1. Introduction

The growth of personal communication demand and the popularity of broadband services of terminal drive the development of next-generation passive optical network (PON) [1]. Due to its convenient maintenance, lower power consumption and backward compatibility, PON as the main solution for breaking the bottleneck of the “last mile” [2] is widely used in fiber-to-the-home (FTTH) deployment nowadays. Starting in 1995, the gigabit-capable PON (GPON) and gigabit Ethernet PON (GE-PON) have been standardized in the ITU-T G.984 and IEEE 802.3ah, respectively [3, 4]. Besides, the ten gigabit passive optical network (XG-PON), known as the next-generation passive optical network (NG-PON1), was standardized in 2009 [5]. The next-generation passive optical network stage 2 (NG-PON2) effort was initiated by FSAN in 2011 to investigate on upcoming technologies enabling a bandwidth increase beyond 10 Gb/s in the PON [6]. In order to satisfy the requirements of providing high capacity and cost effectiveness in next-generation access networks, lots of PON technologies have been studied, such as the wavelength-division-multiplexing PON (WDM-PON) [7], the time-wavelength- division multiplexing PON (TWDM-PON) [8, 9], and the orthogonal-frequency- division multiplexing PON (OFDM-PON) [10].

As the bandwidth demand of end-customers further increases, the space-division multiplexing (SDM) [11–19], one promising technique for increasing the capacity scalability, has been intensively investigated for high-speed free-space and optical fiber transmission, including short-reach communications, to address the capacity crunch by use of various spatial modes and multi-core fiber (MCF), few-mode fiber (FMF) and multi-mode fiber (MMF) instead of single mode fibers (SMFs). Moreover, elliptical-core FMF (EC-FMF), a specialty fiber with elliptical core shape helping to lift the degeneracy of fiber modes and enabling reduced mode coupling/crosstalk [20, 21], has attracted great interest in mode-division multiplexing (MDM) assisted short-reach communications [22]. Beyond the multiplexing techniques such as MDM, advanced modulation formats have also been widely used in short-reach communications to increase the transmission capacity, such as the quadrature amplitude modulation (QAM) [23], the orthogonal frequency division multiplexing (OFDM) [24], and the carrier-less amplitude/phase modulation (CAP) [25, 26]. Overall, CAP, as one kind of intensity modulation and direct detection (IM-DD) technology, is considered to be a promising candidate to achieve optical transmission over 100 Gb/s with improved performance of system complexity, spectrum efficiency (SE) and cost. Different from QAM, CAP uses digital filtering process to generate quadrature signal without additional carriers, thus the implementation of CAP system is comparably simple, and no electrical complex-to-real-value conversion, complex mixer, radio frequency (RF) source and optical IQ modulator are required for CAP [27]. It is worth noting that if one does not know the detailed information of the digital filters, one cannot recover the real signal. This feature guarantees the communication security. Yet another useful feature is the possibility of providing multiple access by extending the dimension of CAP, similar to code division multiple access (CDMA) [28]. Therefore, that is particularly suitable for parallel optical interconnections.

In this paper, we propose and demonstrate the implementation of MDM-CAP-PON architecture based on two modes (LP₀₁ and LP_11a) multiplexing in an EC-FMF. A 1.1-km trench-assisted EC-FMF is employed for anti-bending short-reach multiplexing transmission and IM-DD based optical network unit (ONU) is adopted in the experiment. 2.5-Gbaud CAP-16 signal is used as downstream transmission data. The experiment successfully demonstrates two MDM channels carrying CAP-16 signals, for four users with net data rate of 5 Gb/s per user, in the downstream transmission.

2. Concept and principle of MDM-CAP-PON architecture

Figure 1 shows the concept and principle of a novel downstream transmission for MDM-PON based on CAP-16. For the downstream link from optical line terminal (OLT) to optical network unit (ONU), the multiple data streams from OLT ports are modulated onto the Gaussian mode to achieve multiple optical CAP-16 signals, which are then respectively converted into multiple spatial modes. For simplicity two spatial modes (e.g. LP₀₁ and LP_11a modes) are considered here. After that, the two modes are combined by mode multiplexer at the OLT side and propagate through a 1.1-km EC-FMF. After the multiplexing transmission, the two modes are demultiplexed and then coupled into single mode fibers (SMFs) and received by each ONU.

