QPSK) signal based on XPM using HNLF has been shown. QAM-8 signals have been generated by aggregating QPSK and OOK signals using XPM and XGM in SOA²⁰ and from binary phase shift keying (BPSK) inputs using XPM and PA in HNLF.^17,18 Optical aggregation of four OOK to a single QAM-16 signal using a nonlinear optical loop mirror based on XPM and parametric amplification has been presented in Ref. 5. To demonstrate tunable optical aggregation of two QPSK channels into a QAM-16 channel based on nonlinear wave mixing, many different coherent comb sources like mode-locked lasers (MLL)⁸ and Kerr combs originating from microring resonators¹⁹ have been used. These methods also use nonlinear wave mixing for the coherent vector summation of the input channels. In a recent experiment, aggregation of sinc-shaped Nyquist channels has been reported using the second order nonlinearity of PPLN waveguides.¹⁹ 

Apart from its applicability as a high-performance optical transmitter, the aggregation of low spectral efficiency channels into a higher-order one can also be very promising for network nodes in future networks, where an integrated optical comb source in the central node supplies all network terminations with phase- and frequency-locked carriers. This has the advantage that optical comb sources enable a power reduction in the networks,²¹ and they might be a solution for the so-called capacity crunch problem.²² Optical aggregation also enables promising advances for elastic optical networking (EON)^14,15 and near future fifth-generation (5G) technology¹⁶ with the existing network architecture. It gives an easy option to switch between simple and easy-to-generate modulation formats, including OOK and BPSK in metro and local access networks (M/LAN), and higher modulation formats like QAM with improved transmission capacity in optical backbone networks.

All existing aggregation methods demonstrated thus far necessitate complex electronic-photonic designs or rely on various nonlinear effects, specialized waveguides, pump laser sources, and sufficient optical power to enable nonlinear interactions. These requirements result in increased system complexity, polarization dependence, power consumption, and additional noise, ultimately limiting operational reconfigurability and restricting the range of applications. Therefore, it would be highly advantageous to have a linear approach for channel aggregation that is less complex and can be implemented on an integrated photonic platform such as silicon photonics. Moreover, this method should be applicable to transmitters to reduce complexity and costs and also facilitate channel aggregation in network nodes for future access and aggregation networks utilizing integrated comb sources.

This paper presents a straightforward technique for achieving flexible all-optical channel aggregation without relying on optical nonlinearity, additional pump sources, or specialized optical or electronic components. The approach follows a completely linear process for vector summation of input channels with low modulation formats into a reduced number of channels with higher modulation formats. To facilitate this aggregation, the signal channels to be combined must be modulated on coherent carriers at different wavelengths, which can be achieved using an optical frequency comb generated by an integrated ring resonator, for example.¹⁹ Subsequently, a phase or intensity modulator is employed to modulate the input channels, creating sidebands in such a manner that they superimpose and allow for phase-coherent vector summation. By controlling the phase of the radio frequency (RF) input to the modulator, the phase difference between the superimposed sidebands can be adjusted. Following the aggregation of various channels, the freed-up wavelength channels can be utilized for aggregating other channels or transmitting additional ones, ensuring efficient use of available wavelengths. Experimental results are presented, demonstrating the generation of a 10 GBd QAM-16 signal by aggregating two 10 GBd QPSK signals. Two 10 GBd BPSK signals have been aggregated to one 10 GBd QPSK and one 10 GBd PAM-4 signal. Experimentally, the aggregation of sinc-shaped Nyquist channels of lower modulation formats into a higher modulation format Nyquist channel is also presented.

The introduction of this novel in-line all-optical aggregation concept offers a significant enhancement in operational capacity while significantly reducing complexity. Additionally, the utilization of standard components such as electro-optic modulators, readily available in the process design kits of commercial foundries, opens up possibilities for cost-effective on-chip implementation with high yield.

This article is structured as follows: In Sec. II, we provide a detailed explanation of the operating principle underlying this concept. Section III covers the experimental setup and presents the results obtained. Subsequently, in Sec. IV, we briefly discuss the concept and highlight the key findings from the experimental results. Finally, in Sec. V, we conclude this paper by summarizing the main contributions and discussing potential future directions.

