Expand this Topic clickable element to expand a topic
Skip to content
Optica Publishing Group

Eigen-function division multiplexed coherent optical transmission in time domain by using higher-order Hermite-Gaussian pulses

Open Access Open Access

Abstract

We describe a new multiplexing technique and its application to demultiplexing in the time domain by using higher-order Hermite-Gaussian (HG) pulses, which are solutions of the Schrödinger equation. We call this technique eigen-function division multiplexing (EDM). This method enables us to further increase the total transmission capacity by superimposing many different HG pulses in the same time slot. This technique is different from a conventional optical time domain multiplexing (OTDM) technique using interleaving, where one pulse exists only in one time slot. The transmitted EDM HG pulses can be demultiplexed by adopting the time-domain orthogonality of the HG pulses (eigen-function orthogonality). The information carried by the mth-order HG pulse (HGm pulse) can be coherently detected by a photo detector, where photo-mixing with a phase-locked HGm pulse generated by a local oscillator can realize demultiplexing. The overlap integral with a different HG pulse becomes zero due to the time domain orthogonality. First, we show numerically that such a new EDM transmission scheme in the time domain is possible. We then show experimentally that we could successfully carry out an EDM HG coherent pulse transmission with four different HG pulses (HG0, HG1, HG2, and HG3), where we report a 400∼480 Gbit/s (10 Gbaud x 4 eigen-functions x 2 pol-mux.) 32∼64 QAM EDM transmission over 300∼450 km.

© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement

1. Introduction

The total information capacity worldwide has been rapidly increasing at a rate of 40% per year, which indicates that in 20 years we will need a new communication infrastructure that can handle an information capacity 1000 times larger than today [1]. To realize such an ultrahigh capacity network, 3M technology consisting of multi-level modulation, multi-core fiber, and multi-mode control has been intensively studied [2,3,4]. Recently, as a result of combining all these 3M techniques, peta bit/s transmission experiments [5,6] have been reported although the transmission distances were less than 50 km.

Higher-order transverse modes in few-mode fiber, for example LP11, LP21 and LP02 modes, have been successfully transmitted with the aim of increasing the transmission capacity in the spatial domain as a space division multiplexing (SDM) technique [7,8]. Fan-in and fan-out devices [9,10] and a multi-input and multi-output (MIMO) technique [11] play important roles in SDM.

It is well-known that the transverse higher-order modes in a gas laser cavity are given by the HG shapes in the spatial domain [12], where the HG modes are special eigenmodes that satisfy the paraxial wave equation or a slowly varying envelope approximation of Maxwell’s wave equation in a cartesian coordinate system [13]. It is also independently well-known that a harmonic oscillator or particle trapping in a parabolic potential well can be described by the Schrödinger equation in the spatial domain, where the solutions (eigen-functions) are given by the HG distribution in a spatial plane [14]. Although the physics behind them are different, it is interesting to note that they have the same HG waveforms with the same orthogonality between eigenmodes in space.

On the other hand, in the time domain approach, when the AM or FM mode-locking equation of the laser that was developed by Haus [15,16] for the generation of an optical pulse train was parabolically approximated in the time and frequency regions, the equation also became a Schrödinger equation. This means that there are HG eigen-solutions in the time domain. Although it was difficult to generate higher-order HG pulses directly from a laser, we recently demonstrated that such pulses can be generated from an FM mode-locked laser [17,18]. When performing the experiment, we considered that since these HG pulses are also orthogonal to each other in the time domain, we may be able to use this HG pulse orthogonality for demultiplexing many different HG eigen-functions in the same time slot. This is the key issue in the present paper, where we call this technique the eigen-function division multiplexing (EDM) of higher-order HG pulses and demultiplexing in the time domain. It is also important to note that orthogonality integration can realize both demultiplexing and demodulation simultaneously.

Code division multiplexing (CDM) and the present method are similar in that multiple signals at the same wavelength can be shared in the same time slot, but a fundamental difference from the proposed method is that CDM realized by assigning an individual optical code to each channel uses the code orthogonality rather than the eigen-function orthogonality used in the present case, and therefore demultiplexing is performed simply with code correlation. Although our method can realize complete demultiplexing by using waveform orthogonality, CDM is generally susceptible to interference from other channels, making it difficult to demodulate a signal with a high signal-to-noise ratio.

In addition, in previous studies, the application of HG functions to OFDM subcarriers in a wireless transmission has been proposed and numerically studied [19,20,21], where subcarriers whose shape is given by the HG function are digitally generated and multiplexed. Recently, the use of an HG function in a nonlinear eigenvalue transmission has also been proposed [22,23], where a time-domain waveform is converted to an HG function as a function of the real part of the eigenvalue through an inverse scattering transformation. They are different from the present method where information is directly encoded on HG pulses that are generated, multiplexed and demultiplexed in the optical domain.

It should be noted that the present method is also different from conventional optical time division multiplexing (OTDM), which adopts the interleaving of a pulse where only one pulse exists in one time slot. Even in a Nyquist pulse transmission, although there is a large overlap between adjacent pulses, we use only one pulse in one time slot [24,25,26]. There is Nyquist pulse orthogonality given by the fact that there is output between the Nyquist pulses in the same time slot but there is no output in different time slots. This makes it possible to demultiplex the Nyquist pulse in the different time slot, which we call coherent OTDM Nyquist demultiplexing [27]. However, there are no higher-order Nyquist pulses that have orthogonality between them. In the present paper, it is important to note that there are higher-order HG pulses, which can be superimposed and demultiplexed in the same time slot. This makes EDM possible in the time domain, resulting in a further increase in the maximum transmission capacity even after minimizing the time slot with OTDM. Although HG-EDM in the time domain can thus be combined with an OTDM technique, it will constitute future work.

This paper is composed as follows. In section 2, we describe the fundamentals of the HG pulses and their superimposition in the time domain. In section 3, we propose an EDM technique in the time domain and present numerical analyses of coherent QAM transmission over 450 km, where the time-domain orthogonality is used for demultiplexing the transmitted HG pulses. In section 4, we prove experimentally that the present EDM in the time domain is possible by undertaking 16∼64 QAM coherent EDM transmission experiments over 450 km. Finally, we summarize the work in section 5.

2. Fundamental property of higher-order Hermite-Gaussian pulses

The mth-order HG pulses, aHGm(t), are well-known as eigen solutions of the following Schrödinger equation,

$$\frac{{{d^2}}}{{d{t^2}}}{a_{HGm}}(t) + ({{\lambda_m} - {t^2}} ){a_{HGm}}(t) = 0.$$

Here, λm is the mth eigenvalue, which can be replaced with λm = 2m + 1 due to the energy quantization of the harmonic oscillator. The solution is given by

$${a_{HG\textrm{m}}}(t) = {H_m}(t){e^{ - {t^2}/2}},$$
where Hm(t) is called the Hermite polynomial function. It is known that Hm(t) can be generally expressed in the following polynomial form by using the Maclaurin expansion of the generating function [28]. When m is an even number, Hm(t) is expressed in the form
$$ \begin{aligned} H_m(t) & =m ! \sum_{n=0}^{m / 2} \frac{(-1)^n}{n !(m-2 n) !}(2 t)^{m-2 n} \\ & =2^m t^m-\frac{2^{m-2} m !}{(m-2) !} t^{m-2}+\frac{2^{m-4} m !}{2 !(m-4) !} t^{m-4}-\frac{2^{m-6} m !}{3 !(m-6) !} t^{m-6}+\ldots+(-1)^{m / 2} \frac{m !}{\left(\frac{m}{2}\right) !}. \end{aligned} $$

When m is an odd number, Hm(t) is expressed in the form

$$ \begin{aligned} H_m(t) & =m ! \sum_{n=0}^{(m-1) / 2} \frac{(-1)^n}{n !(m-2 n) !}(2 t)^{m-2 n} \\ & =2^m t^m-\frac{2^{m-2} m !}{(m-2) !} t^{m-2}+\frac{2^{m-4} m !}{2 !(m-4) !} t^{m-4}+\frac{2^{m-6} m !}{3 !(m-6) !} t^{m-6}+\ldots+(-1)^{\frac{m-1}{2}} \frac{2 \cdot m !}{\left(\frac{m-1}{2}\right) !} t.\end{aligned} $$

Thus, we obtain

\begin{align}{H_0}(t) &= 1,\end{align}
\begin{align}{H_1}(t) &= 2t,\end{align}
\begin{align}{H_2}(t) &= 4{t^2} - 2,\end{align}
\begin{align}{H_3}(t) &= 8{t^3} - 12t,\end{align}
\begin{align}{H_4}(t) &= 16{t^4} - 48{t^2} + 12,\end{align}
\begin{align}{H_5}(t) &= 32{t^5} - 160{t^3} + 120t,\end{align}
\begin{align}{H_6}(t) &= 64{t^6} - 480{t^4} + 720{t^2} - 120,\end{align}

As easily confirmed, we have the following important orthogonality relation between aHGm(t) and aHGq(t). When mq,

$$\int\nolimits_{-\infty }^\infty a_{HGm}(t)a_{HGq}(t)dt = 0$$
or
$$\int\nolimits_{-\infty }^\infty H_m(t)H_q(t)e^{-t^2}dt = 0.$$

This orthogonality plays an important role in demultiplexing a single eigen-solution from the multiplexed (superimposed) HG pulses in the time domain as we will see in sections 3 and 4.

