Differential Mode Delay Q&A
|
| ||||||||||||||||||||||||||
What are the current distance specifications in the most recent Gigabit Ethernet draft specification?
It is important to note that the IEEE specifications are conservative distance goals. Operationally, in many customer implementations, Gigabit Ethernet will go much further than the official specifications. For example, at N+I in Atlanta, a Gigabit Ethernet 1000BASE-SX link of the InteropNet exceeded 380 meters. What is DMD? The Gigabit Ethernet specification (IEEE 802.3z) outlines parameters for Ethernet communications at a gigabit-per-second rate. Its initial intent is to offer a higher-speed version of Ethernet for backbone and server connectivity using existing deployed multimode fiber. In order to accomplish this goal, the specification defines the use of laser-based optical components to propagate data over multimode fiber. Lasers were specified because they function at the required baud rates and longer distances for Gigabit Ethernet. The IEEE 802.3z Gigabit Ethernet Task Force has identified a condition known as differential mode delay (DMD) which occurs in certain circumstances with particular combinations of lasers and multimode (MM) fiber. The resulting characteristics create an additional element of "jitter" which can limit the reach of Gigabit Ethernet over MM fiber. DMD is an effect where a single laser light pulse excites a few modes equally within a MM fiber. These modes, or light pathways, then follow two or more different paths. These paths may be of different lengths and may have different transmission delays as the light travels through the MM fiber. With DMD, a distinct pulse propagating down the fiber no longer remains a distinct pulse or, in extreme cases, may become two independent pulses. A string of pulses will tend to interfere with each other such that the data cannot be recovered in a reliable fashion. This jitter effect is the problem that has caused the Task Force to revisit some of the optical parameters of Gigabit Ethernet. Differential mode delay is not experienced in all deployed fibers. It occurs with certain combinations of worst case fibers with worst case transceivers. Based on lab and field testing completed to date, the percentage of fibers was believed to be large enough that the 802.3z Task Force decided to develop a solution. The technical approach to the solution requires conditioning the laser light launched from the source (transmitter) and specifying receiver bandwidth. Differential mode delay will exist for any high-speed technology using a coherent (laser) source for transmission over MM fiber. Gigabit Ethernet is the first technology to experience this problem due to its very high baud rate, its long multimode fiber link lengths, and its rigorous evaluation of the link technology. Neither single mode fiber nor copper Gigabit Ethernet solutions are affected by DMD. What is a conditioned launch? In simple terms, conditioning a laser launch spreads the light evenly across the diameter of the fiber making the launch look more like an LED source to the fiber. The objective is to scramble the modes of light such that the power is distributed more equally in nearly all modes and not concentrated in just a few modes. This is in contrast to an unconditioned launch, which, in the worst case, might concentrate all of its light in the center of the fiber exciting only two or more modes equally. For 1000BASE-SX short wavelength transceivers, the conditioning may be done within the transceiver. For 1000BASE-LX long wavelength transceivers, removable conditioners will be available in order to meet the conditioning specification for MM fiber. With the conditioners being removable, the same -LX transceiver can be used for both multimode and single-mode fiber. How did conditioned launch come to be identified as the correct approach to the resolution to DMD? The 802.3z Task Force specified "conditioned launch" techniques to deal with DMD jitter in multimode fiber. The DMD effect was raised as an issue during an earlier ballot comment. The ballot comment itself was based on research findings from laboratory testing. During the Montreal IEEE plenary, the IEEE 802.3z Task Force and larger 802.3 Working Group voted overwhelmingly to accept a modification in the standards draft to address DMD by making a "conditioned launch" mandatory for 1000BASE-SX and 1000BASE-LX transceivers on multimode fiber. Laboratory and field verification of conditioned launch for 1000BASE-LX with multimode fiber using a hybrid jumper fiber has been successful. IEEE standards participants will continue verification testing of the conditioned launch specifications. Is the Gigabit Ethernet standard still on target for delivery in March? The IEEE Standards Board approval committee meets every three months. Thus, the next opportunity for the approval of the 802.3z standard is in June. A detailed and aggressive work plan has been approved by the 802.3z Gigabit Ethernet Task Force to investigate combinations of worst case components and to develop conformance tests that ensure that transceiver parts perform conditioned launch. This plan targets formal ratification of the standard in that time frame. If the work is not completed by then, the next opportunity for ratification of the standard will be September 1998. See also the "Status of the Gigabit Ethernet Standard FAQ" Will the distance goals be shortened to 100 meters? No. The identification of conditioned launch as a solution means that there will be no reductions to the link lengths specified for 1000BASE-SX and 1000BASE-LX Gigabit Ethernet interfaces. In fact, the link length for 1000BASE-LX over 62.5 micron MM fiber was increased from 440 to 550 meters. There were several reports in the press regarding a reduction in the fiber link distances. These reports were stimulated by premature and incomplete information about an 802.3z sub-group investigation of the DMD phenomenon. Further analysis and investigation by an IEEE optical sub-group confirms that both -SX and -LX transceivers operate within the specification as previously defined. The distances specified in Question 1 are current. This means that there will be no requirement to limit the distance of Gigabit Ethernet over MM