High-speed photodetectors in optical communication system

    Corresponding author: Jianguo Liu, jgliu@semi.ac.cn
  • 1. State Key Laboratory on Integrated Optoelectronics, Institution of Semiconductors, Chinese Academy of Sciences, Beijing 100083, China
  • 2. School of Electronic, Electrical and Communication Engineering, University of Chinese Academy of Sciences, Beijing 100049, China

Key words: high-speed photodetectorsPIN photodetectorspackagingintegration

Abstract: This paper presents a review and discussion for high-speed photodetectors and their applications on optical communications and microwave photonics. A detailed and comprehensive demonstration of high-speed photodetectors from development history, research hotspots to packaging technologies is provided to the best of our knowledge. A few typical applications based on photodetectors are also illustrated, such as free-space optical communications, radio over fiber and millimeter terahertz signal generation systems.


1.   Introduction
  • The abrupt increase of internet traffic has put tremendous pressure on communication capabilities. Three effective solutions, improving the transmission rate of a single channel[1], increasing the number of channels, and adopting an advanced modulation format technique[2, 3], are usually employed to increase communication capacities. In the future optical fiber communication system in which the transmission rate continues to increase, whether adopting optical time division multiplexing, orthogonal frequency division multiplexing or wavelength division multiplexing (OTDM, OFDM or WDM) technology, high-speed photodetectors (PDs) are crucial devices for achieving the photoelectric conversion of the signal, and it directly determines the performance of the communication system. A high-speed optical signal demodulation system requires specially designed PDs and detection systems, which need to meet the following factors: (1) broad bandwidth to accommodate the instantaneous variation of the incoming signal, (2) large response to the incident optical signal, (3) minimum of noise added by the detection systems.

    In an optical communication link, there are two main ways to carry out light detection, named by direct detection and coherent detection. Direct detection means that the signal light is directly incident on the photosensitive surface of the PD, which only responds to the intensity of the incident light radiation, so direct detection is widely used due to its simple and practical advantages. Coherent detection refers to the process that the signal light and the intrinsic light, which satisfying the phase matching condition are mixed in the PD, and then the difference frequency signal is output. Photoelectric coherent detection has many advantages, such as strong detection capability, high signal to noise ratio, good filterability and high stability and reliability. Generally, coherent detection is widely used in long-haul communications links by applying higher level formats with the increasing data rates[4, 5]. In the short reach regime, direct detection is often adopted in order to reduce the complexity of the system and therefore the cost and power consumption.

    There mainly exist two types of PDs, the PIN PD and the avalanche photodetector (APD), which are available for high-speed transmission systems. In recent years, optical communications have become the primary driving force for research and development of APDs[6]. Compared with the PIN detector, the APD detector has a higher sensitivity in optical receivers and it has been regarded as the better candidate for long haul communications in terms of their internal gain availability[7, 8], but at the cost of more complex bias circuits and epitaxial wafer structures. Recently, with the rapid developments in the technology of both PIN and APDs, the bandwidth of both is up to the 1 THz rate or higher. For the PIN PD, thermal noise is significant in the performance of the receiver. In the APD, both the shot noise and thermal play dominant roles[9]. In order to satisfy various requirements, several types of PIN PDs such as the waveguide photodiode (WGPD)[10], uni-traveling carrier photodiode (UTC-PD)[11], velocity-matched distributed photodetector (VMDP)[12] and traveling-wave photodetector (TWPD)[13] etc. have been developed.

    Packaging of the PD in which many issues must be taken into consideration such as impedance matching, coupling efficiency and structure of the biasing circuit has a significant influence on the high-frequency performance of the device. As a last step in manufacturing of the device, packaging of the PD is required to meet the constrains of cost, size, and environmental operability. In this article, the development history of the PDs are first presented. Then, research hotspots and packaging technologies are provided to the best of our knowledge. Finally, the application and future development prospect is described.

2.   History
  • The PDs, as one of the key components in optical communication systems, have been investigated passionately all the time with respect to the quantum efficiency, bandwidth, saturation current and dark current etc. The major limitations on the speed of PIN PDs are: (1) the diffuse time it takes carriers to get out of undepleted regions, (2) the drift time it takes carriers to get across the depletion region, (3) charge trapping at heterojunctions, and (4) the time it takes to charge and discharge the inherent capacitance of the diode plus any parasitic capacitance. However, quantum and coupling efficiency and bandwidth are mutually constrained in terms of the width of the depletion layer, so much effort has been concentrated on increasing data rates of PD’s for systems operating at high efficiency.

