Optical preamplifier receiver for spectrum-sliced WDM

Abstract

Spectrum-slicing provides a low-cost alternative to the use of multiple coherent lasers for wavelength division multiplexing (WDM) applications by utilizing spectral slices of a single broadband noise source for creating the multichannel system. In this paper we analyze the performance of both p-in and optical preamplifier receivers for spectrum-sliced WDM using actual noise distributions, and the results are compared with those using the Gaussian approximation. This extends prior results of Marcuse for the detection of deterministic signals in the presence of optical amplifier and receiver noise. Although the methodology is similar, the results are considerably different when the signal is itself noise-like. For the case of noise-like signals, it is shown that when an optical preamplifier receiver is used, there exists an optimum filter bandwidth which minimizes the detection sensitivity for a given error probability. Moreover the evaluated detection sensitivity, in photons/bit, represents an order of magnitude (>10 dB) improvement over conventional detection techniques that employ p-i-n receivers. The Gaussian approximation is shown to be overly conservative when dealing with small ratios of the receiver optical to electrical bandwidth, for both p-i-n and preamplifier receivers.

Spectrum-Sliced Wavelength Division Multiplexed Systems with Optical Preamplifiers

“…As stated above, the latter is more accurate and thus generally employed from hereon. The GA, although numerically convenient, is again proved to be conservative by showing consistency with the results of Arya and Jacobs [12] only at large slicewidths.…”

“…Compared to the results of Arya and Jacobs [12], obtained from the pdfs, the GA can provide a convenient and sufficiently accurate estimate of the BER when the optical bandwidth is large (B o is larger than 1nm) and the dispersion is moderate (z/L D less than approximately 0.01), with the effect of excess noise easily included via the variance.…”

“…Instead, the SPA approximation provides much better numerical results for the complete range of values of optical bandwidths and dispersion levels considered. For example, with a bit rate of 2.5Gbps, a FWHM slice width of 0.8nm, and no dispersion, the GA sensitivity prediction (-26.9dBm) differs by some 2dB compared with an extremely computationally intensive calculation based on the accurate pdfs (-28.9dBm) from Arya and Jacobs [12], whereas the SPA (-28.5dbm) is within 0.4dB. Moreover, the SPA has proved highly effective at obtaining agreement with measured results [13] and the results correspond well with those using a Fourier approach with chi square statistics [14].…”

“…For the mathematical convenience, a large extinction ratio (r e >>20dB) is assumed, so that the received optical signal only presents in the "ON" state and is well modeled by a chi squared pdf [12]. Since the closed form expression for the combined signal and noise distribution is difficult to manipulate numerically, it is common to approximate all pdfs by Gaussians, with appropriate means and variances [13].…”

“…Normally the receiver sensitivity can be improved by 10-20dB using an erbium doped fiber amplifier (EDFA) as an optical preamplifier [12]. Hence, the power budget can be improved without changing other system configurations.…”