Skip Content
You are currently on the new version of our website. Access the old version .
PhotonicsPhotonics
  • Article
  • Open Access

17 July 2024

Wavelength Dependence of Modal Bandwidth of Multimode Fibers for High Data Rate Transmission and Its Implications

,
,
,
,
and
Corning Incorporated, 1 Riverfront Plaza, Corning, NY 14831, USA
*
Author to whom correspondence should be addressed.
This article belongs to the Special Issue New Perspectives in Optical Design

Abstract

Vertical-cavity surface-emitting laser (VCSEL)-based transmission over multimode fiber (MMF) has achieved data rates of 100G per lane and is progressing towards 200G per lane. Recently, high-data-rate MMFs derived from OM3 and OM4 have been proposed. These fibers exhibit higher effective modal bandwidths at 910 nm, leading to a different wavelength dependence compared to conventional OM3 and OM4 MMFs. Understanding the wavelength dependence of these fibers is crucial to address their utilization in a broader range of applications. Through Monte Carlo simulations, we have obtained the low-end boundary of the effective modal bandwidths (EMBs) for these fibers, revealing capability improvements over the existing OM3 and OM4. The high-data-rate OM4 performs the same as or better than OM5 from 840 nm to 920 nm, while also showing a high bandwidth for the 850–870 nm wavelength window, favoring VCSELs with center wavelengths shifted toward 860 nm. We also obtained the link bandwidth, which includes both modal bandwidth and chromatic dispersion contributions, and the transmission reaches for various types of transceivers. We find that for both high-data-rate OM3 and high-data-rate OM4, the link bandwidth stays above the value at 850 nm until around 910 nm, delivering a similar transmission performance from 850 to 910 nm without declining towards longer wavelengths, unlike the standard OM3 and OM4. This characteristic favors a wider range of wavelength choices for VCSELs and enables optimal deployments for various applications.

1. Introduction

Multimode fibers (MMFs) have been extensively employed for short-reach communications, such as within buildings, campuses, or data centers [1,2]. MMFs, used with vertical-cavity surface-emitting lasers (VCSELs), form a system suitable for short-reach applications spanning distances up to 100 m or more. VCSELs offer cost-effective manufacturing and a lower power consumption when compared to traditional single-mode lasers, making them an economical and energy-efficient choice for short-range communications. In contrast to single-mode fibers, MMFs boast a large core diameter and high numerical aperture, enabling the use of lower-cost VCSEL light sources and connectors between fibers.
The past decade has witnessed a rapid advancement in optical transmission technology, fueled by the explosive growth in data traffic and the soaring demand for cloud computing services in data centers [3]. The single-lane data rates have undergone a remarkable evolution, transitioning from 10G to 25G and further to 100G. The rise of artificial intelligence (AI) and graphics processing unit (GPU)-based computing has indeed catalyzed this acceleration in data rate increases, and the latest GPU architectures are now paving the way towards 200G speeds [4].
To achieve higher overall data rates for transceivers, several approaches have been adopted, including higher data rates per lane, higher modulation formats such as four-level pulse amplitude modulation (PAM4), parallel optics utilizing multiple fibers, and wavelength division multiplexing (WDM). In particular, WDM technology has emerged over the past decade in two types of transceivers. One such transceiver is the bidirectional (BiDi) transceiver, named for its capability of transmitting data in both directions [5]. It primarily utilizes two wavelengths, nominally around 850 nm and 910 nm. The data rate of such transceivers has reached 100G per lane, enabling transceiver-level data rates of 800G and 1.6 T/s [6]. Another competing technology is short-wavelength division multiplexing (SWDM) [7,8], which leverages four wavelengths, nominally around 850 nm, 880 nm, 910 nm, and 940 nm. To date, SWDM transceivers have been offered with 40G and 100G data rates, with each wavelength operating at 10 Gb/s and 25 Gb/s, respectively.
VCSEL technology has advanced significantly over the years, with 100G and 200G transmission demonstrated [9,10,11,12,13]. As the data rate moves to 100G per lane, as standardized by IEEE 802.3db [14], the transmission reach limitation becomes more severe. For data rates based on 25Gbaud or lower, OM3 and OM4 MMFs have been widely used for transmission reaches of 70 m and 100 m, respectively. Many data centers have been designed around the transmission reach capabilities. Starting at 100G per lane data rate, for 400G SR4 applications, the OM3 transmission reach has been reduced to 60 m from 70 m. In Terabit BiDi applications, such as 800G SR4.2, standard OM3 and OM4 fibers can only reach distances of 45 m and 70 m, respectively. Only OM5 fiber [15,16], mainly defined for SWDM applications, is specified for 100 m. In response to the bandwidth limitation for many existing MMFs, we have proposed a sub-category of OM3 and OM4 fibers that can achieve longer transmission reaches than standard OM3 and OM4 fibers at high data rates (HDRs) [17,18,19,20]. In this work, we refer to these fibers as ‘HDR OM3’ and ‘HDR OM4’, both of which have been launched as products [21]. One version of the proposed MMFs, ‘HDR OM4’, has been adopted by another fiber manufacturer more recently with the same 850/910 nm EMBs [22]. The ‘HDR MMFs’ are specified to have effective modal bandwidth (EMB) at both 850 nm and 910 nm to address the needs of these two wavelengths. For wider use in WDM applications, the VCSEL wavelengths are not limited to these two wavelengths. Therefore, it is necessary to understand the wavelength dependence of these fibers to determine the transmission reaches for various wavelengths that may be used, which motivated the current work.
In this paper, we conduct detailed modeling to determine the wavelength dependence of EMB for ‘HDR MMFs’. A Monte Carlo simulation is used to generate the refractive index profiles of 10,000 MMFs with a range of variations. The EMB values of these fibers at various wavelengths are calculated. Using the worst-case values, we can establish the low-end boundary of these fibers. By utilizing the EMB values, we are able to further calculate the link bandwidth, which includes both the contributions from EMB and chromatic dispersion (CD) interaction with the laser linewidth. The system reach can be determined for a wavelength of interest based on a required link bandwidth value. In Section 2, we present the information on new ‘HDR MMFs’, i.e., their EMB values that define such types of fibers. In Section 3, we provide details of the Monte Carlo simulation of 10,000 MMFs. A comparison is made with known EMB values for OM3 and OM4. In Section 4, we demonstrate the link bandwidth using the wavelength dependence of EMB and CD of the MMFs. We also convert the link bandwidth into transmission reach for a few transmission applications to illustrate how we can benefit from the wavelength dependence of the EMB information. Finally, we draw conclusions in Section 6.