Since the Fourier transformation of the Schrödinger equation of Eq. (1) is also the Schrödinger equation, the Fourier transformation of an HG pulse is also the HG function with a certain coefficient change. That is, the Fourier transformation of ${H_m}(t){e^{ - {t^2}/2}}$ is given by ${H_m}(\omega ){e^{ - {\omega ^2}/2}}$ multiplied by $\sqrt {2\pi } {( - i)^m}$. A factor of (−i)m means that the spectrum of an HG pulse changes with a coefficient of 1, −i, −1, and i periodically. That is, the spectrum alternates between real and imaginary values. Therefore, when we use the definition ${C_m} = \sqrt {2\pi } {( - i)^m}$, a general HG spectrum can be rewritten as

$${A_{HGm}}(\omega ) = {C_m}{H_m}(\omega ){e^{ - {\omega ^2}/2}}.$$

Figure 1(a)∼(c) show aHGm(t) and the corresponding AHGm(ω) for m =0, 1, and 2, respectively, where the waveforms in the time domain given on the left hand side have the same shape as the corresponding spectral profile given on the right hand side with changes in the ± sign and the real and imaginary values. Similarly, aHGm(t) and the corresponding AHGm(ω) for m =3, 4, and 5 are given in Fig. 2(a)∼(c), respectively. From these figures, we see that the waveform in the time domain alternates between even and odd functions, where the corresponding spectrum of the even function in the time domain is given by a real even function, and that of the odd function by a pure imaginary odd function.

 figure: Fig. 1.

Fig. 1. Hermite-Gaussian waveforms aHGm(t) in the time domain (left) and the corresponding spectra AHGm(ω) (right) for m = 0, 1, and 2. (a)∼(c) correspond to m = 0, 1, and 2, respectively.

Download Full Size | PDF

 figure: Fig. 2.

Fig. 2. Hermite-Gaussian waveforms aHGm(t) in the time domain (left) and the corresponding spectra AHGm(ω) (right) for m = 3, 4, and 5. (a)∼(c) correspond to m = 3, 4, and 5, respectively.

Download Full Size | PDF

In the mth-order HG pulses, aHGm(t) has m + 1 peaks in its waveform. The ringing repetition does not change sinusoidally, which occurs in Nyquist pulses [24], where the period of the inner zero-crossing is shorter than that of the outer zero-crossing. When we aim to use aHGm(t) as a data pulse for optical transmission, it should be noted that the higher-order aHGm(t) becomes broader than lower-order HG pulses both in the time and spectral domains, which may degrade the transmission performance. These features can be clearly seen in Figs. 1 and 2. Therefore, the pulse width of higher-order HG pulses has to be well confined within the one time slot of the data transmission.

It is also important to note that the EDM waveform and the corresponding spectrum in the time domain have unique features at the input of the transmission fiber depending on the EDM number. Figure 3 shows the input EDM waveforms and the corresponding spectral profiles, where (a-1), (b-1), and (c-1) correspond to the waveforms in 4-, 6-, and 10 eigen-function EDM (superimposing) and (a-2), (b-2), and (c-2) are the corresponding spectra, respectively. The blue and red curves correspond to cases with the same average input power and the same peak input power, respectively. When the average power for each eigen-function was kept constant as shown by the blue curve, the lower eigen-function had a larger peak power. That is, the highest peak power occurred at HG0. More importantly, we see that the superimposed waveforms are asymmetric and differ greatly depending on the number that are superimposed, which may lead to a different Kerr nonlinearity in a fiber transmission. When the peak input power of each pulse is set at the same power as shown by the red curves, we see that although there is a small difference from the blue curves, the overall shapes are very similar. HG0 has the minimum average power in this case.

When we consider IQ modulation of such superimposed HG pulses in a coherent transmission, the transmitting waveform becomes more complicated as the number of EDM increase. Therefore, it is important to note that the transmission system is linear, where complete demultiplexing can be possible by using the orthogonality between the HG pulses. When we include a certain fiber nonlinearity, the demultiplexing performance will be degraded, which we describe in detail in the next section.

3. Proposal of eigen-function division multiplexing and demultiplexing in the time domain by using HG pulses

In this section, we present numerical analyses of HG pulse transmission to evaluate the performance of EDM and demultiplexing in the time domain.

3.1 Numerical model

Our EDM HG pulse transmission simulation model is shown in Fig. 4. At the transmitter, as shown in Fig. 4(a), 10 GHz HGm pulses are modulated individually at an IQ modulator with 10 Gbaud QAM data. After polarization multiplexing, the HGm pulses are multiplexed and launched into a transmission line. The transmission line is composed of dispersion-managed fibers with a span length of 75 km. Each span consists of a 50 km super large area fiber (SLAF) with a dispersion of 20 ps/nm/km, a dispersion slope of 0.07 ps/nm2/km and an effective area of 110 μm2, and a 25 km inverse dispersion fiber (IDF) with a dispersion of − 40 ps/nm/km, a dispersion slope of − 0.14 ps/nm2/km and an effective area of 30 μm2. The fiber loss is 0.2 dB/km and compensated for with an erbium-doped fiber amplifier (EDFA) with a noise figure (NF) of 4.5 dB. These conditions were adopted from our experiments, which are described in section 4.

 figure: Fig. 3.

Fig. 3. EDM waveforms and the corresponding spectral profiles. (a-1), (b-1), and (c-1) correspond to the waveforms in 4-, 6-, and 10 eigen-function EDM (superimposing), respectively, and (a-2), (b-2), and (c-2) are the corresponding spectra. The blue and red curves correspond to cases with the same average input power and the same peak input power, respectively.

Download Full Size | PDF

 figure: Fig. 4.

Fig. 4. Setup for numerical analyses of EDM HG pulse transmission. (a) is the transmission setup and (b) is the receiver, where the transmitted EVM can be obtained by calculating the orthogonality integration given by Eq. (12).

Download Full Size | PDF

The pulse propagation is simulated using a split-step Fourier method with the Manakov equation in the following form [29].

$$ \left\{\begin{aligned}&i \frac{\partial u_x}{\partial z}-\frac{\beta_2}{2} \frac{\partial^2 u_x}{\partial t^2}-i \frac{\beta_3}{6} \frac{\partial^3 u_x}{\partial t^3}+\frac{8}{9} \gamma\left(\left|u_x\right|^2+\left|u_y\right|^2\right) u_x=-i \frac{\alpha}{2} u_x \\&i \frac{\partial u_y}{\partial z}-\frac{\beta_2}{2} \frac{\partial^2 u_y}{\partial t^2}-i \frac{\beta_3}{6} \frac{\partial^3 u_y}{\partial t^3}+\frac{8}{9} \gamma\left(\left|u_x\right|^2+\left|u_y\right|^2\right) u_y=-i \frac{\alpha}{2} u_y \end{aligned}.\right. $$

Here, ux and uy represent the x and y polarizations of the complex amplitude of the electric field, β2 and β3 are the second- and third-order dispersion coefficients, γ = 2πn2/λAeff is the nonlinear coefficient with n2, λ, and Aeff defined as the nonlinear index, wavelength and effective area, respectively, and α is the loss coefficient. At the receiver, as shown in Fig. 4(b), after passing through a 90-deg. optical hybrid device, the transmitted EDM signals are homodyne-detected with an HGm local oscillator (LO) pulse whose phase is locked to the transmitted data, where EDM signals are demultiplexed and demodulated simultaneously.

