Space-division multiplexing focuses on premises networks
Space-division multiplexing focuses on premises networks
At this year`s European Conference on Fiber Optic Communications, held in Brighton, UK, researchers from Siemens Research Laboratories in Munich and the Swiss Federal Institute of Technology in Zurich demonstrated a campus-wide multichannel optical network with high data throughput using space-division multiplexing. This technology is competitive with wavelength-division multiplexing.
Space-division multiplexing is believed to be a more economical solution for campus-wide networks by these European researchers. "This parallel optical link is likely to replace today`s standard local area computer networks such as Ethernet," according to H. Karstensen, head of the Siemens research group.
However, there are no precise plans for the commercialization of space-division multiplexing. Siemens still wants to improve its prototype and find useful specifications to enter the market cost-effectively. Both research groups are supported by their respective governments and by the Commission of European Communities.
The space-division multiplexing 12-channel fiber ribbon link transfers data at 24 gigabits per second (2 Gbits/sec per channel) and allows the flexible design of scalable and survivable networks. All components have closely spaced multiple active regions to reduce size. The size of the opto-electronic parts (12 channels, without electrical connectors) is 5䁾ٶ millimeters.
The transmitter module--a laser-diode array--consists of indium gallium arsenide/gallium arsenide quantum-well strained layer lasers operating at 850 nanometers. This established, low-cost technology works well in short-distance networks. The quantum wells or active layers of the lasers comprise GaAs, and the barrier layers consist of InGaAs.
"Long-wavelength semiconductor lasers emitting at 1.3 or 1.55 microns based on indium phosphide are ideal for optical telecommunications because these wavelengths meet fiber dispersion and attenuation minimums," says Karstensen. InP-based lasers are more sensitive to temperature changes; therefore, the structure has to be carefully structured for good performance. But it is more economical to employ 850-nm lasers of GaAs than lasers of InP technology, if possible.
"Indeed, 850 nm is a well-known and low-cost technology and sufficient for a short-distance network," he adds. In addition, the electrical characteristics of GaAs-based lasers are less sensitive to temperature changes.
A threshold current of only 3.5 milliamperes and a differential efficiency of 0.5 watts/ampere guarantee a low-power operation of 110 milliwatts per channel for a supply voltage of -4.5 volts. The threshold current deviation for the laser diodes of the 24 channels from the mean value is lower than ۪.2 mA. An operation lifetime of 100,000 cumulated hours without failure is specified by the researchers.
The receiver module consists of a photodiode array and an amplifier array that was fabricated at the Swiss Federal Institute of Technology. The 12 top illuminated pin photodiodes of InGaAs/GaAs grown by low-pressure metal-organic vapor phase epitaxy are integrated on a single chip forming a linear 12-channel photodiode array.
According to J. Wieland of the Swiss group, the arrays--with a photodiode pitch of 250 microns and a diameter of 70 microns--are suitable for coupling to multimode or singlemode fiber ribbons. The responsiveness of the photodiodes is 0.40 ampere/watt, and the signal rise and fall times are less than 150 picoseconds.
A direct-current-coupled 12-channel peak-detection-amplifier array is used as the receiver. It comprises a transimpedance input stage, peak amplitude detector for decision level adaptation and differential amplifier and comparator, including output buffer. This receiver array is developed using a semi-custom 0.8-micron silicon bipolar technology by Siemens. The peak-detection unit of the so-called burst-mode receiver triggers the decision level during the first incoming bit of a burst using a peak detector.
According to Karstensen, for a 10-decibel dynamic range, all the transitions at the digital output occur in a window of 250 psec for a continuous 231-PBRS at 1 Gbit/sec, and 400 psec, including the first transition of the burst. "The circuit design emphasizes ease and stability of operation," he says. Each channel operates to 1 Gbit/sec with a 10-dB dynamic range and 2 Gbits/sec with a 6-dB range. The prototype operates from a single standard power supply in the range of 3.2 to 5.2V, and no additional bias or control voltage is necessary.
The optical link is based on space-division multiplexing and is fully compatible in all wavelength-division multiplexing systems, which are commonly deployed for this type of network. The two techniques differ slightly from a systems viewpoint. The former network technique corresponds directly to the virtual wavelength path concept of a corresponding wavelength-division multiplexed network, which can be interference-free.
A wavelength-division multiplexed network corresponds to a wavelength path concept that has the disadvantage of blocking connections if two signals with the same wavelength are transmitted in one fiber. To overcome this problem, wavelength conversions in the crossconnects are needed.
High flexibility, scalability, modularity and survivability can be achieved by switching, for example, asynchronous transmission mode and space- or wavelength-division multiplexed routing. For wavelength-division multiplexing, only one fiber is used. Different wavelengths from electro-optical converters are combined by means of selective optical couplers.
The signal can be transmitted in one or both directions. Siemens prefers time-division multiplexing: Wavelengths are transmitted in parallel at different times by means of 12 fiber channels.
Researchers chose space-division multiplexing because the components of such a network are less expensive than those of a wavelength-division multiplexed system. In most networks, cost depends on length; if longer lengths are required, wavelength-division multiplexing should be applied because only one fiber is used. Space-division multiplexing is more economical for distances to a few kilometers, such as for the campus area of a customer premises network.
For wavelength-division multiplexing, expensive tunable distributed-feedback lasers are necessary to transmit beams at different wavelengths. Furthermore, the multiplexer and demultiplexer in the wavelength-division multiplexed system are costly and can cause high losses. Therefore, expensive amplifiers are needed to compensate for the loss of light intensity. All these components are not needed for space-division multiplexed switched networks.
A drawback of space-division multiplexing is that a fiber ribbon is needed instead of a single fiber, which makes the corresponding systems more expensive than wavelength-division multiplexed links after a certain length. But a space-division multiplexed link is an attractive and economical alternative to wavelength-division multiplexing for distances to a few kilometers. q
Achim Strass writes from Munich, Federal Republic of Germany.