Optical modulation goes external.
Digital or analog modulation of continuous-wave laser sources forms the basis of encoding and transmitting of information through optical fiber link systems. In digital systems, data are formatted in a simple periodic two-bit configuration, represented by high or low light intensities, whereas in analog systems data are represented by selective portions of a time-varying electronic waveform (sinusoidal, for example) applied to the optical carrier. High speed optical communications and the distribution of cable television (CATV) signals are just two examples of digital and analog systems, respectively, that involve the transmission of data, voice and video over fiber networks.
The basic layout of a fiber-optic link system is shown in Figure 1. The optical source wavelength is determined by the characteristics of the optical fiber. For example, the lowest signal attenuation achievable in silica-based fibers is for wavelengths 1.5 pm, whereas minimum signal dispersion occurs at wavelengths around 1.3 pm. Much of the optical communication highway is based on 1.3 [[micro]meter] fiber, although 1.5 [[micro]meter] systems may become more widespread.
If the optical source used is a semiconductor laser diode, information can be imprinted on the optical output by directly modulating the laser drive current with a radio frequency (RF) signal. In digital systems, the low (off) state generally corresponds to a position just below the lasing threshold on the characteristic intensity-current curve of the diode. This position is preferred to the zero current locus because turn-on delays are then minimized. Analog systems require a bias current in addition to the threshold current in order to push the modulation into the linear region of the power-current curve, as shown in Figure 2.
The main disadvantages associated with the direct modulation approach are the available low dynamic range, low modulation speeds and more importantly, the occurrence of frequency chirp, that is, a broadening of the spectral line width of the laser cavity, and hence, the optical carrier. When coupled with the chromatic dispersion of optical fiber and long propagation distances, frequency chirp can effectuate pulse spreading and overlap between successive bits of information, distorting the encoded optical signal, As the use of long haul optical data links and ever increasing bit-rates becomes more widespread, there will be a requirement for optical sources, which in comparison with semiconductor diode lasers offer high output powers and lower noise characteristics in addition to being chirp-free.
Using compact diode-pumped solid-state lasers, such as Nd: YAG, all these requirements can be met. However, the main disadvantage of the solid-state approach is its inability to modulate directly the laser at the data rates nominally entailed in optical communications. This inability causes further limitations associated with the inherently long excited state lifetime of the lasing species (100s [[micro]second]). External modulation overcomes this drawback by modulating the optical output from the laser rather than the material properties of the laser itself, and consequently, is set to play an increasingly important role in fiber-optic systems and applications.
The basic building block of an external modulator is the optical waveguide, which is analogous to a microwave transmission line, but on a much smaller scale. The optical waveguide tracks are formed in a suitable substrate material and are compatible with single-mode optical fibers.
Commercially available guided-wave optical modulators are predominantly manufactured using single-crystal ferroelectric lithium niobate (LiNb[O.sub.3]) substrates. Although other ferroelectric materials, such as LiTa[O.sub.3] or KTiOP[O.sub.4], have been shown to be suitable for fabricating guided-wave optical devices, the processing technologies associated with LiNb[O.sub.3] have been developed over a period of 20 years and are well understood. High quality three-inch diameter substrates are commercially available from a number of crystal growth companies, and several firms worldwide already offer a wide range of both bulk- and guided-wave LiNb[O.sub.3] components.
LiNb[O.sub.3] is normally grown using the Czochralski technique, a process that involves immersing a LiNb[O.sub.3] seed crystal into a high temperature (1250 [degrees] C) melt, which contains the compounds L[i.sub.2]O and [Nb.sub.2][O.sub.5], and simultaneously pulling and rotating the seed such that a boule of stoichiometric material is drawn out. The boule, which takes on the crystal orientation of the seed crystal, is allowed to cool gradually to room temperature during the pulling stage. As the material cools, the constituent Li and Nb ions become randomly displaced along the c-axis of the resultant rhombohedral crystal lattice. By applying an electrical poling field, typically 2 V/cm, across the crystal boule at an elevated temperature (1140 [degrees] C), and cooling under its influence, the ions can be made to displace along the direction of the electric field, thus inducing a well-ordered degree of spontaneous polarization (single ferroelectric domains), which is maintained at room temperature. It is the spontaneous polarization of the crystal that gives rise to properties such as the electro-optic effect.
