Advanced
Planar Waveguide Devices for Communication and Sensing Applications
Planar Waveguide Devices for Communication and Sensing Applications
Journal of the Optical Society of Korea. 2010. Dec, 14(4): 290-297
Copyright ©2010, Optical Society of Korea
  • Received : October 10, 2010
  • Accepted : November 11, 2010
  • Published : December 25, 2010
Download
PDF
e-PUB
PubReader
PPT
Export by style
Share
Article
Author
Metrics
Cited by
TagCloud
About the Authors
Katsunari Okamoto
katsu.okamoto@aidicorp.com
Abstract
The paper reviews progress and future prospects of two kinds of planar waveguide devices; they are (a) silica and silicon photonics multi/demultiplexers for communications and signal processing applications, and (b) a novel waveguide spectrometer based on Fourier transform spectroscopy for sensing applications.
Keywords
I. INTRODUCTION
Silicon photonics is widely regarded as a promising technology to meet the requirements of rapid bandwidth growth and energy-efficient on-chip communication while reducing cost per bit. In all potential application areas it is widely recognized that use of wavelength division multiplexing (WDM) techniques will be critical in achieving the required high levels of data transmission. Si photonics devices will have to deal with several tens of different wavelengths of light in the next-generation multi-core CMOS chips.
There are mainly four kinds of devices capable of multi/demultiplexing tens of WDM signals; they are ring resonators,lattice-form filters, arrayed waveguide gratings (AWG) and planar Echelle gratings. The former two are cascaded devices relying on temporal multi-beam interference effect and the latter two utilize spatial multi-beam interference effect. In order to achieve good crosstalk characteristics in the temporal and spatial multi-beam interference effects, uniformity of effective index nc (= β /k ), where β and k denote propagation constant and wave number, is critically important.Filter characteristics of four kinds of devices will be investigated and performance limitations of silicon photonics filters are discussed.
In the latter part of the paper, an integrated-optic spectrometer based on Fourier-transform spectroscopy will be described. A novel planar waveguide spectrometer consists of interleaved Mach-Zehnder interferometer (MZI) array.Practical importance of Fourier-transform spectrometer is the ability to correct for interferometer defects caused by effective index fluctuations (phase errors) in data processing stage. Successful measurement results of the signal spectrum with 20-GHz resolution by the spectrometer implemented in silica-based planar waveguide will be presented.
II. SILICON PHOTONICS WDM FILTERs
Flat-top passband characteristics are required for filters in most of the WDM applications [1 - 4] . Crosstalk degradation in coupled ring resonators and lattice-form filters are caused by both gap width error in directional couplers and optical path length error in the waveguides.
Effective-index fluctuation δnc in silica Planar Lightwave Circuits (PLCs) have been measured to be about δnc ~1×10 ?6 by using Fourier transform spectroscopy [5] . Good crosstalk characteristics are obtained in silica-based AWGs as shown in Fig. 1 . The relation between the effective-index fluctuationδnc and crosstalk (XT) has been investigated theoretically and experimentally [6] . Based on the investigations, the empirical expression for the relation between δnc and XT has been obtained as
PPT Slide
Lager Image
where Lctr and λ denote array waveguide length in the central region and the wavelength of light signal, respectively. Lctr of AWG in Fig. 1 is Lctr ~ 15 mm. Substituting
PPT Slide
Lager Image
Demultiplexing properties of 64 ch-50 GHz spacingsilica AWG.
δnc ~ 1×10 -6 , Lctr ~ 15 mm, and λ = 1.55 ㎛ into Eq.(1), we obtain XT ~ -40 dB. Although Eq. (1) gives good agreement with the experimental results, the author believes that it should be proved analytically. Crosstalk of the planar Echelle gratings is obtained based on the similar investigations as
PPT Slide
Lager Image
where ns and δζ denote the effective index in the slab region and the facet position fluctuation in the reflection mirrors. It is known from Eq. (2) that δζ should be about 5 nm in order to obtain -40 dB crosstalk in silica Echelle grating ( ns ~ 1.5). Such a small δζ would be possible in the laboratory experiments [7] . However, the mask resolution for the commercial PLC products is normally about 25 nm~ 50 nm. This is the reason why commercial planar Echelle gratings could not achieve sufficient crosstalk value to be used in the current WDM systems.
