W2A.

22 OFC 2023 © Optica Publishing Group 2023

600-Gbit/s/λ Mode-multiplexed Bit-loading DMT Signal

Transmission for Short-reach Optical Interconnect
Xinkuo Yu, Jianping Li*, Yuwen Qin*, Ou Xu, Meng Xiang, Songnian Fu
1
Institute of Advanced Photonics Technology, School of Information Engineering, Guangdong University of Technology, Guangzhou, 510006,
China
2
Guangdong Provincial Key Laboratory of Information Photonics Technology, Guangdong University of Technology, Guangzhou, 51006, China
Author e-mail address: [email protected]; [email protected]

Abstract: Single wavelength 600-Gbit/s bit-loading MDM-DMT signal transmission over 20m
OM2 fiber with BER below SD-FEC threshold (2.7×10-2) is demonstrated without MIMO-DSP
for mode de-multiplexing. This report shows the potential for future large-capacity short-reach
optical interconnects.
OCIS codes: (060.2330) Fiber optics communications; (060.2360) Fiber optics links and
subsystems

1. Introduction
With the broad applications of online services such as 5G, Internet of Things (IoT), and cloud computing, global
data traffic is exploding. Optical transport systems and networks, as the key basis of the global communication
infrastructure, are facing unprecedented challenges. Many advanced technologies, such as wavelength division
multiplexing (WDM), polarization division multiplexing (PDM), higher-order modulation with coherent detection,
are proposed to make the capacity of single-mode fiber (SMF) based systems rapidly approach the nonlinear
Shannon’s limit [1]. To break through the bottleneck, the space division multiplexing (SDM) has been regarded as
the potential solution. Mode division multiplexing (MDM), as one type of SDM technologies, opens an opportunity
to short-haul optical interconnects in terms of the widely deployed multimode fiber (MMF) in the existing networks
[2, 3]. Recently, some demonstrations of MDM over MMF using coherent detection have been reported [4, 5]. To
meet the demand for cost-effective and low power consumption of the short-haul interconnects, the intensity
modulation with direct detection (IM/DD) combined with MDM and digital signal processing (DSP) technology has
been widely demonstrated [6-10]. Among them, K. Benyahya et al have done a lot of research on high-speed IM/DD
MDM transmission system based on multiplane light conversion technology [6, 7]. L. Grüner-Nielsen et al have also
demonstrated 20 Gbit/s MDM on Standard 50/125 µm MMF using Photonic Lanterns [8]. In our previous works, 4λ
× 405Gbit/s data transmission utilizing discrete Fourier transform spread discrete multi-tone (DFT-S DMT)
modulation were implemented [10].
In this paper, we experimentally demonstrate 600-Gbit/s/λ data transmission over 20m OM2 fiber using mode-
multiplexed bit-loading DMT modulation. With simple time domain Volterra equalization (TDVE) for nonlinear
compensation, 1.11 dB received optical power (ROP) gain can be obtained under the 20% soft-decision forward
error correction (SD-FEC) criterion (2.7E-2) [11]. To the best of our knowledge, it is the first report to carry out up
to 600-Gbit/s/λ data transmission over standard MMF based on mode-multiplexed bit-loading DMT modulation
without multiple in multiple out (MIMO) de-multiplexing. The results have also shown the potential in the large-
capacity short-reach optical interconnects, such as the rack-to-rack optical interconnects within data centers.
2. Experimental Setup
Experimental setup and DSP diagram of MDM transmission system is shown in Fig. 1. At the transmitter, the Chow
Cioffi Bingham (CCB) algorithm [12] is adopted to allocate the right bits to different subcarriers according to the
system response obtained by training sequence (TS) with quadrature phase shift keying (QPSK) modulation format
in the system calibration stage. Note that the TS with QPSK signal can obtain a more accurate channel response[13].
An efficient frame packed with 430 data symbols, 30 TS symbols and their complex conjugate is encoded onto 1024
subcarriers. Then 1024-point inverse discrete Fourier transform (IDFT), 32-point cyclic prefix (CP) padding,
parallel-to-serial conversion (P/S) and hard clipping are sequentially executed. The DMT signal generated off-line is
fed into an arbitrary waveform generator (AWG, Keysight M8194A) with a sampling rate of 115.2 GSa/s to perform
digital analog conversion (DAC). The electric signal out of DAC is amplified to 16dBm by an electrical amplifier
(EA) with 3-dB bandwidth of 30GHz. The optical carrier from an external cavity laser (ECL) operating at 1550nm
with 15.5 dBm power is modulated by a 30GHz bandwidth Mach-Zehnder modulator (MZM) which is driven by
16dBm electric DMT signal. Then the generated optical signal is amplified by the high-power erbium doped optical

