Quasi-light Storage for Optical Data Packets

The data transport in the telecommunications networks is optically, since only optical fibers offer the capacity required for today’s data traffic transmitted around the world. However, in every node of the network the optical signal has to be transferred into the electrical domain in order to process it. After processing the signal is converted back to the optical domain for further transmission. This double-transfer between the domains is both time and power consuming. In order to use an all-optical processing of the data, the problem of the intermediate storage has to be solved. Thus, lots of methods for the storage or buffering of the optical signals have been suggested. The simplest way is to send the signals into a matrix of waveguides with different lengths². However, these matrices are bulky and the storage time cannot be tuned since it is predefined by the waveguide length.

The “Slow-Light” method relies on a tunable change of the group refractive index of a medium to slow down the propagation speed of optical signal pulses². Several physical effects and material systems can be used for this purpose^3-6. However, with these methods the signal can be slowed down by just a few bit-lengths, which is by far not sufficient for optical network nodes^7,8.

Another approach uses wavelength conversion and dispersion for the generation of tunable delays. Thereby, the center wavelength of the input signal is shifted via nonlinear optical conversion. Afterwards, the signal is fed into a highly dispersive fiber. The difference in the group velocity in the dispersive fiber leads to a delay that is proportional to the product of the wavelength shift and the group-velocity dispersion (GVD) in the fiber. With a second conversion the wavelength is shifted back to the original value. For the wavelength shift techniques like four-wave mixing or self phase modulation can be used. With the conversion and dispersion method storage times up to 243 nsec of tunable delay, which correspond to 2,400 bit, were reported¹⁰. However, wavelength conversion and dispersion methods in general need special components and setups for producing a large wavelength shift and/or large GVD. Additionally, they are among the most complex and power-hungry delay methods².

Other methods store the optical signal into an excitation of a material system. A probe beam is then used to read out the information. Usually these systems cannot be used in the area of telecommunications since they require ultrahigh or -low temperatures¹¹, will not work with telecommunications bandwidths, or require rather complicated setups and high power^12-14.

Here we show how a basic property of signals (the time-frequency coherence) can be exploited for the storage of optical data packets. Since no excitation of a material system is used, we have called the method Quasi-light Storage (QLS)^15-17. The QLS is independent of the modulation, data format and data rate of the packets and can store optical packets for several thousand bit lengths¹⁸.

The basic idea can be seen in Figure 1, here rectangular shaped pulses are shown. However, the method works for every pulse shape and for packets of pulses. The only restriction is that the signals have to be time-limited.

Figure 1. Time-frequency coherence for an intensity modulated signal²³. A single rectangular signal in the time domain (a) is represented by a sinc-function in the frequency-domain (b). Here the normalized intensity is shown, since it is not possible to measure the fields with optical equipment. The time domain representation for a sequence of rectangular signals is shown in (c). This sequence has still the same spectral shape. But, it consists of equidistant single frequencies under the sinc-envelope (d). The time axis are normalized to half the duration of a single signal and the frequency axis to the first zero crossings, respectively. Click here to view larger image.

A rectangular pulse in the time domain (Figure 1a) has a “sinus cardinalis” or sinc function sin(px) / px shaped spectrum (Figure 1b), where all frequencies under the envelope are present. A train of rectangular pulses in the time domain (Figure 1c) has still a sinc function shaped spectrum (Figure 1d) with the bandwidth Δf. But due to the periodicity, not all frequencies are present anymore. Instead, the spectrum consists of equidistant frequencies and the inverse of the frequency spacing defines the time separation between the pulses ΔT = 1/Δv.

The basic idea of the QLS is now simply to extract equidistant frequencies out of the spectrum of the input packet. Due to time-frequency coherence this results in a copying of the packet in the time domain. The copy with the desired delay can be extracted by a time domain switch.

The principle of our experiment is shown in Figure 2. A time-limited input signal is multiplied with a frequency comb in the frequency-domain. For the multiplication the nonlinear effect of stimulated Brillouin scattering (SBS) is used. The results are equidistant copies of the input signal in the time-domain. One of the signals is extracted with a switch driven by a rectangular function. Thus, at the output of the memory in principle a distortion-free copy of the input pulse can be expected.

Figure 2. Basic idea of the Quasi-light Storage¹⁵. A time limited input signal (a) is multiplied with a frequency comb (b) in the frequency-domain, which is denoted with a X. This leads to various copies of the signal in the time domain (c). From the generated pulse train one of the copies (d) is extracted with a time-domain switch by a rectangular read signal (e). The switch can be a modulator. The result is a storage of the optical signal. The storage time is defined by the frequency spacing between the comb lines and the read signal. Click here to view larger image.

SBS itself is a nonlinear effect that can occur in standard single mode fibers (SSMF) at low powers. Thereby, the signal interacts with an optical density change which is generated by a counter propagating pump wave. If the signal wave is downshifted in frequency, a gain region is formed in which the signal will be amplified. If it is up-shifted the signal will be attenuated in the corresponding loss region. The frequency shift between pump and signal is defined by the acoustic wave, which depends on the material properties. The biggest advantage of SBS for the presented application is the narrow bandwidth Δf_SBS of the gain region. Thus, practically SBS forms a narrow linewidth optical filter. The narrow bandwidth of the gain region depends on the effective length and area of the fiber as well as on the used pump power¹⁹. The natural full-width at half-maximum (FWHM) bandwidth of the SBS gain in a SSMF is around 30 MHz. In special waveguides, such as AllWave fibers, and with high pump powers, the bandwidth can be reduced down to 10 MHz²⁰. Due to the filter bandwidth the different copies are covered with an envelope. Therefore, the maximum storage time of the QLS inversely depends on the SBS bandwidth. A bandwidth of 10 MHz would result in a maximum storage time of 100 nsec. Click here to view larger image.

For very high bit-rate transmission the information has to be encoded into the phase of the carrier instead of its amplitude, since this offers a lot of advantages. Thus, contrary to pulses, the signals in these optical networks have constant amplitude. Figure 3 shows such a phase modulated signal in the time (left) and frequency domain (right). This spectrum can be sampled in the same way as that of the amplitude modulated signal²¹. In fact the spectrum of the rectangular function for intensity- and phase-modulated signals is filtered due to the transmission, which limits the spectrum.

Figure 3. Time-frequency coherence for a phase modulation²¹. In a phase modulated signal the phase of the carrier is changed by the signal which has to be transmitted. If each symbol consists of 1 bit, the phase is changed between 0 and π, for instance. The left side of the figure shows the resulting time-domain representation for such a binary phase shift keyed (BPSK) signal. The resulting frequency-domain signal is shown on the right side. By comparison with Figure 1 it can be seen that the spectrum of the phase modulated signal is qualitatively the same as that of the intensity modulated signal. Thus, the QLS can be applied in the same way.