Effective-index fluctuation δnc in Si-wire waveguides has been evaluated from the measurements on the variation of resonant wavelength in the cascaded ring resonators [8] . Peak-to-peak resonance wavelength variation is 2 δλ ~ 2.4 nm. Then, δnc is obtained as δnc=ncδλ/λ0~2×10-3 , where nc = 2.519 and λ0 = 1.55 ㎛. Effective-index fluctuationδnc in Si-wire waveguides can also be calculated by the vectorial finite element method [3] . Fig. 2 (a) shows nc of Si-wire waveguide with thickness 2t = 0.22 ㎛ (blue) and its dependence on the core width 2a (red). dnc / d (2 a )at the typical core width 2a = 0.5 ㎛ is 1.3 ㎛ -1 = 1.3×10 -3 nm -1 . Fig. 2 (b) shows nc of Si-wire waveguide with width 2a =0.5 ㎛ (blue) and its dependence on the core thickness 2t(red). dnc / d (2 t ) at the typical core thickness 2t = 0.22 ㎛ is 3.4 ㎛ -1 = 3.4×10 -3 nm -1 .
Core side-wall roughness of Si-wire waveguide is measured to be about σwidth ~ 2 nm [9] . Top silicon thickness uniformity of 6-inch SOI (silicon on insulator) wafer is measured to be about +/- 2 nm [10] . However, the photonic device occupies just ~1 mm square or less.
PPT Slide
Lager Image
(a) Effective index nc of Si-wire waveguide with 0.22 μmthickness (blue) and its dependence on the core width 2a (red).(b). Effective index nc of Si-wire waveguide with 0.5 μmwidth (blue) and its dependence on the core width 2t (red).
PPT Slide
Lager Image
Thickness variation in 6-inch SOI wafer.
Then, the thickness fluctuation would be less than σthickness ~ 1 nm as shown in the central region of Fig. 3 [6] .Effective-index fluctuation δnc in Si-wire waveguide is then estimated to be δnc =[ dnc / d (2 a )]× σwidth ~2.6×10 -3 ,or δnc =[ dnc / d (2 t )]×σthickness ~3.4×10 -3 , respectively. Both of the theoretical estimations for the effective-index fluctuation δnc agree well with the measured value. Effective index fluctuation δnc against core width variation can be reduced by enlarging core width in the straight regions. For example, dnc / d (2 a ) becomes to 0.35 ㎛ -1 =3.5×10 -4 nm -1 at 2a = 0.8 ㎛ [11] as shown in Fig. 2 (a).
Si-rib waveguide structure allows us to obtain much smaller δnc as shown in Fig. 4 , where core width W, core height H, and etch depth h are W = 0.5 ㎛, H = 0.3 ㎛,and h = 0.15 ㎛, respectively. dnc / dW at W = 0.5 ㎛ is 6.4×10 -4 nm -1 . However, the minimum bending radius becomes much larger compared to Si-wire waveguides and makes the device size larger. Therefore, Si-wire waveguide is still very attractive in making compact photonic devices.
Effective-index fluctuations of the order of δnc =3×10 -4 ~ 2×10 -3 determine the crosstalk values of the current Si-wire AWGs to be about -20 dB [11 - 13] .
Echelle grating has been believed to be advantageous over AWG because only the facet position error causes crosstalk degradation. But, this is not true in Si Echelle grating since the effective-index fluctuation δns in the slab waveguide is substantially large as shown in Fig. 5 . dns / d (2 t ) at the typical core thickness 2t = 0.22 ㎛ is 3.3×10 -3 nm -1 . Then, Eq. (2) for Si Echelle grating should be rewritten as
PPT Slide
Lager Image
Effective index nc of Si-rib waveguide (blue) and itsdependence on the core width W (red).
PPT Slide
Lager Image
Effective index ns of Si slab waveguide (blue) and itsdependence on the core width 2t (red).
PPT Slide
Lager Image
where Lctr denotes the total path length in the central region of the slab. Eq. (3) explains why large channel count(30 ch-3.2 nm spacing) Echelle grating has about -15 dB crosstalk, though small channel count (4 ch-20 nm spacing)one had about -30 dB crosstalk [14 , 15] . The total path length Lctr in the large channel count Echelle grating is almost 5 times longer than that of small channel count one.Therefore, the second term in Eq. (3) becomes dominant in the large channel count Echelle grating.
