定义：可以降低光束功率的器件。

光衰减器可以使光束衰减，即降低其光功率。衰减的成都通常由光密度来描述，单位为分贝，有时描述为透射或者阻挡光功率的百分比。 

目录

• 典型要求
• 光衰减器类型 
• 2.1 吸收滤波器
• 2.2 反射型衰减器
• 2.3 偏振型衰减器
• 2.4 光纤衰减器
• 量子噪声效应

典型要求 

• 有些情况下，一定程度的衰减就足够了（例如，10分贝），而有的情况下需要采用可变光衰减器（Variable optical attenuator, VOA），其中衰减程度是可以改变的，例如可以通过把手手动调节。另一个这种的方案是具有阶梯式的可变衰减。有些衰减器中衰减是由电信号控制的。如果衰减调节非常快速，该器件实际上就可以看做光调制器。（任意的光学强度调制器都可以看做可变衰减器。） 
• 在衰减功率的同时不影响其横切面也非常重要。例如，当采用只能工作在有限光功率范围内的光束质量分析仪来分析高功率激光光束时就非常重要。显然的，如果光束形状因衰减器而产生畸变的话就会存在问题。 
• 另一个方面就是其波长依赖性。通常需要在特定波长范围内具有近似相同的衰减，例如，在整个可见光波段。也有中性密度滤波器是基于此目的的器件。 
• 在有些应用中，需要光学损耗与偏振无关，或者更准确的说，偏振相关损耗非常低。而有些情况下，例如，采用线偏振激光光束时，这就不太重要。 
• 尤其当激光光束为高功率的情况，衰减器的功率使用范围是另一个很重要的方面。不仅关系到避免对衰减器的损伤，还会对衰减光束不利，例如光束畸变，或者发热导致衰减程度的改变。 
• 很多光衰减器适用于自由空间光束，还有些是光纤耦合类型或者工作在波导结构中等。 

光衰减器类型 
由于不同应用对衰减器的要求不同，如上所述，因此实际中存在很多种类的光衰减器，采用了不同的物理机制。以下是集中最重要的类型。 

吸收滤波器
光学吸收是实现光学衰减的一种方式。例如，在掺杂玻璃种，可以通过掺杂的类型和掺杂浓度来控制在特定波长范围内光的吸收量。
通常来说，吸收滤波器都是片状形式。几个片状吸收滤波器可以实现很大程度的衰减。如果移除或者改变某些滤波器，则可以实现阶梯式可变衰减器。 
采用滤波器转盘可以实现连续可变的衰减，其中吸收量随着绕旋转轴的角度而变化。
还有些可变衰减器是利用合适的机械装置得到变化。任意一种情况下，在光束区域衰减都有一定程度的变化。该效应在阶梯型改变的衰减器情况下回避免。 
为了避免干涉效应和背向反射，通常使滤波片倾斜与入射光束有一定的角度。 
当激光脉冲很强时，滤光片的吸收会在脉冲前或者后一段时间内发生饱和（参阅饱和吸收器）。有些情况下，需要将此效应考虑在内，可能无法采用饱和吸收器。 
采用吸收滤波器非常方便，成本低，所需空间小，并且几乎不会引入偏振相关损耗。
但是，由于衰减的功率被转化成热，在高功率情况下会存在热效应。这会引起光束的畸变，甚至引起衰减器的损伤。 
中性密度滤波器也可以制作成吸收滤波器，需要采用合适的掺杂来覆盖较大的波长范围。 

反射型衰减器 
任何不具有涂层表面的透明材料都会存在一定的菲涅尔反射，可以用来进行衰减。通常利用的是反射光束，功率是入射光束的一部分。如果采用的滤波片是平行的表面，由于两个表面都会发生反射，就会出现问题。这时就需要使入射光书的入射角非常大，这样两个反射光束就可以分离开。另外，还可以采用非平行的表面，例如棱镜。 
对于不适用于吸收原理的高功率情形，中性密度滤波器可以利用反射表面的反射光。 
入射角度越大，表面反射率的偏振相关性越强（根据菲涅尔方程计算）。如果采用p偏振光，入射角接近布儒斯特角的话，偏振相关性最强。由于s偏振光的反射率远高于p偏振光，那么反射光可能主要是入射光中很小部分的s偏振光的反射光。因此，采用反射型的滤波器就不适用了。比较稳定的得到强衰减同时没有很强的偏振相干性的方法是利用小角度入射时的多重反射，但是这需要采用更多的光学元件，也需要对准。 

另一种方法是采用电介质涂层来改变表面的反射率。通过使用相对高反射率的涂层，透射光可以具有很大程度的衰减。但是，并不建议采用很高反射率的涂层来得到强衰减，因为涂层很小的透射率空间变化很大，因此衰减会发生变化，从而会引起光束畸变。因此，采用多个这种类型的滤波器可以实现更强的衰减。 
从原理上看，通过改变入射角度也可以得到可变的衰减。但是，即使是透射型滤波器，也会引起可变的光束偏移，通常这是不能接受的。一种解决办法是采用两个这种滤波器，通过精确的机械调节旋转角度从而经过这一器件不会在纵向产生变化的光束偏移。但是这种情况下，不可避免的衰减会有很大的偏振相关性，除非采用两片这种滤波片。 
由于反射型衰减器中存在不用的输出光束，因此需要束流收集器，主要是考虑到人眼安全，并且避免对器件的影响。 