The crystallographic orientation of single-crystal LiNb[O.sub.3] gives rise to two principle optical refractive indices, the ordinary index [n.sub.o] (2.22 at [Lambda] = 1.3 [[micro]meter]), and the extraordinary index [n.sub.e] (2.15 at [Lambda] = 1.3 [[micro]meter]). Therefore, the optical birefringence of the material, defined as ([n.sub.e]-[n.sub.o]), is relatively large at ^0.07. The ferroelectric nature of LiNb[O.sub.3] enables its birefringence to be altered by the application of an electric field via the electro-optic effect. The relationship between the refractive index change [Delta][n.sub.(ij)k] and the applied field [E.sub.k] is defined via a third-rank electro-optic tensor, which includes a set of coefficients [r.sub.(ij)k] that relates the direction of the externally applied voltage V to both the crystallographic axes and the refractive indices of the material
[Mathematical Expression Omitted]
[E.sub.k] = V/d
d = electrode spacing
[Gamma] = overlap between the optical and electrical fields
This overlap is theoretically modelled and ideally equal to unity. The largest coefficient in the electro-optic tensor for LiNb[O.sub.3] has a magnitude of 31 x [10.sup.-12] m/V. For a refractive index of 2.15, [Gamma] = 1, and an electric field of 5 x [10.sup.5] V/m (5 V over 10 [[micro]meter]), the induced refractive index change is 7.6 x [10.sup.-5], or [approximately] 0.003 percent.
The backbone of the external modulator is a fiber-compatible single-mode optical waveguide structure fabricated in a suitable substrate material. Optical waveguiding is achieved if the structure has a higher refractive index than the surrounding substrate and superstrate, allowing the optical signal to be totally internally reflected within the high refractive index region. The situation is akin to the propagation of light in optical fibers, where the core of the fiber has a higher refractive index than the cladding. Fiber pigtails attached to the optical waveguide facilitate coupling from the laser source and allow the modulated signal to be transmitted down the fiber system.
The majority of guided-wave devices in LiNb[O.sub.3] are based on photolithographically defined waveguides produced by high temperature (^1050 [degrees] C) diffusion of predeposited thin films ([approximately] 1000 [angstrom]) of titanium metal into the substrate. The diffusion process, which takes place over several hours in a controlled ambient, increases both [n.sub.e] and [n.sub.o]. Consequently, the resultant waveguide can support both transverse electric (TE) and transverse magnetic (TM) modes. However, an undesirable consequence of the dual-polarization nature of titanium-indiffused waveguides can be the occurrence of depolarization and/or polarization crosstalk in the waveguide modulator, which without adequate polarization control, can cause some degree of signal degradation in the system.
Such problems can be avoided by using the proton-exchange technology to fabricate the wave-guides,[3,4] a relatively low temperature ([approximately] 200 [degrees] C) ion-exchange process that involves immersing the LiNb[O.sub.3] substrate in an acid solution for several minutes. The resultant exchange between lithium from the substrate and protons from the acid increases only [n.sub.e], a consequence of which is that either pure TE or pure TM modes are supported, depending on the crystallographic orientation of the substrate. In order to obtain wave-guides that are comparable with titanium-indiffused waveguides in terms of being electro-optic and of low attenuation, the proton-exchange process must be followed by a thermal annealing step, typically a few hours at [approximately] 300 [degrees] C in ambient air. Electro-optically active waveguides have been produced with this means[5-7] with propagation losses on the order of 0.1 dB/cm, which is acceptable for most applications. By way of a comparison, the typical attenuation in single-mode optical fiber is in the region of 0.2 dB/km.
Modulation of the optical carrier involves changing the phase of the propagating light by varying the refractive index of the waveguide medium. Such variations can be achieved by applying an electric field across the waveguide or by locally changing the temperature in the vicinity of the waveguide (thermo-optic effect), depending upon the nature of the substrate material. The data rates required for transmitting information determine which physical effect can be used, the highest data rates being attainable by electro-optic modulation.
