Multimode fiber (MMF) is increasingly acquiring market share in high-end data centers, replacing the copper-based cabling that has been the predominant media for high speed communication.1 Currently, there are six major areas that affect data rates in such fiber links:
- Fiber design (numerical aperture [NA], core diameter, core and cladding index profile, modal dispersion, and fiber end-face geometry)
- Fiber material (chromatic dispersion, imperfections in the bulk material, and birefringence)
- Maximum modulation frequency/data rate at which the driving electronics and transmitters can be driven to the required extinction (electronics and VCSEL bandwidth)
- Frequency response of the receiver and associated electronics (detector and amplification electronics)
- Fiber connectors (alignment, end-face surface imperfections, and physical contact)
- Fiber deployment geometries (band radii, vibration, tension, and fiber twist).
Here we will present a case for why the industry should make modifications in areas 1 through 4.
MMF: Better than expected
The continuous advancement of data rates in MMF has been due to the common effort of all industry participants (fiber, transceiver, and equipment manufacturers) in cooperation with the respective industry associations (IEEE, ISO/IEC, and TIA for Ethernet protocols and FCIA for Fibre Channel protocols). For example, about 10 years ago vertical-cavity surface-emitting lasers (VCSELs) started to replace LEDs. The VCSEL spectrum is typically less than 2 nm, while that of the LED is typically about 20 nm. LEDs have a wide, spatially smooth, and continuous radial emission pattern; VCSELs produce a spatially structured and narrow pattern (directional). Also, a VCSEL can be driven on/off faster than an LED. The VCSEL’s spectrum and spatial distribution are time dependent.
Meanwhile, the recent OM4 standard specifies fibers with a bandwidth distance product of 4700 MHz.km. These fibers are designed to support emerging 10-Gigabit Ethernet transmission requirements to about 500 m.2
The work of standards bodies in particular has established metrics upon which to standardize and evaluate the physical characteristics of optical communications components. Some of these measurements involve techniques that have been used in other areas of communication, like eye diagrams and bit-error-rate tests (BERTs). Other methods are more ad hoc, such as differential mode delay (DMD) and effective modal bandwidth (EMB).3,4
But even with the combined efforts of many industries, no standard could predict that fibers and transceivers eventually would be manufactured with such quality and refinement and work together so perfectly that some assumptions about the physical phenomena expected to limit optical link performance would become obsolete. But MMF communication has now reached this point.
At the same time, one anomaly became apparent in 2008: Why did data rate characterization methods poorly correlate to the expected performance of MMF optical links? It turned out that each fiber and transceiver vendor qualified their products with whatever test method(s) was (were) available to them. Efforts to calibrate measurement techniques among vendors invariably involved round robin measurements among competitors to ensure that everyone was measuring the same attributes with the same accuracy. Not surprisingly, this proved unsatisfactory.
As a result, the average designer and installer of fiber systems and fiber upgrades were left without a clear answer to a very important question: What are the chances, given the specified fiber in my network, that my design will support the required reach when I switch it on?
At the time, fiber and cable manufacturers were attempting to correlate eye-diagrams, BER, DMD, and EMB. From experience with these correlations, let’s look at what is relevant from the basic physics involved in a MMF link.5
Consider a narrow temporal pulse launched at one end of a multimode fiber that in currently available high-grade MMF will spread to about 100 ps after traveling 500 m. (This is called “impulse response,” widely used in electronics and optics testing measurements.) This is the total impulse response (Tir) for the MMF with components modal (Mir) and chromatic (or material; Cir) impulse responses where
Tir = [Mir2 + Cir2]1/2 (1)
Using a typical VCSEL (under the same conditions) that narrow temporal pulse would spread to 84 ps due only to chromatic effects. In summary we have this: (Tir, Mir, Cir) = (100, 54, 84) ps. The corresponding bandwidths are (BWir, BWir, BWir) = (10, 19, 12) GHz where 1/BW replaces T in (1). Notice that the lower value of Tir (or a higher BWir) correlates to a higher performance fiber.
Tir and Cir have nearly similar magnitudes, which has been experimentally confirmed. It was realized by one of us (GT) that Mir and Cir could be engineered in such a way that much of Cir could be compensated by Mir and then Tir significantly reduced. This was accomplished by using the fact that the VCSEL’s wavelength emission strongly depends on the angle of emission relative to the VSCEL’s optical axis. These physical phenomena have been reported in the past but not completely taken into consideration in the MMF optical links.9,10 Through compensation of this wavelength dependency in the MMF, Tir can be significantly reduced, increasing the bandwidth of the fiber.
What is ahead for MMF optics communications systems?
