# Optical Transmitter Design

We have discussed the properties of optical sources. Although an optical source is a major component of optical transmitters, it is not the only component. Other components include a modulator for converting electrical data into optical form (if direct modulation is not used) and an electrical driving circuit for supplying current to the optical source. This tutorial covers the design of optical transmitters with emphasis on the packaging issues.

#### 1. Source-Fiber Coupling

The design object for any transmitter is to couple as much light as possible into the optical fiber. In practice, the coupling efficiency depends on the type of optical source (LED versus laser) as well as on the type of fiber (multimode versus single mode). The coupling can be very inefficient when light from an LED is coupled into a single-mode fiber. As we saw before, the coupling efficiency for an LED changes with the numerical aperture, and can become < 1% in the case of single-mode fibers. In contrast, the coupling efficiency for edge-emitting lasers is typically 40-50% and can exceed 80% for VCSELs because of their circular spot size. A small piece of fiber (know as a pigtail) is included with the transmitter so that the coupling efficiency can be maximized during packaging; a splice or connector is used to join the pigtail with the fiber cable.

Two approaches have been used for source-fiber coupling. In one approach, known as direct or butt coupling, the fiber is brought close to the source and held in place by epoxy. In the other, known as lens coupling, a lens is used to maximize the coupling efficiency. Each approach has its own merits, and the choice generally depends on the design objectives. An important criterion is that the coupling efficiency should not change with time; mechanical stability of the coupling scheme is therefore a necessary requirement.

An example of butt coupling is shown in the figure (a) below, where the fiber is brought in contact with a surface-emitting LED. The coupling efficiency for a fiber of numerical aperture NA is given by

nc = (1 - Rf)(NA)2

where Rf is the reflectivity at the fiber front end. Rf is about 4% if an air gap exists between the source and the fiber but can be reduced to nearly zero by placing an index-matching liquid. The coupling efficiency is about 1% for a surface-emitting LED and roughly 10% for an edge-emitting LED. Some improvement is possible in both cases by using fibers that are tapered or have a lensed tip. An external lens also improves the coupling efficiency but only at the expense of reduced mechanical tolerance.

The coupling of a semiconductor laser to a single-mode optical fiber is more efficient than that of an LED. The butt coupling provides only about 10% efficiency, as it makes no attempt to match the mode sizes of the laser and the fiber. Typically, index-guided InGaAsP lasers have a mode size of about 1µm, whereas the mode size of a single-mode fiber is in the range 6-9 µm. The coupling efficiency can be improved by tapering the fiber end and forming a lens at the fiber tip. The above figure (a) shows such a butt-coupling scheme for a commercial transmitter. The fiber is attached to a jewel, and the jewel is attached to the laser submount by using an epoxy. The fiber tip is aligned with the emitting region of the laser to maximize the coupling efficiency (typically 40%). The use of a lensed fiber can improve the coupling efficiency, and values close to 100% have been realized with an optimum design.

The above figure (b) shows a lens-coupling approach for transmitter design. The coupling efficiency can exceed 70% for such a confocal design in which a sphere is used to collimate the laser light and focus it onto the fiber core. The alignment of the fiber core is less critical for the confocal design because the spot size is magnified to match the fiber's mode size. The mechanical stability of the package is ensured by soldering the fiber into a ferrule which is secured to the body by two sets of laser alignment welds. One set of welds establishes proper axial alignment, while the other set provides transverse alignment.

The laser-fiber coupling issue remains important, and several new schemes have been developed in recent years. In one approach, a silicon optical bench is used to align the laser and the fiber. In another, a silicon micromirror, fabricated with the micro-machining technology, is used for optical alignment. In a different approach, a spot-size converter is employed for maximizing the coupling efficiency. Coupling efficiency of up to 80% were realized in 1997 by integrating a spot-size converter with InP semiconductor lasers. A lensed, graded-index, oval-core fiber has also provided higher coupling efficiency compared with conventional lensed fibers.

