Getting the noise out

I was interested to read the article by J. J. Pan and Y. Shi (Laser Focus World, Jan. 1996, p. 93) in which they describe a technique for reducing amplified spontaneous emission (ASE) in an erbium-doped fiber amplifier (EDFA) using a polarizer controlled by a feedback loop. Unfortunately this is misleading.

Getting the noise out

I was interested to read the article by J. J. Pan and Y. Shi (Laser Focus World, Jan. 1996, p. 93) in which they describe a technique for reducing amplified spontaneous emission (ASE) in an erbium-doped fiber amplifier (EDFA) using a polarizer controlled by a feedback loop. Unfortunately this is misleading.

The measurement of noise figure on an HP spectrum analyzer assumes the number of fiber modes that the ASE is in, namely two. Using a polarizer removes the ASE from one of these modes, and hence the measurement produces a noise figure that is a factor of 2 (3 dB) lower than it should be. If the correct number of modes (one) had been used in the calculation, there would have been no change in the noise figure.

The system does remove half of the ASE that would result in spontaneous-spontaneous beat noise at the receiver, but the dominant noise contribution in a well-designed system should be signal-spontaneous beat noise. This is generated by the signal coherently beating with the ASE. This only happens between signal power and ASE power in the same polarization state. The ASE in the same polarization state as the signal is not affected by the scheme presented, thus, there would be no reduction in the signal-spontaneous beat-noise term.

Other points worth noting are that in long-haul communication systems the signal can be depolarized at the transmitter to improve the system performance. Using a polarizer in the EDFAs is then problematic. Also if the signal is not purposely depolarized it may become depolarized as it is transmitted.

In wavelength-division multiplexed systems the signals at different wavelengths will have varying polarization states. The scheme presented requires all the signals to be in the same polarization state, making its use in these systems problematic also.

Nigel Jolley

Nortel-Northern Telecom

Harlow, Essex, UK

The authors reply:

Our experiments show that using a variable polarization beamsplitter with auto-tracking capability can reduce randomly polarized ASE noise by half. This reduces the EDFA noise figure and also has a positive impact on system applications. The noise-reduced EDFA can be characterized by using EDFA measurement features of an optical spectrum analyzer.

For EDFA measurement using an optical spectrum analyzer, optical gain is measured by comparing the measured output and input signal intensities with the amplified-spontaneous-emission background noise from the EDFA. The EDFA noise figure is calculated from the measured ASE power PASE in the signal wavelength band.

For a conventional EDFA, output signal noise or invariance can be calculated from statistical photon distribution of signal and ASE noise as well as gain. The ASE noise is randomly polarized with optical power equally distributed in two polarization modes of the fiber. For a noise-reduced EDFA, the ASE noise is reduced almost by half and the amplified signal is almost unchanged by using a variable polarization beamsplitter.

And finally, the EDFA is followed by a depolarizer to destroy polarization of the output, so it imposes no problem of polarization-dependent gain or loss in the transmission system.

More in Test & Measurement