All-optical converters promise improved networks
By Jeff Hecht, Contributing Editor
The rapidly increasing complexity of optical networks requires efficient wavelength conversion. Several technologies are under development, as the search for the ideal wavelength converter continues.
Interest in wavelength conversion arises from the growing need to manage multiple optical channels in complex optical networks. No conversion is needed if a system merely connects two points, even if it carries dozens of wavelengths for thousands of miles. However, switching signals at many wavelengths among many points can create conflicts when transmitters at different points try to send signals to the same destination at the same wavelength. The obvious solution is to shift optical channels to different wavelengths.
A wavelength converter takes an input optical signal at one wavelength and replicates that signal at a different wavelength. Ideally, the conversion should be independent of bit rate or signal format, entirely optical, require little power, not degrade the signal, and have tunable output. Converters also should be cascadable, so signals can be converted multiple times, like cars switch lanes on a freeway. Technologies in development range from simple but cumbersome optoelectronic converters to sophisticated all-optical devices based on nonlinear interactions and semiconductor optical amplifiers.
Optoelectronic converters are essentially updated versions of the optoelectronic repeaters displaced by optical amplifiers. A receiver converts the input signal to electronic form, and that electronic signal drives a transmitter at the desired wavelength (see Fig. 1). The technology is straightforward, and already established for wavelength conversion at 2.5 Gbit/s.
Although optoelectronic converters are not all-optical, they have some advantages. They can generate different wavelengths either by using a tunable laser transmitter, or by adapting plug-in transmitter modules for different outputs. Optoelectronic receivers do not require high input power, and are insensitive to input polarization and wavelength. Electronic regeneration can reshape and retime the signal.
On the other hand, like traditional repeaters, optoelectronic wavelength converters are inherently dependent on input transmission format and bit rate, so a change in data rate requires changing the converter. Their need for a receiver-transmitter pair also makes them complex.
ALL-OPTICAL WAVELENGTH CONVERTERS
Like optical amplifiers, all-optical wavelength converters promise better response, greater simplicity, and transparency to the signal format. Many types of all-optical wavelength converters have been proposed and demonstrated, and some are in active development. They fall into a few broad classes.
Laser converters direct a strong input signal at one wavelength into a continuous wave single-frequency laser oscillating at another wavelength. The input signal causes gain saturation, draining light energy from the oscillation wavelength. As a result, when the input signal is "on," the laser generates a much lower power at its normal oscillation wavelength than when the input is off. A filter blocks the input wavelength, leaving only the laser wavelength in the output. This effectively converts the signal from the input wavelength to the laser oscillation wavelength. One obvious effect of the conversion is that the output signal at the laser wavelength is the inverse of the original input—low when the input signal was high, and high when it was low. Other drawbacks include speed limited by internal resonances of the laser to about 10 Gbit/s, and requirements for input power of 0 to 10 dBm.
Coherent or nonlinear converters are based on coherent nonlinear processes in which two or more wavelengths interact to generate other wavelengths. Four-wave mixing converters combine a pump signal at frequency νpump with the input signal at νinput to generate an output at 2νpump -νoutput = νout. Difference frequency mixing converters generate the output signal as the difference frequency between the pump light and input signal. Cross-phase modulation converters use the input signal to modulate the phase of a signal at a second wavelength passing through a long fiber or optical loop mirror, then converts that phase modulation into intensity modulation.
As with other nonlinear devices, success depends heavily on availability of a suitable nonlinear medium. Some approaches use long lengths of fiber; others use semiconductor optical amplifiers. Good results have been obtained with parametric conversion in periodically poled lithium niobate waveguides. The nonlinear nature of the processes gives them very good time response, making them attractive for very high-speed systems. Pumping schemes are relatively complex and conversion efficiency is often low, but some offer very good signal-to-noise ratio as well as high speed.
OPTICALLY CONTROLLED AMPLIFIERS
Another family of devices is based on optically controlled gates, in which a relatively weak input signal modulates output of a second wavelength from a semiconductor optical amplifier. Input at the second wavelength is from a continuous-wave source, and the effect is to transfer the signal from the input wavelength to a modulation of second wavelength. Typically the two wavelengths are transmitted in different directions. The input signal can modulate the signal at the second wavelength by at least three different mechanisms. The output wavelength can be tuned simply by tuning the continuous wave source.
In a cross-gain modulated (XGM) converter, the input signal is strong enough to saturate gain of the semiconductor amplifier, reducing the output power at the wavelength. As in the laser converter, this has the effect of modulating the second wavelength by reducing its intensity when the input signal is strong, producing a wave form that is the inverse of the input signal. This approach has some other drawbacks, including a high wavelength chirp, but it is attractively simple, and can be extremely fast, with demonstrated speeds reaching 100 Gbit/s.
