|
1.INTRODUCTIONThe innovative optical imaging system concept based on interferometry known as spider*1 could bring substantial gains in size and weight compared to a conventional focal plane imager by combining the following ideas: The concept of spider was presented in 2013,1 by researchers at Lockheed-Martin, and several preliminary experimental demonstrations followed.3, 4Several aspects of such a concept have been studied since then.5–9 A Si3N4 PIC demonstrator with 12 baselines and 18 spectral channels dispersed by an Arrayed Waveguide Grating or AWG was then developed by Lockheed-Martin and UC Davis.10 The experimental laboratory demonstration of the near-infrared SPIDER imager based on this PIC resulted in a publication in 2018,2 which is the most comprehensive publication by these two teams. The diagram in Figure 1 shows that at their entry in the PIC, the beams in each guide meet first the phase shifters, then the couplers, which combine the apertures two by two, then only the dispersive elements (AWG), and finally the detectors. The flux recorded by the detectors allows, via a modulation produced by the phase shifters, to reconstruct the contrast and the phase of the interference fringes of each pair of recombined apertures. The contrast and phase are grouped into a quantity called complex visibility, which gives a sample of the Fourier Transform of the object via the Van Cittert-Zernike theorem. Then, the object can be estimated through an image reconstruction, using a method among those developed in astronomical interferometry.11, 12 The telescope uses a large number of baselines to sample the complex visibility of the observed object on a rather large number of spatial frequencies. The implementation proposed by Lockheed-Martin is presented in Figure 2, taken from Ref. 2. 2.BASIC DESIGN RULES2.1Field of viewWith spider, the observed object is apodized by the “antenna lobe” in intensity L of the waveguide inside which the light is coupled, so that the effective object whose complex visibilities are measured by the interferometer is: L is the intensity of the back-propagation towards the object of the waveguide’s propagation mode in the aperture plane: L = |FT(P)|2. L is typically a Gaussian 2D function of FWHM λ/D, where λ is the wavelength and D is the diameter of the apertures, as studied in details in Refs. 13,14. Finally, the field-of-view (FOV) of spider is: 2.2Visibility measurement modelLet us consider a pair of apertures separated by B measuring the spatial frequency (u, v) = B/λ. Let us call C1 (respectively C2) the coupling coefficient of aperture 1 (respectively 2) towards the output, ϕk the modulation phase of the k-th measurement, and Nph the mean number of photons received in each of the two apertures, during a given exposure time. Assuming balanced channels i.e., , a model of the recorded raw data (intensities) is: where the complex visibility of the observed object is: The mean number of photons received and detected per aperture and exposure time is: By using typically four measurements (ϕk = (k – 1)π/2, for k ∈ {1, 2, 3, 4}) followed by a demodulation†, one obtains a complex measurement y(u, v) proportional to the mean number of photons received in each of the two apertures and to the complex visibility of the object at the measured spatial frequency: Thanks to the Van Cittert-Zernike theorem: so γ(0, 0) = 1. Equation 6 then yields y(0, 0) = Nph, i.e., the collection y of complex measurements y(u, v) is a set of samples of the Fourier transform of oeff(αx, αy) of integral (i.e., of total flux) Nph. 2.3Resolution and number of resolution elementsAs for any interferometer, the maximum spatial frequency recorded by a compact interferometric imager is given by the maximum distance separating two apertures, or maximum baseline Bmax, and is Bmax/λ. The angular resolution of such an instrument is thus: The angular field being λ/D, the number of resolved elements of a compact interferometric imager in the observable angular field is Bmax/D. 2.4Spectral widthIn case of point-like apertures separated by B interfering at a wavelength λ with zero bandwidth, the visibility is measured at the spatial frequency B/λ. As soon as the bandwidth is non-zero, the detected interferogram is an average of inteferograms at the spatial frequencies given by all the wavelengths in the band. Therefore, what spectral bandwidth can be tolerated? To answer this question, it is useful to realize that the recombined beams contain interferences between apertures of diameter D spaced by B and thus between pairs of points separated by distances ranging from (B−D) to (B+D). Such an interferometer with waveguide injection intrinsically measures an average of spatial frequencies of the object over a typical width of D/λ. It is then reasonable to specify that the spatial frequency averaging due to the spectral width dλ is less than the intrinsic averaging due to the non-zero aperture size. Differentiating f = B/λ we obtain the following condition:‡ For the shortest wavelength and the longest baseline, this yields: The inequality 9 can also be read as follows: the spectral resolution λ/dλ must be greater than the number of (spatially) resolved points of the interferometer, Bmax/D. 3.NOISE MODELING AND COMPARISON WITH A CONVENTIONAL IMAGERA study of the compact interferometric imager concept must include the modeling of noise, and if possible comparatively to a classical i.e., focal plane imager. The propagation of noise in the algorithm that reconstructs the object from complex visibilities is delicate because these algorithms can be notably non-linear. In order to avoid dealing with such noise propagation, we have chosen to model the noise of a compact interferometric imager and of a classical instrument in the Fourier i.e., spatial frequency domain. An in-depth analysis may be found in.15 Here we assume that (only) two raw measurements are obtained simultaneously using the two outputs in phase opposition of each coupler. Two other measurements must then be obtained by introducing a phase shift of π/2. Each complex measurement thus uses 4Nph photons (2 raw interferometric measurements, each receiving Nph photons per aperture). We assume in the following that the measurement noise is predominantly photon noise, which is reasonable for many scenarios at least in the visible and near-infrared. In the noise analysis, we additionally assume that the compact interferometric imager measures each spatial frequency only once, as in Ref. 16. 