 

Fig. 1 Concept and principle of a MDM-CAP-PON architecture. OLT: optical line terminal; ODN: optical distribution network; ONU: optical network unit; Mux/Demux: multiplexing/demultiplexing; MDM: mode-division multiplexing; CAP: carrier-less amplitude/phase; SMF: signal-mode fiber.

Download Full Size | PDF

Remarkably, the MDM-CAP-PON architecture shown in Fig. 1 can be easily scaled to a large number of end users. In addition to the MDM scalability offered by more spatial modes, larger transmission capacity and more users are expected when further employing higher dimension CAP signals. Such feature is more suitable for PON which is a point-to-multipoint optical network and CAP signal has obvious advantage in the downstream transmission [29–32]. Meanwhile, the MDM-CAP-PON for downstream is compatible with other PON architectures for upstream. Now there are many mature technologies for the upstream transmission and the bandwidth requirement for upstream is much less than the downstream. For example, on-off keying (OOK) can be combined with time-division multiplexing (TDM) [14] and the upstream is OOK modulated on the same carrier with lower data rate. An alternative scheme is to use different wavelengths but the same modulation format for the downstream and upstream transmissions [32]. Thus, in the following sections, for simplicity we show only the setup and results of proof-of-concept experiments for the downstream transmission of the MDM-CAP-PON.

Actually, CAP can be considered to be a kind of pulsed amplitude modulation (PAM) which is modulated with square root raised cosine filter pair at the transmitter and receiver [28]. The main reason why CAP code is preferred over PAM code is more of an implementation issue, namely, the CAP transceiver needs to only operate at half the baud rate of PAM transceiver in order to produce the same symbol rate [33]. Figure 2 shows the digital signal processing (DSP) flow chart for CAP-16 signal. At the transmitter side, two independent pseudo random bit sequences (PRBS) are used as the original bit sequences of the data streams 1 and 2. The sequences are mapped into real symbols of PAM-4 (four-level encoding [-3, −1, 1, 3]). Before the code sequences are sent into two shaping filters, the up-sampling process here is needed to match the rate of shaping filter and obtain the output analog signal. Thus, the code sequences are up-sampled by a factor M, i.e. M-1 zeros are inserted between two consecutive input symbols [34]. In order to guarantee high quality signal and achieve large transmission capacity, the up-sampling factor is chosen as low as 4 for the CAP-16 signal. The impulse responses of two shaping filters are expressed as $f_1(t)=g(t)\cdot \cos (2\pi f_{c}t)$ and $f_2(t)=g(t)\cdot \sin (2\pi f_{c}t)$ in order to form a Hilbert pair [35] which have the same amplitude-frequency characteristic and the different phase-characteristic of 90 degree. A square-root-raised-cosine shaping filter$g(t)$is employed as the baseband impulse response with a roll-off coefficient $\alpha$. From the expression form of the formula, it can be regarded as a digital format to realize QAM and$f_c$is center frequency of carriers. After the shaping filters, the signals are combined as the CAP-16 signal and sent to an arbitrary waveform generator (AWG). Let $T$ be the symbol period. Note that the digital filters and AWG operate at a rate of M/T, but the bandwidth of the signal is decided by the symbol rate of the system and the roll-off coefficient which can be written as $B\text{=}(1\text{+}\alpha )/T$.

 

Fig. 2 Digital signal processing (DSP) flow chart for CAP-16 signal. (a) The transmitter of the CAP signal. (b) The receiver of the CAP signal.

Download Full Size | PDF

At the receiver side, direct detection is used and the received signal after analog-to-digital (A/D) conversion are resampled to a rate of M/T so that it can be recovered to two streams by two matched filters. The two filters are the time-reversed version of the shaping filters at the transmitter in order to separate the in-phase and quadrature components. The orthogonality of the shaping filters and matched filters can enable transmission without intersymbol interference (ISI) and intercarrier interference (ICI) in the ideal channel. But in reality, the signal is always destroyed by the sampling timing offsets and the unknown channel impairments. Therefore an adaptive equalizer is needed to find the accurate sampling points and recover the CAP signal. Here a modified cascaded multi-modulus algorithm (MCMMA) [36] is used as the equalization algorithm after the signal is resampled to be 2 samples per symbol. It has a smaller steady-state error and a faster convergence speed because it can deal with two independent signals separately. After the equalization, the signal is decoded and the bit-error rate (BER) is calculated by error counting.