    From 1987 to 1990, a series performance enhancement of the optical receiver was reported including increased bandwidth and optimized structure. For example, a high-impedance pin FET receiver with an 8 GHz bandwidth and a similar design with a 16 GHz bandwidth but using HEMTs was demonstrated by the Gimlett group[14, 15], a 10 GHz PIN detector design comprised by GalnAs which followed by a three-stage, high-impedance amplifier and interconnected by impedance matching networks was demonstrated[16] and in Ref. [17], a 14.5 GHz bandwidth with a transimpedance design was also obtained.

    Generally, there exists tradeoff between the quantum efficiency and bandwidth, which are two important figures of merit for high-speed PDs. Faced with such a situation, an edge-coupled waveguide photodiode (WGPD) has been widely studied for overcoming the efficiency-bandwidth tradeoff in the vertically illuminated photodetectors (VPD). As the name implies, the WGPD has an optical waveguide structure which nearly permits the efficiency and bandwidth to be specified independently. In 1986, the first high speed edge-coupled WGPD with a coupling efficiency of 25% and a bandwidth of 28 GHz was demonstrated[18]. However, the coupling efficiency of the edge-coupled WGPD is relatively low. To improve the coupling efficiency, Kato et al. proposed a double-core multimode waveguide structure in 1991, which was the first breakthrough for the WGPD[19]. Then, WGPDs using the double-core multimode waveguide structure were reported successively with the bandwidth of 40 and 50 GHz, respectively[20, 21]. The CR time constant is another limiting factor for PD’s bandwidth. To overcome the degradation introduced by the parasitic effect, a mushroom-mesa waveguide structure was reported in 1994 by the Kato group[22], which realized a bandwidth above 100 GHz and the bandwidth-efficiency product was 70–90 GHz, this structure was considered as the second breakthrough for large-bandwidth WGPD’s.

    However, WGPD has not made a breakthrough in terms of high power output. With the increasing of signal speed, ultra-bandwidth PDs with a high saturation power become more significant for applications such as ultrafast measurement[23] and high-speed optical fiber communications[24]. In 1997, a novel structure called uni-traveling carrier photodiode (UTC-PD) was first proposed by the NTT Photonics Laboratories in Japan[25]. In the UTC-PD structure, only an electron is used as the active carrier, which is an essential difference from the traditional pin PD. The UTC-PD has two main advantages, the one is a higher operation speed due to the higher electron velocity compared with hole velocity in the depletion layer, and the other one is the higher output saturation current because of much less space charge effect in the depletion layer. For UTC-PD research, the NTT Corporation in Japan has been in a leading position. In Ref. [26], a UTC-PD at a wavelength of 1.55 μm which had a 3 dB bandwidth of 220 GHz was fabricated by the Ito group in 1999, only in the next year, they realized an InP/InGaAs UTC-PD with a bandwidth of 310 GHz[27]. In 2003, a packaged receiver module with a 1 mm connector and a chip with 310 GHz bandwidth has been fabricated; the module has a 3 dB bandwidth of 80 GHz at an average photocurrent of 10 mA[28]. In the same year, they presented a UTC-PD with a monolithically integrated matching circuit, which showed a high saturation output power of 20.8 mW at 100 GHz[29]. In 2007, Wu et al. discussed a PD that was composed of a UTC-PD and an improved evanescently coupled optical waveguide. The integrated UTC-PD realized a high responsivity of 1.04 A/W and a wide electrical 3 dB bandwidth up to 60 GHz[30].

    Compared with PIN PD, APDs can achieve higher 5–10 dB sensitivity. Therefore, many works have been reported on APDs. In 1994, a first waveguide superlattice APD was reported by the Shishikura group[31]. Then, an APD with a thinner photoabsorption layer was fabricated, which reached a 3.2 multiplication factor and 0.9 A/W responsivity at 20 GHz[32] in 1997. Using a similar approach, a broader bandwidth of 35 GHz was realized for APD by Nakata et al. in 2001, its gain-bandwidth product was 140 GHz[33]. In Ref. [34], the gain-bandwidth product of the APD was greatly improved up to 320 GHz, while the quantum efficiency only was 16% in the same year. In 2003, a broad bandwidth of 48 GHz, very high responsivity (> 1 A/W), and 11 mA saturation current were demonstrated[35].

    In the high capacity communication system, a single discrete device is not conducive to the system cost reduction, therefore, an integrated optoelectronic device is a nice choice to reduce system cost, make the system small and easy to achieve. In 1992, M. Tabasky et al. reported a four-channel hybrid receiver with a modulation bandwidth up to 400 MHz and the corresponding data rate in the NZR arrived at 800 Mb/s using a silicon substrate for packaging[36]. The 10-channel transmitter and receiver modules with the rate of 800 Mb/s/ch were fabricated in 1996[37], which were composed of a low capacitance pin PD array, a low threshold current LD array and high-speed Si-bipolar array ICs. In 2003, the NTT Corporation demonstrated a 1.25 Gb/s × 8 channel photoreceiver using silica-based AWG[38].

