Chapter 2

Arrayed Waveguide Gratings, Application and Design

2.1 2.1.1

Introduction Optical Communications Optical fibre is a popular carrier of long distance communications due to its

potential speed, flexibility and reliability. Attenuation and dispersion problems in fibre, which limit the practical speed and distance of communication, were partially resolved with the advent of the Erbium Doped Fibre Amplifier (EDFA)[2.1], eliminating problems caused by attenuation. However, the dispersion qualities of an optical fibre still force a compromise between transmission distance and bandwidth, making it necessary to refresh high-speed signals at intervals using opto-electronic repeaters. Solving the dispersion problem in this manner is expensive, due to the additional cost of high-speed electronics, and maintaining and upgrading the link is made more difficult and costly (especially with a buried or under-water link). A more elegant solution is found using Dense Wavelength Division Multiplexing (DWDM), which effectively increases the useable bandwidth in a system without electronic repeaters, and allows realisation of a true photonic network. 2.1.2 Arrayed Waveguide Gratings Dense Wavelength Division Multiplexing (DWDM) is an efficient method where several channels, each carried by a different wavelength, are transmitted through a single optical fibre, utilising more of the available bandwidth without increasing the effects of dispersion. Each channel, since it is effectively separated from the others, can be independent in protocol, speed, and direction of

16

communication. DWDM also helps realise an all-optical network architecture where signals are routed according to wavelength without the need for electro-optical conversion. As a result, this type of network is potentially faster and more flexible, and can be less costly to maintain when compared to other methods.

Demux

Mux

Figure 2.1 : (from [2.2]), An Add/Drop Multiplexer (ADM). Made reconfigurable by using space division switches (top in crossed state)

H M U X

M U X

Figure 2.2 : (from [2.2]) An Optical Cross Connect (OXC) employing a space division switch for each wavelength. Switch settings determine where each wavelength is routed.

Arrayed

Waveguide

Gratings

(AWGs)

are

optical

wavelength

(de)multiplexers used in DWDM. As well as performing basic (de)multiplexing functions, they can be combined with other components to create add/drop multiplexers, Figure 2.1 [2.2], used to pipe single wavelengths on and off the

17

network, and Cross Connects, Figure 2.2, used for routing. These devices can be passive, where the signal routing is fixed according to wavelength, or active, as in Figures 2.1 and 2.2, where optical switches are utilised to dynamically route the signals. Both circuits shown are transparent to the data format, can allow bidirectional transfer of information, and function entirely in the optical domain. These functions allow the construction of different transparent optical network topologies, examples of the three major types of these are described in the following subsection. 2.1.3 Forms of Photonic Network The simplest form of optical network is the point-to-point network. Optical multiplexers and de-multiplexers are required at each end of the link. In this configuration, DWDM simply increases the number of channels available through one fibre. Passive Optical Networks (PONs) (Figure 2.3) [2.2] use a wavelength (de)multiplexer as a passive optical router, each wavelength servicing an Optical Network Unit (ONU). This allows the ONUs to share a single long optical fibre link back to the central office (CO).
ONUs Central Office Passive Router
1 + 2 + 3 + 4 3 4 1 2

Figure 2.3: A Passive Optical Network

Long haul networks tend to have more than one point where channels are added to and removed from the system, for example to provide a bi-directional channel to an optical network unit. To add and remove channels optical add drop multiplexers (OADMs) and Optical Cross Connects (OXCs) are utilised (Figures 2.1 and 2.2),

18

[2.2], to allow single channels to be individually piped off the network, and to route channels between sections of the network respectively. Figure 2.4 shows an example configuration of a Long Haul network.

ADM

ADM

ADM ADM OXC

ADM ADM

Figure 2.4 : A Long Haul Network, utilising Add-Drop Multiplexers (ADMs) and Optical Cross Connects (OXCs)

2.1.4

Summary

The Arrayed Waveguide Grating (AWG) plays a crucial role in the realisation of modern optical networks. The next section introduces the principles of operation of the AWG, and then examines the design process.