3. Experiment setup

Figure 3 shows the experimental setup of downstream transmission in a MDM-CAP-PON architecture. At the OLT side, the 2.5-Gbaud CAP-16 signal is generated offline by MATLAB and up-sampled by a factor of 4. A square-root-raised-cosine shaping filter with a roll-off coefficient of$\alpha \text{=}0.15$is used as the baseband impulse response. The center frequency$f_c$is given by$f_c=(1+\alpha )/2T+\Delta f$. Here $\Delta f$ is frequency offset and is set to be 0.1 GHz. Figures 4(a) and 4(b) show the time domain and the frequency domain impulse responses of the filter pairs. The electric spectrum of the CAP-16 signal at the transmitter is displayed in Fig. 4(c). Obviously, the CAP is a passband system and the bandwidth is 2.875 GHz. An arbitrary waveform generator (Tektronix AWG 70002) is used to produce the RF signal at 10 GSa/s to generate 2.5-Gbaud CAP-16 signal because of the up-sampled factor. The signal at 1550 nm is modulated by an optical intensity modulator (IM) with CAP-16. Remarkably, in order to guarantee the transmission performance and power budget, the CAP-16 signal is pre-amplified by an erbium-doped optical fiber amplifier (EDFA1). A variable optical attenuator (VOA) and another EDFA2 are used to measure the BER as a function of the received optical signal-to-noise ratio (OSNR) to evaluate the communication system performance. The OSNR is changed by adjusting the VOA followed by EDFA2. Note that the EDFAs are only used in the OLD side which can be regarded as the central office [37]. After that, there are no in-line amplifiers between the ODN and ONU, i.e. the ODN and ONU of the PON architecture remain passive. It is noted that similar PON configurations employing the EDFA in the OLT side were also reported in previous experimental works [7, 8, 38]. After the pre-amplification, the signal light is split into two equal parts by an optical coupler (50:50) and one is connected with a 200-m SMF to remove the correlation between the two signals. Two independent signals are converted into LP₀₁ and LP_11a modes using spatial light modulators (SLM1 and SLM2) and combined by a polarization beam splitter (PBS). The SLMs employed in the experiment are Holoeye PLUTO phase-only SLMs based on reflective liquid crystal on silicon (LCOS) microdisplays enabling$0\text{~2}\pi$phase modulation at 1550 nm. These SLMs have a spatial resolution of $1920\times 1080$ pixels and a small pixel pitch size of 8$\mu m$. After that, the combined light is focused by a 10X objective lens (OL) with a working distance of 10 mm and coupled into the EC-FMF for short-reach transmission.

 

Fig. 3 Experimental setup of downstream transmission in a MDM-CAP-PON architecture. OLT: optical line terminal; ODN: optical distribution network; ONU: optical network unit; PC: polarization controller; IM: intensity modulator; AWG: arbitrary waveform generator; EDFA: erbium-doped fiber amplifier; VOA: variable optical attenuator; OC: optical coupler; SMF: single-mode fiber; Col: collimator; SLM: spatial light modulator; HWP: half-wave plate; PBS: polarization beam splitter; OL: objective lens; EC-FMF: elliptical-core few-mode fiber; PD: photodetector; DSP: digital signal processing.

Download Full Size | PDF

 

Fig. 4 (a) The time domain impulse responses of the shaping filter pairs. (b) The frequency domain impulse responses of the shaping filter pairs. (c) The electric spectrum of the CAP-16 signal at the transmitter.

Download Full Size | PDF

As shown in Fig. 5(a), the employed 1.1-km EC-FMF in the experiment has an elliptical core and a low index ring (trench) is added next to the core to improve the performance under bending. Remarkably, the trench-assisted fiber can reduce not only the macro-bending loss [39] but also the micro-bending loss [40]. The bending loss originates from a power leakage that occurs at the mode-field edge, where the effective index of a bent fiber exceeds the equivalent index of a propagating mode in the fiber. The trench in the cladding suppresses the increase of the effective index at the cladding around the edge of the mode field. This insensitivity to bending radius for the trench-index profile fiber is effective for suppressing additional loss by unexpected strong bending occurrences in fiber installation.

 

Fig. 5 (a) Cross-section view of the trench-assisted EC-FMF. (b) The effective refractive indices of four modes in EC-FMF (LP_01x, LP_01y, LP_11ax, LP_11ay). (c) Relative refractive index profile of the EC-FMF along the long axis. (d) Relative refractive index profile of the EC-FMF along the short axis.