2. Multimode Fibers for High Data Rate Transmission

MMFs with a 50-micron core have been predominantly used in data communication systems. These fibers are categorized as OM2, OM3, and OM4, based on their effective modal bandwidth (EMB) at a wavelength of 850 nm. The abbreviation ‘OM’ stands for ‘optical multimode’. The OM3 fiber has an EMB of 2000 MHz·km, while the OM4 fiber has a higher EMB of 4700 MHz·km at 850 nm. They were standardized in 2002 and 2009, respectively. OM5 is a sub-category of OM4, sharing the same EMB at 850 nm but with an additional EMB requirement of 2470 MHz·km at 953 nm, which is the longest wavelength supported by shortwave-wavelength division multiplexing (SWDM) technology [8]. OM5 was first specified by Telecommunications Industry Association(TIA) in 2016 as TIA-492AAAE [15], followed by International Electrotechnical Commission(IEC) in 2017 [16].
In recent years, the data transmission rate of VCSEL-based transceivers has reached 25 Gbaud. Among these transceivers, 100G SR4 and 100G SWDM utilize a 25G non-return-to-zero (NRZ) modulation format, while 100G BiDi transceivers employ a 50G PAM4 modulation format across two wavelengths. Meanwhile, higher speed 400G transceivers, such as 400G SR8 and 400G SR4.2, all utilize the PAM4 modulation format.
As mentioned in the Introduction, to address the needs of high data rate transmission, we have proposed the concept of ‘HDR MMFs’ for future-ready 100G-per-lane multimode applications. Essentially, ‘HDR MMFs’ are defined by re-categorizing a subset of OM3 or OM4 fibers. The EMB values of ‘HDR MMF’ are specified in Table 1 and are defined at two wavelengths of interest: 850 nm and 910 nm for 100G-per-lane transmission. Therefore, these fibers can support VCSEL transceivers involving either an 850 nm wavelength only or both 850 nm and 910 nm wavelengths, with targeted transmission reaches of 80 m and 100 m, respectively.
Table 1. EMB values of ’HDR MMFs‘ for 100G-per-lane applications.