19

2.2 2.2.1

Arrayed Waveguide Grating (AWG) Operation Principles Basic Operation of The AWG In the previous section we established that AWGs are essential components

for the realisation of DWDM and optical networks. In this section, the basic operation of an AWG as a de-multiplexer is described.
(c) Arrayed Waveguides

(a) Input Waveguide

(e)

(b) Input FPR

(d) Output FPR

Figure 2.5 :[2.2] The structure of an Arrayed Waveguide Grating de-multiplexer.

Figure 2.5 shows the structure of an AWG. The input (a) consists of several channels, typically between 8 and 40 in commercial devices, carried on separate frequencies. Channel spacings of 100GHz or 50 GHz are common in commercial devices, although 25GHz [2.3] and 10GHz [2.4] spacings have been achieved under laboratory conditions. The operational wavelength is commonly around 1.55m where attenuation is lowest in optical fibres. All waveguides in the AWG tend to be single-moded to ensure predictable propagation through the device. Light couples from the input waveguide (a) into the Free Propagation Region (FPR) (b) and disperses to illuminate the arrayed waveguides (AWs) starting on a curved plane as shown in Figure 2.6, along which the light exhibits a constant phase profile. Each AW increases in length by L compared to the previous one in the array, where L = m w / neff, m is an integer number, w is the central operational

20

wavelength, and neff the effective refractive index, (/k0), of the single mode supported by each waveguide forming the AWG. Consequently, at the central wavelength, a constant phase profile is exhibited at the end of the AWs, an integer number of cycles out of phase, along the same plane, as shown in Figure 2.6. Therefore, at the central frequency, the light focuses at the centre of the plane (e), where an output waveguide is positioned to capture the focussed light. Different wavelengths of light will exhibit different amounts of phase change and, due to the increments in length of each waveguide, the phases will change along the AW output plane, causing the focal point to move along the focal plane (e) at the end of the FPR. An output waveguide is positioned on the output plane to pick up each input frequency (channel). Arrayed Waveguides Apeture Width Input / Output Waveguides

RFPR

Focal Plane
Figure 2.6 : The Input / Output Free Propagation Region.

2.2.2

AWG tolerances An AWG is a very large and inherently complex structure, typically several

square centimetres in area and comprised of multiple waveguides. The AWG is clearly also very sensitive to any phase error, particularly at the end of the Arrayed Waveguides. Consequently, careful and rigorous design is critical for its correct 21

operation. A design process based on prototyping is prohibitively expensive, both in terms of cost and time, so it is highly desirable to develop modelling tools that can provide a detailed analysis of the device operation. However, due to the component size, even relatively detailed analyses have previously resorted to approximate semianalytical methods, e.g. [2.5][2.6], which do not fully characterise the structure (Chapter 1). Some of the current design methods are reviewed in the following section. 2.3 2.3.1 Current Design Process of an AWG (Basic De-multiplexer) Introduction This section looks at the analytical methods used to design an AWG. The design of an AWG is covered in detail in [2.2], [2.5], [2.7], and [2.8]. An AWG is specified by the following characteristics: Number of Channels Central Frequency fc, and Channel spacing fch Free Spectral Range fFSR Channel bandwidth Maximum insertion loss Maximum non-uniformity Maximum crosstalk level Polarisation dependence