Download Full Size | PDF

The geometry parameters and mode properties of the trench-assisted EC-FMF are as follows. The radii of the long and short axes of the ellipse are $r_{long}=7.6\mu m$ and$r_{short}=4.9\mu m$, respectively. The cladding of the fiber is a circle and its radius is$61.4\mu m$. Figures 5(c) and 5(d) show the relative refractive index distribution along the long and short axes of the elliptical core. The relative refractive index difference ($\Delta \text{=}(n_1-n_2)/n_2$) between the fiber core ($n_1$) and cladding ($n_2$) is$\Delta \text{=}0.4768\text{\%}$. Figure 5(b) lists the effective refractive indices of four modes (LP_01x, LP_01y, LP_11ax, LP_11ay) guided in the EC-FMF employed in the experiment. The elliptical core shape of EC-FMF helps to lift the degeneracy of fiber modes. The effective refractive index difference of the LP₀₁ and LP_11a modes in Fig. 5(b) is ~2.23 × 10⁻³, which is beneficial to greatly reduce the mode crosstalk between them. The EC-FMF used in the experiment can only support LP₀₁ and LP_11a modes, while the LP_11b mode is cut off. The mode crosstalk including mode (de)multiplexing and 1.1-km EC-FMF transmission is less than −18 dB and multiple input multiple output (MIMO) digital signal processing (DSP) technique is not used in the MDM transmission experiment.

After the 1.1-km EC-FMF transmission, the light is collimated by another 20X objective lens with a working distance of 2.1 mm. A half-wave plate (HWP) is used to change the polarization of light beam. We split the coupled light into two parts and send them to SLM3 and SLM4 for demodulation. At the ONU side, a photo-detector (PD) is used to detect the optical signal, then the electrical signal is sampled using a real-time oscilloscope (Tektronix DPO72004B) operating at 80 GS/s for offline processing. After digital signal processing, the signals are assigned to users.

4. Experimental results

We characterize in detail the mode transmission performance in the 1.1-km trench-assisted EC-FMF. The phase patterns combined with blazed grating are used for pure generation of spatial modes. Figure 6(a) shows measured intensity profiles of the generated input LP modes. After 1.1-km EC-FMF transmission, the output modes are recorded by a camera (HAMAMATSU InGaAs Camera C10633), as shown in Fig. 6(b). One can clearly see from Figs. 6(a) and 6(b) that LP modes transmission in a 1.1-km EC-FMF is successfully demonstrated with favorable transmission performance. Figure 6(c) displays the intensity profiles of the output LP modes after demodulation by the SLM3 and SLM4.

 

Fig. 6 Measured intensity profiles of the input and output LP modes in EC-FMF. (a) Intensity profiles of input LP modes. (b) Intensity profiles of output LP modes. (c) Intensity profiles of output LP modes after demodulation by the SLMs.

Download Full Size | PDF

We also measure and evaluate the transmission performance of trench-assisted EC-FMF under extreme bending conditions. Firstly, we measure the output power after demodulation under different fiber bending radii (4.5 mm, 6.3 mm, 12.6 mm) and loops (1, 3, 6, 10), respectively. The measured results for LP₀₁ and LP_11a modes are shown in Fig. 7. The red dashed line represents the situation without bending for reference. Bending-induced relative power variations are recorded under different fiber bending radii and loops. As shown in Figs. 7(a) and 7(b), the measured relative power variations are less than 0.3 dB for LP₀₁ mode and 0.5 dB for LP_11a mode. That is, negligible power variations are observed for the LP₀₁ and LP_11a modes using the trench-assisted EC-FMF. Secondly, we also record the intensity profiles of LP₀₁ and LP_11a modes under different fiber bending radii (4.5 mm, 6.3 mm, 12.6 mm) and loops (1, 3, 6, 10), respectively. As shown in Fig. 8, one can clearly see stable intensity profiles of LP₀₁ and LP_11a modes with negligible changes under different bend radii and loops. The obtained results shown in Figs. 7 and 8 indicate favorable anti-bending performance using the trench-assisted EC-FMF.

 

Fig. 7 Measured relative power variations under different fiber bending radii (4.5 mm, 6.3 mm, 12.6 mm) and loops (1, 3, 6, 10) for (a) LP₀₁ mode and (b) LP_11a mode. Red dashed line: without bending.

Download Full Size | PDF

 

Fig. 8 Recorded intensity profiles under different fiber bending radii (4.5 mm, 6.3 mm, 12.6 mm) and loops (1, 3, 6, 10) for (a) LP₀₁ mode and (b) LP_11a mode.

