Erbium-Doped Fiber Amplifier Displays Low Noise

Breck Hitz

Most fiber optic systems operate in the 1530- to 1565-nm C-band because the gain of erbium-doped fiber amplifiers peaks in this spectral region. But as the demand for bandwidth grows, many telecom engineers look covetously at the 1565- to 1625-nm L-band for expansion possibilities. Unfortunately, the gain of these amplifiers is much lower in that region.

One technique that has produced positive results is to double-pass the erbium-doped amplifier: Reflect the signal back through the device a second time, effectively doubling its length. However, a problem with erbium-doped fiber amplifiers -- in the C- or L-band -- is the noise they add to the signal. Amplified spontaneous emission is unavoidable over the long lengths of fiber in the amplifier.

Figure 1. Researchers at the University of Malaya have developed an erbium-doped fiber amplifier for the L-band that displays improved noise characteristics. Here, Nizam Tamchek measures the amplifier's parameters.

Because the first few meters of amplifier that an incoming pulse passes through are also the last few meters for an outgoing, amplified pulse, double-passing the erbium-doped fiber amplifier significantly worsens the problem. The powerful outgoing pulse saturates the gain so that the incoming pulse experiences dramatically less gain than it would experience in a single-pass amplifier. Thus, the incoming pulse, when it exits those first few meters of the double-pass amplifier, has a much lower signal-to-noise than it would have in the single-pass amplifier. Moreover, that noisy pulse is amplified through the rest of the amplifier, so that when it finally emerges, its noise figure is decidedly poorer than the noise figure of a pulse emerging from a single-pass amplifier.

Now, a research group at the University of Malaya in Kuala Lumpur, Malaysia, has demonstrated a solution to this problem. The team placed a few meters of single-pass, L-band erbium-doped fiber in front of the double-pass fiber (Figure 2). Because the first few meters of the fiber are shielded from the high-power output pulse, their gain is not saturated, and the pulse emerging from that fiber has a good noise figure.

Figure 2. The amplifier stages at center and right are double-pass erbium-doped fiber. The 4-m fiber in the stage on the left is single-pass, so the amplified pulse does not saturate the gain.

Experimentally, the researchers used 4 m of fiber in the single-pass portion and 50 m in each of the double-pass stages. An optical circulator coupled radiation from the single-pass portion into the double-pass stages, and then coupled the returning, high-power pulse out of the erbium-doped fiber amplifier. Another optical circulator, its first and third ports connected with a short length of fiber, served as a mirror to reflect the signal into the setup for the second pass.

The researchers compared three cases: a single-pass erbium-doped fiber amplifier, a conventional double-pass amplifier and the modified double-pass configuration shown in the figure. For small input signals (lower than <–15 dBm), they found that both double-pass configurations had much greater gain than the single-pass one, but that the modified double-pass configuration exhibited a much better noise figure. The advantage of the modified configuration diminished, however, as the input power approached 0 dBm.

They also compared the three cases across the spectral region from 1560 to 1610 nm (most of the L-band) and observed superior performance of the modified double-pass configuration with respect to the conventional double-pass one.