3.   Research hotspots of photodetectors
  • After the success of 10 Gb fiber ethernet, the 100 Gb ethernet is considered as an ideal choice for the next generation of high-speed data transmission systems. For PDs, the bandwidth is usually increased by reducing the area of the photosensitive surface, but this tends to bring the difficulties for optical coupling in the package progress. As a consequence, the reliability will be reduced and the production cost will be increased. Faced with this situation, waveguide PDs and PD array modules are still attractive choices and they can be integrated with different components to improve the performance of the communication system.

    In 2013, a packaged WGPD integrated with a vertically-tapered spotsize converter and a 1 × 4 MMI coupler was fabricated[39]; this device was composed of an array of 4 PDs in order to increase the saturation current and it had 30 GHz 3 dB bandwidth and high output power of 7.5 dBm at 20 GHz. In the same year, a PD with above 90 GHz 3 dB bandwidth was demonstrated[40]. In Ref. [41], a waveguide integrated PIN PD chip with a 3 dB bandwidth of 130 GHz has been fabricated in 2015 by Runge et al. For microwave photonic applications, for example, photonic generation of terahertz signals and radio over fiber (ROF) links, the performance of high power is another demand for PDs. In 2014, Chenhui Jiang group presented a high linearity and high power PD module using the UTC-PD chip, which can be used for microwave photonic applications[42]. The Gan Zhou group in 2017 demonstrated a monolithic high-speed high-power waveguide integrated balanced UTC-PD which can generate a maximum 8.9 dBm of RF output power and the PD chip has a 3 dB-bandwidth of 80 GHz[43]. For an integration array of PDs, the research work is mainly concentrated in Japan, especially the NTT Corporation. In Table 1, we review the status of PD array modules in terms of their structure and performance. In 2016, a 12.5 Gb/s × 8 channel receiver array module is fabricated by the Zhu group shown in Ref. [47]. The receiver module is composed of three portions: an eight-PIN-chip array, a TIA array, and an RF circuit board, which were packaged in butterfly housing.

    At a loss-limited optical transmission condition such as in the wavelength division multiplexed passive optical network (WDM-PON) system, high sensitivity of the receiver is seriously needed to cover the power consumption with respect to the transmission distance and optical components. Therefore, many researchers have provoked considerable efforts to improve the sensitivity performance of PDs. In 2012, Nada et al. reported that an APD together with a commercially available transimpedance amplifier (TIA) realized −21.7 dBm sensitivity at a data rate of 25.8 Gb/s which was assembled in the ROSA[48]. However, commercial APD above 10 Gb/s is scarcely available, and because of the low ionization coefficients ratio of the avalanche material, it is leading to low gain × bandwidth for 40 Gb/s applications. Therefore, optical preamplification is an attractive solution to increase the sensitivity for high speed applications. In 2014, a 40 Gb/s receiver comprising a detector and a transimpedance amplifier (TIA) with a high sensitivity of −23 dBm was achieved by the Caillaud C group[49]. In the same year, Maria Anagnosti et al. presented a high-speed PD which consists of a UTC-PD monolithically integrated with a semiconductor optical amplifier (SOA). This photoreceiver operates with 3 dB bandwidth up to 110 GHz and a dark current of 1 nA at 1.55 mm wavelength[50]. From 2012 to 2015, an integrated SOA-PIN photonic circuit with a high responsivity of 88 A/W and a large −3 dB bandwidth of 50 GHz[51], a receiver with an integrated SOA-PIN chip and a TIA with a broad bandwidth of 35 GHz and a very high sensitivity of −23 dBm at 25 Gb/s and −21 dBm at 40 Gb/s[52, 53] is achieved in succession by the Caillaud group shown in Fig. 1.

4.   Packaging for photodetectors
  • As we know, the bandwidth of the PD chip has exceeded 100 GHz; at such high operating frequency, the efficient design of packages and interconnects is quite challenging in terms of the loss due to the resonances in transmission line structures, launch transitions, impedance mismatch as well as electrical loss of electrodes. Moreover, to achieve the target of high capacity, multi-channel modules is a nice choice for the WDM system. Currently, hybrid integration packaging has been adopted for high-speed modules instead of the traditional packaging design, and it is a trend to integrate many functional components on one single substrate. In this section, some technologies of the packaging are introduced in the following.

