Remote Optical Control of an Optical Flip-Flop Demonstrated
Important step toward all-optical signal processing is realized.
All-optical switching — the switching of one beam of light by another — is an absolutely essential technology for transparent fiber optic networks and for all forms of optical signal processing. Various types of optical switches and switching techniques are being investigated in laboratories around the world (see, for example, “Semiconductor Optical Amplifiers Light Up All-Optical Signal Processing,” Photonics Spectra, August 2007, p 62).
Those switches whose output power can be latched in one of two states, commonly referred to as optical flip-flops, promise to be important for signal-processing applications that require a memory of the past signals. Recently, Drew N. Maywar and his colleagues at the University of Rochester’s Laboratory for Laser Energetics in New York demonstrated a means of optically controlling a flip-flop in a remote fashion.
In this research, the signals that provide control of the flip-flop (that is, turn it “on” and “off”) are envisioned to come directly from an optical communications system or data network, without converting them into electrical signals. As such, a key aspect of the demonstrated flip-flop action is that the control signals are low power (below 1 mW), that they occur over a wide range of wavelengths (>40 nm) and that they operate independent of the polarization state.
Figure 1. The transmission of the holding beam through the flip-flop is remotely controlled by the signal beam. One type of signal programs the holding beam to turn itself off at the flip-flop, while another type programs the holding beam to turn itself on — that is, to transmit itself through the flip-flop.
The control signals are sent through a dedicated semiconductor optical amplifier (SOA) that acts as a programming (or encoding) device by transferring the control signal data onto a continuous-wave holding beam that then passes through the flip-flop (Figure 1).
The control signal pulses might be at a wavelength within the SOA’s gain (the “reset” signal), or they might be at a shorter wavelength (the “set” signal) (Figure 2). Photons in the “set” signal are energetic enough to generate free carriers and increase the SOA’s population inversion — in other words, they increase the power in the holding beam that emerges from the SOA. On the other hand, photons in the “reset” signal stimulate emission in the SOA, diminishing the population inversion. So an incoming “reset” signal decreases the power in the holding beam that emerges from the SOA.
Figure 2. The holding beam and the “reset” signal are both within the gain spectrum of the semiconductor optical amplifier (SOA), while the “set” signal is at a shorter wavelength. The exact values of these wavelengths are not critical as long as they meet these conditions. Reprinted with permission of Optics Letters.
Thus, the holding beam emerging from the SOA experiences an uptick corresponding to a “set” signal pulse and experiences a down-tick corresponding to a “reset” pulse (Figure 3a and b). How does this cause the flip-flop to … well, to flip or to flop? In this experiment, the flip-flop is another SOA, placed inside a Fabry-Perot resonator. The SOA is biased at a value just below the laser threshold (a value such that the round-trip gain in the resonator is slightly less than the round-trip loss), making it a highly nonlinear resonator that exhibits optical bistability.
Initially, the Fabry-Perot’s resonance is slightly detuned from the incoming holding beam’s wavelength. Thus, as with any nonresonant Fabry-Perot interferometer, it re-flects the incoming beam; very little is transmitted through the interferometer.
However, when the uptick in power arrives at the Fabry-Perot, it momentarily increases the photon flux inside, generating free carriers in the semiconductor. The free carriers, in turn, change the refractive index of the semiconductor just enough to bring the interferometer into precise resonance with the wavelength of the holding beam. Now, as with any resonant Fabry-Perot, it transmits the incoming beam. Presto, the flip-flop has flipped!
And because the Fabry-Perot is resonant, a large circulating power builds up inside it, which is sufficient to keep it resonant even after the uptick has passed through the interferometer. The flip-flop is latched into its “flip,” or transmit, state.
Figure 3. The signal consists of pulses at the “set” wavelength and at the “reset” wavelength (a). A “set” signal causes an uptick in the holding-beam power, while a “reset” signal causes a down-tick (b). As explained in the text, the uptick and down-tick cause the flip-flop to switch between its “transmit” and “reflect” states (c).
But when the down-tick arrives at the Fabry-Perot, the photon flux inside momentarily drops, causing the interferometer to go out of resonance with the holding beam. Once it is off resonance, it again reflects the incoming beam. The flip-flop has flopped. And because there is no longer a large circulating power inside the interferometer, it is latched into its “flop” (i.e., reflect) state until the next uptick arrives. The entire process is shown schematically in Figure 3.
Figure 4. The difference between the experimental data shown here and the conceptual drawing in Figure 3 is discussed in the text. (R = reset, S = set.) Reprinted with permission of Optics Letters.
Experimental data with this setup resembles the conceptual drawing in Figure 3, but there are several differences (Figure 4). The Rochester scientists used a much smaller “reset” signal because it was amplified in the SOA, while the “set” signal was absorbed. Even so, the uptick in the holding beam was smaller than the down-tick because the saturation mechanism in the SOA was stronger than the gain mechanism. Although the response time in this experiment was limited by the equipment available in the lab, ultimately the speed is limited by the carrier lifetime and the SOA recovery time. Other experiments have demonstrated SOA’s capability at speeds in excess of 40 Gb/s, so the scientists believe that their technique can be effective at such speeds.
The demonstrated control techniques work in a remote fashion — the flip-flop is separate from the nonlinear SOA through which the control signals passed. One benefit of this arrangement is that an amplified holding beam can be split and fed into an array of flip-flops for signal-processing applications requiring many devices. In addition, this remote control technique is applicable to any flip-flop that is enabled by a holding beam.
Optics Letters, Nov. 15, 2007, pp 3260-3262.
MORE FROM PHOTONICS MEDIA