The waveguide structure most likely to see widespread deployment is the Mach-Zehnder interferometer configuration, shown in Figure 3. In this device, amplitude modulation is effectuated by splitting the optical signal equally at the first Y-branch, and phase modulating the light in either or both of the two interferometer arms by applying a voltage across the electrodes. Branching angles typically vary between 1 [degrees] and 2 [degrees]. Recombination of the two components at the second Y-branch, results in either constructive or destructive interference, depending on the phase difference between the two. Consequently, the light output of the device can be made to vary between a maximum and minimum state, that is, on or off. The degree of phase change [Delta][Phi], induced by an applied voltage V is given by the relationship
[Delta][Phi] = 2[Pi]L [Delta]n/[Lambda] (2)
L = electrode length
[Delta]n = field-induced change in the refractive index
[Lambda] = wavelength of the propagating light
For [Lambda] = 1.3 [[micro]meter], an electrode length of 5 mm and [Delta]n = 7.6 x [10.sup.-5], [Delta][Phi] is [approximately] 0.6 [Pi].
The Mach-Zehnder structure is nominally fabricated in an electro-optic substrate so that the phase of the propagating light can be modulated by the application of a voltage across the waveguide via a set of coplanar surface electrodes, Gold is normally used as the electrode material since it exhibits low RF propagation loss and enables the microwaves to travel at high electrical phase velocities, a useful prerequisite for broad bandwidth applications.
The modulation speed or bandwidth of the modulator is determined by the electrode design. Low bandwidths ([less than] 1 GHz), such as those used in fiber sensor or spatial rerouting applications, can be achieved by using a simple lumped electrode design, as shown in Figure 4. In the case of the lumped configuration, the electrodes act like a capacitor, the bandwidth being limited by both the product of capacitance and load resistance, and the time taken for the light to propagate along the interaction length.
For implementation over a wide bandwidth range ([greater than] 1 GHz), as required for optical communications, the electrode structure must be of a traveling-wave format, as show, in Figure 5. The traveling-wave design, which has its origin in transmission line theory, ensures that the modulating RF voltage wave travels at the same speed as the optical carrier, in the same direction; thus a constant and uniform overlap between the optical and electrical fields is maintained along the direction of propagation. Additionally, the electrodes are designed to have an impedance close to that of the 50 [Omega] RF coaxial cable that drives it. The electrodes then appear to be an extension of the microwave RF line. This approach facilitates delivery of the electrical signal to the electrodes.
Although the relationship between the degree of phase modulation and the applied RF voltage is linear, the intensity-modulated output takes the form of a sinusoidal intensity/voltage characteristic, shown in Figure 6, simply defined as
[I.sub.out] = [I.sub.in][cos.sup.2] ([Delta][Phi]/2) (3)
where [Delta][Phi] can vary between 0 and 2 [Pi]. This inherent nonlinearity in the Mach-Zehnder transfer function can create harmonics of the driving frequency in the modulator output. Although relatively unimportant in digital systems, harmonics are undesirable in analog systems because they can distort the imprinted optical signal.
A variety of both electronic and optical techniques have been developed to provide some degree of linearization of the transfer function, compensating for the nonlinearity of the Mach-Zehnder configuration. These techniques include, but are not limited to, electronic predistortion of the applied analog signal and optical polarization mixing. The electronic predistortion method relies on prior knowledge of the Mach-Zehnder transfer function. If this function is known, the RF input signal can be tailored, through the use of electronic circuitry, to compensate for the nonlinearity anticipated from the modulator. The optical polarization mixing method is an example of optical compensation where, subsequent to adjusting the amount of TE and TM optical components propagating in the modulator, the bias points for orthogonal polarizations are set to obtain distortion cancellation. Such linearization techniques have enabled television pictures to be transmitted through fiber systems with a quality that far exceeds industry standards.
