外文翻译--10Gbs的高性能PIN和APD光接收器.docxVIP

high performance 10 gb/s pin and apd optical receiversabstract the increasing market demand for high-speed optical- transmission systems at rates of 10 gb/s has resulted in technical challenges for suppliers of high-performance, manufacturable opto-electronic components and systems. in particular, the performance of the inp semiconductor devices, integrated circuits (ics) and hybrid ic modules strongly influences the achievable transmission capability.an optical receiver design is presented which incorporates an inp-based p-i-n (positive-intrinsic-negative) photodetecto(pd) or avalanche photodetector (apd) and a gaas high electron mobility transistor (hemt) pre-amplifier integrated circuit. several aspects of the receiver design are presented, including the p-i-n pd and apd structures and performance, pre-amplifier performance, hybrid module layout and electrical simulation and results. the use of analytical techniques and theory commonly used in the design of microwave amplifiers and circuits is emphasized. receiver test results are included which are in close agreement with predicted theoretical performance.introduction over the past 15 years the demand has continued to increase for higher speed and higher performing opto- electronic components. components designed to operate at data rates of 155 mb/s through 1 gb/s are now used in high volume, are manufactured with high yields, and are available from several suppliers. components designed for 2.5 gb/s are fast approaching this manufacturing status as well. the emphasis now for new opto-electronic product development centers around performance requirements at transmission rates of 10 gb/s and higher.the optical receiver represents one of the key components in optical-fiber based communication systems, and is generally considered as a component, or module, which is available with specified levels of electrical functionality or integration. the basic elements of an optical receiver module are a photodetector, pre-amplifier, limiting or agc(automatic gain control) amplifier, and clock and data recovery circuitry. at data rates of 2.5 gb/s and below, the system designer can currently purchase the optical receiver elements in various levels of integration, from a discrete photodetector module to a fully integrated clock and data recovery module.in many multi-element systems and circuits the performance is strongly influenced by those elements which are located near the input of the system or circuit. this is certainly true in a digital optical receiver where the performance of the photodetector and pre-amplifier elements will have a strong impact on receiver and system performance. in addition to the individual performance of these two elements, the electrical and physical design of the interface between them is equally critical. at speeds of 10 gb/s, the current focus for suppliers of optical receivers is the development of modules, which incorporate the photodetector and pre-amplifier elements. naturally as time progresses, the additional electrical functions will be incorporated into the modules as well. this paper focuses on the design and characterization of 10 gb/s optical receiver modules that incorporate the photodetector and pre-amplifier elements.optical receiver basics before considering 10 gb/s receiver design a brief review is presented of optical receivers for digital applications. a basic schematic of an optical receiver front-end is shown in figure i. the schematic includes the photodetector and pre- amplifier elements.the key perfomance requirements of an optical receiver are high sensitivity, wide dynamic range and adequate bandwidth