An important problem that needs to be addressed in designing an optical transmitter is related to the extreme sensitivity of semiconductor lasers to optical feedback. Even a relatively small amount of feedback (< 0.1%) can destabilize the laser and affect the system performance through phenomena such as linewidth broadening, mode hopping, and RIN enhancement. Attempts are made to reduce the feedback into the laser cavity by using antireflection coatings. Feedback can also be reduced by cutting the fiber tip at a slight angle so that the reflected light does not hit the active region of the laser. Such precautions are generally enough to reduce the feedback to a tolerable level. However, it becomes necessary to use an optical isolator between the laser and the fiber transmitters designed for more demanding applications. One such application corresponds to lightwave systems operating at high bit rates and requiring a narrow-linewidth DFB laser.

Most optical isolators make use of the Faraday effect, which governs the rotation of the plane of polarization of an optical beam in the presence of a magnetic field: The rotation is in the same direction for light propagating parallel or antiparallel to the magnetic field direction. Optical isolators consist of a rod of Faraday material such as yttrium ion garnet (YIG), whose length is chosen to provide 45º rotation. The YIG rod is sandwiched between two polarizers whose axes are tilted by 45º with respect to each other. Light propagating in one direction passes through the second polarizer because of the Faraday rotation. By contrast, light propagating in the opposite direction is blocked by the first polarizer. Desirable characteristics of optical isolators are low insertion loss, high isolation (> 30 dB), compact size, and a wide spectral bandwidth of operation. A very compact isolator can be designed if the lens in figure (b) above is replaced by a YIG sphere so that it serves a dual purpose. As light from a semiconductor laser is already polarized, a single polarizer placed between the YIG sphere and the fiber can reduce the feedback by more than 30 dB.

#### 2. Driving Circuitry

The purpose of driving circuitry is to provide electrical power to the optical source and to modulate the light output in accordance with the signal that is to be transmitted.

Driving circuits are relative simple for LED transmitters but become increasingly complicated for high-bit-rate optical transmitters employing semiconductor lasers as an optical source. In the case of direct modulation, semiconductor lasers are biased near threshold and then modulated through an electrical time-dependent signal. In this case, the driving circuit is designed to supply a constant bias current as well as modulated electrical signal. Furthermore, a servo loop is often used to keep the average optical power constant.

The figure below shows a simple driving circuit that controls the average optical power through a feedback mechanism. A photodiode monitors the laser output and generates the control signal that is used to adjust the laser bias level. The rear facet of the laser is generally used for the monitoring purpose. In some transmitters a front-end tap is used to divert a small fraction of the output power to the detector. The bias-level control is essential, since the laser threshold is sensitive to the operating temperature. The threshold current also increases with aging of the transmitter because of gradual degradation of the semiconductor laser.

The driving circuit shown in the figure above adjusts the bias level dynamically but leaves the modulation current unchanged. Such an approach is acceptable if the slope efficiency of the laser does not change with aging. As discussed before, the slope efficiency of the laser generally decreases with an increase in temperature. A thermoelectric cooler is often used to stabilize the laser temperature. An alternative approach consists of designing driving circuits that use dual-loop feedback circuits and adjust both the bias current and the modulation current automatically.

The electrical components used in the driving circuit determine the rate at which the transmitter output can be modulated. For lightwave transmitters operating at bit rates above 1 Gb/s, electrical parasitics associated with various transistors and other components often limit the transmitter performance. The performance of high-speed transmitters can be improved considerably by using monolithic integration of the laser with the driver. Since optical and electrical devices are fabricated on the same chip, such monolithic transmitters are referred to as optoelectronic integrated-circuit (OEIC) transmitters. The OEIC approach was first applied to integration of GaAs lasers, since the technology for fabrication of GaAs electrical devices is relatively well established. The technology for fabrication of InP OEICs evolved rapidly during the 1990s. A 1.5-µm OEIC transmitter capable of operating at 5 Gb/s was demonstrated in 1988. By 1995, 10-Gb/s laser transmitters were fabricated by integrating 1.55-µm DFB lasers with field-effect transistors made with the InGaAs/InAlAs material system. Since then, OEIC transmitters with multiple lasers on the same chip have been developed for WDM applications.