Aside from the linearity of the response, the main characteristics to be considered when incorporating an external modulator into a fiber-optic link are the amount of optical loss introduced into the system, the optical power handling capability of the waveguide, the degree of modulation that can be attained, the switching voltage required to effectuate full modulation, the amount of RF drive power consumed and the modulation speed or bandwidth. Optical insertion loss is determined by the spatial mis-match in the optical fields at the fiber/waveguide coupling interfaces, Fresnel reflection losses at the fiber/waveguide interfaces, attenuation due to absorption and scattering within the waveguide material, and scattering losses at each of the two Y-branches. Careful design of the waveguide can ensure a mode profile that closely matches that of the pigtailed optical fibers. Fresnel losses at each interface arise from the different refractive indices of the optical fiber (n = 1.5) and waveguide (n = 2.15 to 2.22, for LiNb[O.sub.3], [Lambda] = 1.3 [[micro]meter]), and can be minimized by depositing antireflection coatings on the waveguide facets, nominally based on Si[O.sub.2], Mg[F.sub.2] or [Y.sub.2][O.sub.3] films. Fiber-to-fiber insertion losses on the order of 3 dB or less are acceptable for most applications.
In systems that involve multiple branching networks, for example the distribution of CATV, the signal becomes weaker every time the optical path is split. In this situation, a high power laser source such as Nd: YAG, which can deliver 10s of mW of optical power, is beneficial. When focused down to the dimensions of optical wave-guides, the resulting high power densities ([10.sub.4] to [10.sub.5] W/[cm.sup.2]) can sufficiently modify the properties of the waveguide material (photorefractive effect). Therefore, the modulator should be capable of handling high power densities without being damaged. Photorefractive effects can be particularly widespread in titanium-indiffused waveguides, even at low powers ([less than] 1 mW), and especially at wavelengths of less than 1 [[micro]meter]. However, proton-exchanged waveguides possess a much superior power handling capability.
The degree of modulation, commonly referred to as the extinction ratio and defined as the ratio of the maximum and minimum intensities transmitted, is determined by the electrode and waveguide designs. The better the overlap between optical and electrical fields, the higher the extinction ratio. An acceptable figure for most applications is 20 dB. Optical loss and extinction ratio are both wavelength dependent, although the waveguide design is usually optimized for the system's operating wavelength.
The switching voltage [V.sub.[Pi]], defined as the voltage required to effectuate full modulation, that is, to induce a phase difference of [Pi] between the two signals in the Mach-Zehnder configuration, is dependent upon the extent of overlap between the optical and electrical fields and the geometry of the coplanar electrode structure. The use of long electrodes enables lower drive voltages to be used since the effective interaction length is longer. The length used is often restricted to 5 mm by packaging limitations. Ideally, the distance between the active and ground electrodes should be small, necessitating a small electric field. There should be no interaction of the optical field with the metal electrodes since this would result in excess attenuation via the skin effect. The use of a dielectric buffer layer, such as Si[O.sub.2], between the waveguide and metal electrodes can provide a large degree of optical isolation, minimizing the skin effect, although the buffer layer should be thin enough to maintain a sufficiently large overlap between the optical and electrical fields. Using Equation 1, the switching voltage can be written as
[V.sub.[Pi]] = [Lambda]d/[n.sup.3][r.sub.(ij)k][Gamma]L (4)
d = electrode spacing, typically 10 [[micro]meter]
For previously given values [V.sub.[Pi]] is ^8.5 V. [V.sub.[Pi]] can be reduced by one-half by using the push-purl electrode configuration, that is, applying a voltage to the central electrode and grounding the two outer electrodes. This approach utilizes both arms of the Mach-Zehnder configuration, allowing the induced phase difference between the two propagating signals to be doubled effectively.
The RF drive power consumed is dependent on the voltage required to effectuate the degree of extinction required (generally maximum to minimum) and the characteristic impedance of the electrode structure R via the relationship
[Mathematical Expression Omitted]
If the electrode has a characteristic impedance of 50 [Omega], and the switching voltage is 5 V, the RF drive power is 250 mW.
Bandwidth is an important figure of merit for optical modulators, and mainly is limited by the degree of velocity mismatch or walk-off introduced between the copropagating optical and electrical fields. This velocity mismatch is dependent on the transmission line parameters, for example, electrode width and gap and buffer layer thickness. Defining the difference in propagation constants as [Delta]k, and neglecting RF loss in the electrodes, the 3 dB bandwidth can be written as
[Delta]f = c/2L[Delta]k
c = speed of light in vacuum L = electrode length
Titanium-indiffused LiNb[O.sub.3] modulators capable of operating at bandwidths of up to 20 GHz can be obtained commercially. Research designs have been proposed that are capable of modulation speeds of up to 150 GHz.