3. Wavelength Dependence of EMB for ‘HDR MMF’ through Monte Carlo Simulation

3.1. Generation of Refractive Index Profiles for Monte Carlo Simulation

The relative refractive index profile, or delta profile of an MMF with a graded-index profile, is often referred to as the alpha profile. It takes the following form:
r = 0 1 r / a α ,
where Δ0 is the relative refractive index in the center of the core defined as Δ 0 = ( n 0 2 n 1 2 ) / ( 2 n 0 2 ) , wherein n 0 is the refractive index in the center of the core, n 1 is the refractive index of the cladding, a is the core radius, and α is the profile shape parameter, also called the alpha parameter or the alpha shape parameter. As illustrated in Figure 1, the fiber has a 1% core delta, and a ~50 µm core diameter. Note that when the alpha value is two, the profile takes a parabolic shape. The MMF is most optimized when the alpha parameter has a value of around 2.1, at which point the MMF’s modal bandwidth peaks at a wavelength near 850 nm. A trench with a negative delta value is placed outside the core to enhance the bending performance of the fiber [23]. The cladding is typically pure silica and is used as a reference with zero delta value.
Figure 1. Delta profile of an MMF with 50 μm-diameter core.
Due to the nature of MMF fabrication, each individual MMF differs from another. The idea behind the current work is to generate an ensemble of MMFs with a range of variation types that can occur in real fibers. Figure 2 schematically shows the wavelength dependence of modal bandwidth for an individual MMF. The MMF follows a curve that can reach a peak or maximal modal bandwidth at a wavelength called the peak wavelength. Each MMF can have its own maximum modal bandwidth value and peak wavelength value. Therefore, they can have different modal bandwidth values at a given wavelength of interest, such as 850 nm. An MMF with an ideal refractive index profile can have a higher maximum bandwidth than other MMFs, but its peak wavelength can vary over a range, resulting in different wavelength dependence for each individual MMF. Below, we describe several types of variations that can affect the fiber modal bandwidth and its wavelength dependence.
Figure 2. The schematic of wavelength dependency of modal bandwidth of an individual MMF.
One of the most important parameters to consider is the alpha parameter, as it determines where the fiber modal bandwidth peaks, as illustrated in Figure 2. When the alpha parameter is altered, the wavelength at which the modal bandwidth peaks also changes. Therefore, to emulate actual fiber, we would alter the alpha value within a range of over 200 nm. Figure 3 shows the relationship between alpha and peak wavelength. When the alpha value is decreased, the peak wavelength shifts to a higher value.
Figure 3. The peak wavelength as a function of alpha for MMF with 1% delta and 50 μm core diameter.
The second type of variation is the fluctuation of the alpha profile itself on top of the perfect alpha profile, which is referred to as non-alpha error. In the modeling, we expand such errors using Fourier series, as follows:
n r = i = 1 N a i · sin i π r a + b i · cos i π r a .
In our actual generation of the Monte Carlo data, N is chosen to be six.
Another type of error that can occur in the center portion of the core involves a small spike or dip in the refractive index, leading to split pulses, which can degrade system performance [24]. In our modeling, we introduce such errors within the first several microns of the core radius. One last type of error to consider is the variation of the trench. In this case, we generate random offsets in the Monte Carlo data set. Alternatively, one can alter the depth of the trench to achieve similar performance variations.