fiber to 100 meters. Just as is the case today with FDDI, fiber optic Gigabit Ethernet solutions will be specified to support current building wiring standards. There will be no special 802.3z requirement to replace multimode fiber with single mode fiber or run new multimode fiber. What is the impact of these issues on customers? Most of the products shipping today are equipped with 1000BASE-SX optics. It is premature to draw definitive conclusions about the impact of the additional specification work on existing 1000BASE-SX parts. These parts could very well end up being conformant with the proposed additional specifications that are being written. The standard is written to cover cases involving worse case optics operating over worse case fiber at the maximum specified distance. Conditions outside of these may not exhibit any problems. Early implementers of any networking technology are sometimes faced with an element of uncertainty when dealing with pre-standard products. The industry has faced this challenge before. These issues do not necessarily impact the functionality of the majority of installed networks or the majority of pre-standard Gigabit Ethernet products that have shipped to date. The Gigabit Ethernet products shipped to date will not stop working because of the outcome of the February meeting. Ethernet vendors have an excellent track record in taking care of their customers and successfully migrating customers from pre-standard to standards-based equipment. Customers should discuss this issue with their equipment suppliers and work with them to understand their vendors' specific transition plans. Why are these issues appearing now with Gigabit Ethernet? First, the challenges associated with using lasers on multimode fiber have become more apparent as the operating speeds have increased. The Gigabit Ethernet standard includes both high data rates and long link lengths. It is one of the first standards planning large-scale deployment using lasers over multimode fiber at distances of 260 meters or more. Fibre Channel, from a theoretical standpoint, has exactly the same problem. Any link technology that uses lasers to illuminate multimode fiber can potentially be affected by DMD. Though subject to DMD, other high-speed technologies that specify laser launch over a long distance using multimode fiber may not have observed this condition yet. Furthermore, the IEEE 802.3z Task Force is particularly attentive to the characteristics of the installed base of network cabling, be it copper or fiber optic cabling. The group will not allow a draft standard to progress unless it is sure that the underlying signaling technology will work on the vast majority of installed cabling. Thus, the IEEE is applying far greater scrutiny to the operating characteristics of lasers on multimode fiber than has ever been applied before. This is the first time that laser launch has been extensively tested over multimode fiber. What is the background on Gigabit Ethernet and high-speed LAN fiber-optic transmission? Historically, multimode fiber has only been tested for use with light emitting diode (LED) sources. LEDs create a condition within a fiber that is referred to as overfilled launch condition. The overfilled launch condition describes the way LED transmitters couple light into the fiber. Like a light bulb radiating light into a dark room, the generated light shines in multiple directions in a way that "overfills" the existing space which is greater in size than the original diameter of the source. This departing light "excites" a large number of modes, or light pathways from the single source that illuminates the entire room. Lasers launch light in a more concentrated fashion and use the fiber in a different way. A laser transmitter typically couples light into only a fraction of the existing modes, or optical pathways, present in the fiber. Because single mode fiber has only one mode or pathway for the transmission of light, DMD is only an issue with multimode fiber. Because of the differences in transmission characteristics, the performance of lasers on fiber differs from the performance of LEDs. The performance of lasers on fiber is also different from the performance observed under the overfilled launch conditions used to characterize the fiber. Since both the Fibre Channel and Gigabit Ethernet standards projects include laser sources for multimode fiber transmissions, joint interest studies were initiated to evaluate the maximum link distances possible. It was originally thought that laser launch would be at least as good as LED launch within the multimode fiber. The new test procedures developed by the IEEE 802.3z committee that was used for these studies resulted in worst-case link lengths much shorter than the experts had originally predicted. The differential mode delay issue was discovered because the industry is now focusing on new combinations of technology and new testing procedures. Can you explain in more detail what DMD is? DMD stands for Differential Mode Delay. It is a phenomenon that can appear in many forms of communications, acoustical, electronic, and optical. The term describes the simple fact that different rays of light, emanating from the laser source in the transceiver, traverse different paths through the large core of the multimode fiber. As these rays travel through the fiber, they spread out in time, because some take a slower, longer path, and others take a faster, shorter path. This phenomenon is seen only in multimode fiber - because it supports many modes. 62.5 micron fiber supports more modes than 50 micron, because it has a larger core. Single mode fiber, which accommodates only a single mode, or ray of light, does not exhibit this effect.
Because almost all ATM OC-12 622 Mbps implementations are LED-based, they are not affected by DMD. Any 622 Mbps implementation using short wavelength lasers over multimode fiber would be subject to DMD. At higher ATM rates, DMD will likely be a significant problem with the use of multimode fiber. The ATM Forum can use the data developed in Gigabit Ethernet studies to determine if any changes are required in their specification.
For more information, please visit the IEEE 802.3 Working Group Website
| |||||||||||||||||||||||||||