for the intended application. the purpose of the pd is to convert the incident optical signal to an electrical current. the photodiode should have the following performance characteristics: high responsivity (quantum efficiency), low dark current, low capacitance and wide bandwidth. for applications at optical wavelengths of 1310nm and 1550 nm, high quantum efficiency ingaas / inp type photodetectors are commonly selected. the purpose of the pre-amplifier is to convert the photocurrent from the pd into a usable voltage that can be further processed. the common pre-amplifier technology used in optical receivers is transimpedance amplification, (tia) due to its optimum trade-off between noise, dynamic range and bandwidth. other types of pre-amplifiers include high-impedance and low-impedance (e.g. 50) designs.p-i-n photodetectorsour p-i-n photodiode is a double-heterojunction structure grown on an n+-inp substrate and consists of an n+ -1np buffer layer, an n-ingaas active layer, and an n inp cap layer. the buffer growth precedes the active layer growth to provide a surface with fewer defects than exist on the bare substrate surface. the in0.53ga0 .47as active layer is lattice-matched to inp and, with a bandgap g 0.75 ev, is sensitive to light with wavelengths shorter than 1.65 m. the device exhibits a short-wavelength cutoff at 0.90m since more energetic short-wavelength light is absorbed in the inp (g 1.35 ev) before it reaches the ingaas. the larger-bandgap inp cap layer reduces surface leakage (relative to ingaas) and is passivated using si3n4. using etched patterns in the si3n4 as a mask, high-reliability planar diodes are created by diffusing a p-type dopant (zn) to form one-sided p+-n- junctions just below the inp-ingaas (cap-active) heterojunction (see fig. 5a). contact metallization alloyed to the diffused junction allows electrical contact to the p-side of the junction. after thinning the substrate to 120 m, the back side of the wafer is metallized to provide electrical connection to the n-side of the junction. apertures in the backside metallization allow optical coupling to the active region in a back-illuminated geometry, and an anti-reflecting (ar) si3n4 coating is present in the aperture to eliminate reflection from the air-inp interface.several of the critical device characteristics pose conflicting design constraints that must be optimized for good high frequency performance. of primary importance is the ability to achieve sufficient 3-db bandwidth. the standard p- i-n diode has two fundamental bandwidth limitations: (i) finite carrier transit time and (ii) rc roll-off. the finite transit time taken by photon-induced carriers to traverse the active region can be shortened by reducing the thickness of the active region, but only at the expense of increased capacitance per unit area and lower quantum efficiency (which results in lower responsivity). the tendency towards increased capacitance for thinner active layers can be offset by reducing the total junction area, but this leads to greater difficulties in achieving high optical coupling efficiency and reliable electrical connections (e.g., by wire bonding).for 10 gb/s performance, the conflicting requirements just described can be adequately resolved using a device diameter of 30 um. in this case, an active layer width wa 2.3 um gives rise to average transit times of about 25 ps implying a maximum bandwidth f3-db 18 ghz. the resulting capacitance of 0.15 pf contributes a bandwidth limitation of f3-db 21 ghz assuming a 50load. (note that low contact resistance is yet another device requirement necessary for minimizing rc bandwidth limitations.) direct measurement of