At NTT Photonics Laboratories (Kanagawa, Japan), researchers have developed a way of producing a broad-wavelength source from a single-wavelength pulse train. The technique uses a synchronized phase change in an electroactive medium to shift the wavelength of pulses in the initial train up, down, or both. By combining sets of raw pulses with previous generations that have been shifted by varying amounts, a broad spectrum—more than 100 channels with wavelength peaks separated by just 0.1 nm (12.5 GHz)—has been produced. In addition, the signal-to-noise ratio can be kept high, despite the amplification of spontaneous emission in the device, by using a temporal filtering technique.
The easiest way to perform wavelength-division multiplexing is to produce pulse trains that include all the wavelengths that the system can handle, and then filter out what is not required. Producing a broad continuous spectrum is not ideal, however, because it places too much burden on the precision of the filters. Instead, the ideal source contains well-defined peaks at each of the desired wavelengths and a return to zero in between. Since lasers generally operate at a single wavelength or a small (and somewhat arbitrary) set of wavelengths, designing the ideal source has not been easy. In the past, wavelength-shifting techniques have generally involved some kind of optical nonlinearity (most notably the Kerr nonlinearity, in which the refractive index of a medium is a function of the light intensity). Producing the required effect took high power, which, in turn, necessitated high-power components and so made the systems expensive.
What the NTT team has done is to exploit and enhance the so-called serrodyne modulation technique.1 Conventionally, a beam is exposed to a sawtooth-shaped (hence "serrodyne") phase modulation. If the phase in the linear portion is rising, the wavelength increases by an amount related to its gradient. Isao Tomita and his colleagues realized that, for pulses, the sawtooth shape was unnecessary: the linear gradient was only required at the temporal location of the pulse. So they switched to sinusoidal modulation, which is not only easier to implement (no fast return-to-zeros), but is also more efficient.
Making the technique work requires careful synchronization between the pulse and modulator to ensure that the pulse always falls on the linear portion of the sine wave. Synchronization turns out to be easy to implement by using unwanted pulse side lobes as a trigger, an approach that has another advantage: both the rising and falling gradients can be used at once, producing a symmetric spectrum (see Fig. 1). This technique is what enabled the 100-channel result.
As the pulses travel through the ring (being wavelength-shifted through each loop), however, the signal must be boosted using an erbium-doped fiber amplifier. This is the only way to maintain a reasonably constant peak intensity across the spectrum. Unfortunately, with this comes the amplification of stimulated emission (essentially white noise) and a corresponding drop in the signal-to-noise ratio. Researchers solved the problem by using an optical-gating technique that had been developed for a previous project (see Fig. 2).2 With light only allowed to emerge at the required pulse times, stimulated emission actually enhances the signal-to-noise ratio temporally (though it can still contribute to crosstalk between channels). According to Tomita, a high signal-to-noise ratio of more than 25 dB was achieved.
- Isao Tomita et al., IEEE Photon.Tech. Lett. 15(9), 1204 (September 2003).
- Hiroaki Sanjoh et al., OECC/IOOC Conf. Incorporating ACOFT, Sydney, Australia (July 2001).