Disclaimer: Preliminary paper, subject to publisher revision

Authorized licensed use limited to: FUDAN UNIVERSITY. Downloaded on December 04,2024 at 10:27:50 UTC from IEEE Xplore. Restrictions apply.
 W2A.22 OFC 2023 © Optica Publishing Group 2023

Coupler
DAC HP- (i)

Conjugate

1:3
TS MZM

Mapping
@115.2GS/s EDFA LP01

QPSK-
Bits EA
S/P
VOA

1024 pt. IDFT

Hard Clipping

DL
Adding CP
ECL@1550nm

P/S

PC
(ii) LP01 LP11b LP21a

Encapsulate
-24.5 -25
Bit/Power-

Data
Mapping
-22.9

Frame
LP11b
Load

Bits -21.6 -20.8 -20.6

S/P

-18.4 -20

Crosstalk(dB)

Mode
Mux
-15

-10

OM2 Fiber
-5
Channel Estimation

N/A N/A N/A 0

SNR Calculation

FD-Equalization

Synchronization
Optional TDVE
LP21a
Removing CP
1024 pt. DFT

LP01 LP11b LP21a

De-Mapping

Data
Input mode
Bits
ADC Mode
PD
@160GS/s DeMux
VOA

Fig. 1. Experimental setup and DSP diagram of MDM transmission over OM2 fiber. (i) The pattern of selected modes and (ii) inter-mode
crosstalk among them.

fiber amplifier (HP-EDFA) and split into 3 uncorrelated optical signal paths to realize the MDM transmission. The
entire MDM fiber link consists of a mode multiplexer, 20m OM2 fiber and a mode de-multiplexer. In the
experiment, the LP01, LP11b and LP21a modes are selected as independent channels to implement the high speed
and large capacity data transmission experiment. Insets (i) and (ii) shown in Fig. 1 display the pattern of selected
modes, and the inter-mode crosstalk among them respectively. The mode isolation of used Mux/DeMUX (Cailabs)
is less than -18dB that allows us to achieve MDM transmission without MIMO processing. At the receiver side, a
PD with a 3dB bandwidth of 50GHz successively performs the photoelectric conversion of bit-loading DMT signals
on different modes. Note that the PD used the experiments lacks the trans-impedence amplifier (TIA). And a
variable optical attenuator (VOA) is used to adjust the received optical power. Then, the detected electric signal is
sampled by Lecory real time oscilloscope with 160GSa/s and processed offline in Matlab. The offline DSP
procedure includes synchronization, optional TDVE, removing CP, 1024-point DFT, channel estimation, 1-tap
frequency domain equalization (FDE) with and de-mapping. Here, a simple optional TDVE are employed to
mitigate nonlinear effects, which may be induced from the cosine modulation function of MZM, the electrical
amplifier, clipping of the signal or the signal-signal beating interference (SSBI). The model of m-th order TDVE is
given by
L L L L L L
1
(1) 2 2 m m m

y n   h  l1 x  n  l1     h  l1, l2 x  n  l1 x  n  l2         h  l1, l2， lm x  n  l1  x  n  l2  x  n  lm 

l1  L1 l1  L1 l2 l1 l1  L1 l2 l1 lm lm1
where x(n) and y(n) are the input and output signal, respectively. Lm is the memory length and h(l1, l2,…, lm) is the
m-th order Volterra kernel. In this paper, we use 3rd order TDVE, and L1, L2 and L3 are set to 30, 5 and 2
respectively after sweep optimization.
3. Result and Discussion
8
210 The optimum point (b) 25
(a) 210 (c) 7
Max data rate(Gbit/s)

The optimum point

Max data rate(Gbit/s)

205 20
6
200 205 15
5
SNR(dB)

10
Bit

195 4
200
190 5 3
0 2
185 195
-5 1
180
190 -10 0
200 300 400 500 600 700 800 0 5 10 15 20 25 30 0 10 20 30 40 50
AWG output amplitude(mV) Clipping rate (%) Frequency (GHz)
Fig. 2. (a) Max transmission data rate vs. AWG output amplitude, (b) Max transmission data rate vs. clipping rate and (c) Channel
response and bit allocation for LP01 mode channel.

