50-Gbit/s error-free operation is demonstrated by a high-speed avalanche photodiode for the first time. The APD exhibits 3-dB bandwidth of 35 GHz and excellent receiver sensitivity of −10.8 dBm at a BER of 10−12 against non-return to zero input optical signals. These results indicate our APD is promising for the systems with serial baud rate of 50 Gbit/s such as 400-Gbit/s Ethernet systems.
© 2014 Optical Society of America
The rapid growth of the transmission capacity of optical fiber communications systems leads to an increase of the baud rate. For example, large capacity 400-Gbit/s Ethernet (400 GbE) is a candidate of an upcoming system . In the standardization of 400 GbE, some transmission formats are discussed. Some candidates are employing a serial baud rate of 25 Gbit/s, and using 16 lanes of wavelength-division multiplexing (WDM), or combination of pulse amplitude modulation (PAM) with 8 lanes of WDM. The other is employing 8 lanes of WDM with a serial baud rate of 50 Gbit/s. Using a serial baud rate of 25 Gbit/s is technologically acceptable for optical devices because the serial baud rate is same as that of the conventional 100 GbE. However, there remains an issue of power consumption due to using the large number of lanes, which requires many light sources, drivers, and amplifiers, or an electrical signal processing for PAM. Another issue with 400 GbE is how to extend the transmission distance because a large number of lanes results in lower launch power of a transmitter due to the limit of an optical power in a fiber. One attractive solution for those issues is using a serial baud rate of 50 Gbit/s so as to decrease the number of lanes, and improving sensitivity of an optical receiver. A photodiode is a key element for achieving both high-speed operation and high sensitivity, and an avalanche photodiode (APD) is a good candidate for satisfying these demands. One approach to obtaining such performance is to employ a waveguide-type structure because the structure allows the use of a very thin absorption layer and results in a short carrier-transit time, and bandwidth of as large as 27 GHz has been successfully demonstrated with this type of APD . So far, optical receivers using waveguide-type APDs have achieved 40-Gbit/s operation [3,4]; however, 50-Gbit/s error-free operation has not been demonstrated yet. One difficulty with the waveguide-type is the optimization of waveguide length: a longer waveguide provides higher responsivity but increases device capacitance, which results in the degradation of high-speed characteristics. In addition, the waveguide-type APD requires tight optical tolerance because of the narrow waveguide. On the other hand, the vertical-illumination APD has larger optical tolerance. Therefore, if the tradeoff between the carrier transit time and responsivity in the vertical-illumination structure can be solved, it will be promising for realizing large-capacity communications systems with a high baud rate.
This paper is the first report of a high-sensitivity and high-speed vertical-illumination APD demonstrating 50-Gbit/s error-free operation. The APD structure was designed to relax the tradeoff between responsivity and carrier-transit time. The APD assembled in receiver module together with a trans-impedance amplifier (TIA) exhibits excellent receiver sensitivity of −10.8 dBm at a serial baud-rate of 50 Gbit/s.
2. Device design and structure
Figure 1 shows a schematic cross sectional view of the fabricated APD with the inverted p-down configuration . The epitaxial structure was grown on a semi-insulating InP substrate by using the MOCVD method. It consists of a p-type contact, p-type InGaAs absorption, undoped InGaAs absorption, p-type field control, InAlAs avalanche, n-type field control, edge-field buffer, and n-type contact layers. The thickness of the InAlAs avalanche layer was set to be 90 nm in order to obtain a large gain-bandwidth (GB) product. To obtain both a large 3-dB bandwidth (f3dB) and high responsivity, we employed a hybrid (p-type and undoped-) InGaAs absorption layer . In our previous work, we employed a 1-µm-thick absorption layer for 25-Gbit/s APDs [7–9]; however, the total thickness of the absorption layer needed to be reduced for operation at 50 Gbit/s. When the total thickness of the absorption layer gets smaller, the ratio of p-/undoped (namely, neutral/depleted) absorption layers providing the maximized intrinsic f3dB can be changed because neutral and depleted layers have different carrier transport mechanisms. In the neutral absorption layer, the carrier transit time depends on the square of its thickness, while that for the depleted absorption layer simply depends on the thickness. Thus, when the total absorption layer thickness is decreased, the contribution of the neutral absorption layer becomes dominant in the carrier-transit time.