3.2. Wavelength Dependence of EMB of MMFs

Using the refractive index profiles, the EMB of a particular fiber can be obtained from its differential mode delay (DMD) chart, which is the sequence of output pulses resulting from single-mode pulse launches at various radial offsets at the input. In reality, the DMD chart is measured using a DMD bench, while, in our case, it is calculated based on the refractive index profile [25]. In the standard, 10 VCSEL launch conditions were used [26], as the modal bandwidth depends on the launch condition. The DMD weight functions associated with the 10 launch conditions are shown in Figure 4. Using the measured DMD chart, the output pulses over ten VCSEL launch conditions are assembled as weighted by the ten DMD weight functions. The modal bandwidth associated with each weight or launch condition can be calculated from the 3 dB drop of the frequency response, S21, and is referred to as calculated effective modal bandwidth (EMBc). The frequency response of a particular launch condition is calculated from the Fourier transform of the input pulse ( P i n ( t ) ) and output pulses P o u t , i t as follows:
S 21 , i = F P o u t , i t F P i n t ,
where F ( P o u t , i t ) and F ( P i n t )denote the Fourier transform operations of the output pulse and input pulse respectively. The range of launch conditions ensures that essentially all conditions are covered, so the EMB reflects not just the outcome of one launch condition but the modal bandwidth capability of the specific fiber. With the 10 EMBcs obtained, the EMB can be further obtained. We also note that the integration of the DMD weight function over radius, which represents integrated power over radius, is referred to as encircled flux. The standard requires that, at the 4.5 μm radius, the integrated power is less than 30% of the total power and, at the 19 μm radius, the integrated power is over 86% of the total power.
Figure 4. Ten weight functions associated with ten VCSEL launch conditions.
In the Monte Carlo study, we generated 10,000 cases of refractive index profiles to cover a wide range of profile variabilities, as described in Section 3.1, with a mixture of all types of profile errors. The magnitude of each type of error was also randomized within a range to yield MMFs with EMB spread over a range of interest. The purpose of the Monte Carlo set is to generate refractive index profiles with various possibilities that can affect the modal bandwidth under different launch conditions. For each wavelength, we calculated the EMB from the profiles as described above. By combining all the wavelengths, we can generate the wavelength dependence curve of one individual MMF. However, MMFs are primarily categorized as OM3 and OM4, so the interest lies in the low-end values of each type of MMF. IEC 60793-2-10 [16] published the low-end EMB values for OM3 and OM4 over the range of 840 nm to 953 nm. In the simulation presented in the current paper, we have repeated the work to ensure that the methodology yields consistent results. Our main objective, as described in Section 1, is to study the EMB wavelength dependence for ‘HDR MMFs’, which are further broken down into ‘HDR OM4’ and ‘HDR OM3’. For most of the calculations, the low-end EMB values were obtained at the 99.9th percentile, meaning there is a 0.1% chance that the values can fall below the threshold. We believe this is a reasonable practice, as the truly worst case or lowest value can be overly pessimistic and, therefore, not a good representation of the product’s practical capability. Later in the paper, we will discuss the impact of choosing such a percentile to put the results in perspective. In some cases, we will refer to the EMB of a particular type of MMF at its low-end value or ‘worst’ value without explicitly stating that these are the low-end values.
We first show the EMB of OM4 and OM3 in Figure 5a,b, referring to them as ‘IEC OM4’ and ‘IEC OM3’, respectively, as they are based on IEC guidance. The curves with discrete dots were obtained from Monte Carlo simulations for OM4 and OM3. They were labeled as ‘Monte Carlo OM4’ and ‘Monte Carlo OM3’ to differentiate them from the IEC-guided OM4 and OM3 EMBs. As the wavelength moves away from 850 nm, the EMB values for both OM4 and OM3 decrease, because the EMB value at 850 nm can be achieved by MMFs with peak wavelengths higher or lower than 850 nm. The agreement between the IEC guidance and the current Monte Carlo simulations is excellent. Note that the low-end EMB values in the simulation were obtained from the 99.9th percentile instead of using the 100th percentile or absolute lowest value, in order to be more realistically aligned with real-world situations. The agreement also gives us confidence that the methodology used in the current work would be valid when applied to ‘HDR OM4’ and ‘HDR OM3’.
Figure 5. (a) The EMB vs. wavelength for OM4; (b) the EMB vs. wavelength for OM3.
In Figure 6, we show the wavelength dependence of EMB for ‘HDR OM4’ obtained from the Monte Carlo simulation with a comparison to two related types of MMFs, namely OM4 and OM5. The EMB values for OM4 and OM5 are based on IEC guidance. As mentioned in Section 1, OM5 was established for supporting SWDM-based VCSEL transceivers with the highest wavelength at 953 nm. SWDM is a type of WDM that includes more wavelengths and, therefore, more data traffic within one transceiver, which is in contrast to BiDi transceivers, which use 850 nm and 910 nm. However, in the past decade, BiDi transceivers have seen more widespread use and have now reached the Terabit age [6]. MMFs that can align with the underlying transceivers would allow for more economical deployments of such applications in larger volumes. It can be found that the EMB values of ‘HDR OM4’ are roughly equal to or above those of OM5 from 840 nm to 920 nm. Therefore, it has the same capability as OM5 for the wavelengths of interest for BiDi transceivers. Even for wavelengths from 920 to 953 nm, where the EMB of ‘HDR OM4’ falls below that of OM5, the EMB values are still significantly better than those of OM4, enabling better transmission reach when SWDM wavelengths are used. On the other hand, OM4’s EMB decreases rapidly as the wavelength increases from 850 nm, which explains the limitations encountered for transmission at longer wavelengths, such as those used in BiDi transmission.
Figure 6. The EMB as a function of wavelength for ‘HDR OM4’ as compared to those for OM4 and OM5 defined by IEC.
In Figure 7, we show the wavelength dependence of EMB for ‘HDR OM3’ obtained from the Monte Carlo simulation, as compared to those of OM3 and OM4 based on IEC guidance. The EMBs of ‘HDR OM3’ over the whole wavelength range are higher than those of OM3, since the EMB at 850 nm is chosen to support an 80 m reach for 100G-per-lane transmission. Coupled with the corresponding EMB at 910 nm, the EMB over the entire wavelength range has been increased, not just for the 850 nm and 910 nm wavelengths where the fiber EMBs are defined. Also, when compared to OM4, the EMB of ‘HDR OM3’ becomes higher, starting at 883 nm. The overall higher EMB over the wavelength range can enable ‘HDR OM3’ to achieve a system reach of 80 m for 800G BiDi applications, even exceeding the specified reach for OM4 at 70 m, as will be shown in the next section.
Figure 7. The EMB as a function of wavelength for ‘HDR OM3’ as compared to those for OM3 and OM4 defined by IEC.
In the Monte Carlo simulation presented above in Figure 6 and Figure 7, we have used the 99.9th percentile for obtaining the low-end EMB. We would note that the low-end boundary can be affected by how EMB is distributed. To illustrate how the change in percentile can affect the EMB, we show the results obtained from two thresholds in Figure 8. Some differences are shown for wavelengths between 850 and 900 nm, but smaller differences are observed at higher wavelengths. In particular, in the 99.6th percentile, the EMB of ‘HDR OM4’ exceeds 4700 MHz·km over the range of 850 nm to 870 nm. The actual low-end EMB values can be fine-tuned to align with realistic situations with the support of actual measurement results.
Figure 8. The EMB as a function of wavelength for ‘HDR OM4’ with two different percentiles.