In order to achieve good crosstalk characteristics (< -20 dB) in both Si-wire AWG and Echelle grating, (a) improvement in core width and thickness uniformity and/or(b) size reduction to make Lctr small are strongly required.
Cascaded ring resonators and lattice-form filters are also attractive candidates for filter applications in Si photonics. Fig. 6 shows coupled resonator optical waveguide (CROW) [1] . Kappa’s are amplitude coupling coefficients in the directional couplers. Ring perimeter L = 27.7 ㎛ for the ring diameter of 4.4 ㎛. Free spectral range (FSR)between the adjacent resonance order is about 2,600 GHz,which allows us to make 13 ch-200 GHz(1.6 nm) spacing WDM filter. FSR is inversely proportional to L.Therefore, ring resonator using other type of waveguide(such as silica PLC with bend radius of 5 mm) cannot obtain sufficiently wide FSR. Fig. 7 (a) shows a filter response without effective index fluctuation (dotted lines)and with index fluctuation δnc = 5×10 -4 only in the third ring (solid lines). It is shown that δnc less than 10 -3 degrades crosstalk characteristics drastically. Fig. 7 (b) is a filter response when κ 0 is deviated from its ideal value 0.47 to 0.42. This amount of κ 0 deviation can be caused by about +/- 13 nm gap width variation in the directional coupler.
Fig. 8 shows schematic configuration of three-stage latticeform filter [2] . c 1 to c 4 are power coupling coefficients in the directional couplers. Channel spacing between the two
PPT Slide
Lager Image
Coupled resonator optical waveguide (CROW).
PPT Slide
Lager Image
(a) Filter response without effective index fluctuation(dotted lines) and δnc = 5x10-4 only in the third ring (solid lines).(b). Filter response when κ0 is deviated from 0.47 to 0.42.
PPT Slide
Lager Image
Three-stage lattice-form filter.
output ports is given by
PPT Slide
Lager Image
where ΔL, Nc and c are path length difference, group index ( Nc=nc-λdnc/dλ ) of the waveguide and light velocity. In the typical Si-wire waveguide, ΔL = 93 ㎛ gives the channel spacing Sch = 400 GHz. Fig. 9 shows a filter response without effective index fluctuation (dotted lines) and with random index fluctuation ' δnc ' <5×10 -4 (solid lines). It is shown that δnc less than 10 -3 degrades crosstalk characteristics drastically. It is also shown by other numerical simulation that when c 2 is deviated from
PPT Slide
Lager Image
Filter response without effective index fluctuation(dotted lines) and random index fluctuation 'δnc' < 5x10-4(solid lines).
its ideal value 24.8% to 17.8%, crosstalk becomes about-20 dB. This amount of c2 deviation can be caused by about +/- 20 nm gap width variation in the directional coupler. Experimentally, crosstalk level of about -10 dB has been obtained in Si-wire lattice-form filters [16] .
III. PLANAR FT SPECTROMETER
Spatial heterodyne spectroscopy (SHS) is an interferometric Fourier-transform technique based on a modified Michelson interferometer with no moving parts and relying on analysis of stationary interference patterns [17] . In the bulk-optic SHS, the mirrors of the Michelson interferometer are replaced by diffraction gratings to offer scanning capability for two wavefronts. The SHS concept has been generalized into a waveguide Mach-Zehnder interferometer (MZI) array based on cosine fast Fourier transform (FFT) [18] . The path length difference in each asymmetrical MZI increases by ΔL , similar to AWG [3] . For any input wavelength, the entire output of a MZI array creates spatially oscillatory patterns, where the oscillation period is wavelength dependent due to the dispersive nature of the asymmetrical MZI.Each input wavelength has a unique fringe representation,which is the basis of the Fourier-transform relation between the input spectra and the output interferogram. Employing multiple input waveguides in the waveguide SHS is an important advantage over the existing AWG-type spectrometer,since it greatly increases the light capturing capability in proportion to the number of MZIs by using, for example,photonic lantern technology [19] . A photonic lantern is a device that efficiently converts light from a multimode fiber tip for light capturing to single-mode fibers connected to MZI input waveguides. The practical importance of the Fourier-transform spectrometer is the ability to correct for interferometer defects (phase errors) in data processing stage. The first waveguide SHS have been fabricated in silicon-on-insulator rib waveguides using stacks of multiple MZIs [20] . However, a measurement result on the spectrum has not been reported so far to our knowledge.