The calculation results in Fig. 2 show how the intrinsic f3dB depends on the thickness ratio of p-/undoped absorption layers for various total absorption layer thickness with estimated responsivities at unity gain for 1310 nm as parameters. We calculated the f3dB based on the charge-control model of the absorption layers , with a hole drift velocity (vh) of 5.0 × 106 cm/s, and electron diffusion coefficient (De) of 2.0 × 102 cm2/s. The ratio of zero corresponds to a conventional fully depleted absorption layer, or PIN-type absorption layer.
The ratio of one corresponds to a uni-travelling carrier, or UTC-mode absorption layer. The responsivity gets smaller as the total thickness of the absorption layer decreases, while the intrinsic f3dB bandwidth increases. The ratio of p-/undoped absorption layers giving maximum intrinsic f3dB increases when the total thickness decreases, meaning that the contribution of the neutral absorption layer to in the carrier transit time becomes larger. The total absorption layer thickness of 0.6 µm provides the best intrinsic f3dB of 70 GHz at the ratio of 0.40, which corresponds to 30-GHz improvement compared with the conventional PIN absorption layer along with high responsivity at unity gain of 0.74 A/W at a wavelength of 1310 nm. Thus, we employed the absorption layer thickness of 0.6 µm as an optimized absorption layer design for high-speed and high-responsivity operation. Actually, f3dB of the APD is not determined only by the intrinsic f3dB of the whole absorption layer but also by the total thickness, materials, resistance, capacitance, and GB Product. We will discuss about the actual f3dB of the fabricated APD in the next chapter.
3. Chip and module characteristics
Figure 3 shows the measured responsivities against the applied voltage of the fabricated APD, which has an active area with a 14-µm-diameter under 1310-nm illumination of −20 dBm. In order to define the responsivity at unity gain, the fitting is also shown. Here, we assumed the responsivity at unity gain of 0.69 A/W. At the fitting of multiplication characteristics, we used electron- and hole-ionization coefficients, which include a correction for local ionization model by considering dead-space effect, described in . The fitting well explains the measurement, indicating that the high responsivity at unity gain of about 0.7 A/W with the 0.6-µm absorption layer is obtained as expected [see Fig. 2]. Figure 4 shows gain-bandwidth (GB) characteristics of the fabricated APD. The calculated characteristics are shown with a solid line. For the calculation, a GB product of 270 GHz, an intrinsic f3dB of 70 GHz, and a device capacitance of 30 fF were used. Here, we estimated the device capacitance by the total thickness of the depletion layer and active area. The calculation well explains the measurement, indicating that our APD has a large GB product of 270 GHz thanks to employing the thin InAlAs avalanche layer. The maximum bandwidth reaches 35 GHz at M = 3, and the large bandwidth of over 30 GHz is achieved up to a gain of 4.7. The obtained performance of f3dB, gain and responsivity are sufficient for highly sensitive detection for 50-Gbit/s optical signals.
We assembled the fabricated APD to an optical receiver module together with a TIA, and measured bit-error rate (BER). The module has a butterfly-type configuration with GPPO electrical outputs as shown in Fig. 5. A commercially available InP-based TIA, which was designed for 40-Gbit/s applications, was used. For the BER measurement and observation of an eye diagram, we used a 1310 nm light source, LN modulator, and a driver for 40 Gbit/s applications. The input optical signal was NRZ signal with an extinction ratio of 7.7 dB and a baud rate of 50.0 Gbit/s as shown in Fig. 6(a). The pseudo-random bit sequence (PRBS) was set to 231-1. Figure 6(b) shows an electrical output signal of the APD receiver module at M = 4.36 under the back-to-back condition. We obtained clear eye-opening against the 50-Gbit/s optical input signal. Figure 7 shows the BER characteristics under the back-to-back condition with the same M as in the eye-diagram observation. The minimum receiver sensitivity of −10.8 dBm at a BER of less than 10−12 is successfully achieved. In this operation condition, the power consumption of the APD receiver module is less than 500 mW, and more than 98% of the power is consumed by the TIA. Based on the results, the obtained minimum receiver sensitivity will be enough for 20-km transmission at 50 Gbit/s when we assume a launch power of 0 dBm and transmission loss in the optical fiber of 0.5 dB/km.