The proposed all-optical aggregation of lower bit rate channels is based on optical vector summation, as shown in Fig. 1. In Fig. 1(a), the symbol vectors from BPSK₁ (blue) are summed up with the symbol vectors from BPSK₂ (red) to build the symbol vectors of the QPSK signal (green). The two vectors should have the same amplitude (α = 1, with α as the peak-to-peak amplitude ratio) but a phase difference of ϕ = π/2. If the two vectors have the same phase but a different amplitude (α = 0.5), a PAM-4 signal can be generated, as illustrated in Fig. 1(b). Similarly, Fig. 1(c) shows the generation of a QAM-16 signal by vector summation of two QPSK signals.

Optical aggregation using coherent vector addition^7,8,19 allows for the superposition of two signal spectra by appropriately adjusting the amplitude and phase difference of the carriers.²³  Figure 2(a) illustrates the operational principle of coherent spectral superposition using optical modulators in a transmitter. This approach offers the advantage of replacing the high-quality electronics required for generating higher-order modulation formats with lower-quality alternatives. Furthermore, this method enables the aggregation of multiple wavelength channels, not just signals modulated on a single carrier wavelength.^9–13 

The proposed coherent superposition technique does not rely on nonlinear optical effects; instead, it utilizes electro-optic modulation. Combining several channels with lower spectral efficiencies makes it possible to construct fewer channels with higher spectral efficiency. Although Fig. 2(a) illustrates the aggregation of two wavelength channels into one output channel for simplicity, it should be noted that this approach can be applied to aggregate multiple wavelength channels while preserving the spectral width. This requires a sequential aggregation operation. At first, four wavelength channels can be aggregated to create two channels with twice the spectral efficiency. In the second stage, these two channels can be further combined into an aggregated channel with four times the spectral efficiency of the four inputs. It is important to note that the system is not limited to input channels with the same spectral efficiency, and an odd number of channels can also be aggregated.

To begin the process, two specific input channels at frequencies ω_c1 and ω_c2 are selected using a wavelength-division multiplexing (WDM) filter. Next, one of the channels undergoes amplitude weighting using a suitable factor α, which is determined based on the desired higher modulation format. It’s worth noting that this weighting can be achieved by employing adjustable attenuation or amplification techniques. The two channels are then combined and fed into an electro-optic modulator. The modulator is driven by a sinusoidal radio frequency (RF) signal with a frequency of ω_m, which can be generated by either a radio frequency generator (RFG) or a voltage-controlled oscillator. Through electro-optic modulation, the lower sideband (LSB) originating from the channel around ω_c2 overlaps with the higher sideband (HSB) resulting from the channel around ω_c1. As depicted in Fig. 2(a), the frequency spacing between the parent channels can reach as high as 2ω_m.

A superposition of the spectral amplitudes is insufficient to achieve the required vector summation for the intended modulation format. The optical carrier phases of the superimposing spectra are also essential factors to be considered. For the presented concept, the optical carrier phases of the superposing spectra are manipulated by the phase of the input sinusoidal RF signal to the modulator. Following, we will present a mathematical analysis regarding the phase relationship of the two optical sidebands resulting from the electro-optic modulation process, determining the proposed concept’s feasibility.

A complex quadrature amplitude-modulated carrier can be decomposed into two amplitude-modulated signals: in-phase (I) and quadrature (Q). These two components are related by a constant phase offset of π/2 between the respective carrier wave (ω_c).

Similar to other state-of-the-art all-optical techniques,^{5–8,14–20} the proposed method for coherent vector addition relies on maintaining a constant optical phase relation between different comb lines. When the carrier frequencies of the different channels at the transmitter are generated by a frequency comb, the frequencies and phases become locked.^24,25 The utilization of integrated frequency combs brings numerous advantages to future optical transmission systems and significantly reduces electrical power consumption.²¹ In a microcomb-based transmission system, the comb lines remain correlated over long transmission distances.²⁴ Although the absolute phase of the lines may change over time, the phase relationship remains conserved. Hence, frequency- and phase-locking are inherent to the comb sources. Recent advancements in microcombs have revolutionized high-capacity coherent communication.^21,26–28 Given their multitude of advantages, microcombs are expected to be extensively used in next-generation optical communications. Besides enhancing the overall capacity while maintaining low-quality electronic infrastructure, the presented method can as well be implemented in a central network node in an optical frequency comb-based, cost-effective wavelength division multiplexed passive optical network (WDM-PON), where source-free optical network units (ONUs) utilize optical carriers coming from a comb source situated at the central station.^29,30 In this case, this method can aggregate the channels with low spectral efficiency into one or several channels with higher spectral efficiency.

