The future of optically multiplexed SF transceivers
Mark E. Heimbuch MRV Communications
The best-known optically multiplexed transceiver is the single-fiber (SF) transceiver. The SF transceiver has begun to appear as a commodity from many suppliers, but its current use is in niche applications. The benefits of an SF transceiver are obvious: It allows a transceiver to use one fiber instead of two. This simple fact makes the SF transceiver attractive for businesses that are renting fiber, trying to conserve their available dark fiber, or running at maximum capacity and would need to pull new fiber (or upgrade their equipment). These applications will not disappear in the future and neither will the SF transceiver. Whether optically multiplexed transceivers remain a niche component, or become a commonly used standard will depend on future applications and future developments.
One future use for optically multiplexed transceivers is to extend the reach of higher-speed data links to transmission distances that equal today's slower-speed transceivers. This "distance data-rate" product is a very real limitation for fast data links. With optical multiplexing, a two-fiber transceiver can be built that is effectively made up of multiple transceivers. Using multiple slower transceivers to do the work of one faster transceiver can result in greater distance data rates. This next-generation transceiver could be referred to as a "2xTransceiver," "4xTransceiver," or "NxTransceiver," depending on the number of optically multiplexed channels.
Before delving into this application, we need to explore the current SF transceiver in more detail.
A SF transceiver uses the same electronics as a standard dual SC transceiver. The key component that enables SF operation is the internal optical multiplexing. For short distances, the optical multiplexing can be full duplex using a single wavelength (transceiver pair (a) in the Figure). Although this method is the least expensive, reflection sensitivity makes it inferior for long transmission distances. The "standard" SF transceiver consists of two wavelengths (transceiver pair (b), consisting of transceivers B and C, in the Figure) and is a more robust design that is better suited for long transmission distances. The disadvantage of the standard SF transceiver is that a different transceiver is required on each side of the link.
There are two primary methods for packaging the optical components of a SF transceiver. The first is to use bulk optics similar to current laser packaging. The other method is to use an integrated package that consists of a laser and receiver, with a waveguide layer to couple these elements together. The advantage of the integrated package has yet to be confirmed, but lower assembly cost and a smaller package size are expected to be the primary benefits.
The optical components necessary for a standard SF transceiver make it more expensive than the current low-cost Fabry-Perot (FP) transceivers. The standard SF transceiver requires one 1.3-micron FP laser and one 1.5-micron distributed feedback (DFB) laser to be used on one link instead of two 1.3-micron FP lasers. Traditionally, the cost of a DFB has been significantly higher than the cost of the 1.3-micron FP. This is due to three factors: higher DFB manufacturing costs, lower volumes, and the requirement for additional expensive optical components in the package.
However, the cost of the 1.5-micron DFB element can be reduced for short-distance applications, and with high volumes, the cost can begin to approach a 1.3-micron FP device. For higher-speed future applications such as 10 Gbits/sec, the industry will need to move to DFB devices in order to maintain a 2-km solution, so the issue of additional cost will be eliminated.
For businesses that are renting fiber or do not have fiber to spare, any additional cost for a standard SF transceiver will be less than purchasing a faster switch and much less than installing or renting more fiber. For businesses that are not fiber limited, it is not cost effective to use SF transceivers, but it might be prudent to do so in order to reserve dark fiber for future installations. If a business finds that next-generation transceivers will not reach their required distances, then more careful allocation of the fiber becomes necessary.
This effect of reduced distance has already been observed for Gigabit Ethernet transmission over multimode (MM) fiber. Currently, the IEEE Standards Committee 802.3 High Speed Study Group is working to establish the standards for 10-Gigabit Ethernet. The preliminary distance specifications are 100 and 300 m on MM fibers and 2, 10, and 40 km on singlemode (SM) fibers. The physical layer components used to achieve these distances are still being discussed but optically multiplexed transceivers are one of the primary components being considered.
Why is the distance an issue for next-generation transceivers? It is a basic limitation of fiber that with each increase in speed, a decrease in transmission distance is realized. This is due to dispersion limitations and/or the requirements for higher powers at the receiver. Dispersion limitations are the dominant effect and for components that are dispersion limited (such as the current low-cost transceivers), a factor of four increase in speed will result in a factor of four decrease in the transmission distance. The best new MM fiber is limited to 2 km at 1.25 Gbits/sec and an uncooled 1310-nm FP laser operated at 1.25 Gbits/sec over SM fiber is limited to around 10 km (depending on the ambient temperature range). At 10 Gbits/sec, these existing low-cost components will have very limited distances: around 250 m for MM fiber and 1500 m for SM fiber. Therefore, industry will have to accept new solutions to address the required distances.