We concentrate first on the basic design rules that govern AWG dimensions. 2.3.2 Receiver Waveguide Spacing The spacing of the receiver waveguides affects the adjacent-channel crosstalk of the system and is covered in detail in [2.5] and [2.6]. A calculated crosstalk of 22

around -30dB is normally considered sufficient, where other sources of crosstalk, resulting from flaws in design and manufacture, normally become dominant. 2.3.3 Free Propagation Region (FPR) length The length of the Free Propagation Region is determined by the maximum acceptable channel non-uniformity (expressed in dB). Channel non-uniformity is defined in [2.5] as the difference in intensity of the central and edge channels of the AWG, and is the result of the variation of the waveguide mode far field with angle. Channel non-uniformity can be estimated analytically, as described in [2.5], or determined through numerical simulation (as will be described in Chapter 8). By specifying the maximum channel non-uniformity, a value for the maximum dispersion angle (MAX) can be obtained. If the distance to the outermost output waveguide, smax, is known, then the minimum length of the Free Propagation Region, RFPR, is calculated by using RFPR = smax / MAX. 2.3.4 Arrayed Waveguide Length Increment, L The increment in length of the Arrayed Waveguides (AWs) is determined from the required dispersion D = ds / df, where ds is the displacement of the focal spot on the image plane, and df is the change in frequency to cause this displacement. Minimum dispersion is related to the receiver waveguide spacing and the free propagation length. Dispersion is calculated from [2.7]:

D=

ds d = RFPR df df

(2.1a)

where,

= a sin

( m2 ) / FPR m2 d da FPR a

(2.1b)

23

where = L, is the propagation constant in the waveguides (in rads m-1), and

FPR that in the Free Propagation Region. The waveguide spacing, da, is chosen to be
as small as possible to maximise coupling efficiency from the FPR to the AWs. m is the order of the phased array. 2.3.5 Arrayed Waveguide Aperture Width The aperture width, shown on Figure 2.6, is the effective capture width of the AWs. The number of AWs, is determined by the aperture width and the crosstalk characteristics of the waveguides used in the AWG, which affects the separation between the AWs. It is covered in depth in [2.5] and [2.7]. The aperture size also affects the amount of light captured by the grating and is normally chosen to capture the majority of the expanded field at the end of the input FPR. 2.3.6 Free Spectral Range (FSR) The required frequency range for the arrayed waveguide grating is a fundamental design factor. FSR is defined as the frequency shift, for which the phase-shift, , equals 2, i.e.:

2f FSR ~ NgL = 2 c

(2.2a)

~ where Ng is the group index of the waveguide mode and c the velocity of light in free space. The FSR has to be sufficiently large to accommodate the required operating frequency range. Two frequencies separated by the Free Spectral Range FSR and input into an AWG de-multiplexer will focus and leave though the same output waveguide, since their phase at the outputs is the same. It follows from 2.2a that dN g ~ Ng = N g + f df
(2.2b)

24

so

c f FSR = ~ N g L

(2.2c)

Since the FSR is not constant with frequency the analysis above is only approximate, but in most cases it is sufficient.
2.3.7 Summary

The parameters described in this section determine the basic dimensions for the AWG. Although the procedures outlined provide a basic guidance for a working design, a good design requires the consideration of many other issues, as detailed in the next section.
2.4 2.4.1 Issues Affecting the Performance of Arrayed Waveguide Gratings Crosstalk

The causes of inter-channel crosstalk are many, and may be a result of the design or imperfect fabrication of the AWG. The primary source taken into account in initial design is the inter-channel crosstalk, caused by the overlap of the focussed spot in the output FPR with adjacent output waveguides, as covered in [2.5]. Approaches that are more rigorous take into account potential AW phase inaccuracies, caused by design or fabrication anomalies, which cause the spot to spread, increasing crosstalk. This form of crosstalk, however, is easily controlled by increasing the separation of the output waveguides. Often, as evaluated in [2.9], there are design tradeoffs between crosstalk and other desirable characteristics, such as insertion loss or channel spacing. Crosstalk is also likely to occur as a consequence of more complex effects in the AWs, such as through light propagating in the AWs in modes other than the

25

single-waveguide fundamental modes. This would adversely affect the phase and amplitude distributions at the output of the Arrayed Waveguides. Due to the assumptions inherent within AW design using current design tools, these effects are difficult to evaluate completely. This is an area where an improved understanding is presented in this Thesis, particularly in Chapters 6 and 7, gained through the application of advanced numerical design tools.
2.4.2 Insertion Loss