It is important to address potential amplitude or power fluctuations in the channels being aggregated, as these can impact the performance of the aggregated channel by altering the symbol vectors. To mitigate this, a power monitoring and adjustment mechanism can be employed. A variable optical attenuator can be used to adjust the weighting factor associated with channel aggregation. Silicon photonics has seen significant advancements in integrated monolithic variable optical attenuators and photodiodes for power monitoring,^31,32 making them suitable for on-chip implementation of this technique.

Furthermore, relative delays between the channels being aggregated can introduce distortions in the transmitted information. However, delays corresponding to integer multiples of the symbol period do not affect the information. To compensate for stochastic delay variations, tunable delay lines in silicon photonics^33,34 can be integrated into the aggregation device to enable delay compensation. In cases where far-away ONUs result in a phase change between the optical carriers from different positions, a constant phase difference does not affect the aggregation since it can be compensated by the RF phase of the sinusoidal signal driving the modulator. However, stochastic phase changes introduce random angle variations in vector addition. To address this, an active optoelectronic phase stabilization mechanism can be implemented.³⁵ 

By considering and addressing these factors, the presented method can be effectively utilized in central network nodes and WDM-PON systems, further demonstrating its practical applicability.

As depicted in Fig. 2(a), the aggregated channel is positioned at a different wavelength compared to the input channels. This characteristic offers the opportunity to optimize bandwidth allocation and improve spectral efficiency. When employing the method at a central network node, multiple spectral bands can be simultaneously freed.

To illustrate this, let us consider the aggregation of four adjacent WDM channels into two higher spectral efficiency channels, with each pair of channels being aggregated in separate branches [please refer to Fig. 2(b)]. A WDM filter selectively extracts channels 1 and 3 in the first branch. These two channels are combined and aggregated to occupy the position of the second WDM channel, thereby freeing up channels 1 and 3. Similarly, the second parallel branch filters out channels 2 and 4, which are aggregated to occupy position 3, releasing channels 2 and 4. Consequently, if the outputs of the two branches are multiplexed using a power combiner, channel 2 carries the information of channels 1 and 3, while channel 3 carries the information of channels 2 and 4. Importantly, channels 1 and 4 remain available for use, allowing for adding new data channels or aggregating other channels. In a subsequent step, channels 2 and 3 can be further aggregated into a single channel, resulting in even higher spectral efficiency.

Following the aggregation process, it is necessary to implement a bandpass filter [BPF, as shown in Fig. 2(a)] to suppress any unwanted spectral components. Standard WDM filters can be utilized for this purpose, or integrated silicon photonics filters can also be employed. Notably, integrated silicon photonics filters have been demonstrated to achieve more than a 25 dB extinction ratio.^36–38 It is important to highlight that the bandwidth of the aggregated output channel remains the same as that of a single input channel. However, the increase in spectral efficiency leads to a higher number of bits per symbol. It should be noted that the aggregated signal is at a different carrier frequency compared to the input channels. For wavelength-transparent conversion, a second modulation followed by a filtering stage can be employed to shift the signal back to its original wavelength channel. Alternatively, other wavelength conversion methods can be utilized for this purpose if required.^39–41 Therefore, by incorporating a bandpass filter and employing appropriate wavelength conversion techniques, the aggregated signal can be effectively filtered and restored to its original wavelength channel, ensuring efficient utilization of the allocated bandwidth.

Another notable advantage of the method is that it only uses standard equipment and is compatible with integrated photonics platforms. The phase difference between the corresponding optical carriers associated with the superimposing spectra can be arbitrarily regulated by controlling the RF phase to get the required vector summation for the intended modulation format. Tunable RF phase shifter devices are widely available as integrated circuits and embedded modules. However, the proposed method aims to be fully integrated into a chip. Tunable RF phase shifters can be integrated using a quadrature vector summation technique.⁴² The input RF is first fed to a quadrature signal generator, which provides differential 0° and 90° phase-shifted signals. This can be achieved through frequency division⁴³ or 90° hybrid couplers.⁴⁴ Then, the two phase-shifted signals are fed to tunable vector combiners that generate the RF signals through weighted signal summation.^42,44

Finally, a high-capacity optical transport system utilizes polarization multiplexing to double the throughput without sacrificing bandwidth. Therefore, the aggregation module has to be polarization diverse in real-world applications. As an electro-optic modulator is often polarization-dependent, a polarization rotator device must be incorporated to align the input polarization of the channels to be aggregated. For a system-on-chip implementation of the presented method, integrated polarization controllers can be used.^45–48 

To verify the feasibility of the proposed aggregation method, proof-of-concept experiments were carried out. In this section, the experimental details are discussed first, and then the results are presented.