We reported a novel planar waveguide SHS configuration with an interleaved MZI array [21] . The interleaved configuration allows us to put 2~3 times more MZIs than the simple stacking layout when we compare two layout methods for the same spectrometer parameters. We made an array of 32 MZIs in silica-based PLC. The path length difference increment was ΔL = 162 ㎛, which resulted in 20 GHz resolution at the 1.55 ㎛ region. Phase errors in the asymmetrical MZIs were measured by the thermo-optic phase scanning method [22] . The spectrum of the light source was successfully retrieved by correcting the MZI output data with the phase error values.
Figure 10 shows the configuration of the novel waveguide SHS with the interleaved MZI array. The total number of MZIs is N = 32. The waveguide core size is 4.5㎛× 4.5 ㎛ with 1.5% refractive-index difference.The minimum bend radius is 2 mm. White boxes indicate 3 dB couplers consisting of either directional couplers or multimode interference couplers. The input and output waveguide separations at the chip ends are 15 and 50 ㎛,respectively. Waveguide arms in the MZI are intentionally inclined to both sides so that the waveguides intersect by more than 45º with each other. It is known that the excess loss of the waveguide crossing can be reduced as low as~ 0.02 dB/intersection when crossing angle is larger than 45º [23] . Dummy crossing waveguides are placed to make the total number of waveguide crossing equal for all MZIs. The chip size, including input and output fan-out regions, is 43mm× 25 mm . We measured both cross-port and through-port outputs p (k ) and q(k) in the k -th (k = 0~ N-1) MZI so that the spatial nonuniformity of the input light distribution can be corrected. The output power of each port was measured by shifting the single-mode fiber in the current preliminary experiments. Fiber-to-fiber loss of MZIs are measured to be 4.2 ~ 5.4 dB.
The operational principle of a SHS spectrometer based on the MZI array has been elaborated in [18] . For a signal s ( f ) passing through the k -th MZI, a normalized cross-port output is given by assuming negligible waveguide loss as
PPT Slide
Lager Image
Planar waveguide SHS configuration with interleavedMZI array.
PPT Slide
Lager Image
where β is a propagation constant, FSR is a free spectral range determined by ΔL , and S = ∫ f0 f0+ FSR s ( f ) df . f0 is denoted as the Littrow frequency at which phase delays in different MZIs become integer multiples of ( β ( f 0 ) ΔL = 2 ). Since MZI response repeats periodically with FSR, one needs to block the unnecessary spectral range by a bandpass filter. Based on the discrete cosine Fourier transform, the input spectrum s ( fn ) ( fn = f 0 + n ∙ FSR /□, where □=2 N =64) can be calculated from the measured output power P ( k ) as
PPT Slide
Lager Image
In the above equation, A is a constant and P ( k ) for n = N ~ □ ?1 is assumed to be P (□ ? k ). Although FSR was sampled with N points in [18] , we sample it with 2N points in order to correctly represent the MZI outputs. Because MZI responses for the signal in the upper half of FSR, s ( fn ) ( n = N ~□ ?1), have identical spatial fringe representation to those of the signal in the lower half, only the lower half of the signal spectrum can be measured. Resolution of the spectrometer is given by δ f=c /( □ncΔL )= 20 GHz. Phase errors caused by effectiveindex fluctuations in the MZI array deteriorate the accuracy in the retrieved signal by Eq. (6).Phase error δφk in the k -th MZI is expressed as δφk = ( 2π/λ0 ) δnc ( k ) Lk ,where δnc(k) and Lk denote effective-index fluctuation and MZI arm length, as shown in Fig. 11 (a). A heater with length l was placed from outside of the chip on either one of the MZI arms to measure δφk . The through-port transmittance q (k) under the thermo-optic effect is given by
PPT Slide
Lager Image
Here, H is a heater power applied to the phase shifter, α is a coefficient of thermo-optic refractive-index change per unit heater power and λ0=c/ f0 . Figure 11 (b) shows an example of the thermo-optic phase scanning measurement.The first extinction point indicated by H0 corresponds to the point at which the phase error is compensated for. The power between two adjacent extinction points HT corresponds to an optical path length change with λ0. δφk is then given by δφk=2π ∙H0/HT . Effective-index fluctuation is obtained as δnc ( k ) =( δφk / Lk ) λ0 / . Measured
PPT Slide
Lager Image
(a) Mach-Zehnder interferometer in the kth stage and(b) through-port response versus heater power.