The primary cause for insertion loss in the AWG is due to inefficient coupling at the interface between the first FPR and the AWs. Due to reciprocity [2.2], identical loss occurs at the second AW - FPR interface into higher diffraction orders. Coupling efficiency, and therefore insertion loss is largely determined by the separation of the AWs at these interfaces, where smaller separations increase the coupling efficiency [2.2]. However, at small separations, coupling between the AWs becomes significant. This effect has to be carefully quantified through the Finite Difference- Beam Propagation Method (FD-BPM) or another simulation method to avoid phasing errors in the AWs. Other areas that cause loss may include

Material losses Scattering due to fabrication errors and waveguide roughness De-focussing of the spot on the output plane due to phase errors, decreasing coupling efficiency into the output waveguide.

2.4.3

Polarisation Dependent Dispersion

Unless specifically engineered, waveguide boundary conditions cause quasiTE and quasi-TM polarised modes to propagate at different speeds (birefringence), particularly in the case of strongly confining waveguides. As well as birefringence

26

due to waveguide geometry, stresses within the structure may occur due to fabrication processes that can cause anisotropy and stress birefringence [2.2]. Birefringence causes a second shadow spot on the output plane of the FPR, where the TE- and TM- like polarisations have experienced different phase shifts, potentially coupling with the wrong output waveguide and causing inter-channel crosstalk. Several methods have been presented to reduce this polarisation dependence, such as making the Free Spectral Range equal the difference between the phase change between TE and TM polarised modes, hence overlapping the TE/TM spots [2.10], or using a polarisation converting lambda half-plate half way along the AWs [2.11], causing both polarisations to undergo the same phase change. Stress birefringence in the waveguides is treated to some extent by coating the waveguides with a stress inducing film or, as covered in [2.12], in the case of silicon on Silica waveguides, stress-relieving grooves can be cut either side of the waveguide.
2.4.4 Polarisation Rotation

Curved waveguides by their nature will exhibit a certain amount of Polarisation Rotation, where the light energy is transferred from one polarisation to the other. This effect is exploited in polarisation conversion devices in [2.13][2.14]. Polarisation rotation in the curved waveguides of the AWG is an effect that has not presently been covered by any paper on the subject of AWGs even though it may be in some cases a contributor to dispersion in the device.
2.4.5 Passband Shape

A sharp passband, such as the one illustrated schematically in Figure 2.7a, allows very little error in laser frequency and AWG wavelength tolerance. It is

27

desirable in most circumstances to flatten the passband, as illustrated in Figure 2.7b so that the device produces a similar output for small changes in laser wavelength. As stated in [2.2], the ideal shape for the passband of an AWG is to have a flat top, with a deviation of less than 1dB, for over 70% of the channel separation, and as wide a -3dB bandwidth (Full Width Half Maximum (FWHM)) as possible without increasing crosstalk. (a)
Gain Frequency

3dB Passband
Gain

(b)
Frequency

-3dB

-3dB

Figure 2.7 : (a) a typical AWG channel passband & 2.7b, with flattened, widened, passband

Various methods are used to flatten the pass band [2.15], such as broadening the capture width of the receiver waveguide by using wide, multimode waveguides. Another approach is to create a spot that has a broad flat centre with steep cut-off at either side. The latter is achieved by using Multi-Mode Interference Devices (MMIs) [2.16], y-junctions or a parabolic tapered horn [2.17]. Using these methods, a pass band of over 50% of the channel separation of 200GHz is achieved in [2.18]. Alternative methods of flattening the spectral response of the AWG are covered in [2.19], where the spectral response is manipulated by altering the respective lengths of each arm of the AWG and their positions at the edge of the Free Propagation Regions (FPRs). The authors of [2.20] take a similar approach through de-focussing the phase profile in the AWG, to make it near parabolic, again widening