5. Discussion

In this paper, we have studied the wavelength dependence of modal bandwidth for the ‘HDR MMFs’ we have recently proposed [17,18,19,20,21]. We have shown detailed results in terms of EMB and link bandwidth, which are directly tied to transmission performance. The ‘HDR MMFs’ were proposed to address the needs for VCSEL-based 100G-per-lane transmission over a wide wavelength range, with the highest performance from 850 nm to 910 nm. To better understand the benefits as they relate to wavelength, we discuss and highlight the capabilities from the following perspectives.
  • Benefit for the 850 nm wavelength window: By definition, ‘HDR OM3’ and ‘HDR OM4’ are defined at 850 nm and 910 nm wavelengths. However, since we have taken into consideration the limitation of OM3 for 100G-per-lane transmission with a reach of 60 m, we have chosen the 850 nm EMB for ‘HDR OM3’ to support an 80 m reach. This ensures that ‘HDR OM3’ can also benefit 850 nm-only applications. It alleviates the issue of OM3’s reach dropping below 70 m, a distance that many customers have grown accustomed to and around which many data centers have been designed.
  • Addressing the wavelength shifting toward 860 nm in 850 nm window: As the VCSEL data rate increases to 100G per lane and will reach 200G per lane very soon, some implementations of VCSELs have shown a shift in the wavelength, more centered around 860 nm instead of 850 nm. The ‘HDR MMFs’ offer the benefit of having enhanced performance for the entire 850–870 nm wavelength window. Taking into account the CD contribution, the link bandwidth over this wavelength range stays above the level at 850 nm, meeting the high bandwidth demands despite the wavelength shift.
  • Comparison to OM3 and OM4: One major limitation of OM3 and OM4 is that their EMBs decrease rapidly as the wavelength moves away from 850 nm, particularly when moving toward higher wavelengths up to 910 nm. Even though the 850 nm EMBs of these fibers remain adequate in many cases, the limitation at 910 nm restricts their applications for longer wavelength use, including 860 nm-centered applications and BiDi applications involving 850/910 nm dual wavelengths.
  • Enabling VCSELs with more wavelengths: Currently, most VCSELs are implemented at 850 nm and 910 nm, with the exception of SWDM VCSELs, which work up to 953 nm but are so far limited to 100G data rates at the transceiver level. The current study shows that ‘HDR MMFs’ can have uniform link bandwidths from 850 nm to 910 nm. This opens up the feasibility of implementing VCSELs at any wavelength from 850 nm to 910 nm to achieve similar performance.
  • Benefits for SWDM applications: even though SWDM applications have not been deployed in large volumes like other types of VCSEL transceivers and have not reached higher data rates, ‘HDR MMFs’ still demonstrate their advantages over OM3 and OM4 by delivering significantly better transmission reaches, although they trail somewhat behind OM5.
  • Comparison to OM5: Unlike OM5, which was defined based on EMBs at 850 nm and 953 nm, ‘HDR MMFs’ were defined based on EMBs at 850 nm and 910 nm. ‘HDR OM4’ has the same EMBs as OM5 at both 850 nm and 910 nm. The wavelength dependence study in Section 3 shows that ‘HDR OM4’ has an EMB, from 850 nm to 910 nm, equal to or higher than that of OM5 guided by IEC. This implies that ‘HDR OM4’ can cover all the needs of OM5 from 850 nm to 910 nm and can be a more cost-effective and higher-volume solution for a wide range of applications using the relevant wavelengths.

