In this paper we summarize production data from serial 10 Gb/s devices and report on 850 nm VCSEL arrays with
channel speeds up to 25 Gb/s. The production data demonstrates that robustness of the basic technology as well as its
suitability for cost effective, high volume production. The >10 Gb/s measurements on two dimensional arrays show that
850 nm VCSEL technology can be extended well beyond the 10 Gb/s links currently beginning to be deployed by
volume field users.
In this paper, we present the design and manufacturing of next-generation 850 nm 10 Gb/s vertical-cavity
surface-emitting lasers (GenX VCSELs). They were developed to provide a 10 Gb/s solution that meets
Class-1 eye safety limits, IEEE 802.3ae standards, 10G Fiber Channel standards, and corresponding multisource
agreement requirements for emerging low-cost, high-volume, and high-performance data
communication applications in local and storage area networks (LANs and SANs). The paper covers GenX
device designs, manufacturing processes, DC and AC characteristics, equivalent circuit models,
recommended operating conditions, as well as reliability studies. As a simple drop-in replacement, we
have successfully demonstrated that GenX VCSELs work well with all existing Emcore 10G transmitter
optical sub-assembly (TOSA) products.
In this paper we report the results from on-going performance enhancements of Emcore's comprehensive line of data communication VCSEL products in cost effective hermetic TO packages. Data are presented on the -20 to 100°C temperature range operational characteristics of our offerings at 1.25, 2.5, 4, and 10 Gb/s. The discussion covers high-speed parameters, fiber coupling efficiency, and other important features of the packaged devices.
In this paper we describe the processes and procedures that have been developed to ensure high reliability for Emcore’s 850 nm oxide confined GaAs VCSELs. Evidence from on-going accelerated life testing and other reliability studies that confirm that this process yields reliable products will be discussed. We will present data and analysis techniques used to determine the activation energy and acceleration factors for the dominant wear-out failure mechanisms for our devices as well as our estimated MTTF of greater than 2 million use hours. We conclude with a summary of internal verification and field return rate validation data.
In this paper, we describe 850 nm oxide VCSEL array technologies being developed at Emcore Optical Devices. We demonstrate the excellent performance, uniformity and reliability of oxide VCSEL arrays operating at 2.5Gb/s per channel which are entering into high volume production. Due to the ever-increasing demand for bandwidth by high-bit-rate data communications links, VCSELs operating at even higher bandwidths are needed. We discuss the development of oxide VCSELs capable of transmitting 10 Gb/s for application in the 10 gigabit Ethernet and other emerging high-aggregate bandwidth standards.
The theory, structure, and current manufacturing technologies for InGaAlP high brightness light emitting diodes (HB-LED) emitting in the range of 650 to 585 nm are described in this paper. A state-of-the-art HB-LED MOCVD reactor designed for high volume manufacturing (42 - 2' or 16 - 3' wafers) is demonstrated. Data for thickness and compositional uniformity and reproducibility are presented showing the material quality and reactor stability that can currently be achieved. In addition, device data for InGaAlP HB-LEDs is reported, including brightness, forward voltage, and emission wavelength with excellent intra and inter wafer uniformity and run-to-run reproducibility.
The bulk of LEDs sold today are still fabricated using older epitaxial techniques such as LPE and VPE, but have relatively low brightness and a limited color range. The newer high brightness LEDs are fabricated from the InGaAlP and III-Nitride systems, with MOCVD being the preferred growth technique for manufacturing. While these new materials represent a significant increase in performance, they are also more expensive to grow. In this paper we consider the reasons for this, which include a less mature growth technology, lower production volumes, expensive starting materials, process efficiency, equipment throughput and cost, and safety and environmental concerns. Addressing each of these issues in turn, we examine what has already been accomplished, and what may be improved by further advances in equipment and process. A realistic COO model is of great utility in comparing product cost for different device structures, staffing schemes, reactor sizes, etc. We demonstrate that for a dedicated LED manufacturing facility, the lowest epitaxial cost is achieved by running around the clock with the highest throughput reactor that is fully utilized for the desired production level. When maintenance tasks such as cleaning and test or calibration runs are minimized, then materials costs will dominate the epi cost, which leads to the desirability of achieving both the best reproducibility and increasing the process efficiency. We show how in-situ control techniques are now capable of increasing preproducibility and thereby lowering product costs for manufacturing scale MOCVD reactors.