In the experimental setup depicted in Fig. 3, two ancillary carriers were generated from a primary carrier through electro-optic modulation. A laser diode (LD) emitting at a frequency of 193.4 THz served as the primary carrier. The output of the LD was modulated by an electro-optic Mach-Zehnder modulator (MZM-1, Optilab IM-1550-20-A) in carrier suppression mode. An 18 GHz RF input was applied to MZM-1, generating two sidebands spaced 36 GHz apart. The MZM-1 bias was adjusted to the minimum transmission point, achieving a 20 dB suppression of the primary carrier at 193.4 THz. This arrangement was necessary due to the unavailability of any optical comb source.

A programmable WaveShaper (WS, Finisar-1000s) was employed as a filter to independently adjust the power in the two ancillary carriers without altering their phases. Thereby, the WS allowed for different weights to be added for vector addition. Please note that in our proof-of-concept setup, the WaveShaper was used for experimental convenience only. For the actual system, only wavelength separation and amplitude adjustment are necessary. A WDM filter, such as an arrayed waveguide grating, can be used for wavelength selection, while a variable optical attenuator is sufficient for amplitude adjustment. These components can be integrated into photonic platforms for practical implementation.

The carriers were then modulated with the desired lower spectral efficiency modulation formats using an I-Q optical transmitter module (Tektronix OM5110). A bias control mechanism is inherent to this transmitter module for the chosen modulation format. The same data were modulated on both channels. To introduce relative symbol delays between the two channels and decorrelate them, a Teraxion dispersion-compensating fiber module (D) was used. The dispersion induced by the module could be tuned to a resolution of ±10 ps/nm. It is important to note that the dispersion module also changed the relative phase between the two channels. However, since the relative phase for aggregation was adjusted using the RF signal driving MZM-2 (Optilab IM-1550-20-A), the phase changes introduced by the dispersion module did not affect the experiment.

The aggregation of the two channels, or their coherent superposition, was carried out by MZM-2, which is driven with the same 18 GHz RF as MZM-1. This was performed for experimental convenience. For the method, locking between the phases of the optical sidebands and the RF source is unnecessary. As we have studied in extensive system simulations, the optical source can be a completely independent comb source. The electrical RF phase compensates for any phase difference. As mentioned earlier, for stochastic variation in phase, a simple optoelectronic compensation can be used.³⁵ 

The upper sideband from the lower carrier was superimposed on the lower sideband from the carrier with a higher frequency at the laser output position of 193.4 THz. The spectrum, as measured by an optical spectrum analyzer (OSA), is shown in Fig. 4. The two superimposed spectra correspond to different sidebands and,thus, they exhibit a relative phase relationship. A tunable RF phase shifter (PS) was used to achieve the required phase tuning between the overlapping spectra. Since the RF phase alters the phase of the optical waves in opposite directions, as shown in Sec. II, any phase relationship between the optical channels can be adjusted to generate the required vector sum. It should be noted that erbium-doped fiber amplifiers (EDFAs) along with a 3 nm bandpass filter (BPF) were employed to amplify the signals and suppress out-of-band amplified spontaneous emission noise.

In the experimental setup, a coherent optical detector was used with another laser source as the local oscillator (LO) to demodulate the aggregated optical channel. For experimental simplicity, the required bandpass filtering of the aggregated optical channel was alternatively performed in the electrical domain via lowpass filtering after demodulation. A coherent modulation analyzer (Tektronix-OM1106) performed the required digital signal processing (DSP) of the recorded waveforms in a real-time oscilloscope (Tektronix DPO73304) for the visualization of symbol constellations and the measurement of other performance metrics like Q-factor and error vector magnitude (EVM). The bit-to-bit mapping of the signals was confirmed as the bit error rate testing was carried out concerning the expected pattern. In addition, the delay due to dispersion was considered for the bit-to-bit mapping.

In the experiments, two types of signal formats were investigated. The first signal format is a standard lower-order non-return to zero (NRZ) modulation format, such as binary phase shift keying (BPSK) or quadrature phase shift keying (QPSK). These modulation formats were generated with raised cosine spectral pre-shaping, which corresponds to a roll-off factor of 1.0. The spectral width of each channel matches the symbol rate. The second signal format is Nyquist BPSK or QPSK. In this format, symbols were modulated using sinc-shaped Nyquist pulse sequences.^49–51 The use of Nyquist pulses ensures that the spectrum of each channel is confined within a rectangular bandwidth, resulting in a zero roll-off factor. This allows for transmission at the maximum possible symbol rate, optimizing spectral efficiency.