Disclaimer: Preliminary paper, subject to publisher revision

Authorized licensed use limited to: FUDAN UNIVERSITY. Downloaded on December 04,2024 at 10:27:50 UTC from IEEE Xplore. Restrictions apply.
 W2A.22 OFC 2023 © Optica Publishing Group 2023

In the system calibration stage, we first take the LP01 mode as an example to optimize the system parameters based
on the used devices to achieve the best performance. According to the results shown in Fig.2 (a), the optimum Vpp of
AWG output amplitude of 600mV is chosen to realize the maximum data rate. Meanwhile, in view of the higher
peak-to-average power ratio (PAPR) of bit-loading DMT signals, the impact of hard clipping on system
performance is shown in Fig. 2(b) and then 5% hard clipping is executed in this paper. Fig.2 (c) shows the channel
response of the band limited MDM transmission system and allocation of 1834 bits using the CCB algorithm. Then
the total data rate can be reach to 1834/102432/33115.23 = 600-Gbit/s/λ.
The BER performances vs. ROP with and without TDVE of the MDM transmission for LP01, LP11b and LP21a
modes have been shown in Fig. 3. With the simple TDVE for nonlinear compensation, 0.58dB, 0.73dB and 1.11dB
ROP gains can be obtained for the three mode channels under 20% SD-FEC criterion, respectively. Meanwhile, it is
noted that, compared with B2B transmission scenario, LP01, LP11b and LP21a mode channel show a power penalty
of 0.37dB, 1.23dB and 1.64dB respectively. This is attributed to the higher-order mode channel has relatively lower
mode isolation. Fig. 4 displays the received bit-loading DMT signal constellations of LP21b mode with ROP of
7dBm, including 128QAM, 64QAM, 32QAM, 16QAM 8QAM, QPSK and BPSK. The experimental results indicate
that the MDM based transmission system has the potential to upgrade to 800G/1.6T ultra-short reach data center
interconnect (DCI) without replacing the deployed multimode fiber.
B2B W/ TDVE B2B W/ TDVE B2B W/ TDVE
LP01 W/O TDVE (a) LP11b W/O TDVE (b) LP21a W/O TDVE (c)
LP01 W/ TDVE LP11b W/ TDVE LP21a W/ TDVE
0.1 0.1 0.1

0.58dB

BER
BER

BER

0.73dB 1.11dB
FEC-limit FEC-limit FEC-limit

1.23dB 1.64dB
0.37dB
0.01 0.01 0.01
3 4 5 6 7 8 9 10 3 4 5 6 7 8 9 10 3 4 5 6 7 8 9 10
Received optical power(dBm) Received optical power(dBm) Received optical power(dBm)
Fig. 3. BER vs. ROP with/ without TDVE for (a) LP01, (b) LP11b and (c) LP21a mode.

Fig. 4. Received bit-loading DMT signal constellations of LP21a mode with ROP of 7dBm
4. Conclusion
In this paper, we experimentally demonstrate a total data rate up to 600-Gbit/s/λ transmission over conventional
OM2 fiber. Enabled by the simple nonlinear compensation with TDVE, a 1.11dB ROP gain can be obtained under
the 20% SD-FEC (2.7E-2) criterion. To the best of our knowledge, it is the first report to achieve up to 600-Gbit/s/λ
data throughput over standard MMF using bit-loading DMT modulation and IM/DD. The proposed scheme would
be a promising candidate for future 800G/1.6T DCI.
5. Acknowledgements
This work is partly supported by National Key R&D Program of China (2018YFB1800901), the National Natural Science Foundation of China
(62022029), the Guangdong Introducing Innovative and Entrepreneurial Teams of “ The Pearl River Talent Recruitment Program”
(2019ZT08X340) and Guangdong Guangxi Joint Science Key Foundation (2021GXNSFDA076001)

6. References
[1]. R. J. Essiambre, et al., Journal of Lightwave Technology, 28, [7]. K. Benyahya, et al., OFC (2017), paper M2D.2.
662-701 (2010). [8]. L. Grüner-Nielsen, et al.,OFC (2021), paper W6A.41.
[2]. B. J. Puttnam, et al., Optica 8, 1186-1203 (2021). [9]. D. Zou, et al.,OFC (2020), paper W1D.6.
[3]. P. J. Winzer, Nature Photonics 8, 345-348 (2014). [10]. J. Li, et al., IEEE Journal of Selected Topics in Quantum
[4]. G. Rademacher, et al., Nature Communications 12, 4238 Electronics 26, 1-8 (2020).
(2021). [11]. Y. Gao, et al.,OFC (2022), paper M4H.2.
[5]. G. Rizzelli, et al., OFC (2022), paper W1G.2. [12]. P. S. Chow, et al., IEEE Transactions on Communications 43,
[6]. K. Benyahya, et al., Journal of Lightwave Technology 36, 773-775 (1995).
4174-4180 (2018). [13]. F. Li, et al., Optics Express 22, 22962-22967 (2014).

Disclaimer: Preliminary paper, subject to publisher revision

Authorized licensed use limited to: FUDAN UNIVERSITY. Downloaded on December 04,2024 at 10:27:50 UTC from IEEE Xplore. Restrictions apply.