3.2 Transmission performance of a single higher-order HG pulse

First, we consider the propagation of a single HGm pulse. Figure 5 shows the waveforms (a-1)∼(a-4), corresponding spectra (b-1)∼(b-4), and constellations (c-1)∼(c-4) of 16 QAM signals after a 450 km transmission when HG0, HG1, HG2, and HG3 pulses are transmitted individually in a single channel. The red and black curves are analogue HG waveforms after transmission and the theoretical input HG waveforms, respectively. We can see clean time-domain HG pulses, the corresponding spectrum and QAM constellations. Here, the average input power was optimized for each HG pulse to minimize the error vector magnitude (EVM). The power optimization results are shown in Fig. 6, which also includes the result for a conventional QAM transmission with the CW input beam denoted as CW. The average optimum powers that we obtained were 0 dBm for HG0 and HG1, 4 dBm for HG2, and 2 dBm for HG3 and CW transmission. It can be seen that the performance of the HG0 pulse is superior to that of the conventional CW QAM, as can be expected from the well-known fact that an RZ pulse has a better performance than an NRZ pulse. It can also be seen that HG1 and HG3 pulses are superior to the conventional QAM.

 figure: Fig. 5.

Fig. 5. Waveform, spectrum, and constellation of 16 QAM signals after 450 km transmission when HG0 (a-1)∼(c-1), HG1 (a-2)∼(c-2), HG2 (a-3)∼(c-3), and HG3 (a-4)∼(c-4) pulses are transmitted individually in a single channel. (a), (b), and (c) correspond to the waveform, the corresponding spectrum, and the calculated constellation with 4096 symbols, respectively.

Download Full Size | PDF

 figure: Fig. 6.

Fig. 6. Relationship between average input power and EVM after 450 km transmission for 16 QAM 10 Gbaud HG signals (HG0∼HG3) including CW. CW means a conventional QAM modulation with a CW light beam. This figure shows a single channel transmission for each HG eigen-function.

Download Full Size | PDF

Then, we analyzed the optimization of the receiver bandwidth to obtain the ideal time-domain orthogonality calculated from Eq. (12). To ensure perfect digital sampling to maintain the transmitted HG pulses, the sampling rate was set as high as 1280 GSa/s (128 points/100 ps considering 10 Gbaud transmission), which does not introduce any orthogonality errors caused by the sampling speed itself. Figure 7 shows the EVM dependence on the receiver bandwidth in the balanced detection when 10 Gbaud, 16 QAM EDM signals are demultiplexed and demodulated under a back-to-back condition. As the bandwidth increases, the EVM is improved thanks to the higher accuracy of the time-domain orthogonality integration for the demultiplexing and demodulation of the EDM signals. The EVM also has an oscillatory dependence on the bandwidth and is greatly degraded when the bandwidth becomes narrower than 10 GHz. It is important to note that in the experiment, we cannot use an A/D converter with an unlimited sampling rate of, for example, 1280 GSa/s and an unlimited receiver bandwidth. That is, when a sampling rate is reduced in the experiment to a lower sampling rate, there is an optimum bandwidth for generating a better EVM. In such cases, a wider bandwidth is not always the optimum depending on the balance between the sampling error and the degree of the orthogonality integration based on the bandwidth. The details are provided in section 4.

 figure: Fig. 7.

Fig. 7. The EVM dependence on the receiver bandwidth for orthogonality integration when demultiplexing and demodulating 16 QAM 10 Gbaud signals (calculation). The sampling rate was set as fast as 1280 GSa/s to realize an ideal ultrafast sampling. This sampling rate is unrealistic, but we show here the bandwidth dependence for the orthogonality calculation. A wider bandwidth gives a smaller EVM.

Download Full Size | PDF

3.3 Nonlinear effects in the time-domain EDM transmission

In this section, we describe how the Kerr nonlinearity degrades the performance of the EDM in the time domain when using HG pulses. The blue plots in Fig. 8 shows the result of 4 eigen-function 16 QAM transmission over 450 km when four HG pulses are transmitted at the same average power of − 4 dBm. Here, the solid line shows the results with fiber Kerr nonlinearity such as self-phase modulation (SPM) and cross-phase modulation (XPM) implemented in Eq. (15), whereas the dashed line shows the results without the nonlinearity as an ideal case to observe the influence of nonlinearity on EDM transmission. The EDM signals are demultiplexed using a balanced receiver with a 30 GHz bandwidth and A/D converted at a sampling rate of 80 GSa/s based on actual experiments. The BER was estimated from the SNR, which is proportional to 1/(EVM)2. It can be clearly seen that nonlinearity-induced transmission impairment increases in all HG pulses as seen in the increase in EVM by 0.3∼0.7% with larger BERs. A case with the same average input power indicates that a lower HG pulse has a higher peak power than the higher HG pulses, resulting in a larger XPM.

 figure: Fig. 8.

Fig. 8. EVM values of 16 QAM 10 Gbaud, 4 eigen-function HG pulse (HG0-HG3) transmission over 450 km when HG pulses are transmitted at the same average power of - 4 dBm (blue) and the same peak power of 3.4 dBm (red). The solid and dashed lines show the results with and without fiber nonlinearity, respectively. Fiber nonlinearity means SPM and XPM.

Download Full Size | PDF

We then fixed the peak input power of each HG pulse rather than the average input power so that each HG pulse experienced the same amount of nonlinearity. The red plots in Fig. 8 show the results of 4 eigen-function 16 QAM transmission over 450 km with and without fiber nonlinearity, respectively, when HG pulses are transmitted at the same peak input power. The peak power was set at 3.4 dBm, which indicates that the average powers are -5.1 dBm for HG0, -3.8 dBm for HG1, -3.3 dBm for HG2, and -3.0 dBm for HG3. It can be seen that the performance is improved compared with the blue plots in Fig. 8 for higher-order HG pulses. Conversely, the HG0 pulse has a slightly worse EVM, which indicates that we need a higher average power to improve the OSNR of the HG0 pulse.

We further investigated the way in which the fiber nonlinearity degrades the EVM greatly when the EDM number is increased. Figure 9 shows the results for a 6 eigen-function (HG0∼HG5) 16 QAM transmission over 450 km when HG pulses are transmitted at the same peak power (3.4 dBm). The insets show corresponding constellations for convenience. As the number of eigen-functions increases from that in Fig. 8, the transmission performance is more impaired especially for HG2 and HG3, which now suffers from XPM with HG4 and HG5. This phenomenon did not happen in the 4 eigen-function case described in Fig. 8. Figure 9 also shows the results for a 10 eigen-function (HG0∼HG9) 16 QAM transmission over 450 km when HG pulses are transmitted at the same peak power. It can be clearly seen that the transmission performance is greatly degraded compared with 6 eigen-functions. In particular, the performance from HG2 to HG6 were worse than those for HG7∼HG9 due to the XPM effects from the higher-order eigen-functions. These results indicate that although it is possible to superimpose many HG pulses, the total performance is degraded due to fiber nonlinearity such as XPM. Therefore, when the number of eigen-functions is large, the transmission distance may be decreased compared with a small number of superimposed pulses, which may be caused by a high peak-to-average-power (PAPR) effect.

 figure: Fig. 9.

Fig. 9. EVM values of 16 QAM 10 Gbaud, 6 eigen-function HG pulse (HG0-HG5) and 10 eigen-function HG pulse (HG0-HG9) transmissions over 450 km when HG pulses are transmitted at the same peak power of 3.4 dBm.

Download Full Size | PDF

3.4 Comparison of transmission performance of time-domain EDM with different number of eigen-function and QAM multiplicity

Here, we compare the transmission performance with different QAM multiplicities (4, 16, 64, and 256) depending on the number of eigen-functions (4, 6, and 10). The results are shown in Figs. 10(a)∼(d), respectively. Figure 10(a) compares 4-, 6-, and 10 eigen-function QPSK transmission over 450 km when HG pulses are transmitted at the same peak power. The sampling rate and the bandwidth were set at 80 GSa/s and 30 GHz, respectively. EVM values after transmission are plotted for each HG pulse. We can see that even with 10 eigen-function operation, it appears to be quite easy to realize a 10 eigen-function EDM transmission over 450 km when QPSK is adopted.

 figure: Fig. 10.

Fig. 10. Comparison of EVM values in 4-, 6-, and 10 eigen-function 10 Gbaud transmissions. (a) QPSK (450 km), (b) 16 QAM (450 km), (c) 64 QAM (450 km), (d) 256 QAM (150 km). It can be seen that the EVM worsens with an increase in eigen-function number.

Download Full Size | PDF

Figure 10(b) compares 4-, 6-, and 10 eigen-function 16 QAM transmission over 450 km. Although an EVM degradation compared with Fig. 10(a) can be seen when the eigen-function number is increased, it still appears to be possible to transmit over 450 km with 10 eigen-function operation. The results for 64 QAM transmission over 450 km are shown in Fig. 10(c). Here, the sampling rate of the A/D converter was increased to 160 GSa/s from 80 GSa/s to obtain sufficient accuracy in the orthogonality integration required for 64 QAM high-multiplicity demodulation. The bandwidth was increased to 50 GHz from 30 GHz. It appears to be possible to send 6 eigen-functions over 450 km, but 10 eigen-function transmission seems to be difficult. As we describe in section 4, we were able to realize 4 eigen-function EDM 64 QAM transmission over 300 km experimentally.