  • 4.1.   Transmission line design

  • Single-ended PDs usually use coaxial output connectors whose cavity diameters are 1.8, 2.4, or 2.92 mm, depending on their speed. The interface of the connector pin and the internal transmission line are very crucial at such high speeds, because misalignment between them can cause degradation in the electrical return loss. In addition, it is necessary to cautiously control the shape and quantity of the solder used in the connection. The design of transmission lines has a significant impact on module performance. Generally, coplanar waveguides have better signal shielding and lower dispersion than microstrip lines, but a slightly higher loss.

  • 4.2.   Effect of connecting ways on module performance

  • Flip-chip, tape-bonding, and wire-bonding are widely adopted for interconnecting the chips based on different substrates whose transmission performances up to 120 GHz were compared in Ref. [54]. Many works have been conducted into the behavior of wire-bonding[55, 56] and flip-chip[57, 58]. Compared with wire-bonding technology, flip-chip technology generally has a superior performance in terms of small solder bumps and less inductive parasitics. However, the length of bonding wires can be fabricated to be fairly short with the development of wire-bonding technology. So wire-bonding is considered as another attractive connecting method in high-speed applications due to its advantages such as low cost, robustness, high thermal tolerance, and convenient fabrication process.

    The length of the wire bond used in the PD packaging between different components should be precisely controlled respectively. The inductance introduced by wire bonding causes peaking gain, which can be used to extend the bandwidth of the PD module, while the wire bond connecting the PD and TIA should be kept as short as possible. A method to reduce the inductance between the PD and TIA is making the TIA mounted very close to the PD chip and coplanar with it. What is more, the AC current through the PD to the ground must meet minimal inductance, because it causes a ground like an open circuit, which gives rise to a sharp decrease in gain, so the bonding wire connecting to the ground should also be short as possible. It worth noting that introducing additional capacitance that can increase module noise and reduce bandwidth should be avoided.

  • 4.3.   Optical coupling

  • Highly efficient optical coupling is one of the most challenging aspects in the packaging process. This is primarily because the active areas of the high-speed PD are small which makes optical coupling extremely sensitive to any shifts in the optical fiber. Many types of compact ROSAs that were integrated with surface-illuminated PDs adopting various optical systems have been demonstrated. As shown in Fig. 2, a PD module using a catadioptric system was demonstrated by the Kiyohide Sakai group in 2009; this optical system consists of a BK7 ball lens and a plastic-molded offset parabolic mirror[59]. In 2013, two different forms of demultiplexer (DMUX) used for optical coupling for ROSA are presented by the NTT Corporation and Chungnam National University, respectively[60, 61]. As shown in Fig. 3, a waveguide grating (AWG) DMUX is used in array module packaging, another optical DMUX block which is composed of local area network-wavelength division multiplexing (LAN-WDM) thin-film filters and an optically transparent quartz block is adopted shown in Fig. 4. In 2017, Isaac et al. presented a coupling approach, which uses vertical grating couplers to illuminate the III–V modified (UTC-PD) in a silicon PIC, as shown in Fig. 5[62].

5.   Application and prospect
  • PDs must meet various performances in terms of bandwidth, output power, sensitivity, power consumption and dynamic range for different applications such as optical communication systems, radio over fiber systems and millimeter terahertz signal generation systems. For optical communication systems, the capacity is constantly pursued by researchers, which is up to 100 GbE or even 400 GbE. Therefore, high-speed detectors and detector array receiver modules are still research hotspots for optical communication systems. The sensitivity is a significant parameter for free-space optical (FSO) communication systems, because the optical carrier after a long distance of transmission will become divergent and unstable, which makes it difficult for the detector to capture the optical signal. To overcome this problem, a big area of the photo-surface is strictly demanded, whereas this is often accompanied by a decrease in the speed of response. Therefore, a PD with new structure needs to be developed urgently. PDs with high linearity and high power are vital for analog photonic links and they have been used in many various microwave photonic applications such as phased array radars, ROF systems, etc. As a high frequency mixing component, a PD which is able to deliver a very high photocurrent level will bring numerous advantages such as improving link gain, reducing the noise figure, and making the linearity of the PD better. The saturation current characteristic is a significant factor that determines the spurious-free dynamic range (SFDR) and signal to noise ratio (SNR) of the system. In particular, the non-linearity of the PD determines the upper bound of the spurious-free dynamic range of the system where a linear modulator has been used. In addition, for millimeter terahertz signal generation systems, a high-speed PD with big saturation currents can broaden the frequency range of the generated signal and the intensity of the signal produced by the PD is boosted by increasing the incident optical power. As a consequence, it is conducive to transmit signals and reduce the cost of the system. Therefore, the PD with high saturation currents got the enthusiasm of the researchers.

Figure (5)  Table (1) Reference (62) Relative (20)

Journal of Semiconductors © 2017 All Rights Reserved