Download Full Size | PDF

We further measure the BER performance of the MDM-CAP-PON. The measured BER curves for back-to-back (BTB) and MDM-CAP-PON downstream transmission are shown in Fig. 9. Figure 9(a) plots the BER performance for the CAP-16 downstream transmission link with the only LP₀₁ mode and the only LP_11a mode (i.e. single mode transmission without mode crosstalk). The measured OSNR penalties at a BER of $3.8\times {10}^{\text{-}3}$ (7% forward error correction (FEC) threshold) for LP₀₁ mode only and LP_11a mode only are about 1.3 dB and 1.8 dB, respectively. The BER performance for the CAP-16 signals with mode crosstalk (i.e. two modes multiplexing transmission with mode crosstalk) is depicted in Fig. 9(b). The measured OSNR penalties at a BER of $3.8\times {10}^{\text{-}3}$ (7% FEC threshold) for LP₀₁ mode and LP_11a mode are about 2.9 dB and 3.1 dB, respectively. One can observe slight BER performance degradation when two multiplexed LP modes are simultaneously used for transmission. Such performance degradation is mainly due to the crosstalk (less than −18 dB) between the two modes. Remarkably, since the designed and fabricated EC-FMF effectively separates the LP₀₁ mode and LP_11a mode (relatively large effective refractive index difference) with relatively low-level mode crosstalk, the demonstrated MDM-CAP-PON shows favorable operation performance even without using MIMO DSP technique.

 

Fig. 9 Measured BER and constellation performance for MDM-CAP-PON. (a) BER performance for downstream transmission link with the only LP₀₁ mode and the only LP_11a mode; (b) BER performance for the CAP-16 signals with mode crosstalk.

Download Full Size | PDF

5. Discussion

In the experimental setup, we can simply use a beam splitter (BS) (non-polarizing) to combine the LP₀₁ and LP_11a modes together. Although it is straightforward, an extra 3-dB insertion loss is introduced. To increase the power budget and reduce the total insertion loss, we employ a PBS in the experiment. Meanwhile, polarization controllers (PCs) are used to select one polarization (e.g. x polarization) from LP₀₁ and the other polarization (e.g. y polarization) from LP_11a. Although such configuration employing PBS borrows the idea of both mode diversity and polarization diversity, it is actually applicable to conventional MDM transmission only using mode diversity.

Additionally, the transmission capacity can be further increased by fully utilizing both mode diversity and polarization diversity. Remarkably, as shown in Fig. 5(b), since the x and y polarizations are still close to each other (△n<10⁻⁵), MIMO DSP is required when employing four modes (LP_01x, LP_01y, LP_11ax, LP_11ay) multiplexing transmission. In the present experiment, we do not fully use mode diversity and polarization diversity for simplicity (MIMO-free MDM). With future improvement, we may also consider both mode diversity (MDM) and polarization diversity (polarization-division multiplexing (PDM)) assisted by MIMO DSP technique.

As well known, PON is a technology viewed by many as an attractive solution to the last mile problem. A PON is a point-to-multipoint optical network with no active elements in the signals’ path from source to destination. In a PON system there should be no in-line amplifiers. In the proof-of-concept experiment, the coupling loss from free space to fiber is relatively large. Hence, we use a pre-amplifier to burst the signal power in the OLT side so that ODN and ONU can remain passive. The relatively large coupling loss is due to mode profile mismatch between free space and fiber, which, in principle, can be further reduced to a low level via optimized precise coupling. “BER vs OSNR” evaluation method is used to characterize the system performance. Remarkably, the in-line amplifiers are not necessary devices in the proposed PON system. Actually, by reducing the system loss (coupling, transmission, etc.) and employing “BER vs Power” evaluation method, the presented architecture can be, in principle, applied to a PON system without using any in-line amplifiers.

The EC-FMF employed in the experiment is 1.1 km long. The transmission loss of the employed EC-FMF was measured to be ~0.2 dB/km. Tens of km EC-FMF will introduce several dB more transmission loss. Moreover, the EC-FMF employed in the experiment only supports greatly separated LP₀₁ and LP_11a modes with large effective refractive index difference (△n>10⁻³). Hence, the mode crosstalk between LP₀₁ and LP_11a modes is effectively suppressed to a low level. Tens of km EC-FMF might still work but the mode crosstalk will increase. We believe several km long fiber transmission should also work well with negligible performance degradation due to the slightly increased transmission loss and mode crosstalk. Tens of km or even longer fiber transmission may suffer performance degradation due to the non-negligible accumulated transmission loss and mode crosstalk, which however, could be improved by further optimizing the fiber design and fabrication with lower transmission loss and larger separation between LP₀₁ and LP_11a modes.