Figure 10(d) shows the results for 256 QAM transmission over 150 km. Here, the sampling rate of the A/D converter and the bandwidth were also set at 160 GSa/s and 50 GHz, respectively, and the transmission distance was reduced to 150 km as the distortion becomes significantly large after 450 km. It appears to be difficult to send more than 4 eigen-functions. We find that a higher QAM multiplicity with a larger HG eigen-function number is difficult to realize.

Finally, in Fig. 11(a)∼(d) we summarize the relationship between BER and a transmission distance for HG0 ∼ HG5 pulses in a 6 eigen-function QPSK transmission of 16 QAM, 64 QAM, and 256 QAM signals, respectively. (a) and (b) are 80 GSa/s and a 30 GHz bandwidth, and (c) and (d) are 160 GSa/s and a 50 GHz bandwidth. The BER gradually increases in all cases. (a) is QPSK, where the BER has a relationship of HG5 < HG0 < HG4 = HG1 < HG2 < HG3. HG5 and HG0 have a better transmission performance than HG2 and HG3. (b) is 16 QAM, where the BER has a relationship of HG5 < HG0 < HG1 = HG4 < HG2 < HG3. 16 QAM has the same BER dependence as QPSK. (c) is 64 QAM, where the BER has a relationship of HG5 < HG4 < HG0 < HG1 < HG3 < HG2. (d) is 256 QAM, where the BER has a relationship of HG5 < HG4 < HG3 < HG1 < HG2 < HG0. Surprisingly, HG5 performs the best in all multiplicities, which would result from minimizing the total XPM effect.

 figure: Fig. 11.

Fig. 11. BER vs. transmission distance for HG0 ∼ HG5 pulses in a 6 eigen-function 10 Gbaud transmission. (a) QPSK, (b) 16 QAM, (c) 64 QAM, and (d) 256 QAM. HG5 has the highest performance in each case.

Download Full Size | PDF

4. Coherent transmission experiments on eigen-function division multiplexed Hermite-Gaussian pulses in the time domain

In this section, we present experimental results obtained for EDM coherent optical transmission in the time domain with HG pulses.

4.1 Experimental setup for time-domain EDM transmission

The experimental setup for a 4 eigen-function EDM transmission over 450 km is shown in Fig. 12. In the present experiment, we used a combination consisting of a comb generator and a liquid crystal on silicon (LCoS) optical filter to generate various HG pulses since it is not easy to prepare many mode-locked HG lasers [17,18]. An external-cavity laser-diode (ECLD) with an 8 kHz linewidth [30] and a 10 GHz optical comb generator [31] were used as a pulse source, and the output signal was split into 5 arms. The wavelength of the ECLD was 1550 nm. Four LCoS filters and IQ modulators were used simultaneously to generate QAM-modulated HG0-HG3 pulses. The IQ modulators were driven by arbitrary waveform generators 1 and 2 (AWG1 and 2). The 9th harmonic of the optical comb signal, which was 90 GHz down-shifted from the center frequency of the signal, was simultaneously extracted with a narrow optical filter and used as a pilot tone for the optical phase-locking of the LO HG pulse. Furthermore, an intensity modulated laser diode (LD) signal at a wavelength of 1564 nm was used to deliver a 10 GHz clock.

 figure: Fig. 12.

Fig. 12. Experimental setup for a 4 eigen-function 10 Gbaud HG-EDM transmission over 450 km.

Download Full Size | PDF

The transmission line we used consisted of six 75 km spans with a 50 km super large area (SLA) fiber (dispersion: 20 ps/nm/km, dispersion slope: 0.07 ps/nm2/km) and a 25 km inverse dispersion fiber (IDF, dispersion: - 40 ps/nm/km, dispersion slope: - 0.14 ps/nm2/km) so that the second- and third-order dispersions were compensated for simultaneously. The average loss was 17 dB/span including a splicing loss of 2 dB, which was compensated for with an EDFA. These conditions were used for the numerical analyses described in section 3.

At the receiver, the transmitted signal was split into 3 arms, where the EDM, pilot tone, and 10 GHz clock signals, respectively, were extracted. The extracted EDM signal was input into a coherent receiver consisting of a 90-degree hybrid and a balanced photo detector (B-PD) after polarization demultiplexing. In contrast, the extracted pilot tone was frequency-upshifted by 90 GHz with an optical frequency shifter (OFS) to coherently recover the original optical phase and then coupled into a DFB-LD for injection locking. We used the injection-locked LD and a comb generator to generate phase-locked HGm pulses, where the extracted 10 GHz clock was also used as the driving signal of the comb generator. The time-domain EDM signal was homodyne-detected with the LO HG pulse thus prepared by adjusting the timing between the transmitted signal and the LO pulse with an optical delay line (Δτ). Then, the detected signal was A/D-converted by a digital oscilloscope and demodulated by using a digital signal processor (DSP). In the DSP, the EDM signal was demultiplexed by calculating the overlap integral given by Eq. (12), and the demultiplexed QAM signal was demodulated with vector signal analyzer (VSA) software [32]. Since HG pulses are demultiplexed in the optical domain, the DSP performs only an add operation when calculating the overlap integral. Such a low DSP complexity is another advantage of EDM.

In our experiment, we adopted a dispersion-managed fiber instead of electric dispersion compensation, because HG pulses are demultiplexed in the optical domain and thus the dispersion must be compensated for before demultiplexing. If an A/D converter with a sufficiently wide bandwidth is available, it is possible to convert the multiplexed HG pulses entirely into digital signals and perform both dispersion compensation and demultiplexing in a DSP. As for dispersion compensation other than with a DSP in the receiver, an analog or digital prechirp method can be employed in the transmitter. It is also possible to employ a chirped fiber Bragg grating (CFBG) before demultiplexing at the receiver when SMF is used [33].

4.2 Generation of higher-order HG pulses and their IQ modulation

The intensity waveforms of HG0 ∼ HG3 pulses generated by the comb generators are shown in Fig. 13(a-1)∼(d-1), respectively, where an optical sampling oscilloscope (OSO) with an 800 fs resolution was used for the measurement. Figure 13(a-2)∼(d-2), respectively, are the corresponding spectra of HG0 ∼ HG3 pulses with a resolution of 0.02 nm. In Fig. 13, the blue dashed lines correspond to the theoretical curves, which agree well with the experimental curves shown by the black lines. It can be seen that high-quality HG pulses were successfully prepared. Note here that the - 20 dB spectral bandwidth gradually increases in the higher-order eigen-functions, which will be utilized for the precise calculation of the spectral efficiency (SE) of the present EDM system.

 figure: Fig. 13.

Fig. 13. Waveform (upper side) and the corresponding optical spectrum (lower side) of each HG pulse at the transmitter. (a-1) and (a-2) are for HG0, (b-1) and (b-2) are for HG1, (c-1) and (c-2) are for HG2, and (d-1) and (d-2) are for HG3 pulses.

Download Full Size | PDF

In our experiments, we used the non-inverted and inverted output signals from AWG1 and AWG2, respectively, to modulate the four HG pulses independently, where a time delay was applied between the non-inverted and inverted output signals to decorrelate them. The sampling rates and bandwidths of AWG1 and AWG2 were 10 and 90 GSa/s, and 7.5 and 32 GHz, respectively. Figure 14 shows the eye diagram of each IQ modulator in the transmitter, where a CW optical carrier was modulated by a 10 Gbaud QPSK signal with AWG1 or AWG2. In Fig. 14, the rise and fall changes were sharper in AWG2 ((c) and (d)) than those ((a) and (b)) in AWG1 due to the wider bandwidth. It is important to avoid waveform distortions in the data modulation process to maintain the orthogonality of the HG pulses. Therefore, AWG2 was used for the data modulation of the HG2 and HG3 pulses, which are relatively widespread in time compared with the HG0 and HG1 pulses. Thus, the wide flat region of the eye diagram can be used for the data modulation of the high-order HG pulses which have wider pulse widths.

 figure: Fig. 14.

Fig. 14. Eye diagram for each IQ modulator in the transmitter. (a) IQM1 for HG0, (b) IQM2 for HG1, (c) IQM3 for HG2, and (d) IQM4 for HG3.