Table 1 compares reported performance of high-speed APDs and PIN-PD. Our APD exhibits good performance balance in terms of responsivity and f3dB even though it employs vertical illumination structure. These results indicate that our APD is promising for realizing 50-Gbit/s receiver modules and low-power 400-GbE systems.
We demonstrated error-free operation of a 50-Gbit/s APD for the first time. The APD shows large f3dB of 35 GHz and high responsivity at unity gain of 0.69 A/W with the vertical illumination structure. The APD receiver module exhibited minimum receiver sensitivity of −10.8 dBm. These results indicate that our APD is promising for middle-reach and low-power applications that need a high serial baud-rate of 50 Gbit/s such as 400 GbE systems.
The authors thank T. Yoshimatsu, E. Yoshida, M. Nagatani, H. Fukuyama for valuable discussions, and K. Murata for his continuous encouragement.
References and links
2. G. S. Kinsey, C. C. Hansing, A. L. Holmes Jr, B. G. Streetman, J. C. Campbell, and A. G. Dentai, “Waveguide In/sub 0.53/Ga/sub 0.47/As-In/sub 0.52/Al/sub 0.48/As avalanche photodiode,” IEEE Photon. Technol. Lett. 12, 416–418 (2000). [CrossRef]
3. N. Yasuoka, H. Kuwatsuka, and A. Kuramata, “High-speed and high-efficiency InP/InGaAs waveguide avalanche photodiodes for 40 Gbit/s transmission systems,” in Proceedings of OFC (2004), TuM2.
4. S. Shimizu, K. Shiba, T. Nakata, K. Kasahara, and K. Makita, “40 Gbit/s waveguide avalanche photodiode with p-type absorption layer and thin InAlAs multiplication layer,” Electron. Lett. 43(8), 476–477 (2007). [CrossRef]
5. M. Nada, Y. Muramoto, H. Yokoyama, N. Shigekawa, T. Ishibashi, and S. Kodama, “Inverted InAlAs/InGaAs avalanche photodiode with low–high–low electric field profile,” Jpn. J. Appl. Phys. 51(2), 02BG03 (2012). [CrossRef]
6. Y. Muramoto and T. Ishibashi, “InP/InGaAs pin photodiode structure maximising bandwidth and efficiency,” Electron. Lett. 39(24), 1749–1750 (2003). [CrossRef]
7. M. Nada, Y. Muramoto, H. Yokoyama, T. Ishibashi, and S. Kodama, “InAlAs APD with high multiplied responsivity-bandwidth product (MR-bandwidth product) of 168 A/W･GHz for 25 Gbit/s high-speed operations,” Electron. Lett. 48(7), 397–399 (2012). [CrossRef]
8. M. Nada, Y. Muramoto, H. Yokoyama, T. Ishibashi, and S. Kodama, “‘High-sensitivity 25 Gbit/s avalanche photodiode receiver optical subassembly for 40 km transmission,” Electron. Lett. 48(13), 777 (2012). [CrossRef]
9. T. Yoshimatsu, M. Nada, M. Oguma, H. Yokoyama, T. Ohno, Y. Doi, I. Ogawa, and E. Yoshida, “Compact and high-sensitivity 100-Gb/s (4 × 25 Gb/s) APD-ROSA with a LAN-WDM PLC demultiplexer,” in Proceedings of ECOC (2012), Th.3.B. [CrossRef]
10. Y. L. Goh, D. J. Massey, A. R. J. Marshall, J. S. Ng, C. H. Tan, W. K. Ng, G. J. Rees, M. Hopkinson, J. P. R. David, and S. K. Jones, “Avalanche multiplication in InAlAs,” IEEE Trans. Electron. Dev. 54(1), 11–16 (2007). [CrossRef]
11. T. Ohno, Y. Muramoto, K. Sano, S. Kodama, and N. Shigekawa, “A CAN-type MIC-PD ROSA operating at 40-Gbit/s,” in Proceedings of ECOC (2012), Th.10. D.2.