28

the pass band. In all cases, flattening the pass band response also increases the insertion loss of the AWG and often increases crosstalk.
2.4.6 Passband Position

Design or fabrication errors cause a phase error at the end of the AWs which may shift the focal point away from the expected position, hence affecting the passband position. To compensate for this, the position of the focal point may be shifted by adjusting the temperature of the AWG [2.4]. If the phase error for each waveguide is random, causing the spot to defocus, then separate heaters for each individual Arrayed Waveguide may be implemented. However this approach increases the energy consumed by the device and requires additional control circuitry, increasing the cost of manufacture.
2.4.7 Summary

In this section issues that affect the performance of the AWG have been discussed. All issues that occur in Arrayed Waveguide Gratings brought up in this section can be attributed to issues with the Arrayed Waveguide part of the structure. However, due to the complexity of the structure, the Arrayed Waveguides (AWs) have not previously been fully modelled. Consequently, issues such as crosstalk are not accurately predicted, since complex propagation mechanisms in the AWs are not taken into account. It was suggested in Chapter 1 that FD methods, and in particular the FD-BPM, may be adapted to allow a more thorough investigation into AW structures, in order to predict or even reduce the impact of unwanted properties of the AWs, ultimately improving performance and simplifying AWG design. The next section summarises areas which the FD algorithms can improve analysis of the AWG, to allow more thorough simulation of the AWs.

29

2.5

Conclusion

This chapter described the basic operation of the AWG and summarised the methods used for AWG design. However it is highlighted that current methods of analysis do not accurately predict effects observed in fabricated structures, such as increased inter-channel crosstalk. It is concluded that although the semi-analytical analysis methods presented in Section 2.3 are sufficient to design an Arrayed Waveguide Grating (AWG), these techniques, due to the number of assumptions made, limit the analysis so that levels of inter-channel crosstalk, and other effects summarised in Section 2.4, cannot be predicted. Assumptions that are made include:

Propagation of power in the waveguides of the AW structure is to be wholly in the fundamental mode of the waveguide

The field at the end of the first FPR is assumed not to excite modes other than the fundamental mode of each waveguide in the AW set, such as higher order and grating modes.

Coupling, and other propagation effects in the AW structure is normally ignored

The propagation is assumed to be constant in waveguides of varying radius.

To improve the analysis, and allow fewer assumptions to be made, a numerical method may be used, such as the FD-BPM, introduced in Chapter 3, which is popular due to its relative speed and accuracy. However even FD-BPM has hitherto not widely been considered to be a viable simulation tool for the full simulation of large components such as the AWG, due to its memory and processing requirements although successful simulations of the output Free Propagation Regions and coupling areas of AWGs have been conducted 2D, and in [2.6] a 3D model of the coupling regions of the AWs is used to assist in the analysis of coupling and defocusing

30

effects in the AWG. (In this case, for speed of simulation, a Fourier optics modelling approach is used in conjunction with FD-BPM). However, even in [2.6], the modelling of the AW structure relies on semi-analytical methods, with all light energy in the AW structure assumed to propagate in the fundamental mode. Consequently, in this Thesis, the goal is to improve the understanding of the AW region of the AWG through the use of improved FD techniques. To achieve this the basic FD-BPM is extended and improved together with an accompanying FD mode solver, to allow efficient and accurate modelling of the complex mechanisms present in the AWs of the AWG. The next Chapter introduces the theory behind the FD-BPM and FD Mode Solver in Cartesian co-ordinates, and develops the basic algorithms.
2.6 References