These different signal formats were employed to evaluate the performance of the proposed aggregation method under different modulation formats.

In the first set of experiments, the aggregation of two 10 GBd BPSK channels to form one 10 GBd PAM-4 channel and the aggregation of two 10 GBd QPSK channels to form one 10 GBd QAM-16 channel were demonstrated. The dispersion compensating fiber (DCF) module was used to introduce a delay of −270 ps/nm between the two parent channels for decorrelation. Figure 5 shows the symbol constellation diagrams for the output 10 GBd QPSK signal [Fig. 5(a)] and the output 10 GBd PAM-4 signal [Fig. 5(b)]. In Fig. 5(a), the amplitude of both input channels was set to be equal (α = 1), and the carrier phase difference was set to ϕ = 90°. In Fig. 5(b), which represents the PAM-4 modulation, the amplitude of one input channel was reduced to α = 0.5 (the attenuation at the WS was programmed to be 6 dB for one of the channels), and the carrier phase difference was set to ϕ = 0°. The measured average error vector magnitudes and Q-factors are also provided in the figure.

For the aggregation of two 10 GBd QPSK channels to form a QAM-16 signal, the result is shown in Fig. 6(a). The measured eye diagram of the in-phase (I) component of the aggregated signal is displayed in Fig. 6(b). The slight deformation in the eye diagram is a result of the imperfect delay introduced by the dispersion compensating fiber module. The applied dispersion, amplitude difference, and RF phase were identical to the BPSK to QPSK aggregation case.

Since they have rectangular spectra, sinc-shaped Nyquist signals allow transmitting data with the maximum possible symbol rate in a given bandwidth. They enable wavelength division multiplexing without any guard bands between the channels.⁵² However, sinc-shaped Nyquist pulses are unlimited in the time domain and thus just a mathematical construct. Therefore, usually complex digital⁵³ or optical⁵⁴ signal processing is required to generate and detect the signals. Recently, a new method for Nyquist WDM based on sinc pulse sequences generated, modulated, and multiplexed by a single modulator and demultiplexed by another single modulator, drastically reducing the requirements for digital signal processing, has been proposed.^49,51 Here we have followed the same method to generate signals to be aggregated, which are modulated on sinc-sequences with two zero crossings. The experiments were performed using 5 and 8 GBd BPSK and QPSK signals as parent channels. The induced dispersion required for the decorrelation of 5 and 8 GBd Nyquist signals was −520 and −440 ps/nm, respectively.

Figure 7 presents the symbol constellations and measured performance metrics for the aggregated Nyquist QPSK and PAM-4 data signals obtained from two Nyquist BPSK parent channels with symbol rates of 5 and 8 GBd. In the previously reported experiments, it was observed that the reception of 8 Gbaud Nyquist-QPSK signals within a 24 Gbaud orthogonally multiplexed channel resulted in a degradation of ∼5 dB in the Q-factor for 40 km of single-mode fiber transmission.^51,55 The bit error rate (BER) can be estimated from the Q-factor using the relationship $BERest≈(1/2)erfc(Q/2)$⁠.⁵⁶ Applying this equation, the estimated BER after 40 km of transmission is on the order of $≈10−10$⁠. The EVM (error vector magnitude) and Q-factor are related to the optical signal-to-noise ratio (OSNR).⁵⁶ To improve the measured Q-factors, one approach is to use compact integrated devices and stand-alone comb sources that provide better OSNR. These improvements in OSNR can potentially lead to higher Q-factors and better overall system performance.

Figures 8(a) and 8(b) show the symbol constellation diagrams and corresponding performance metrics for Nyquist QAM-16 signals aggregated from two Nyquist QPSK parent channels with symbol rates of 5 and 8 GBd, respectively. Figure 8(c) displays the eye diagram of the in-phase component of the output 8 GBd Nyquist QAM-16 signal. The return-to-zero nature, resulting from the Nyquist pulse sequence associated with the three-line flat phase-locked comb, is evident from the eye diagram.

It is important to note that no digital pre- or post-equalization techniques were employed in these measurements. The use of digital pre- and post-equalization techniques, along with advancements in integrated devices and comb sources, can further enhance the system’s performance. These additional measures can help mitigate impairments and improve the overall quality of the transmitted signals.