Download Full Size | PDF

Figure 15 shows the waveform and corresponding optical spectrum of the 4 eigen-function (HG0-HG3) EDM signal from the transmitter. (a) is the waveform averaged 50 times. Although the eye of the actual waveform is closed since each HG pulse is modulated with a random 16 QAM signal, the trapezoidal envelope can be seen through the averaging process. (b) is the corresponding optical spectrum with a resolution of 0.1 nm. Here, a tone signal for phase-locking the LO pulse is seen on the longer wavelength side. The spectral profile of the EDM signal was also trapezoidal as seen in the waveform, where the -20 dB bandwidth and the OSNR were 150 GHz and 38.5 dB, respectively. Since the Fourier transformation of the time-domain HG pulse has the same shape in the spectral region as we saw in Eqs. (2) and (14), the similarity of the waveform and the spectrum is understandable.

 figure: Fig. 15.

Fig. 15. Waveform (a) and the corresponding optical spectrum (b) for a 4 eigen-function 10 Gbaud EDM signal at the transmitter.

Download Full Size | PDF

After a 450 km transmission, the waveform and optical spectrum of the 4 eigen-function EDM signal have changed as shown in Fig. 16(a) and (b), respectively, where the average transmission power was optimized at - 4 dBm/eigen-function. The OSNR after a 450 km transmission decreased to 22 dB from 38.5 dB in (b).

 figure: Fig. 16.

Fig. 16. Waveform (a) and the corresponding optical spectrum (b) of a 4 eigen-function 10 Gbaud EDM signal after a 450 km transmission.

Download Full Size | PDF

4.3 Polarization-multiplexed 10 Gbaud, 16 QAM, 4 eigen-function EDM transmission

Here, we describe the experimental results for a polarization-multiplexed 10 Gbaud, 16 QAM, 4 eigen-function EDM transmission. We show the EVMs only for X-polarization data since the demodulation performance of the X- and Y-polarization data was almost the same. First, we measured the demodulation performance of the EDM signal by using a standard digital oscilloscope that had a sampling rate of 80 GSa/s with a bandwidth of 32 GHz including the response of the electronic devices. The black plots in Fig. 17 show the EVM of each HG pulse (HG0∼HG3) obtained under a back-to-back condition. The EVMs ranged from 8.3 (HG0) to 10.2% (HG2), where BERs of the order of 10−4 (1.2 × 10−4∼4.8 × 10−4) were obtained.

 figure: Fig. 17.

Fig. 17. EVM values of polarization-multiplexed 10 Gbaud, 16 QAM, 4 eigen-function EDM signal (HG0∼HG3 pulses) obtained under a back-to-back condition (black), after a 450 km transmission when set at the same average power of - 4.0 dBm (green), and after a 450 km transmission when set at the same peak power of 3.4 dBm/ eigen-function (blue). The average powers for HG0, HG1, HG2, and HG3 become - 5.1, - 3.8, - 3.3, and - 3.0 dBm, respectively, in blue line. The sampling rate is 80 GSa/s and the receiver bandwidth is 32 GHz.

Download Full Size | PDF

After a 450 km transmission, the EVM of the 4 eigen-function EDM signal had changed as shown by the green plots in Fig. 17, where the average power of each HG pulse was kept constant at – 4.0 dBm. The EVM and BER of each HG pulse varied in the 10.5 (HG0) ∼ 11.9% (HG1) and 1.3 ∼ 2.7 × 10−3 ranges, respectively. On the other hand, when we set the peak powers so that they were constant at 3.4 dBm/eigen-function as shown by the blue plots in Fig. 17, the EVM of each HG pulse became rather uniform and had slightly better values in the 10.9 ∼ 11% and 1.4 ∼ 1.7 × 10−3 ranges, respectively. In this case, the average powers of the HG0, HG1, HG2, and HG3 eigen-functions were -5.1, -3.8, -3.3, and -3.0 dBm, respectively. As we saw in the numerical simulation in Fig. 8, we could also demonstrate experimentally that it is better to set the peak power of each HG pulse at a constant value to obtain uniform demodulation performance. Thus, we use the same peak power condition in the following experiments.

4.4 Relationship between receiver bandwidth and sampling rate to obtain improved orthogonality integration

Here, we investigate the relationship between the receiver bandwidth and the sampling rate with a view to obtaining better orthogonality integration. As seen in Fig. 7, a higher sampling rate with a wider bandwidth generally gives us a lower EVM, which indicates perfect orthogonality since the HG pulses can be precisely quantized without any distortions. However, when we undertake experiments, the sampling rate depends on the performance of the A/D-convertor we use and certainly does not have unlimited speed. The sampling rate is currently limited to 256 GSa/s.

Therefore, we further investigated numerically the relationship between the bandwidth and the commercially available sampling rate. Figure 18 shows the EVM as a function of bandwidth when the sampling rate was fixed at 80 GSa/s. The result indicates that a bandwidth of 10 or 20 GHz is best for obtaining the smallest EVM, and a bandwidth exceeding 30 GHz or narrower than 10 GHz gives a larger EVM. A bandwidth narrower than 10 GHz is understandable because the baud rate is 10 Gbaud. However, it is very surprising that a larger EVM is obtained by orthogonality integration when the bandwidth is broadened.

 figure: Fig. 18.

Fig. 18. EVM vs. bandwidth characteristics for calculating orthogonality integration with an 80 GSa/s digitizer for a 4 eigen-function 10 Gbaud EDM transmission. The optimum bandwidths were 10 and 20 GHz.

Download Full Size | PDF

This may result from a relationship between the sampling error and the degree of orthogonality based on the bandwidth. To prove this, we have numerically analyzed constellations with their EVMs and the corresponding time-domain waveforms with and without digitization for different bandwidths. The results are shown in Fig. 19, where (a-1), (b-1), and (c-1) are constellations and (a-2), (b-2), and (c-2) are the corresponding time-domain waveforms with different bandwidths of 30, 10 and 5 GHz, respectively. The sampling rate is 80 GSa/s. Here, we describe only the HG0 channel in the 4 eigen-function transmission. The red and black solid lines in Fig. 19(a-2) indicate an 80 GSa/s digitized transmitted waveform with a 30 GHz bandwidth and an original analogue HG0 waveform, where we see a large deviation between them. This deviation gives rise to a sampling error in the orthogonality integration although the calculation itself is correct, where the EVM is 7.4% from (a-1). A better orthogonality can be seen in Fig. 19(b-2) where the bandwidth is reduced to 10 GHz although the digitized waveform becomes rather smooth compared with the original HG waveform in (a-2) because of the narrower bandwidth. The EVM was 5.6% as seen in (b-1). This is because although the HG waveform is somewhat distorted from the original HG pulse, the smoothly digitized waveform still has sufficient information to calculate the orthogonality, while there are no sampling errors between the red and black lines. The worst case is shown in Fig. 19(c-2), where the bandwidth was further narrowed to 5 GHz. The EVM increases to 12.3% as seen in (c-1). We see a large waveform deviation in Fig. 19(c-2) from Figs. 19(b-2) or (a-2), which indicates that the information as a HG0 pulse is completely lost and therefore, the orthogonality calculation has a large error although the sampling has no error.

 figure: Fig. 19.

Fig. 19. The constellations of 10 Gbaud HG0 eigen-function and the corresponding digitized waveforms with different bandwidths. (a-1), (b-1), and (c-1) are constellations and (a-2), (b-2), and (c-2) are HG0 waveforms with 30, 10, and 5 GHz bandwidths, respectively. The optimum bandwidth for the smallest EVM (5.6%) obtained by orthogonality integration was 10 GHz.

Download Full Size | PDF

In accordance with this analysis, we reduced the bandwidth from 32 GHz to 11 GHz. As a result, we could improve the demodulation performance of a polarization multiplexed 10 Gbaud, 16 QAM, 4 eigen-function EDM signal as shown by the black curve in Fig. 20 under a back-to-back condition. Here, the receiver bandwidth was narrowed by inserting a low-pass filter before the A/D convertor. The EVM of each HG pulse thus obtained was 6.9 ∼ 8.1%, which was improved by 1 ∼ 2% compared with the black plots in Fig. 17 as we expected. In the present case, there was no error with transmitted data consisting of 4096 symbols. After a 450 km transmission, the EVM changed as shown by the blue curve in Fig. 20, where the EVM of each HG pulse was 9.6 ∼ 10.3%, which was also 0.7 ∼ 1.4% lower than that the blue plots in Fig. 17. This result derives an important relationship between the sampling rate and the corresponding bandwidth that allows us to calculate the orthogonality between HG pulses and indicates that a wider bandwidth is not always better when determining a sampling rate. There is an optimum balance between the sampling rate and the receiver bandwidth depending on the waveform distortion they cause.

 figure: Fig. 20.