A more complex approach that promises better performance is the cross-phase modulated (XPM) converter. In this device, a low-power input signal modulates the phase of a continuous wave beam at the second wavelength, rather than the intensity. The phase modulation occurs because the input signal depletes carrier density in the semiconductor optical amplifier, changing the material's refractive index and thus shifting the phase of the amplified light. An interferometer stage converts the phase modulation into intensity modulation. The carrier depletion is virtually instantaneous, but the carrier recovery time is around 50 ps, therefore further refinements are needed to raise data rates above 10 Gbit/s.
Refinements to this approach center on ways to convert the phase modulation of the second wavelength to intensity modulation. One way is to fabricate the semiconductor optical amplifier as two parallel arms to form an integrated interferometer. In the design, the input signal illuminates only one arm while the continuous beam being amplified is divided equally between both (see Fig. 2, top). The relative phase shift caused by the input signal is converted to intensity modulation of the second wavelength where the two arms recombine. The Michelson interferometer configuration produces a similar effect, but reflects the CW beam from the facets at the right rather than transmitting it through a single-pass arm (see Fig. 2, bottom).
A simpler variation is the use of a single semiconductor optical amplifier in the delayed interference configuration (see Fig. 3). In this scheme, the input signal and entire CW beam pass through the same semiconductor optical amplifier. That phase-modulated beam then is divided between two arms of unequal length, one of which delays the light—by 10 ps in this example. One arm is a passive waveguide, but the other contains an adjustable phase-shifter. The second coupler directs light out one of two ports, depending on the phase shift between the two signals reaching it. When there is no input signal at the first wavelength, the phase shifter is adjusted to make all the light emerge from one arm of the second coupler. When the input signal is switched on, it causes a phase delay in the amplified light at the second wavelength. When that phase-shifted light reaches the second coupler, it shifts its output to a different port because the two inputs have different phase shift. Once the phase-shifted light arrives through the delayed arm, the output switches back because the inputs have the same phase. In the example, this generates pulses lasting the 10-ps delay between the two input signals. Although this approach sounds complex, it has the advantage of requiring only two adjustments, one to the phase shift and one to the drive current to the semiconductor optical amplifier.
NEW IDEAS AND IMPRESSIVE RESULTS
Developers are pushing to very high speeds, and producing some impressive results. At the European Conference on Optical Communications (ECOC) in 2000, Juerg Leuthold of Bell Labs reported converting signals at speeds up to 100 Gbit/s using a delayed-interference configuration. His group demonstrated error-free transmission at 100 Gbit/s for a string of 231-1 non-return-to-zero pulses.
Wavelength conversion is being extended in other directions as well. At ECOC 2000, Jianjun Yu of the Technical University of Denmark (Lyngby) reported simultaneously generating 40 Gbit/s return-to-zero pulses at five different wavelengths using a nonlinear optical loop mirror. Based on commercial components, it promises "a simple way to generate a 40-GHz multiwavelength source for WDM transmission systems," Yu said.
Optoelectronic wavelength converters are finding near-term applications, but all-optical converters look good for the faster and more sophisticated optical networks coming in a few years. Cross-phase modulation converters "have very good characteristics from a network point of view," says Tony Kelly of Kamelian Ltd. in Glasgow, Scotland. They can readily perform "2R" regeneration, amplifying and reshaping input signals. With clock recovery, they also can retime signals for full "3R" optical regeneration, which Kelly calls "the holy grail of optical transmission." Regeneration is accurate enough that thousands can be cascaded in sequence in the laboratory.
Laboratory developers continue pushing the state of the art; stay tuned for better results and more refined wavelength converters. More work remains to convert cutting-edge laboratory demonstrations to practical technology ready for field use. The wavelength-conversion requirements of future optical networks have yet to be defined. Yet as these requirements evolve over the next few years, it looks like all-optical wavelength converters are likely to become practical realities.
A. Kloch et al., IEICE Trans. on Comm. E82-B, 1209 (August 1999).
D .Nesset, T. Kelly, and D. Marcenac, IEEE Comm. Mag. 36 (13) 56 (December 1998).
K. E. Stubkjaer et al, IEICE Trans. on Comm. E-82B, 390 (February 1999).
Representative Research Results
I. Brener et al, Tech. Dig. OFC 2000 (March 7-10, 2000).
J. Leuthold et al., Proc. 26th Euro. Conf. on Opt. Comm. (Sept. 3-7, 2000).
B. E. Olsson et al., IEEE Phot. Tech. Lett. 12, 846 (July 2000).
T. Simoyama et al., IEEE Phot. Tech. Lett. 12, 31 (January 2000).
J. Yu et al., Proc. 26th Euro. Conf. on Opt. Comm. (Sept. 3-7, 2000).
X. Zheng, F. Liu, and A. Kloch, IEEE Phot. Tech. Lett. 12, 272 (Mar 2000).