a wire-bonded photodiode using microwave probes has confirmed a device bandwidth of 20 ghz. finally, assuming an ar coating reduces surface reflections to negligible levels, the quantum efficiency, of such a device is still reasonably high:= l - exp(-wa) 80% where the absorption coefficient 0.70 um-1 for n- -ingaas and a wavelength of 1.55 um. avalanche photodetectorsthe design of an avalanche photodiode for use at 10 gb/s is considerably more difficult than for a p-i-n diode, but the benefits to receiver sensitivity can be substantial. the utility of the apd is that it provides a means of circumventing the basic quantum limitation of the p-i-n diode, which dictates that each photon can generate only a single electron-hole pair. the apd structure is designed to create a region of electric field sufficiently high that a single carrier is accelerated enough to generate additional electron-hole pairs through impact ionization. newly generated carriers are similarly accelerated, and so a single carrier can trigger an avalanche effect, which provides internal gain resulting in many electron-hole pairs generated per absorbed photon. all ingaas-inp apds employ a separate absorption and multiplication (sam) structure (see fig. 2b) since high fields in the ingaas absorption region would induce large tunneling currents before the onset of the avalanche effect. the low- doped ingaas absorption and inp multiplication regions are spatially separated by a layer of n-doped inp used to maintain low field in the ingaas and high field in the inp. the inp multiplication region is terminated by a p+-n- junction in inp created by a diffusion technique similar to that used in fabricating p-i-n diodes. the polarity of the device is determined by the fact that holes have a higher probability than electrons for ionizing collisions in inp; therefore, the structure is designed to inject photoexcited holes from the ingaas into the inp multiplication region to seed the avalanche process. although there is noise inherent in the avalanche effect (due to stochastic fluctuations in the number of carriers generated per photon), as long as this avalanche noise is no greater than the noise from other components in the receiver (such as amplifiers), the apd can provide a significant increase in the receiver signal-to-noise ratio. this is particularly attractive at higher frequencies at which increased amplifier noise is unavoidable.apd design is complicated by a number of factors. foremost among these is the difficulty in controlling premature avalanche breakdown at the edge of the device.the geometry of planar diffused junctions includes inherent curvature at the junction periphery. this curvature typically causes locally enhanced electric fields, and the consequent enhanced avalanche at the junction periphery leads to an undesirable non-uniformity in the multiplication profile across the device. to solve this problem, we have used a novel double-diffusion technique to shape the diffusion profile so that edge fields are reduced.achieving high bandwidth apd performance involves the same transit time and rc limitations described for the p-i-n diode. however, there is an additional bandwidth constraint imposed by the avalanche process itself in the form of a fixed gain-bandwidth (g-bw) product. the carrier acceleration and impact ionization involved in creating avalanche gain require an avalanche build-up time proportional to the gain, so the higher the operating gain is, the lower will be the device bandwidth. (note that another new bandwidth- limiting process is introduced since all electrons created in the multiplication layer during the avalanche process must traverse