One solution is to run parallel fiber. However, parallel fiber is not a long-term solution, as it will quickly consume existing dark fiber or will require new installations. For SM fiber, the more likely solution is to use higher-cost components. Directly modulated DFB lasers are limited to around 20 km at 10 Gbits/sec due to dispersion for 1.5-micron lasers and attenuation for 1.3-micron lasers. Note that "long-haul" applications of 100 km at 10 Gbits/sec utilize cooled lasers with integrated electro-absorption (EA) modulators and fiber amplifiers. The integrated laser/modulator is a possibility for data-communications transceivers; however, the device cost, cooler cost, cooler size, cooler power consumption, and cooler heat dissipation make it unattractive.
For increased transmission distances on both MM and SM fiber at 10 Gbits/sec and higher, there are two primary competing technologies. One technology is optically multiplexing slower transceivers and the other is modulating the laser with similar methods used in the modem industry. By using different electrical-modulation techniques, one source running at a lower bandwidth can transmit the longer distances while still achieving the desired high bit rates. This can be achieved by using multilevel detection (such as PAM5) and/or phase-shift-keying. For data-communications transceivers, these electrical-modulation techniques are not commercially in use. Currently, the IEEE 803.2 High Speed Study Group is considering both multilevel modulation and wide wavelength-division multiplexing (WWDM) as standards for 10-Gigabit Ethernet transceivers.
The advantage for using different modulation techniques is that the long-term pricing points will be lower than the WDM techniques. The advantage of the WDM technique is that it is currently available and can be used in parallel with future modulation techniques to enable solutions for the next several generations of transceiver speeds. The disadvantage of the higher cost (around 2 to 3 times) associated with optically multiplexed devices will be offset by the fact that these devices will enable distances otherwise unattainable. Rather than 1.3-micron FPs providing the short haul (0 to 10 km) and 1.55-micron DFBs providing the long haul (10 to 50 km), one DFB will provide the short haul and multiple DFBs will provide the long haul.
Multiple versions of optically multiplexed transceivers can be built using the current SF transceiver. Using two standard SF transceivers over two fibers can result in an equivalent 2xTransceiver (transceiver configuration (c) in the Figure). Note that for channel 1 of transceiver 1 to "talk" to channel 1 of transceiver 2 requires different transceivers at each side of the link. Using four transceivers and only 1.3-micron and 1.5-micron components, a low-cost, short-distance 4xTransceiver (transceiver pair (d) in the figure) can be made. This configuration is limited to applications where the data on each channel are independent, since the two fibers can have different lengths, resulting in an uncontrollable amount of skew (or delay) between the channels.
For deserialized data transmitted over parallel wavelength channels, (such as a 10-Gigabit Ethernet transmitted by using four 3.125 Gbaud transceivers) skew between channels becomes a critical parameter. In order to control the skew between channels, all four TX signals must be sent over the same fiber. This type of operation can be achieved by using four different wavelengths (transceiver pair (e) in the Figure). Although this device is a significant deviation from the standard SF transceiver, this four-color 4xTransceiver shares many conceptual and technical similarities.
Selection of the four wavelengths for the 4xTransceiver depends primarily on two criteria. The first criterion is temperature performance. The laser must operate uncooled and the passive insertion loss and crosstalk must be tightly controlled from 0 to 70° C. The second criterion is cost. The active and passive components must be high yield, low-cost technologies. Four-color 4xTransceivers that meet these criteria are becoming commercially available using wide (or coarse) WDM technology, which consists of 20-nm channel spacing.
While this technology allows for control of the skew, it does not eliminate the skew. The overall system design must allow for a maximum tolerable skew in the transmission. Once the maximum skew is fixed, the skew-limited maximum transmission distance is determined from both the channel wavelength separation and the absolute wavelength. For example, 5-nsec maximum skew allows for up to 40-km transmission at channel wavelengths of 1.28, 1.30, 1.32, and 1.34 microns. On the other hand, at 1.5 microns a channel separation of 20 nm and a 5.8-nsec skew result in 10-km transmission lengths.
The emerging 4xTransceivers could prove to be the primary entry point for standardization of WWDM in the data com marketplace. One hurtle to overcome is the perception that since dense WDM (DWDM) exists, it is only a matter of time before it is used throughout the data communications industry. While DWDM does hold the potential of over 100 channels as compared to the four to 16 channels used in a WWDM system, the key limitation is the eventual price point of the DWDM technology. DWDM has inherently higher-cost passive components-and also higher-cost active components.
While data communications applications for DWDM do exist, the high-volume, low-cost transceiver does not fit well with DWDM technology. The SF transceiver, on the other hand, is an excellent, cost-effective, WDM solution for businesses with fiber-poor environments.
Optically multiplexed transceivers using WWDM are also a practical solution for high distance-data rate applications like 10-Gigabit Ethernet over 20 to 50 km on SM fiber (or over 0.3 to 1 km on MM fiber). Additionally, WWDM can be extended to 8, 12, or 16 channels, allowing for future 4xTransceivers to operate on the same fiber as existing 4xTransceivers.
Mark E. Heimbuch received a Ph.D. in electrical engineering from the University of California at Santa Barbara in 1997, under the direction of Prof. L.A. Coldren. He joined MRV Communications (Chatsworth, CA) in 1997 and is currently the director of semiconductor device R&D.