[2.1] P. C. Becker, N.A. Olsson, J.R. Simpson, Erbium-Doped Fibre Amplifiers, Academic Press, 1st edition, May 1999 [2.2] K. A. McGreer, Arrayed Waveguide Gratings for Wavelength Routing, University of Manitoba and TRLabs, IEEE commications Magazine, December 1998. [2.3] Research by Nippon Telegraph and Telephone Corporation, 400-channel arrayed waveguide grating with 25GHz spacing, 2002. [2.4] Hiroaki Yamada, Kazumasa Takada and Seiko Mitachi, Crosstalk Reduction in a 10 GHz Spacing Arrayed-Waveguide Grating by Phase-Error Compensation, Journal of Lightwave Technol., vol. 16, no. 3, March 1998 [2.5] M. K. Smit, C van Dam, Phasar based WDM-devices: principles, design and applications, IEEE J Selected Topics in Quantum Electron., Vol. 2, No. 2, June 1996, pp. 236-250

31

[2.6] A Klekamp and R Munzer, Imaging Errors in Arrayed Waveguide Gratings, IEEE Optical and Quantum Electron., Vol. 35, 2003, pp. 333-345 [2.7] P Munoz, D Pator, and J Capmany Modelling and Design of Arrayed Waveguide Gratings, J. Lightwave Technol., vol. 20, no. 4, April 2002 [2.8] D. Fonseca, R. Lus and A. Cartaxo, Design and performance of AWG multiplexer / demultiplexer in WDM systems, 3Conferncia Nacional de Telecomunicaes, Figueira da Foz, Portugal, 2001, pp. 164-168. [2.9] J Lam and L Zhao Design Trade-offs For Arrayed Grating DWDM MUX/DEMUX, SPIE Proc. Volume 3949, WDM and Photonic Switching Devices for Network Applications, April 2000, pp. 90-98 [2.10] L H Spiekman and M R Amersfoort, Design and Realisation of polarization independent phased array wavelength demultiplexers using different array orders for TE and TM, J Lightwave Technol., Vol. 14, 1996, pp. 991-995. [2.11] H Takahashi, Y Hibino, and I Nishi, Polarization-insensitive arrayedwaveguide grating wavelength multiplexer on silicon Opt. Lett., Vol. 17, No. 7, 1992, pp. 499-501 [2.12] Christoph K Nadler et al, Polarisation Insensitive, Low-Loss, Low-Crosstalk Wavelength Multiplexer Modules, J. Selected Topics in Quantum Electron., Vol. 5, No. 5, 1999, pp. 1407 - 1412 [2.13] C. van Dam et al, "Novel compact polarization converters based on ultra short bents", IEEE Photon. Technol. Lett., vol.8, 1996, pp.1346-1348 [2.14] W.W. Lui, , T Hirono, , K Yokoyama and Wei-Ping Huang, Polarization Rotation in Semiconductor Bending Waveguides : a coupled-mode theory formulation, J. Lightwave Technol., Vol. 16, No. 5 , 1998, pp. 929 -936

32

[2.15] C. Dragone, Efficient Techniques for widening the passband of a wavelength router, J of Lightwave Technol., Vol. 16, No. 10, 1998, pp. 1895 1906. [2.16] P. Munoz, D. Pastor and J. Capmany, Analysis and design of arrayed waveguide gratings with MMI couplers, Optics Express, 2001, Vol. 9, No. 7 [2.17] K.Okamoto and A. Sugita, Flat Spectral response Arrayed Waveguide Grating multiplexer with parabolic waveguide horns Electron. Lett., Vol. 32, 1996, pp. 1661-1662 [2.18] S. Suzuki, Arrayed Waveguide Gratings for dense-WDM systems,

IEEE/LEOS Summer Topical Meeting WDM components Technology, Montreal, Canada, Aug 11-15, 1997, pp. 80 81 [2.19] T Kamalakis and T Sphicopoulos, An Efficient Technique for the Design of an Arrayed-Waveguide Grating with Flat Spectral Response, J Lightwave Technol., Vol. 19, No. 11, 2001, pp. 1716-1725 [2.20] M C Parker and S D Walker, A Fourier-Fresnel Integral-Based Transfer Function Model for a Near-Parabolic Phase Profile Arrayed Waveguide Grating, IEEE Photon. Technol. Lett., Vol. 11, No. 8, 1999, pp. 1018-1020

33