For MDM transmission with mode crosstalk, MIMO DSP is required to mitigate the crosstalk. MIMO-free MDM transmission is applicable only when the mode crosstalk is suppressed to a low level. To enable MIMO-free MDM, we employ an EC-FMF in the experiment. Remarkably, for conventional round-core FMF supporting LP₀₁ and LP₁₁, where LP₁₁ contains almost degenerated LP_11a and LP_11b with strong mode crosstalk, MIMO DSP is needed to mitigate the mode crosstalk. In contrast, EC-FMF has separated LP_11a and LP_11b modes with relatively large effective refractive index difference [22]. The LP₀₁ mode is even separated from LP_11a and LP_11b modes. Hence, EC-FMF can enable MIMO-free MDM transmission of LP₀₁, LP_11a and LP_11b modes, which is difficult to implement using a round-core FMF. In particular, the practically fabricated EC-FMF employed in the experiment only supports LP₀₁ and LP_11a modes with large effective refractive index difference (△n>10⁻³) and the LP_11b mode is cut off. Thus, MIMO-free MDM transmission of LP₀₁ and LP_11a modes is achievable using the fabricated EC-FMF. Note that the round-core FMF always supports both LP_11a and LP_11b modes as they are almost degenerated with small effective refractive index difference. To sum up, EC-FMF can support MIMO-free MDM of LP₀₁, LP_11a and LP_11b modes. EC-FMF can also only support LP₀₁ and LP_11a modes (with LP_11b cut off) for MIMO-free MDM. By comparison, those MIMO-free MDM operations are difficult to be realized using round-core FMF which always simultaneously supports almost degenerated LP_11a and LP_11b modes. From this point of view, for MIMO-free MDM transmission EC-FMF shows some advantages over round-core FMF.

Remarkably, MDM technique employing the space domain of light waves is fully compatible with other well-explored physical dimensions. That is, MDM is fully compatible with different kinds of advanced modulation formats including CAP (complex amplitude) and WDM (wavelength). In this scenario, it is possible to combine MDM-CAP-PON with WDM. Actually, WDM-CAP-PON have already been reported with favorable performance [30, 31, 41]. For example, in [30], a 8-channel 10GHz-spaced UDWDM-PON system was demonstrated with high downstream data rate of 10Gb/s per user over 42.5km SSMF transmission employing CAP modulation and direct detection. In [31], a WDM-CAP-PON based on optical single-side band (OSSB) multi-level multi-band CAP was demonstrated with 11 WDM channels, 55 sub-bands, for 55 users with 9.3-Gb/s per user in the downstream over 40-km SMF. In [41], a WDM PON using CAP modulation was demonstrated with 4 WDM channels and 224-Gbps 32-CAP 60-km transmission for a 64-ONU. Hence, it is also possible to combine WDM with MDM for CAP modulation signals to further increase the number of users in a PON architecture.

6. Conclusion

We propose and experimentally demonstrate a novel MDM-CAP-PON architecture for multiple-user access network. Two spatial modes (LP₀₁ and LP_11a) in an EC-FMF are employed for MDM. CAP-16 signals are adopted, which are generated by the AWG with 10-GSa/s. In the experiment, different OLTs are dealt with different filter pairs to combine CAP signals and transmitted by different spatial modes in the EC-FMF. As a proof of concept, the experiment successfully demonstrates two spatial modes carrying CAP-16 signals for four users with net data rate of 5-Gb/s per user in the downstream transmission over a 1.1-km trench-assisted EC-FMF without using MIMO DSP technique. The trench-assisted EC-FMF shows favorable transmission performance against fiber bending. Negligible power variations and stable mode intensity profiles under extreme bending conditions are observed in the experiment. The mode crosstalk including mode (de)multiplexing and 1.1-km EC-FMF transmission is measured to be less than −18 dB. With future improvement, the transmission capacity and the number of end-customers of the MDM-CAP-PON could be further improved by increasing the level or dimension of the CAP signals and the number of separated spatial modes supported in the EC-FMF.