The first step in the design of an optical communication system is to decide how the electrical data would be converted into an optical signal with the same information. The original electrical data can be in an analog form, but it is invariably converted into a digital bit stream (with the RZ or NRZ format) consisting of a pseudorandom sequence of 0 and 1 bits. Two techniques, know as (a) direct modulation and (b) external modulation, can be used to generate the corresponding optical bit stream. Both of them are discussed in this tutorial.
1. Direct Modulation
In the case of direct modulation, the laser source itself is biased close to its threshold and driven by the electrical bit stream that increases the applied current considerably above the laser's threshold to create optical pulses representing digital bits (the so-called large-signal modulation). The important question is how closely the optical pulse mimics the shape of electrical pulse. To answer this question, the two rate equations must be solved numerically with I(t) = Ib + Imfp(t), where fp(t) represents the shape of electrical pulses. The following figure shows, as an example, the shape of the emitted optical pulse for a laser biased at Ib = Ith and modulated at 10 Gb/s using rectangular current pulses of during 100 ps and amplitude Im = 3Ith. The optical pulse does not have sharp leading and trailing edges because of a limited modulation bandwidth of the laser. It is also delayed considerably because it takes time for the optical power to build up from its initial negligible values. Even though the resulting optical pulse is not an exact replica of the applied electrical pulse, its overall shape and width are such that this semiconductor laser can be used for direct modulation at 10 Gb/s.
As discussed before, amplitude modulation in semiconductor lasers is accompanied by phase modulation and temporal phase variations can be studied. A time-varying phase is equivalent to transient changes in the mode frequency from its steady-state value ν0. Such as pulse is called chirped. The frequency chirp δν(t) is given by
The bottom trace in the figure above shows the frequency chirp across the optical pulse. The mode frequency first shifts toward the blue side near the leading edge and then toward the red side near the trailing edge of the optical pulse. Such a frequency shift implies that the pulse spectrum is considerably broader than that expected in the absence of frequency chirp, a feature that degrades the system performance through excessive broadening of optical pulses during their transmission through a fiber link.
Since frequency chirp is often the limiting factor for lightwave systems operating near 1.55-µm, several methods have been used to reduce its magnitude. These include pulse-shape tailoring, injection locking, and coupled-cavity schemes. A direct way to reduce the frequency chirp is to design semiconductor lasers with small values of the linewidth enhancement factor βc. The use of quantum wells or quantum dots reduces βc by a factor of 2 or so. A further reduction occurs for strained quantum wells. Indeed, βc ≈ 1 has been measured in modulation-doped strained MQW lasers. Such lasers exhibit low chirp under direct modulation.
2. External Modulation
At bit rates of 5 Gb/s or higher, the frequency chirp imposed by direct modulation becomes large enough that direct modulation of semiconductor lasers is rarely used. For such high-speed transmitters, the laser is biased at a constant current to provide the CW output, and an optical modulator placed next to the laser converts the CW light into a data-coded pulse train with the right modulation format.
Two types of optical modulators developed for lightwave system applications are shown in the following figure.
An important class of optical modulators makes use of the electro-optic effect inside a LiNbO3 waveguide such that the effect mode index changes in response to an applied voltage across it. Such a simple device modulates the phase of light passing through it and is useful as a phase modulator. To construct an intensity modulator, phase modulation is converted into amplitude modulation with the help of a Mach-Zehnder (MZ) interferometer. Two titanium-diffused LiNbO3 waveguides form the two arms of a MZ interferometer. In the absence of an external voltage, the optical fields in the two arms of the MZ interferometer experience identical phase shifts and interfere constructively. The additional phase shift introduced in one of the arms through voltage-induced index changes destroys the constructive nature of the interference and reduces the transmitted intensity. In particular, no light is transmitted when the phase difference between the two arms equals π, because of destructive interference occurring in that case. As a result, the electrical bit stream applied to the modulator produces an optical replica of the bit stream.
A LiNbO3 modulator is rarely used in ASK lightwave systems that simply turn the light on and off to encode the information because of considerable insertion loss that invariably occur when CW light from the laser source is coupled into the LiNbO3 waveguide inside an external modulator. The electroabsorption modulator (EAM) shown in the second figure above solves this problem because it is made with the same InP material used to make the laser source, and both of them can be integrated on the same InP substrate.
An EAM makes use of the Franz-Keldysh effect, according to which the bandgap of a semiconductor decreases when an electric field is applied across it. Thus, a transparent semiconductor layer begins to absorb light when its bandgap is reduced electronically by applying an external voltage. An extinction ratio of 15 dB or more can be realized for an applied reverse bias of a few volts at bit rates of up to 40 Gb/s. Although some chirp is still imposed on the coded pulses, it can be made small enough not to be detrimental for the system performance.
An advantage of EAMs is that they are made using the same semiconductor material that is used for the laser, and thus the two can be easily integrated on the same chip. Low-chirp transmission at a bit rate of 5 Gb/s was demonstrated as early as 1994 by integrating an EAM with a DBR laser. By 1999, 10-Gb/s optical transmitters with an integrated EAM were available commercially and were used routinely for WDM lightwave systems. By 2001, such integrated modulators could be operated at a bit rate of 40 Gb/s, and such devices became available commercially soon after. Moreover, EAMs exhibit the potential of operating at bit rates of up to 100 Gb/s.