PPT Slide
Lager Image
Measured effective-index fluctuation in the MZIarray.
δnc ( k ) in the MZI array is shown in Fig. 12 .
A major advantage of the waveguide SHS based on Fourier-transform spectroscopy is that the interferometer defects due to effective-index fluctuations can be corrected in data processing stage. On the contrary, physical refractiveindex correction has been required for AWGs after the phase error measurement [24] . A known signal spectrum was coupled into the waveguide SHS. The normalized cross-port output power distribution P ( k ) and the retrieved signal spectrum s ( λn )( λn = c/ fn ) by the cosine FFT are shown in Figs. 13 and 14 , respectively. The Littrow wavelength λ0 was determined by measuring the normalized cross-port output at the second MZI P ( k =1). P (1) rapidly oscillates with respect to the wavelength change. But, it reaches to stationary and the maximum value at λ max = λ0 [1+ δφ1 /(2 πm )]~ λ0 . λ0 was measured to be 1550.1 nm. Solid curve in Fig. 14 is the original signal spectrum measured by a bulk spectrometer. The Hanning
PPT Slide
Lager Image
Measured cross-port output power distribution inthe MZI array.
PPT Slide
Lager Image
Retrieved signal spectrum by cosine Fourier transform.
window was multiplied to P ( k ) so as to reduce oscillatory features in the retrieved spectrum. The FSR in terms of wavelength is given by λFSR = λ0 /m = 10.24 nm. Signals in the range of λ = 1539.9 ~ 1545.04 nm are not shown in Fig. 14 since they are mirrored replica of the spectrum in λ = 1545.2 ~ 1550.1 nm. Deviation of the measured spectrum from the original one is mainly due to the phase error δφ'ks since they cannot be directly included in the cosine FFT analysis. The horizontal axis in Fig. 13 is understood to be the optical path length difference divided by ncΔL. Then, the measured MZI output should be located at k+δncLk /ncΔL instead of the ideal position, k . The correction can be done by using the measured data and the Lagrange interpolation to replace the unevenly spaced interferogram points by the evenly spaced ones [25] . Fourier-transform algorithms can then be applied to the evenly spaced interferogram points. For a large number of data points, as in bulk-optic Fourier-transform spectroscopy,a FFT-based procedure is a prerequisite. However,for a small number of N, the discretized form of Eq. (1), including phase errors,
PPT Slide
Lager Image
PPT Slide
Lager Image
Signal spectrum corrected by taking account of themeasured phase errors.
can be solved by N×N simultaneous equations. A signal spectrum corrected with the above procedure is shown in Fig. 15 . The main part of the spectrum is accurately retrieved. Some oscillatory features in the peripheral spectral regions may be caused by the imperfection of the 3 dB couplers and differential attenuation in longer asymmetrical MZIs. Correction of these imperfections will be reported in the future.
IV. SUMMARY
Four kinds of silicon photonics WDM filters have been investigated in detail. It is made clear that (a) improvement in core width and thickness uniformity and (b) size reduction to make the device size small are strongly required in order to achieve good crosstalk characteristics(< -20 dB).
We have demonstrated a novel planar waveguide spectrometer implemented in silica PLC. The signal spectrum was successfully retrieved by correcting the MZI output data with the phase error values. Miniature spectrometers will be important for daily health care and environmental sensing applications since they are compact in size and potentially very low cost (disposable).
References
Yariv A , Xu Y , Lee R. K , Scherer A 1999 Coupledresonator optical waveguide: a proposal and analysis Opt.Lett. 24 711 - 713
Jinguji K , Oguma M 2000 Optical half-band filters IEEE J. Lightwave Technol. 18 252 - 259
Okamoto K , Chapter 9 2006 Fundamentals of Optical Waveguides 2nd ed. Elsevier New York USA Chapter 9
Shi Z , He S 2002 A three-focal-point method for the optimal design of a flat-top planar waveguide demultiplexer IEEE J. Select. Topics Quantum Electron. 8 1179 - 1185
Takada K , Inoue Y , Yamada H , Horiguchi M 1994 Measurement of phase error distributions in silica-based arrayed-waveguide grating multiplexers by using Fourier transform spectroscopy Electron. Lett. 30 1671 - 1672
Okamoto K , paper FrA Evolution of planar waveguide devices: communication and sensing applications Proc. ECIO 2010 Cambridge UK April 2010 paper FrA
He J. J , Lamontagne B , Delage A , Erickson L , Davies M , Koteles E. S 1998 Monolithic integrated wavelength demultiplexer based on a waveguide rowland circle grating in InGaAsP/In IEEE J. Lightwave Technol. 16 631 - 638
Xia F , Sekaric L , Vlasov Y 2007 Ultracompact optical buffers on a silicon chip Nature Photonics 1 65 - 71
Romagnoli M Silicon on insulator based integrated flexible optical components Proc. ECOC ’07 Workshop BerlinGermany Sep. 2007
Van Thourhout D , paper OtuB5. , Bogaerts W , paper OtuB5. Silicon photonics Proc. OFC ’10 San Diego CA USA Mar. 2010 paper OtuB5.