6. Conclusions

VCSEL-based transmission over MMFs has been widely deployed. When the data rate for VCSEL-based transmission reaches 100G per lane, the modal bandwidth of OM3 and OM4 becomes limiting for certain wavelengths. In light of such limitations, we have proposed sub-categories of OM3 and OM4 fibers to address the needs of HDR transmission [17,18,19,20,21]. These fibers exhibit a higher EMBs at 910 nm, leading to a different wavelength dependence compared to conventional OM3 and OM4 MMFs. Understanding the wavelength dependence of these fibers is crucial to address their utilization in a broader range of applications. We adopted Monte Carlo simulations as the primary approach for the current study. We have generated refractive index profiles of 10,000 MMFs. By calculating EMB of each fiber over a range of wavelengths of interest, from 840 nm to 953 nm, we obtained the low-end boundary of the EMB at each wavelength. These values were compared with known cases of OM3 and OM4, revealing the capability differences.
We observe that both ‘HDR OM4’ and ‘HDR OM3’ offer higher EMBs than OM4 and OM3 for wavelengths longer than 850 nm. In particular, our findings indicate that the ‘HDR OM4’ performs the same or better than OM5 from 840 nm to 920 nm. Due to their higher bandwidth capability at longer wavelengths, these fibers are more suitable for high-data-rate transmission. We also obtained the link bandwidth that includes both the modal bandwidth contribution and the CD-contributed bandwidth, which is directly tied to the system level performance. We find that, for both ‘HDR OM3’ and ‘HDR OM4’, the link bandwidth stays above the value at 850 nm until around 910 nm. This means these fibers deliver similar transmission performances from 850 to 910 nm without declining towards longer wavelengths, unlike standard OM3 and OM4. In the application space, we further list the transmission reaches for various transceiver applications, both for 100G-per-lane-based and older 25 Gbaud-based transceivers. To better illustrate various aspects of ‘HDR MMFs’, in the discussion section, Section 5, we presented several comparisons. We believe the current study provides insights about the benefits of high-data-rate MMFs, enabling more VCSEL wavelength choices and optimal deployment in diverse scenarios.

Author Contributions

Conceptualization, X.C., H.D., H.C., Z.D.B. and M.-J.L.; methodology, X.C. and M.-J.L.; validation, X.C. and J.E.H.; formal analysis, X.C.; investigation, X.C. and J.E.H.; writing—original draft preparation, X.C.; writing—review and editing, X.C., H.D., H.C. and Z.D.B.; visualization, X.C. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Data Availability Statement

Data are contained within the article, as shown in the figures and as-sociated descriptions.

Conflicts of Interest

Xin Chen, Hao Dong, Hao Chen, Jason E. Hurley, Zoren D. Bullock and Ming-Jun Li were employed by the company Corning Incorporated. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