偏振型衰减器
尤其是线偏振光束会采用偏振型可变衰减器。最常用的方案是采用二分之一波片和偏振片的结合。绕着光束轴向旋转波片，偏振方向会发生变化，因此偏振片的衰减度会相应的发生变化。如果采用合适的偏振片，它可以应用在非常高的功率情况下。反射的光束则会进入束流收集器中。对于非偏振光束，尤其是无特定偏振方向的光束，这种衰减器非常难实现。 

光纤衰减器
很多种类的光衰减器都可以用在光纤应用中，例如，光纤通信。 
在这些器件中，光纤中的光变成准直的自由空间光束，进入自由空间光衰减器，然后进入输出光纤中。还有的情况是，采用光纤器件实现衰减，不涉及自由空间光束。 

一种纯光纤光学的方案是采用一个光纤耦合器，一部分功率不进入使用的输出端口，而是另一个端口。另一种方案是利用可变的耦合损耗，它受光纤端口位置的变化而影响。例如，可以通过改变输出光纤的横向位置或者两光纤见的空气间隙，得到可变的损耗，并且损耗与波长无关。（这一原理在采用单模光纤时最好。）还有采用掺杂光纤的光学衰减器，对某一特定波长区域的光有吸收。 

采用光纤实现可变的衰减还有很多其它的方法。例如，可以利用弯曲损耗或者与衰逝波有关的损耗，例如锥形光纤。方法的选择取决于特定应用的具体要求。 
有些光纤衰减器采用带连接头的光纤，具有光纤连接器的光缆。其它的与光纤连接器集成起来。 
对于多模光纤，衰减器的衰减程度并不十分与传播模式有关。这通常需要采用一些体光学元件实现衰减。 

量子噪声效应 
从宏观方面看，光衰减器的功能类似于简单的使光束功率以一定的因子减小。而微观方面，存在一些其他细节性的过程。简单来说，线性光衰减器从光束中移除一些光子，每一个入射光子被移除的概率是相等的。这种随机性也会引入量子噪声。也正因为此，光学测量的信噪比也会因为衰减存在而降低，并且这种效应不能通过采用光学放大器进行放大而消除，即使放大器不引入附加噪声的情况下.

Definition: devices which can reduce the optical power e.g. of a light beam

More specific terms: variable optical attenuators, fiber-optic attenuators

Opposite terms: optical amplifiers

Optical attenuators are devices which can be used to attenuate a light beam, i.e., to reduce its optical power. The amount of attenuation is often specified in terms of an optical density or in decibels, sometimes in percent of transmitted or blocked optical power.

Typical Requirements

• In some cases, a fixed degree of attenuation (e.g., 10 decibels) is sufficient, whereas in other cases one needs a variable optical attenuator (VOA), where the degree of attenuation can be adjusted, for example manually using some knob. An intermediate solution is to have a stepwise variable attenuation. In some cases, the adjustment of the attenuation can be controlled with an electronic signal. If this adjustment can be made quite fast, the device is actually an optical modulator. (Of course, any optical intensity modulator could be regarded as a variable attenuator.)
• It is often essential that the power can be attenuated without affecting the transverse beam profile. This is the case, for example, when a high-power laser beam needs to be characterized with a beam profiler which can handle only a limited amount of optical power. Obviously, it would be a problem if the beam profile would be distorted by the attenuator.
• Another aspect is the wavelength dependence. Often, one needs an approximately constant amount of attenuation in a certain wavelength range – for example, for all visible light. There are so-called neutral density filters which are made for such purposes.
• For some applications, it is vital that the obtained optical losses are not dependent on the polarization, or more precisely, that the polarization-dependent loss is very low. In other cases, for example when working with linearly polarized laser beams, that dependence may not be relevant.

The limited power handling capability of some optical attenuators can be a problem.

• Particularly when working with high-power laser beams, the power handling capability of an attenuator may be another important aspect. It may be relevant not only to safely avoid damage of the attenuator, but also detrimental effects on the attenuated beam, such as beam distortions, or thermally induced changes of the degree of attenuation.
• Many optical attenuators are applicable to free-space beams, whereas others are of fiber-optic type or work for waveguides of other kinds.

Different Types of Optical Attenuators

Because the requirements can be very different for different applications, as shown above, a wide range of different kinds of optical attenuators is used in practice, which can exploit different physical mechanisms. The most important types are explained in the following.

Absorbing Filters

Optical absorption is one of the possibilities for realizing optical attenuation. One can obtain it, for example, in doped glasses, where the type and concentration of the dopant can be used to control the amount of absorption in a certain wavelength range. Typically, one uses such kind of absorbing filters in the form of plates. Several of such plates can be used in series in order to realize a higher degree of attenuation. By removing or exchanging some of the filters, one can realize a stepwise variable attenuation.