Fig. 20. EVM of a polarization-multiplexed 10 Gbaud, 16 QAM, 4 eigen-function EDM signal under a back-to-back condition and after a 450 km transmission. The sampling rate is 80 GSa/s and the receiver bandwidth is 11 GHz.

Download Full Size | PDF

4.5 High-performance 16∼64 QAM time-domain EDM transmission by incorporating 256 GSa/s digitization with a 40 GHz bandwidth

A higher transmission performance can be expected when we incorporate high-speed digitization with a wider bandwidth, where a much better orthogonality calculation can be achieved compared with that achieved with a low sampling rate and the corresponding low bandwidth. In Fig. 21 we calculated the EVM dependence of a 4 eigen-function EDM 10 Gbaud signal as a function of the receiver bandwidth for orthogonality integration when using a 260 GSa/s A/D converter. In this case, the EVM gradually decreases when the bandwidth increases as already seen in Fig. 7, which indicates that the sampling rate is sufficiently fast to avoid causing any distortions for a 10 Gbaud signal.

We then experimentally demonstrated large improvements in the demodulation performance of the EDM transmission by increasing the sampling rate of the A/D convertor and the receiver bandwidth. Figure 22 shows the EVM of the EDM signal under a back-to-back condition and after a 450 km transmission obtained by using a digital oscilloscope with a sampling rate of 256 GSa/s and a 110 GHz bandwidth including electronic devices. The constellations for HG2 before and after transmission are also shown for comparison. Here, the receiver bandwidth was 40 GHz, which was limited by the BPDs. The EVM of each HG pulse under a back-to-back condition was as small as 5.8 ∼ 6.4%, which indicates that the demodulation performance was greatly improved compared with that obtained with the 80 GSa/s A/D convertor shown in Fig. 20. The EVM of each HG pulse after a 450 km transmission was as small as 8.2 ∼ 8.4%, where an error-free operation was achieved. A wider bandwidth of above 40 GHz will further improve the performance.

 figure: Fig. 21.

Fig. 21. EVM dependence of a 4 eigen-function 10 Gbaud EDM signal on the receiver bandwidth for orthogonality integration when using a 260 GSa/s A/D converter (calculation).

Download Full Size | PDF

 figure: Fig. 22.

Fig. 22. EVM of a polarization-multiplexed 10 Gbaud, 16 QAM, 4 eigen-function EDM signal under a back-to-back condition and after a 450 km transmission. The sampling rate is 256 GSa/s and the receiver bandwidth is 40 GHz. Much better EVMs were obtained than those in Fig. 20.

Download Full Size | PDF

Since we found that a combination consisting of a 256 GSa/s digitizer and a 40 GHz receiver provide excellent performance for time-domain EDM transmission, we increased the QAM multiplicity from 16 to 32 and 64. The blue plots in Fig. 23 show the demodulation performance of polarization-multiplexed 10 Gbaud, 32 QAM, 4 eigen-function EDM (400 Gbit/s) transmission, where the circles are the EVMs under a back-to-back condition and the squares are those after a 450 km transmission. It is possible to realize error-free transmission by employing a 7% overhead FEC threshold (2 × 10−3) for all HG pulses even after a 450 km transmission. There is no difference between the EVM degradation with different HG eigen-functions.

 figure: Fig. 23.

Fig. 23. Experimental results of EVMs for polarization-multiplexed 10 Gbaud, 32 QAM (blue) and 64 QAM (green), 4 eigen-function EDM transmissions. The circles are the EVMs obtained under a back-to-back condition and the squares are those after transmission. The sampling rate is 256 GSa/s and the receiver bandwidth is 40 GHz.

Download Full Size | PDF

In this transmission, since the -20 dB bandwidth of the EDM signal was 150 GHz, the potential SE can be calculated as 400 Gbit/s/150 GHz/1.07 = 2.49 bit/s/Hz. From Fig. 13(a-2), the -20 dB bandwidth is approximately 100 GHz, which results in an SE of 100 Gbit/s/100 GHz/1.07 = 0.93 bit/s/Hz for a single HG eigen-function. For 4 eigen-function EDM transmission, although the transmission capacity is four times larger, it does not mean a four times larger SE. That is, it becomes 2.49/0.93 = 2.67 times since the -20 dB bandwidth of the HG3 eigen-function, which is seen in Fig. 13(d-2), is broadened to 150 GHz.

We further increased the multiplicity to 64. The green plots in Fig. 23 show the demodulation performance of a polarization-multiplexed 10 Gbaud, 64 QAM, 4 eigen-function EDM (480 Gbit/s) transmission, where the circles are the EVMs under a back-to-back condition and the squares are those after a 300 km transmission. It is possible to realize error-free transmission by employing a 20% overhead FEC threshold (2 × 10−2) for all HG pulses after a 300 km transmission. In this case, the SE was 480 Gbit/s/150 GHz/1.2 = 2.67 bit/s/Hz.

The BER characteristics for Fig. 23 are summarized in Fig. 24 as a function of the transmission distance. The normalized SE vs. number of eigen-functions is also depicted in Fig. 25. Here, it should be noted that both the spectral width and the time-bandwidth product increase for higher-order HG pulses. This indicates that, if one higher-order HG pulse is transmitted alone, the SE would become lower [21]. In EDM, this does not cause the SE to increase linearly as shown by the dashed line, but it increases along the solid line. The average increase in the SE per eigen-function is approximately 0.4 bit/s/Hz/eigen-function, which is dominantly determined by the bandwidth of the highest-order HG pulse.

 figure: Fig. 24.

Fig. 24. BER performance of polarization-multiplexed 10 Gbaud, 32 and 64 QAM, 4 eigen-function EDM (400∼480 Gbit/s) transmission as a function of transmission distance.

Download Full Size | PDF

 figure: Fig. 25.

Fig. 25. Normalized spectral efficiency (SE) vs. number of HG eigen-functions. The average SE increase per eigen-function is approximately 0.4 bit/s/Hz/eigen-function.

Download Full Size | PDF

A 10 Gbaud pol. mux. transmission usually gives us 100∼120 Gbit/s when we use 32 ∼ 64 QAM, but its capacity can be improved fourfold (400 or 480 Gbit/s) with the present method, which is a very interesting new technique in the time domain.

5. Conclusion

We proposed and demonstrated eigen-function division multiplexing (EDM) in the time domain and its demultiplexing with the use of higher-order Hermite-Gaussian pulses, which made it possible to increase the transmission capacity in the time domain. The demultiplexing and demodulation were simultaneously achieved by adopting time-domain orthogonality between the higher-order HG pulses in the same time slot. N time EDM gives an N time larger transmission capacity, although the SE is reduced to approximately 0.4N. If our aim is to have the same transmission capacity as before, we can reduce the baud rate to 1/N, which means that we can use low-speed optical devices in the high-speed system. The dispersion tolerance will be also reduced.

We showed in the numerical analysis that the same peak input power for all HG pulses achieves more uniform eigen-function-dependent transmission performance than with the same average power. This is because the SPM and XPM can be mitigated when the peak power is the same.

In experiments, we have shown that 4 eigen-function EDM 16∼64 QAM transmission with HG0∼HG3 can be achieved over 450 km. It is also important to note that a high-resolution A/D convertor is necessary to realize precise time-domain orthogonality with the use of the overlap integral on a high-speed photo detector. However, once a sampling rate has been determined in the experiment, there is an optimum bandwidth, where a wider bandwidth is not always appropriate depending on the balance between sampling error and the degree of orthogonality based on the bandwidth. In our experiment, for example, the photo detector bandwidth was electrically narrowed to 11 GHz for a sampling speed of 80 Gsample/s, which made it possible to have the lowest EVM.

The present EDM system needs to employ different HG pulses as an input and a phase-locked coherent HGm pulse as a local oscillator, which slightly complicates the ordinary system. However, this new technique will be useful not only for ultrahigh-speed EDM optical transmission, but also for signal processing in the time domain.

Disclosures

The authors declare no conflicts of interest.

Data availability

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.

References

1. M. Nakazawa, M. Suzuki, Y. Awaji, et al., eds., Space-Division Multiplexing in Optical Communication Systems: Extremely Advanced Optical Transmission with 3M Technologies (Springer, 2022).

2. M. Terayama, S. Okamoto, K. Kasai, et al., “4096 QAM (72 Gbit/s) Single-Carrier Coherent Optical Transmission with a Potential SE of 15.8 bit/s/Hz in All-Raman Amplified 160 km Fiber Link,” Proc. OFC 2018, paper Th1F.2.