The concept of monolithic integration can be extended to build single-chip transmitters by adding all functionality on the same chip. Considerable effort has been directed toward developing such OEICs, often called photonic integrated circuits, which integrate on the same chip multiple optical components, such as lasers, detectors, modulators, amplifiers, filters, and waveguides. Such integrated circuits had reached the commercial stage by 2008.

#### 3. Reliability and Packaging

An optical transmitter should operate reliably over a relatively long period of time (10 years or more) in order to be useful as a major component of lightwave systems. The reliability requirements are quite stringent for undersea lightwave systems, for which repairs and replacement are prohibitively expensive. By far the major reason for failure of optical transmitters is the optical source itself. Considerable testing is performed during assembly and manufacture of transmitters to ensure a reasonable lifetime for the optical source. It is common to qualify the lifetime by a parameter tF known as mean time to failure (MTTF). Its use is based on the assumption of an exponential failure probability [PF = exp(-t/tF)]. Typically, tF should exceed 105 hours (about 11 years) for the optical source. Reliability of semiconductor lasers has been studied extensively to ensure their operation under realistic operating conditions.

Both LEDs and semiconductor lasers can stop operating suddenly (catastrophic degradation) or may exhibit a gradual mode of degradation in which the device efficiency degrades with aging. Attempts are made to identify devices that are likely to degrade catastrophically. A common method is referred to as burn-in or accelerated aging and is based on the assumption that under high-stress conditions weak devices will fail, while others will stabilize after an initial period of rapid degradation. The change in the operating current at a constant power is used as a measure of device degradation. The following figure shows the change in the operating current of a 1.3-µm InGaAsP laser aged at 60ºC under a constant output power of 5 mW from each facet.

The operating current for this laser increases by 40% in the first 400 hours but then stabilizes and increases at a much reduced rate indicative of gradual degradation. The degradation rate can be used to estimate the laser lifetime and the MTTF at the elevated temperature. The MTTF at the normal operating temperature is then extrapolated by using an Arrhenius-type relation tF = t0exp(-Ea/kBT), where t0 is a constant and Ea is the activation energy with a typical value of about 1 eV. Physically, gradual degradation is due to the generation of various kinds of defects (dark-line defects, dark-spot defects) within the active region of the laser or LED.

Extensive tests have shown that LEDs are normally more reliable than semiconductor lasers under the same operating conditions. The MTTF for GaAs LEDs easily exceeds 106 hours and can be > 107 hours at 25ºC. The MTTF for InGaAsP LEDs is even larger, approaching a value of ~109 hours. By contrast, the MTTF for InGaAsP lasers is generally limited to 106 hours at 25ºC. Nonetheless, this value is large enough that semiconductor lasers can be used in undersea optical transmitters designed to operate for a period of 25 years. Because of the adverse effect of high temperatures on device reliability, most transmitters use a thermoelectric cooler to maintain the source temperature near 20ºC even when the outside temperature may be as high as 80ºC.

Even with a reliable optical source, a transmitter may fail in an actual system if the coupling between the source and the fiber degrades with aging. Coupling stability is an important issue in the design of reliable optical transmitters. It depends ultimately on the packaging of transmitters. Although LEDs are often packaged nonhermetically, an hermetic environment is essential for semiconductor lasers. It is common to package the laser separately so that it is isolated from other transmitter components. Previous figures showed two examples of laser packages. In the butt-coupling scheme, an epoxy is used to hold the laser and fiber in place. Coupling stability in this case depends on how epoxy changes with aging of the transmitter. In the lens-coupling scheme, laser welding is used to hold various parts of the assembly together. The laser package becomes a part of the transmitter package, which includes other electrical components associated with the driving circuit. The choice of transmitter package depends on the type of application; a dual-in-line package or a butterfly housing with multiple pins is typically used.

Testing and packaging of optical transmitters are two important parts of the manufacturing process, and both of them add considerably to the cost of a transmitter. The development of low-cost packaged transmitters is necessary, especially for local-area and local-loop applications.