Most fiber-optic attenuators exhibit a relatively high return loss (at least several dozens of decibels), i.e., there is not much light which is reflected back into the input fiber. For some sensitive applications, e.g. when using an attenuator before. Most fiber-optic attenuators exhibit a relatively high return loss (at least several dozens of decibels), i.e., there is not much light which is reflected back into the input fiber. For some sensitive applications, e.g. when using an attenuator before or after a high-gain fiber amplifier, one may have two use attenuators with particularly high retu. Generally, the obtained insertion loss has some dependence on the optical wavelength. Some attenuators have a relatively strong wavelength dependence and are made for working in narrow wavelength regions, e.g. with a bandwidth of only 20 nm around a center wavelength of 1550 nm. Others are optimized for a weaker wavelength dependence, making them u. As light in fibers often does not have a well defined polarization state, it is important that a fiber-optic attenuator exhibits only a minimum amount of polarization dependence.For single-mode devices, the insertion loss can not depend on the direction of propagation, as long as no non-reciprocal parts are used, as e.g. in a Faraday isolator. For multimode devices, however, some loss difference is possible in conjunction with a mode dependence.For many applications, it will not be a problem if the obtained insertion loss slightly deviates from the specification (e.g. by 1 dB), or if it slightly changes over time. Example cases, however, one may require a higher precision.