3. Y. Sasaki, K. Takenaga, K. Aikawa, et al., “Single-Mode 37-Core Fiber with a Cladding Diameter of 248 μm,” Proc. OFC 2017, paper Th1H.2.

4. G. Rademacher, R. S. Luís, B. J. Puttnam, et al., “1.53 Peta-bit/s C-Band Transmission in a 55-Mode Fiber,” Proc. ECOC 2022, paper Th3C.3.

5. D. Soma, Y. Wakayama, S. Beppu, et al., “10.16-peta-b/s Dense SDM/WDM Transmission over 6-mode 19-core Fiber across the C + L Band,” J. Lightwave Technol. 36(6), 1362–1368 (2018). [CrossRef]  

6. G. Rademacher, B. J. Puttnam, R. S. Luís, et al., “10.66 Peta-Bit/s Transmission over a 38-Core-Three-Mode Fiber,” Proc. OFC 2020, paper Th3H.1.

7. N. K. Fontaine, R. Ryf, H. Chen, et al., “30×30 MIMO Transmission over 15 Spatial Modes,” Proc. OFC 2015, paper Th5C.1. [CrossRef]  

8. K. Shibahara, M. Hoshi, and Y. Miyamoto, “10-spatial-mode 1300-km Transmission over 6-LP Graded Index Few-Mode Fiber with 36-ns Modal Dispersion,” Proc. OFC 2023, paper M2B.2.

9. M. Yoshida, T. Hirooka, and M. Nakazawa, “Low-loss and Reflection-free Fused Type Fan-out Device for 7-core Fiber based on a Bundled Structure,” Opt. Express 25(16), 18817–18826 (2017). [CrossRef]  

10. T. Takahata, A. Kaya, Y. Ozawa, et al., “High Reliability Fan-in / Fan-out Device with Isolator for Multi-core Fibre Based on Free Space Optics,” Proc. ECOC 2021, paper We1A.5.

11. R. Ryf, S. Randel, A. H. Gnauck, et al., “Mode-Division Multiplexing Over 96 km of Few-Mode Fiber Using Coherent 6 × 6 MIMO Processing,” J. Lightwave Technol. 30(4), 521–531 (2012). [CrossRef]  

12. H. Kogelnik and T. Li, “Laser Beams and Resonators,” Appl. Opt. 5(10), 1550–1567 (1966). [CrossRef]  

13. H. A. Haus, Waves and Fields in Optoelectronics (Prentice-Hall, 1983).

14. L. I. Schiff, Quantum Mechanics, (McGraw-Hill Kogakusha, 1968), pp. 66–76.

15. H. A. Haus, “A Theory of Forced Mode Locking,” IEEE J. Quantum Electron. QE-11(7), 323–330 (1975). [CrossRef]  

16. H. A. Haus, “Mode-locking of Lasers,” IEEE J. Sel. Top. Quantum Electron. 6(6), 1173–1185 (2000). [CrossRef]  

17. M. Nakazawa and T. Hirooka, “Theory of Higher-Order Hermite-Gaussian Pulse Generation from an FM Mode-Locked Laser,” IEEE J. Quantum Electron. 59(2), 1600325 (2023). [CrossRef]  

18. M. Nakazawa, M. Yoshida, and T. Hirooka, “Experiments on the Generation of Higher-order Hermite-Gauss Pulses from an FM Mode-Locked Laser,” IEEE J. Quantum Electron. 59(6), 1300216 (2023). [CrossRef]  

19. R. Haas and J. C. Belfiore, “A time-frequency well-localized pulse for multiple carrier transmission,” Wirel. Pers. Commun. 5(1), 1–18 (1997). [CrossRef]  

20. T. Kurt, M. Siala, and A. Yongaçoglu, “Multi-carrier signal shaping employing Hermite functions,” 13th European Signal Processing Conference, 1–4 (2005).

21. T. Moazzeni, “On the compactness of OFDM and Hermite signals,” IEEE Comm. Lett. 20(7), 1313–1316 (2016). [CrossRef]  

22. M. Balogun and S. Derevyanko, “Hermite-Gaussian Nonlinear Spectral Carriers for Optical Communication Systems Employing the Nonlinear Fourier Transform,” IEEE Comm. Lett. 26(1), 109–112 (2022). [CrossRef]  

23. M. Balogun and S. Derevyanko, “Enhancing the Spectral Efficiency of Nonlinear Frequency Division Multiplexing Systems via Hermite-Gaussian Subcarriers,” J. Lightwave Technol. 40(18), 6071–6077 (2022). [CrossRef]  

24. M. Nakazawa, T. Hirooka, P. Ruan, et al., “Ultrahigh-speed “Orthogonal” TDM Transmission with an Optical Nyquist Pulse Train,” Opt. Express 20(2), 1129–1140 (2012). [CrossRef]  

25. T. Hirooka, R. Hirata, J. Wang, et al., “Single-channel 10.2 Tbit/s (2.56 Tbaud) Optical Nyquist Pulse Transmission over 300 km,” Opt. Express 26(21), 27221–27236 (2018). [CrossRef]  

26. M. Yoshida, K. Kimura, T. Iwaya, et al., “Single-channel 15.3 Tbit/s, 64 QAM Coherent Nyquist Pulse Transmission over 150 km with a Spectral Efficiency of 8.3 bit/s/Hz,” Opt. Express 27(20), 28952–28967 (2019). [CrossRef]  

27. K. Harako, D. O. Otuya, K. Kasai, et al., “High-performance TDM Demultiplexing of Coherent Nyquist Pulses using Time-domain Orthogonality,” Opt. Express 22(24), 29456–29464 (2014). [CrossRef]  

28. G. B. Arfken, H. J. Weber, and F. E. Harris, Mathematical Methods for Physicists (Academic, 2013).

29. D. Marcuse, C. R. Menyuk, and P. K. A. Wai, “Application of the Manakov-PMD Equation to Studies of Signal Propagation in Optical Fibers with Randomly Varying Birefringence,” J. Lightwave Technol. 15(9), 1735–1746 (1997). [CrossRef]  

30. K. Kasai, M. Nakazawa, Y. Tomomatsu, et al., “1.5 μm, Mode-hop-free Full C-band Wavelength Tunable Laser Diode with a Linewidth of 8 kHz and a RIN of -130 dB/Hz and its Extension to the L-band,” Opt. Express 25(18), 22113–22124 (2017). [CrossRef]  

31. T. Sakamoto, T. Kawanishi, and M. Izutsu, “Asymptotic Formalism for Ultraflat Optical Frequency Comb Generation using a Mach–Zehnder Modulator,” Opt. Lett. 32(11), 1515–1517 (2007). [CrossRef]  

32. Keysight Technologies, “Infiniium oscilloscopes with 89600 VSA software” Application Notes, https://www.keysight.com/us/en/assets/7018-02776/application-notes/5990-6819.pdf

33. K. Kimura, M. Yoshida, K. Kasai, et al., “Experimental and Numerical Analysis of Ultrahigh-Speed Coherent Nyquist Pulse Transmission with Low-Nonlinearity Dispersion Compensator,” IEICE Trans. Comm. E105-B(9), 1014–1022 (2022). [CrossRef]  

Data availability

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.

Cited By

Optica participates in Crossref's Cited-By Linking service. Citing articles from Optica Publishing Group journals and other participating publishers are listed here.

Alert me when this article is cited.