the ingaas absorption region to reach the n-contact.) higher g-bw products result when thinner, higher-field multiplication regions are used. with a multiplication layer thickness of -0.2 m, we have achieved g-bw products of about 90 ghz (see fig. 3).a very attractive attribute of our apd design is the fact that it is based on well-established processes identical to those used in fabricating planar p-i-n diodes. this can be expected to result in favorable production yields and extremely high- reliability devices. we have confirmed that our 2.5 gb/s apds (based on a structure similar to that described above for the 10 gb/s device) have reliability performance comparable to p-i-n diodes, and initial lifetesting on our 10 gb/s apds has provided similar results. conclusionsthe design of manufacturable optical receiver modules has been presented for 10 gb/s applications. both a p-i-n or apd detector can be used, depending on sensitivity requirements. the design and fabrication of the planar ingaas-inp photodetectors was presented along with a physical description of the optical module. a detailed electrical analysis based on microwave cad simulation was presented with an emphasis on the identification of the critical circuit elements that effect the microwave performance. finally, prototype p-i-n and apd receiver test results were presented and compared to the simulated results, showing a relatively strong correlation. 作者：jim rue, mark mer, nitish agrawal, stephen bay and william sherry国籍：美国出处：electronic components and technology conference，1999. 10gb/s的高性能pin和apd光接收器摘要随着市场需求对传输速率为10gb/s的高速光纤系统日益增长，使之对生产高性能，制造光电元件和系统的供应商的提出了更高的技术要求。特别是inp半导体器件，集成电路（ic）电路和混合ic模块的性能，对实现传输能力有着的强烈影响。在一种光接收机的设计上，人们提出结合采用基于inp基脚的p-i-n结光电探测器或雪崩光电探测器（apd）和砷化镓高电子迁移率晶体管（hemt）的前置放大器集成电路。光接受器的设计，包括在p-i-n结光电探测器和apd的结构和性能，前置放大器性能，混合动力模块布局和电气模拟结果的几个方面的设计。经常强调在微波放大器及电路设计中使用的分析技术和理论。使接收器的测试结果与预测的理论性能接近一致。引言在过去15年中，人们对于速度更高和性能更高的光电组件的需求一直在不断增加。设计出来的数据传输速率为155mb/s与1gb/s的光电组件现在正在在被大批量使用，并且其容易生产且产量高，容易从几个供应商中得到。设计的传输速率为2.5千兆/秒的光电组件的生产同样快速达到了这个状况。现在光电产品研发中心新的重点是研发传输速率性能要求约为10gb/s和更高的元件。光接收器是基于光纤通信系统的关键部件之一，一般被认为这是装备在指定的电气功能或一体化水平设备的组件或模块。光接收器模块的基本元件是光探测器，前置放大器，限制或agc（自动增益控制）放大器，时钟和数据恢复电路。如果从一个离散探测器模块到完全集成的时钟和数据恢复模块的数据传输速率要求在2.5gb/s及以下的，系统设计师目前可以购买在各种水平的集成光接收机的基本元件。 在许多元件系统和电路的性能受到那些位于附近的输入系统或电路的元件的强烈影响。这是确实存在的，在数字光接收机上的光电探测器和前置放大器元件的性能将会对接收器和系统性能产生强烈影响。除了这两种元件的独特影响外，它们之间的接口的电气和物理设计也是同样重要。在传输速度为10gb/s的基础上，光接收器供应商目前开发的重点是发展结合光探测器和前置放大器元件的模块。自然随着时间的推移，更多的电气功能同样也将被纳入该模块。本文重点是描述和设计结合光探测器和前置放大器的传输速率为10gb/s的光接收模块研究。光学接收器的基础在考虑10gb/s接收器设计的之前,简要回顾以前提出的数字化应用中的光接收器。光接收器前端的一个基本原理如图1所示，示意图包括光探测器和前置放大器的元件。图1 光接收器前端的一个基本原理图光接收器的关键要求是在实际应用中要有很高的灵敏度，宽动态范围和足够的频宽。光电二极管（pd）的目的是将入射的光信号转换成电流。光电二极管应该具有以下性能特点：高响应（量子效率），低暗电流，低电容和宽频带。对于应用在1310nm和1550nm波长的光的波长，量子效率高的ingaas / inp的类型探测器是常用的选择。前置放大电路的目的是把从光电二极管（pd）的光电流转换成可用的电压，从而能进一步操作。由于对噪声、动态范围和频宽之间的最佳权衡，常见的用于光接收器前置放大电路具有高阻抗。其他类型的前置放大器，包括高阻抗和低阻抗（例如50）的设计。pin光电探测器我们的pin光电二极管是以一个n+-inp为衬底的双异质结结构和一个n+-inp缓冲层，一个n-ingaas积极层和n型inp的帽层。在积极层增长之前，缓冲区增长提供的表面比存在的裸露材料表面缺陷少。in0.53ga0 0.47活性层匹配的电势差为0.75 ev的能隙的inp晶格，对波长小于1.65微米的光很敏感。该器件在遇到0.90m波长以下的光时才截止接收，因为更有活力的短的波长的光在到达ingaas之前被inp（电势1.35 ev）吸收。较大能隙的inp帽层可减少表面泄漏电流（相对的ingaas），并使用钝化si3n4 。使用口罩的si3n4的蚀刻模式，用掺杂（锌）的扩散p型去形成片面的p-n结创造的可靠性高的平面二极管在inp-gaas（帽活跃）异质结的正下方（见图2a）。使金属合金与扩散结连接，以便允许电跟结区的p端接通。在把衬底减薄到120微米后，薄片的背后的金属是用来提供电气与n端连接的。背面金属上的光圈允许光耦合到处在被光的活跃的几何区，在光圈中的防反射（ar）si3n4涂层是来消除来自air-inp接口的反射。设计时一些关键设备造成的冲突必须得到约束，并得到优化从而具有为良好的高频性能。最重要的是能够实现足够的3db带宽。标准pin二极管有两个基本的带宽限制：（1）有限的渡越时间；（2）信号衰减。只有在增加单位面积电容和较低的量子效率（导致响应较低）的费用时，才可以通过缩短减少厚度活跃的地区来减少光子传输到活跃地区的时间。对减薄活性层而靠电容增加的趋势，可以通过降低总交界处被来抵消，但是这将使实现高光耦合效率和可靠的电气连接（例如，焊线）具有更大的困难。为实现10gb/s的性能，并解决描述的相互矛盾，只要使用直径30微米的设备就行了。在这种情况下，有源层宽度2.3 um 的wa产生的平均运输时间约25ps，这就意味最大带宽（f3-db ）为18ghz 。假定有50负载