References

  1. Chen, X.; Bickham, S.R.; Abbott, J.S.; Coleman, J.D.; Li, M.-J.; Peng, G.D. Multimode fibers for data centers. In Handbook of Optical Fibers; Springer Nature Singapore Pte Ltd.: Berlin/Heidelberg, Germany, 2018; pp. 1–57. [Google Scholar]
  2. Li, M.-J. MMF for High Data Rate and Short Length Applications. In Proceedings of the Optical Fiber Communication Conference, OFC 2014, San Diego, CA, USA, 24–28 March 2024; pp. 1–3. [Google Scholar]
  3. Cisco Annual Internet Report (2018–2023) White Paper (Updated: March 9, 2020). Available online: https://www.cisco.com/c/en/us/solutions/executive-perspectives/annual-internet-report/index.html (accessed on 14 July 2024).
  4. InfiniBand Network Application in DGX Cluster. Available online: https://www.naddod.com/blog/infiniband-network-application-in-dgx-cluster (accessed on 14 July 2024).
  5. Cisco 40GBASE QSFP Modules Data Sheet. Available online: https://www.cisco.com/c/en/us/products/collateral/interfaces-modules/transceiver-modules/data_sheet_c78-660083.html (accessed on 14 July 2024).
  6. Terabit BiDi MSA. Available online: https://terabit-bidi-msa.com/ (accessed on 14 July 2024).
  7. Tatum, J.A.; Gazula, D.; Graham, L.A.; Guenter, J.K.; Johnson, R.H.; King, J.; Kocot, C.; Landry, G.D.; Lyubomirsky, I.; MacInnes, A.N.; et al. VCSEL-based interconnects for current and future data centers. J. Light. Technol. 2015, 33, 727–732. [Google Scholar] [CrossRef]
  8. Shortwave Wavelength Division Multiplexing (SWDM) Alliance. Available online: https://www.swdm.org/ (accessed on 14 July 2024).
  9. Chorchos, L.; Ledentsov, N.; Makarov, O.; Shchukin, V.A.; Kalosha, V.; Turkiewicz, J.P. Multi Aperture High Power 100G Single Mode 850nm VCSEL for Extended Reach 800G Ethernet. In Proceedings of the Optical Fiber Communications Conference and Exhibition (OFC), San Diego, CA, USA, 5–9 March 2023; pp. 1–3. [Google Scholar]
  10. Zuo, T.; Zhang, T.; Zhang, S.; Liu, L. 850-nm VCSEL-based single-lane 200-Gbps PAM-4 transmission for datacenter intra-connections. IEEE Photonics Technol. Lett. 2021, 33, 1042–1045. [Google Scholar] [CrossRef]
  11. Murty, M.V.R.; Wang, J.; Jiang, S.; Dolfi, D.; Wang, T.K.; Vaughan, D.; Feng, Z.; Leong, N.; Sridhara, A.; Joyo, S.T.; et al. Toward 200G per Lane VCSEL-based Multimode Links. In Proceedings of the Optical Fiber Communication Conference (OFC) 2024, San Diego, CA, USA, 24–28 March 2024; Technical Digest Series; Paper M2D.3; Optica Publishing Group: Washington, DC, USA, 2024. [Google Scholar]
  12. Tirelli, S.; Corti, E.; Duda, E.; Pissis, A.; Hönl, S.; Hoser, M.; Paul, M.; Zibik, E. Lithographic Aperture VCSELs Enabling Beyond 100G Datacom Applications. In Proceedings of the Optical Fiber Communication Conference (OFC) 2024, San Diego, CA, USA, 24–28 March 2024; Technical Digest Series; Paper M2D.1; Optica Publishing Group: Washington, DC, USA, 2024. [Google Scholar]
  13. Wettlin, T.; Lin, Y.; Stojanovic, N.; Calabrò, S.; Wang, R.; Zhang, L.; Kuschnerov, M. 200 Gb/s VCSEL transmission Using 60 m OM4 MMF and KP4 FEC for AI computing clusters. arXiv 2024, arXiv:2403.17275. [Google Scholar]
  14. IEEE Standard for Ethernet—Amendment 3: Physical Layer Specifications and Management Parameters for 100 Gb/s, 200 Gb/s, and 400 Gb/s Operation over Optical Fiber using 100 Gb/s Signaling. Available online: https://standards.ieee.org/ieee/802.3db/10283/ (accessed on 14 July 2024).
  15. TIA-492AAAE; Detail Specification for 50-µm Core Diameter/125-µm Cladding Diameter Class 1a Graded-Index Multimode Optical Fibers with Laser-Optimized Bandwidth Characteristics Specified for Wavelength Division Multiplexing. TIA: Arlington, VA, USA, 2016.
  16. IEC 60793-2-10:2019; Optical Fibres—Part 2–10: Product Specifications—Sectional Specification for Category A1 Multimode Fibres. Edition 7.0. IEC: Geneva, Switzerland, 2019.