In spite of its relative maturity and attractive physical properties, LiNb[O.sub.3] as a waveguide modulator material may yet be surpassed by other waveguide technologies. Although relatively expensive and complicated to process, semiconductor waveguides offer the prospect of monolithic integration, that is, a combination of drive electronics, modulators, interconnects (waveguides) and detectors on a single chip. Polymeric waveguide materials exhibit electro-optic properties comparable to LiNb[O.sub.3], and offer the potential of low cost production, although problems with temporal and thermal stability are yet to be fully resolved. Therefore, LiNb[O.sub.3] seems to be fairly secure, at least for the next few years.
1. M. Lawrence, "Lithium Niobate integrated Optics," Reports on Progress in Physics, Vol. 56, No. 3, 1993, pp. 363.
2. I.P. Kaminov, L.W. Stulz and E.H. Turner, "Efficient Strip-Waveguide Modulator," Applied Phys. Letters, Vol. 27, 1975, pp. 555.
3. M.L. Shah, "Optical Waveguides in Lithium Niobate by Ion-Exchange Techniques," Appl. Phys. Lett., Vol. 26, No. 11, 1975, pp. 652.
4. J.L. Jackel, C.E. Rice and J.J. Veselka, "Proton-Exchange for High Index Wave-guides in LiNb[O.sub.3]", Appl. Phys. Lett., Vol. 51, 1982, pp. 607.
5. A. Loni, R.M. De La Rue and J.M. Winfield, "Very Low Loss Proton-Exchanged Waveguides with a Substantially Restored Electro-Optic Effect," Topical Meeting on Integrated and Guided-Wave Optics, Santa Fe, Technical Digest Series, Vol. 5, 1988 (OSA), pp. 85.
6. M. Rottschalk, A. Rasch and W. Karthe, "Electro-Optic Behaviour of Proton-Exchanged LiNb[O.sub.3] Waveguides," J. Opt. Commun., Vol. 9, 1985, pp. 19.
7. P.G. Suchoski, P.K. Findalky and F.J. Leonberger, "Stable Low Loss Proton-Exchanged LiNb[O.sub.3] Waveguides with no Electro-Optic Degradation," Opt. Lett., Vol. 13, No. 11, 1988, pp. 1050.
8. R.B. Childs and V.A. O'Byrne, "Predistortion Linearization of Directly Modulated DFB Lasers and External Module Transmission," Conference on Optical Fiber Communications, San Francisco, Technical Digest paper WH6 (OSA/IEEE), 1990.
9. L.M. Johnson and H.V. Roussell, "Linearization of an Interferrometric Modulator at Microwave Frequencies by Polarization Mixing," Photonics Tech. Lett., Vol. 2, 1990, pp. 810.
10. J.L. Jackel, A.M. Glass, G.E. Peterson, C.E. Rice, D.H. Olson and J.J. Veselka, "Damage-Resistant LiNb[O.sub.3] Wave-guides," Journal of Applied Phys., Vol. 55, 1984, pp. 269.
11. K. Noguchi and K. Kawano, "A Proposal for a Ti:LiNb[O.sub.3] Optical Modulator with Modulation Bandwidth of More Than 150 GHz," Electron Letters, Vol. 28, 1992, pp. 1759.
Armando Loni received his BSc degree with honors in natural philosophy and his PhD degree in lithium niobate integrated optics from the University of Glasgow, Scotland in 1984 and 1988, respectively. Since 1992, he has been at the Defense Research Agency in Malvern, UK as a senior scientific officer, where he researches the light-emitting properties of porous silicon. Previously, Loni served as research fellow at the University of Glasgow, where he investigated nonlinear optical properties of lithium niobate waveguide. Additionally he worked as a project manager at BT & D Technologies (now Hewlett-Packard), developing lithium niobate integrated optical devices. Loni is a member of the Institute of Physics and a chartered physicist.
|Printer friendly Cite/link Email Feedback|
|Date:||Feb 1, 1995|
|Previous Article:||Scalar measurement error in microwave and RF test configurations.|
|Next Article:||HBT amplifiers break $1 price barrier.|