These experimental results demonstrate the successful aggregation of different modulation formats, achieving higher spectral efficiency while maintaining acceptable signal quality. The generation of a PAM-4 signal from two BPSK channels involves attenuating one of the channels by 6 dB. This attenuation can result in a degradation of the overall optical signal-to-noise ratio (SNR) compared to the BPSK to QPSK aggregation, where both channels were equal in amplitude. The reduced SNR can lead to a decrease in signal quality and potentially impact the performance of the system. However, the attenuation has been used for experimental convenience; instead, an amplification can be used. Additionally, it is mentioned that the PAM-4 signal arrived at only the I-branch of the coherent receiver and the corresponding oscilloscope channel. This suggests that the signal was detected using a single detection path, unlike the QPSK signal, which utilized two signal detection paths (the I and Q branches). The utilization of two detection paths in the QPSK scenario can provide improved performance and better signal quality compared to a single detection path. Therefore, the observation of a slight degradation in the signal quality in the PAM-4 scenarios [Figs. 5(b) and 7(d)] compared to the QPSK scenario is expected, given the differences in signal attenuation and detection configuration.

Most real-world applications of wavelength channel aggregation and modulation format conversion require reconfigurability and scalability regarding data rates, bandwidth, and dynamic range. These criteria depend on the modulator and RF source used for the presented technique. In contrast to all other aggregation methods presented so far, no optical nonlinearity, unique electronic and photonic components, or pump laser sources are required. Integrated modulators offer broad operational wavelength ranges with very high bandwidth (⁠$≥100$ GHz) and energy efficiency in many commercially viable material platforms like LiNbO₃, silicon, and InP.^57–61 Therefore, integrating on silicon or other material platforms is straightforward. Although the presented theory and proof-of-concept experiment refer to an MZM, another intensity modulator, like a ring modulator or a phase modulator, will be sufficient to achieve similar functionality. By cascading the aggregation stages, any number of channels with the aggregated symbol rate can be aggregated to one single channel or any smaller number of channels with the same symbol rate.

The experimental results show the feasibility of achieving Nyquist orthogonal time-division multiplexing of aggregated higher modulation format channels to achieve higher spectral efficiency.^51–53 Moreover, more than two channels can also be aggregated by cascading the aggregation stages.

This technique offers reconfigurability, scalability, and the potential to achieve higher spectral efficiency in wavelength channel aggregation and modulation format conversion applications. By ensuring the stability of the optical source, addressing signal-to-noise ratio requirements, and minimizing the jitter of the RF source, it is possible to extend the proposed technique to support higher-order modulation formats. It is worth noting that while the possible sources of imperfections in the generated bit stream have been discussed in Sec. II, further investigations are needed to address the effects of various impairments specific to comb-sources. It is also important to note that comb-based transmission systems are still an evolving field, and system performance against the source specifications has recently been analyzed.^62,63 The authors plan to continue their efforts beyond this proof-of-concept demonstration to quantify the various optical and electrical sources of signal impairments.

In conclusion, we have proposed and experimentally demonstrated a linear optical signal processing method to aggregate lower bit rate channels into fewer higher bit rate channels. The method relies only on electro-optic modulators to manipulate the optical phase by controlling the electrical phase to accomplish a phase-coherent vector summation. In the first proof-of-concept experiments, we have shown aggregation of 10 GBd NRZ and 8 GBd Nyquist signals of lower modulation formats (e.g., BPSK, QPSK) to spectrally efficient higher-order modulation formats (e.g., QPSK, PAM-4, QAM-16) while conserving the symbol rates. No optical nonlinearity, pump lasers, or specialized electronics and photonics are needed. Therefore, integrating the method into any material platform is possible, which might make it a viable solution for future access and aggregation networks based on integrated optical frequency combs.

This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Grant Nos. 322402243, 403154102, 424608109, 424608271, 424607946, and 424608191, and in part by the German Federal Ministry of Education and Research (BMBF) under Grant No. 13N14879.

The authors have no conflicts to disclose.

Arijit Misra: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Investigation (lead); Methodology (lead); Validation (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead). Stefan Preußler: Conceptualization (lead); Investigation (equal); Methodology (equal); Supervision (supporting); Writing – original draft (supporting); Writing – review & editing (supporting). Karanveer Singh: Data curation (equal); Visualization (supporting); Writing – review & editing (supporting). Janosch Meier: Conceptualization (equal); Methodology (equal); Writing – review & editing (supporting). Thomas Schneider: Conceptualization (lead); Funding acquisition (lead); Project administration (lead); Resources (lead); Supervision (lead); Writing – review & editing (lead).

The data that support the findings of this study are available from the corresponding author upon reasonable request.