Automated testing speeds Raman amplifiers into the field


by Christian Stauter and Wallace Clements

Techniques to measure gain and gain flatness are essential to faster field deployment of optical amplifiers. Successful measurements can be taken with OTDRs, or broadband ASE sources in conjunction with OSAs, depending on the progress of the installation.

With the expanding deployment of Raman amplification, network managers are looking for procedures to optimize the field commissioning of Raman amplifiers. The sensitivity of Raman amplification to any variability in the properties of fibers, inline passive components, or pump parameters is such that tuning of the whole system in the field is essential. Coarse tuning is performed before carrying any signal traffic over the link to optimize the Raman amplifier parameters and evaluate the performance. These measurements provide valuable information on transceiver specification requirements, as well as on the maximum number of channels that could be supported by the link. The tests involve setting the Raman gain at the target value, optimizing the gain flatness, and then measuring the noise figure.

To provide a relatively flat gain over a wide wavelength range, Raman amplifier pump sources (either wavelength- and/or polarization-multiplexed laser diodes or Raman fiber lasers) typically emit multiple wavelengths simultaneously. Adjustable amplifier pump parameters include the total output power as well as the output power at each individual pump wavelength. These parameters directly affect the gain and gain flatness. They require tuning in the field since gain and gain flatness depend on the particular characteristics of the individual span, such as the loss spectrum and the Raman gain coefficient of the installed fiber, and the position and nature of lumped losses.

The definition of Raman on/off gain is:

G (λ) = (P3-P2)/P1,

where P1 is the measured signal power received when amplification is off, P2 is the measured background amplified spontaneous emission (ASE) level at the signal wavelength when amplification is on, and P3 is the measured power at the signal wavelength received when amplification is on (all quantities in linear units).

Neglecting the ASE level in the above formula would give the on/off gain as P3/P1, a more commonly found definition for the gain. The on/off gain is simply the difference P3-P1 when P1 and P3 are expressed in dBm. Frequently, optical signal-to-noise ratios (OSNRs) in Raman amplification are on the order of 20 to 30 dB. In such cases, neglecting the ASE level introduces only a slight error when measuring the gain with an OSA. However, if the testing engineer measures the gain using a broadband detector, a substantial error can result from the integration of the ASE over the entire amplifier bandwidth.

The Raman gain must be measured in the field to obtain an estimate of the link loss budget. Generally, the Raman gain varies significantly with the type of the fiber used, and variations can even occur for fiber from the same production batch in the case of a single fiber type. These result from variability in dopant concentration or in absorption at the pump wavelengths. In addition, network specifics such as the position of a lumped loss relative to the pump injection point can result in Raman gain variability (see Fig. 1).

The Raman gain in silica fibers peaks in a region downshifted in frequency by 12 to 15 THz from the pump (approximately 100 nm at 1550 nm) and the gain curve, when plotted in dB, exhibits a full width at half maximum (FWHM) of approximately 7.5 THz (59 nm at 1550 nm). To improve the uniformity of the gain across a wide wavelength range, pump sources emitting at multiple wavelengths are used (see Fig. 2). However, a residual gain ripple always remains. In multispan systems, it is necessary to accurately control the gain ripple or it can systematically accumulate along the link.

In DWDM systems where the application of Raman amplification results in relatively high signal powers approaching the end of the span, another phenomenon can become significant. The shorter-wavelength signals can act as Raman pumps for the longer-wavelength channels. This leads to a transfer of power from the short wavelength signals to those at longer wavelengths and an on/off gain curve exhibiting a positive tilt with wavelength. As with the magnitude of the Raman gain, the gain flatness achievable with a given selection of pump wavelengths can vary significantly with fiber type.

Prior to the system commissioning phase, field measurement of the Raman gain flatness allows the network designer to be better informed when he or she is considering sacrificing some net gain for a better gain uniformity by inserting a gain-flattening filter. There are several ways to measure Raman gain and gain flatness. The measurement techniques and their accuracy differ according to the test equipment used. Each has its own set of advantages and disadvantages.

In the case of optical time-domain reflectometry (OTDR), a measurement is performed solely at the pump wavelength.2 This method makes use of a mathematical model to interpret the loss profile along the link at the pump wavelength. It can provide an estimate for the Raman gain at a specific signal wavelength Vs. launched pump power, assuming the Raman gain coefficient for the particular fiber type at that wavelength is known. The measurement can be interpreted as a deviation or penalty in comparison to an ideal link. The advantage of this technique is that it is not necessary to actually connect a pump source to the link. The main use of this technique is the quick screening of links to assess their suitability to support Raman amplification.

Another technique for measuring Raman gain is the use of a tunable laser source (TLS) in conjunction with an optical spectrum analyzer (OSA). This combination enables the testing of very long links owing to the large available signal power. This advantage applies mainly to long repeaterless links. In multispan systems, the individual spans are relatively short and the tuning of Raman amplifiers is usually performed one span at a time. Simultaneous testing of multiple spans including several amplifiers would be a very difficult task because of the large number of parameters involved.

This method., however, has several drawbacks. The high coherence of the source can cause interference effects and produce stimulated Brillouin scattering. The high degree of polarization of the TLS means that the polarization-dependent loss or polarization-dependent gain of the link can affect the repeatability measurement. More important, measuring gain one wavelength at a time is far from normal operating conditions where tens of channels are present simultaneously. With single-wavelength measurements, it is impossible to correctly appreciate the gain tilt from interchannel Raman pumping and therefore to adjust the amplifier parameters to correct for this effect. It is also impossible to simulate the resulting across-the-band gain saturation in this way.

The use of a broadband ASE source and an optical spectrum analyzer is another way to measure Raman gain and gain flatness. A broadband ASE source covering the same wavelength range used in normal operation is a good substitute for the actual DWDM signals (see Fig. 3). Even for cases where interchannel Raman pumping tilts the on/off gain profile or where significant gain saturation occurs, this technique will produce a gain profile similar to that obtained with the actual signals, provided that the output power of the source is similar to the planned composite signal launch power and is passed through a pre-emphasis filter to produce a spectral profile similar to the planned channel launch spectrum.

When combined with a high-acquisition-rate OSA with automatic trace subtraction, this method allows real-time tuning of amplifier parameters (see Fig. 4). An automated routine can measure both the gain and gain flatness and adjust the amplifier parameters in a continuous control loop until the gain and gain flatness reach target values.

For cases where the OSNR is high, the measurement consists of taking an OSA scan of the amplified source signal on a logarithmic scale and then subtracting from it a reference scan of the received signal taken while the pump is off. This immediately gives the Raman gain and gain flatness. If the amplifier ASE background level is not negligible compared to the signal power, then a scan of the amplifier ASE, taken with the source turned off, is also required.

For this measurement technique, an ASE source with a high spectral stability is highly recommended to allow taking measurements over long periods without having to shut down the pump to take a fresh reference scan of the source. The maximum testable fiber length with this technique is typically 150 to 200 km (depending on the output power of the ASE source), which is larger than typical distances between two successive amplifiers. It is also important to evaluate the noise figure of a span because it will induce a penalty on the transmission budget.

Christian Stauter is a product manager, in the Industrial and Scientific Division of EXFO, 465 Godin Ave., Vanier, Québec, Canada G1M 3G7; and Wallace Clements is the director of engineering and development at MPB Communications Inc., 147 Hymus Blvd., Pointe-Claire, Québec, Canada H9R 1E9. Christian Stauter can be reached at

1. T. Hoshida et al, OFC 2000, paper M14.

2. Movassaghi et al., J. Lightwave Tech. 16, 812.

More in Home