At this point, most efforts to increase data rates on MMF have focused on increasing the number of lanes that carry data (typically via multiple strands of MMFs and using ferrules that mate several fibers at the same time, such as MTP/MPO connectors) or increasing the modulation speed of the transceivers. The manufacturing and assembly cost of MTP/MPO-terminated cables leads to the conclusion that alternatives to this multi-lane approach need to be considered.
One limiting factor for MMF links that could be dealt with relatively easily is that the same bandwidth distance products are specified for 1300 nm (for backward compatibility) as 850 nm. Limiting the nominal wavelength to 850 nm would result in higher bandwidth distance products – a reasonable move given that 1300 nm is mostly obsolete in high-end data centers.
Yet today, at maximum bandwidth, modal and chromatic dispersion are the major limiting factors in the longest reach OM4 fibers. At this time there are almost no obvious paths toward overcoming these factors to increase fiber bandwidth other than electrical or optical dispersion compensation techniques. To compensate for insertion loss (IL), regenerators could be used to re-time, re-shape, and re-transmit the optical signals (otherwise known as “3R regeneration”). But such regenerators are costly and thus unsuited to cost-sensitive MMF applications.
So what could be done to continue advancing the MMF communication field? It appears some sort of bold move is required, as happened when changing the core diameter from 62.5 µm to 50 µm decreased the NA of MMF and created an increase from 500 MHz.km to the current OM4 standard of 4700 MHz.km. A similar move is needed at this time to be able to continue increasing the reach of MMF communications at 850 nm.
In Reference 6 it is shown that the modal dispersion is proportional to the fourth power of the NA. That is, reducing the NA of the fiber by a factor of 2 could lead to a factor of 16X lower modal dispersion. If that is possible, chromatic dispersion will become the leading impairment in fiber links at 850 nm -- if the 2-nm spectral bandwidth of VCSELs remains unchanged. We suspect that VCSELs with much lower spectral bandwidth could be produced; when combined with an 850-nm-only standard, such VCSELs would significantly improve the current bandwidth distance product.
We see an opportunity for transceiver and MMF vendors to specify optimal (narrower spectral line width) VCSELs and produce the best fiber compatible with those VCSELs. This will result in a completely new product with higher capacity (with a possible tenfold increase in the communication capacity of a single MMF). Such a technology would take the market by storm.
But this approach obviously would require new technology. The other alternative is to initiate the long march toward a political agreement among all industry players for a new standard that would lay the groundwork for a new optical fiber communication system platform as an alternative to the current MPO/MTP approach.
We suggest that readers make their voices heard at the next IEEE, ISO, TIA/IEC, or FCIA meeting and demand that working being on a new standard that will pave the way for ever-expanding MMF optical communication systems. A new standard that increases tenfold the capacity of MMF communications systems by decreasing the NA of the fiber and the line width of the VCSELs should be immediately started.
- Bell, P. D., Fontaine, N. H., Kouzmina, I. I., “Evolution of 50/125 µm Fiber Since the Publication of IEEE 802.3ae.” http://www.corning.com/assets/0/433/573/637/645/950BA748-8F9A-457E-A5CA-6A2166B78A60.pdf
- Document ISO/IEC 11801
- Lane, B., Pimpinella, R., Brunsting, A., “Correlation of BER Performance to EMBc and DMD Measurements for Laser Optimized Multimode fiber,” 2007, Panduit, Inc. http://www.panduit.com/groups/MPM-OP/documents/TechnicalPaper/CMSCONT_033699.pdf
- Brunsting, A., Pimpinella, R., “Correlations between Eye Diagram and BER Measurements for Multimode Fibers,” 2007, Panduit, Inc., http://www.panduit.com/groups/MPM-OP/documents/TechnicalPaper/109606.pdf
- Brunsting, A., Pimpinella, R., “Certifying Multimode Fiber Channel Links for 10 Gb/s Ethernet” 2005, Panduit, Inc., http://www.panduit.com/groups/MPM-OP/documents/TechnicalPaper/103110.pdf
- “Optical Waveguides and Fibers”, SPIE document by Ajoy Ghatak and K. Thyagarajan, Module 1.7
- Pimpinella, R., Tudury, G., “Certifying MMF for 100G Ethernet transmission,”Lightwave, November 2008
- Tudury, G.E and Pimpinella, R. J., US2011/0044594 A1. Tudury, G.E and Pimpinella, R. J., US 2011/0037183
- “snomvcsel.pdf” brochure on www.Witec.com
- “Wave Chaos in real world Vertical Cavity Surface Emitting Lasers,” Phys. Rev. Let. 94, 233901 (2005)
Gaston E. Tudury, Ph.D, has five published patent applications in MMF and several pending (unpublished) and 8 years of applied research in fiber-optic communications.
Al Brunsting, Ph.D., has 15 issued USA patents and 12 years’ experience in applied research for fiber optics in telecom systems.