Bogaerts W , Dumon P , Van Thourhout D , Taillaert D , Jaenen P , Wouters J , Beckx S , Wiaux V , Baets R. G 2006 Compact wavelength-selective functions in siliconon-insulator photonic wires IEEE J. Select. Topics Quantum Electron. 12 1394 - 1401
Kim D. J , Lee J. M , Song J. H , Pyo J , Kim G 2008 Crosstalk reduction in a shallow-etched silicon nanowire AWG IEEE Photon. Technol. Lett. 20 1615 - 1617
Fang Q , Liow T. Y , Song J. F , Ang K. W , Yu M. B , Lo G. Q , Kwong D. L 2010 WDM multi-channel silicon photonic receiver with 320 Gbps data transmission capability Opt. Express. 18 5106 - 5113
Brouckaert J , Bogaerts W , Dumon P , Van Thourhout D , Baets R 2007 Planar concave grating demultiplexer fabricated on a nanophotonic silicon-on-insulator platform IEEE J.Lightwave Technol. 25 1269 - 1275
Bogaerts W , Selvaraja S. K , Dumon P , Brouckaert J , Vos K. D , Van Thourhout D , Baets R 2010 Silicon-oninsulator spectral filters fabricated with CMOS technology IEEE J. Select. Topics Quantum Electron. 16 33 - 44
Horst F , Green W. M. J , Offrein B. J , Vlasov Y. A 2009 Silicon photonic WDM devices: simulation design and implementation Photonics North 2009 Proc. SPIE 7386 L1 - L9
Harlander J. M , Roesler F. L , Cardon J. G , Englert C. R , Conway R. R 2002 SHIMMER: a spatial heterodyne spectrometer for remote sensing of Earth’s middle atmosphere Appl. Opt. 41 1343 - 1352
Florja?czyk M , Cheben P , Janz S , Scott A , Solheim B , Xu D. X 2007 Multiaperture planar waveguide spectrometer formed by arrayed Mach-Zehnder interferometers Opt. Express. 15 18176 - 18189
Noordegraaf D , Skovgaard P. M. W , Maack M. D , Hawthorn J. B , Haynes R , Lagsgaad J 2010 Multi-mode to single-mode conversion in a 61 port photonic lantern Opt.Express. 18 4673 - 4678
Florja?czyk M , Cheben P , Janz S , Lamontagne B , Lapointe J , Scott A , Solheim B , Xu D. X 2010 Development of slab waveguide spatial heterodyne spectrometer for remote sensing Proc. SPIE 7394 R1 - R9
Okamoto K , Aoyagi H , Takada K 2010 Fabrication of Fourier-transform integrated-optic spatial heterodyne spectrometer on silica-based planar waveguide Opt. Lett. 35 2103 - 2105
Goh T , Suzuki S , Sugita A 1997 Estimation of waveguide phase error in silica-based waveguides IEEE J. Lightwave Technol. 15 2107 - 2113
Kominato T , Kitoh T , Katoh K , Hibino Y , Yasu M Loss characteristics of intersecting silica-based waveguides Proc. Optoelectronics Conf. OEC’ 92 Makuhari Japan Jul. 1992 138 - 139
Takada K , Tanaka T , Abe M , Yanagisawa T , Ishii M , Okamoto K 2000 Beam-adjustment-free crosstalk reduction in 10GHz-spaced arrayed-waveguide grating via photosensitivity under UV laser irradiation through metal mask Electron. Lett. 36 60 - 61
Fan G , Liu Q.H 2004 Fast Fourier transform for discontinuous function IEEE Trans. on Antennas and Propagation 52 461 - 465