Continuously variable attenuation can be realized with a filter wheel (optical attenuator wheel), where the amount of absorption varies along a circle around the axis of rotation. There are also variable attenuators which are translated linearly by some suitable mechanics. In any case, the attenuation may be somewhat variable within the area of a light beam. That effect is avoided in devices where the attenuation varies in steps.

In order to avoid interference effects and problems with back reflections, one often slightly tilts such plates against the incoming beam.

When working with intense laser pulses, the absorption of a filter may be saturated (→ saturable absorbers) for some time during and after the pulse. This effect has to be taken into account; in some cases, it makes absorbing filters unusable.

Using absorbing filters can be a convenient and low-cost approach, requiring little space and introducing essentially no polarization dependence. However, as the removed optical power is converted into heat, thermal effects will occur at high power levels. These can easily lead to distortions of the beam profile or even to damage of the attenuator.

Neutral density filters can also be made as absorbing filters by using an appropriate combination of dopants to cover a wide wavelength range.

Attenuators Based on Reflection

Any uncoated surface of a transparent material exhibits some amount of Fresnel reflection, which may be exploited for attenuation. Often, one uses a reflected beam, carrying a few percent of the incoming power. When using a plate with parallel surfaces, problems can result from the fact that reflections occur at both surfaces. It may then be necessary to operate such a plate at a sufficiently large angle of incidence, so that the two reflected beams are clearly separated from each other. Alternatively, one may use a device with nonparallel surfaces, such as a prism.

Neutral density filters can also work based on reflecting surfaces in order to make them usable for higher optical powers than devices based on absorption.

Working near Brewster's angle can provide a high attenuation in a single reflection, but may easily lead to problems due to the very strong polarization dependence of the loss.

The larger the angle of incidence is, the stronger is the polarization dependence of the reflectivity of the surface (as can be calculated with Fresnel equations). This dependence becomes extreme when trying to obtain a strong attenuation by using a p-polarized beam with an angle of incidence near Brewster's angle. As the reflectivity for s-polarized light is then far higher, the reflected light may actually be dominated by some small unwanted content of s-polarized light in the input beam. Therefore, this approach is often not suitable. A more reliable way of getting strong attenuation without a strong polarization dependence is to use multiple reflections at small angles in sequence, even though this may require more optical elements and more alignment.

Dielectric coatings with a high reflectivity are also often not well suitable for attenuating a light beam.

Another possibility is to use a dielectric coating to modify the reflectivity of the surface. By using a relatively highly reflecting coating, one can obtain a high degree of attenuation for the transmitted light. However, it is again not advisable to use a very high reflectivity (e.g., well above 99%) for obtaining strong attenuation, because the small transmissivity of the coating may then exhibit substantial spatial variations, which can lead to inconsistent attenuation and also to beam distortions. Therefore, strong attenuation is better realized by a combination of several such filters.

In principle, a variable attenuation is possible by varying the angle of incidence. Even for filters working in transmission, however, this will lead to a variable transverse beam offset, which may often not be acceptable. A solution can be to use two such filters in sequence, where the rotation angles are coordinated by some precision mechanics such that no variable transverse beam offset occurs after the device. A substantial polarization dependence of the attenuation can hardly be avoided with that approach, except with a second pair of such plates.

For the non-used output beam of an attenuator based on reflection, one often requires a beam dump for reasons of eye safety or for avoiding disturbing effects in the device.

Attenuators Based on Polarization

Particularly for linearly polarized beams, variable attenuation is often obtained based on polarization. The most common approach is to use a λ/2 waveplate in combination with a polarizer. By rotating the waveplate around the beam axis, the direction of polarization and therefore the degree of attenuation at the polarizer can be varied. Using a suitable type of polarizer, this principle can be realized at very high power levels. The power of the rejected beam is then sent into a robust beam dump. For unpolarized beams and particularly for beams with an unspecified polarization state, such attenuator devices are more difficult to realize.

Fiber-optic Attenuators

Various types of optical attenuators can be used in the context of fiber optics – for example, in optical fiber communications. Some of those just contain some type of bulk-optical attenuator, placed between two lenses for collimating the beam coming from one fiber and launching it the output fiber. Others are purely fiber-optic devices, e.g. containing some piece of fiber where the fiber core is doped with some absorbing material.

See the article on fiber-optic attenuators for details.

Quantum Noise Effects

On the quantum level, the effects of optical attenuators are not trivial to understand.

On a macroscopic level, the function of an optical attenuator is simply to reduce the optical power of the light beam by a certain factor. On a microscopic level, however, additional non-trivial details come into play. In a simplified picture, a linear optical attenuator removes some of the photons from a beam, where the probability of removal is equal for each incoming photon. The involved randomness implies that additional quantum noise is introduced. For that reason, the signal-to-noise ratio of optical measurements can be degraded as a result of attenuation, and that effect can generally not be undone by subsequent optical amplification in some kind of optical amplifier, even if that amplifier would not introduce any excess noise.