Figures (25)

Fig. 1.
Fig. 1. Hermite-Gaussian waveforms aHGm(t) in the time domain (left) and the corresponding spectra AHGm(ω) (right) for m = 0, 1, and 2. (a)∼(c) correspond to m = 0, 1, and 2, respectively.
Fig. 2.
Fig. 2. Hermite-Gaussian waveforms aHGm(t) in the time domain (left) and the corresponding spectra AHGm(ω) (right) for m = 3, 4, and 5. (a)∼(c) correspond to m = 3, 4, and 5, respectively.
Fig. 3.
Fig. 3. EDM waveforms and the corresponding spectral profiles. (a-1), (b-1), and (c-1) correspond to the waveforms in 4-, 6-, and 10 eigen-function EDM (superimposing), respectively, and (a-2), (b-2), and (c-2) are the corresponding spectra. The blue and red curves correspond to cases with the same average input power and the same peak input power, respectively.
Fig. 4.
Fig. 4. Setup for numerical analyses of EDM HG pulse transmission. (a) is the transmission setup and (b) is the receiver, where the transmitted EVM can be obtained by calculating the orthogonality integration given by Eq. (12).
Fig. 5.
Fig. 5. Waveform, spectrum, and constellation of 16 QAM signals after 450 km transmission when HG0 (a-1)∼(c-1), HG1 (a-2)∼(c-2), HG2 (a-3)∼(c-3), and HG3 (a-4)∼(c-4) pulses are transmitted individually in a single channel. (a), (b), and (c) correspond to the waveform, the corresponding spectrum, and the calculated constellation with 4096 symbols, respectively.
Fig. 6.
Fig. 6. Relationship between average input power and EVM after 450 km transmission for 16 QAM 10 Gbaud HG signals (HG0∼HG3) including CW. CW means a conventional QAM modulation with a CW light beam. This figure shows a single channel transmission for each HG eigen-function.
Fig. 7.
Fig. 7. The EVM dependence on the receiver bandwidth for orthogonality integration when demultiplexing and demodulating 16 QAM 10 Gbaud signals (calculation). The sampling rate was set as fast as 1280 GSa/s to realize an ideal ultrafast sampling. This sampling rate is unrealistic, but we show here the bandwidth dependence for the orthogonality calculation. A wider bandwidth gives a smaller EVM.
Fig. 8.
Fig. 8. EVM values of 16 QAM 10 Gbaud, 4 eigen-function HG pulse (HG0-HG3) transmission over 450 km when HG pulses are transmitted at the same average power of - 4 dBm (blue) and the same peak power of 3.4 dBm (red). The solid and dashed lines show the results with and without fiber nonlinearity, respectively. Fiber nonlinearity means SPM and XPM.
Fig. 9.
Fig. 9. EVM values of 16 QAM 10 Gbaud, 6 eigen-function HG pulse (HG0-HG5) and 10 eigen-function HG pulse (HG0-HG9) transmissions over 450 km when HG pulses are transmitted at the same peak power of 3.4 dBm.
Fig. 10.
Fig. 10. Comparison of EVM values in 4-, 6-, and 10 eigen-function 10 Gbaud transmissions. (a) QPSK (450 km), (b) 16 QAM (450 km), (c) 64 QAM (450 km), (d) 256 QAM (150 km). It can be seen that the EVM worsens with an increase in eigen-function number.
Fig. 11.
Fig. 11. BER vs. transmission distance for HG0 ∼ HG5 pulses in a 6 eigen-function 10 Gbaud transmission. (a) QPSK, (b) 16 QAM, (c) 64 QAM, and (d) 256 QAM. HG5 has the highest performance in each case.
Fig. 12.
Fig. 12. Experimental setup for a 4 eigen-function 10 Gbaud HG-EDM transmission over 450 km.
Fig. 13.
Fig. 13. Waveform (upper side) and the corresponding optical spectrum (lower side) of each HG pulse at the transmitter. (a-1) and (a-2) are for HG0, (b-1) and (b-2) are for HG1, (c-1) and (c-2) are for HG2, and (d-1) and (d-2) are for HG3 pulses.
Fig. 14.
Fig. 14. Eye diagram for each IQ modulator in the transmitter. (a) IQM1 for HG0, (b) IQM2 for HG1, (c) IQM3 for HG2, and (d) IQM4 for HG3.
Fig. 15.
Fig. 15. Waveform (a) and the corresponding optical spectrum (b) for a 4 eigen-function 10 Gbaud EDM signal at the transmitter.
Fig. 16.
Fig. 16. Waveform (a) and the corresponding optical spectrum (b) of a 4 eigen-function 10 Gbaud EDM signal after a 450 km transmission.
Fig. 17.
Fig. 17. EVM values of polarization-multiplexed 10 Gbaud, 16 QAM, 4 eigen-function EDM signal (HG0∼HG3 pulses) obtained under a back-to-back condition (black), after a 450 km transmission when set at the same average power of - 4.0 dBm (green), and after a 450 km transmission when set at the same peak power of 3.4 dBm/ eigen-function (blue). The average powers for HG0, HG1, HG2, and HG3 become - 5.1, - 3.8, - 3.3, and - 3.0 dBm, respectively, in blue line. The sampling rate is 80 GSa/s and the receiver bandwidth is 32 GHz.
Fig. 18.
Fig. 18. EVM vs. bandwidth characteristics for calculating orthogonality integration with an 80 GSa/s digitizer for a 4 eigen-function 10 Gbaud EDM transmission. The optimum bandwidths were 10 and 20 GHz.
Fig. 19.
Fig. 19. The constellations of 10 Gbaud HG0 eigen-function and the corresponding digitized waveforms with different bandwidths. (a-1), (b-1), and (c-1) are constellations and (a-2), (b-2), and (c-2) are HG0 waveforms with 30, 10, and 5 GHz bandwidths, respectively. The optimum bandwidth for the smallest EVM (5.6%) obtained by orthogonality integration was 10 GHz.
Fig. 20.
Fig. 20. EVM of a polarization-multiplexed 10 Gbaud, 16 QAM, 4 eigen-function EDM signal under a back-to-back condition and after a 450 km transmission. The sampling rate is 80 GSa/s and the receiver bandwidth is 11 GHz.
Fig. 21.
Fig. 21. EVM dependence of a 4 eigen-function 10 Gbaud EDM signal on the receiver bandwidth for orthogonality integration when using a 260 GSa/s A/D converter (calculation).
Fig. 22.
Fig. 22. EVM of a polarization-multiplexed 10 Gbaud, 16 QAM, 4 eigen-function EDM signal under a back-to-back condition and after a 450 km transmission. The sampling rate is 256 GSa/s and the receiver bandwidth is 40 GHz. Much better EVMs were obtained than those in Fig. 20.
Fig. 23.
Fig. 23. Experimental results of EVMs for polarization-multiplexed 10 Gbaud, 32 QAM (blue) and 64 QAM (green), 4 eigen-function EDM transmissions. The circles are the EVMs obtained under a back-to-back condition and the squares are those after transmission. The sampling rate is 256 GSa/s and the receiver bandwidth is 40 GHz.
Fig. 24.
Fig. 24. BER performance of polarization-multiplexed 10 Gbaud, 32 and 64 QAM, 4 eigen-function EDM (400∼480 Gbit/s) transmission as a function of transmission distance.
Fig. 25.
Fig. 25. Normalized spectral efficiency (SE) vs. number of HG eigen-functions. The average SE increase per eigen-function is approximately 0.4 bit/s/Hz/eigen-function.

Equations (15)

Equations on this page are rendered with MathJax. Learn more.

d 2 d t 2 a H G m ( t ) + ( λ m t 2 ) a H G m ( t ) = 0.
a H G m ( t ) = H m ( t ) e t 2 / 2 ,
H m ( t ) = m ! n = 0 m / 2 ( 1 ) n n ! ( m 2 n ) ! ( 2 t ) m 2 n = 2 m t m 2 m 2 m ! ( m 2 ) ! t m 2 + 2 m 4 m ! 2 ! ( m 4 ) ! t m 4 2 m 6 m ! 3 ! ( m 6 ) ! t m 6 + + ( 1 ) m / 2 m ! ( m 2 ) ! .
H m ( t ) = m ! n = 0 ( m 1 ) / 2 ( 1 ) n n ! ( m 2 n ) ! ( 2 t ) m 2 n = 2 m t m 2 m 2 m ! ( m 2 ) ! t m 2 + 2 m 4 m ! 2 ! ( m 4 ) ! t m 4 + 2 m 6 m ! 3 ! ( m 6 ) ! t m 6 + + ( 1 ) m 1 2 2 m ! ( m 1 2 ) ! t .
H 0 ( t ) = 1 ,
H 1 ( t ) = 2 t ,
H 2 ( t ) = 4 t 2 2 ,
H 3 ( t ) = 8 t 3 12 t ,
H 4 ( t ) = 16 t 4 48 t 2 + 12 ,
H 5 ( t ) = 32 t 5 160 t 3 + 120 t ,
H 6 ( t ) = 64 t 6 480 t 4 + 720 t 2 120 ,
a H G m ( t ) a H G q ( t ) d t = 0
H m ( t ) H q ( t ) e t 2 d t = 0.
A H G m ( ω ) = C m H m ( ω ) e ω 2 / 2 .
{ i u x z β 2 2 2 u x t 2 i β 3 6 3 u x t 3 + 8 9 γ ( | u x | 2 + | u y | 2 ) u x = i α 2 u x i u y z β 2 2 2 u y t 2 i β 3 6 3 u y t 3 + 8 9 γ ( | u x | 2 + | u y | 2 ) u y = i α 2 u y .
Select as filters


Select Topics Cancel
© Copyright 2024 | Optica Publishing Group. All rights reserved, including rights for text and data mining and training of artificial technologies or similar technologies.