  17. Dong, H.; Chen, X. The Road to 800G with Multimode Fiber. Corning Incorporated. Available online: https://www.corning.com/catalog/coc/documents/articles/CO8000_Article_Road%20to%20800G%20multimode%20fiber.pdf (accessed on 14 July 2024).
  18. Chen, H.; Wang, C.; Chen, X.; Dong, H.; Bullock, Z. Optimized Multimode Fiber for 100G/land and BiDi technologies. In Proceedings of the 2023 China Communication Academic Annual Conference, Dalian, China, 10–12 August 2023. [Google Scholar]
  19. Chen, X.; Dong, H.; Chen, H.; Mu, J.; Zheng, L.; Yang, S.; Hurley, J.E.; Wood, W.A.; Bullock, Z.; Li, M.-J. Concept of BiDi Optimized OM4 Multimode Fiber for High Date Rate Short Reach VCSEL Transmission. In Proceedings of the 2023 Opto-Electronics and Communications Conference (OECC), Shanghai, China, 2–6 July 2023; pp. 1–5. [Google Scholar]
  20. Chen, X.; Dong, H.; Huang, O.; Mu, J.; Chen, H.; Hurley, J.E.; Bullock, Z.; Li, M.-J. Concept of OM3 Multimode Fiber Optimized for 100G/lane and BiDi Based VCSEL Transmission. In Proceedings of the 2024 Opto-Electronics and Communications Conference (OECC), Melbourne, Australia, 30 June–4 July 2024. Paper #45. [Google Scholar]
  21. Corning® ClearCurve® OM3 XT and OM4 XT Optical Fibers—Product Information. Available online: https://www.corning.com/media/worldwide/coc/documents/Fiber/product-information-sheets/PI-1568.pdf (accessed on 14 July 2024).
  22. LaserWave® Dual-Band OM4+ Multimode Optical Fiber. Available online: https://fiber-optic-catalog.ofsoptics.com/LaserWave-Dual-Band-OM4-Multimode-Optical-Fiber-10033 (accessed on 14 July 2024).
  23. Li, M.-J.; Tandon, P.; Bookbinder, D.C.; Bickham, S.R.; Wilbert, K.A.; Abbott, J.S.; Nolan, D.A. Designs of bend-insensitive multimode fibers. In Proceedings of the 2011 Optical Fiber Communication Conference and Exposition and The National Fiber Optic Engineers Conference, Los Angeles, CA, USA, 6–10 March 2011; pp. 1–3. [Google Scholar]
  24. Bourdine, A.V.; Praporshchikov, D.E.; Yablochkin, K.A. Investigation of defects of refractive index profile of silica graded-index multimode fibers. Proc. SPIE 2011, 7992, 799206. [Google Scholar]
  25. Lenahan, T.A. Calculation of modes in an optical fiber using the finite element method and EISPACK. Bell Syst. Tech. J. 1983, 62, 2663–2694. [Google Scholar] [CrossRef]
  26. IEC 60793-1-49; Optical Fibres—Part 1-49: Measurement Methods and Test Procedures—Differential Mode Delay. Edition 3.0. IEC: Geneva, Switzerland, 2018.
  27. Bottacchi, S. Multi-Gigabit Transmission over Multimode Optical Fibre: Theory and Design Methods for 10 GbE Systems; John Wiley & Sons, Ltd: Hoboken, NJ, USA, 2006. [Google Scholar]
  28. Li, M.-J.; Li, K.; Chen, X.; Mishra, S.K.; Juarez, A.A.; Hurley, J.E.; Stone, J.S.; Wang, C.-H.; Cheng, H.-T.; Wu, C.-H.; et al. Single-Mode VCSEL Transmission for Short Reach Communications. J. Lightwave Technol. 2021, 39, 868–880. [Google Scholar] [CrossRef]
  29. Ledentsov, N.N.; Makarov, O.Y.; Shchukin, V.A.; Kalosha, V.P.; Ledentsov, N.; Chrochos, L.; Sanayeh, M.B.; Turkiewicz, J.P. High Speed VCSEL Technology and Applications. J. Lightwave Technol. 2022, 40, 1749–1763. [Google Scholar] [CrossRef]
  30. Understanding the Differences Between OM4 and OM5 Multimode Fiber. Available online: https://www.cisco.com/c/en/us/products/collateral/interfaces-modules/transceiver-modules/diff-om4-om5-multimode-fiber-wp.html (accessed on 14 July 2024).
  31. Available online: https://blog.siemon.com/infrastructure/om5-no-reach-advantage-cisco-multimode-transceivers#:~:text=Most%20of%20Cisco's%20multimode%20transceivers,realize%20significant%20benefits%20from%20OM5 (accessed on 14 July 2024).
  32. Chang, Y.-C.; Coldren, L.A. Optimization of VCSEL structure for high-speed operation. In Proceedings of the 2008 IEEE 21st International Semiconductor Laser Conference, Sorrento, Italy, 14–18 September 2008; pp. 159–160. [Google Scholar]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.