The following figure shows schematically the basic idea behind modulator-integrated DFB laser. The DFB laser on the left provides the CW signal at a fixed wavelength (determined by the grating) that is modulated by the EAM on the right. The middle section is designed to isolate the two devices electrically while inducing minimum losses. The facets of the whole device are coated such that the left facet has a high reflectivity (> 90%), while the right facet has as low reflectivity as possible (<1%).
The fabrication of modulator-integrated lasers requires attention to many details. In general, the active core layers in the laser and modulator sections should be made using different compositions with different bandgaps so that they can be optimized for each device separately. Two different approaches are used for this purpose. In one scheme, the waveguides for the laser and modulator are butt-joined using separate epitaxial growth steps for each of them. First, the layers are grown for one device, say the laser. Then a mask is used to remove the epitaxial layers from the modulator region, and new layers are regrown. Although this approach offers the maximum flexibility for optimizing each device separately, the vertical alignment of the layers in the two sections is relative difficult and affects the yield. In the much simpler, selective-area growth technique, both devices (laser and modulator) are formed in a single epitaxial growth, but the oxide pads placed on the wafer prior to growth allow one to shift the laser wavelength toward the red side by more than 100 nm. The wavelength shift results from a change in the Bragg wavelength of the laser grating occurring because of changes in the effective mode index induced by the oxide pads. This technique is commonly used in practice for making modulator-integrated lasers.
The performance of modulator-integrated lasers is limited by the optical and electrical crosstalk between the laser and modulator sections. Typically, the separation between the electrical contacts used for the two devices is less than 0.2 mm. Any leakage from the modulator contact to the laser contact can change the dc bias of the laser in a periodic manner. Such unwanted laser-current changes shift the laser wavelength and produce frequency chirp because the laser frequency changes with time. Since laser frequency can shift more than 200 MHz/mA, the middle section should provide an isolation impedance of 800 Ω or more. Although such values are easily realized at low modulation frequencies, this level of isolation is difficult to achieve at microwave frequencies approaching 40 GHz. In one approach, the FM efficiency of the laser is controlled by reducing the chirp parameter βc for the laser.
The optical crosstalk between the laser and modulator results from the residual reflectivity of the output facet. This residual reflectivity is seen by the laser only when the modulator is on because in the off state, the laser light is totally absorbed by the modulator before it reaches the output facet. As a result, the laser gain, and hence the emission wavelength, are slightly different during each on-ff cycle of the modulator. This is an additional source of frequency chirping. It can be nearly eliminated if the front facet has a residual reflectivity of less than 0.01%. However, it is hard to realize antireflection coatings of this quality in practice.
In general, frequency chirp associated with the laser and modulator sections is a limiting factor for modulator-integrated DFB lasers. Typically, the chirp parameter βc exceeds 2 in the on state and changes to below -2 in the off state when the reverse voltage of about 3 V is applied. In a novel approach, chirp was reduced by designing the quantum wells of the modulator to be relatively shallow. More specifically, the bandgap difference between the barrier and quantum-well layers was reduced from 0.2 to close to 0.1 eV. The measured values of βc were below 0.7 for such devices in the entire 0-3 V range of the reverse bias, resulting in improved performance when the device was used in a lightwave system operating at 10 Gb/s. Physically speaking, frequency chirp is due to changes in the refractive index occurring because of the pileup of electrons and holes inside the quantum wells. Since the barrier escape time is reduced considerably in shallow quantum wells, the carrier density does not build up to high values, resulting in a low chirp.
Integration of DBR lasers with an EAM provides certain advantages and is being pursued for realizing tunable optical sources. In one 2002 experiment, a four-section DBR laser fabricated with a sampled grating was integrated with a modulator and an amplifier, resulting in a six-section structure shown schematically in the figure below.
Such a monolithically integrated device was tunable over 40 nm, while maintaining an extinction ratio better than 10 dB. Since then, considerable progress has been made in realizing widely tunable transceivers capable of operating at bit rates of up to 40-Gb/s. Such devices integrate an optical receiver with the laser transmitter on the same chip and can be used over a wavelength range covering the entire C band.
Some applications require a transmitter capable of emitting a pulse train at high repetition rates such that a short optical pulse is present in each bit slot. Examples of such applications include optical time-division multiplexing and WDM systems designed with advanced modulation formats. An EAM modulator can be used for generating short optical pulses suitable for such applications. The EAM in this case acts as a saturable absorber and is employed to realize mode locking of the semiconductor lasers. A DFB laser, integrated monolithically with a MQW modulator, was used as early as 1993 to generate a 20-GHz pulse train. The 7-ps output pulses were nearly transform-limited because of an extremely low chirp associated with the modulator. A 40-GHz train of 1.6 ps pulses was produced in 1999 using an EAM. By 2007, such monolithic mode-locked semiconductor lasers were available in a packaged form.