3.1Noise modeling for a compact interferometric imagerUnder the assumption that |γ(u, v)| ≪ 1, the variance of the noise on the complex measurement y(u, v) of Equation 6, resulting from the demodulation of four raw measurements given by Equation 3, is§: where Nph is still the average number of photons received in each of the two apertures contributing to the interference during a raw measurement. This directly follows from the fact that y(u, v) is the demodulation of 4 raw data measurements, where each is corrupted by photon noise, of variance Nph. The variance of these complex measurements, normalized by the square modulus of the signal at frequency (u, v), is therefore simply, according to Equation 5: or: 3.2Noise modeling for a conventional imagerThe total number of elementary apertures of a compact interferometer spider is denoted Nd. Considering each elementary aperture used a single time and connected by pairs, Nd is necessarily even and the number of measured spatial frequencies is Nd/2. We consider an imaging system with the same collecting surface as this spider⁋. The object observed by the imaging system is assumed to be the object seen by the SPIDER device, oeff, to simplify the comparison. The measurement of each frequency requires the detection of 4 signals of average amplitude Nph. The measurement of all these spatial frequencies therefore takes a total of Nphtot = Nd/2×4Nph = 2NdNph photons. Hence, the image model of a conventional imager collecting a total of Nphtot photons writes: where h is the discrete PSF, and * is the discrete convolution operator. We have already assumed that the visibility is negligible with regards to 1. For a classical imaging instrument, we additionally assume that the scene luminance is sufficiently homogeneous, so that the noise can be assumed to be stationary, i.e., so that the variance of the noise of the classical imager can be assumed to be approximately constant over all Npix pixels. Hence, in the Fourier space, the noise spectrum is uniform and equals Nphtot. Since , as pointed out in paragraph 2.2, in the Fourier domain, the spectrum of the noise normalized by the square modulus of the signal at frequency (u, v) is: or: Comparison of Equations 11 and 13 suggests that signal-to-noise ratios are similar for both types of instruments, i.e., proportional to , and that they are identical if: at all frequencies except the zero frequency, at which the MTF is 1 by convention. In other words, with respect to noise propagation, a “standard” compact interferometric imager (standard in the sense that each spatial frequency would be measured once and only once) is equivalent to a conventional imager with a flat MTF of the order of . Note that focal-plane imagers with quite flat MTFs exist, namely phased array instruments or sparse aperture telescopes with minimal redundancy such as the ones discussed in Ref. 17. For these, the MTF is flat and of the order 1/Ntel, where Ntel is the number of apertures. Finally, the above results show that, for the highest spatial frequencies, a compact interferometric imager and a conventional imaging instrument have a very similar behavior with respect to noise. For the former, the propagation of noise is unfavorable for the lower frequencies. However it is quite possible, for a particular task (detection for example), to design the transfer function of a compact interferometric imager to make it higher at the spatial frequencies relevant to the task. 4.FREQUENCY COVERAGEThe set of measured spatial frequencies, i.e., the frequency coverage is a key aspect in the design of the interferometer. Lockheed Martin’s aperture configuration18 of one arm is presented in the top panel of 3. The frequency coverage associated to this aperture configuration is presented at the bottom, and each measured frequency is represented by a blue dot. Since the spatial frequencies are sampled with a step of b/λ, with b the distance between two consecutive apertures, normalized spatial frequencies in units of b/λ are considered for simplicity purposes. Not all apertures are used (the unused ones are shown in red) and missing frequencies start from spatial frequency 7. The bidimensionnal optical transfer function deduced from this aperture configuration is illustrated in Fig. 6a. In case of the Earth observation viewed from a satellite, few priors are available making it necessary to have no gaps in the frequency coverage. A no gap frequency coverage is called compact.19, 20 Harvey and Rockwell21 introduced the “practical resolution limit” (PRL), defining it as the maximum spatial frequency before which no zero occurs in the optical transfer function. In the following, the list of combined aperture pairs in the instrument is called aperture configuration. The monodimensional aperture configurations can be scanned to find the one with the highest PRL. Unfortunately, the number of aperture configurations scales exponentially with the number of apertures. An approach based on combinatorial theory results obtained by several authors22–26 has been elaborated by Debary.27 This paragraph summarizes her main results. 4.1Optimization of the frequency coverage of dense 1D arraysThe total number of apertures in a single arm (1D) is denoted Np. Still considering each aperture used a single time and connected by pairs, Np is even. The number of measured spatial frequencies is consequently Np/2 and the measured frequencies have values in the range [1, Np − 1]. A frequency coverage is defined as buildable if there is an aperture configuration that allows it to be measured (not all frequency coverages are buildable). Finding compact buildable frequency coverages composed of consecutive spatial frequencies {1, 2, …, Np/2} is a problem formulated and solved in the framework of combinatorial theory by Skolem.23 Such a frequency coverage is called a Skolem set and its associated aperture configuration a Skolem sequence. More generally, it is possible to build frequency coverages in the form {nmin, nmin + 1, …, nmin + Np/2 − 1}, where nmin is the first measured spatial frequency. This problem has been introduced by Langford.28 Such a frequency coverage is called a Langford set and its associated aperture configuration a Langford sequence. The existence of buildable compact frequency coverages was proved for nmin = 1 by Skolem in 1958, for nmin = 2 by Davies24 in 1959, then partially generalized by Bermond25 in 1978 and finally completed by Simpson26 in 1983. 4.2Application to a SPIDER-like interferometerApplying the Skolem results, an illustrative aperture configuration is provided in Fig. 4. 4.3Suggestion of a hybrid architecture using Langford configurationsIt is possible to further optimize the PRL obtained for the SPIDER-like design by considering an interferometer based on a Langford sequence, complemented with a small monolithic telescope measuring short spatial frequencies. If buildable, the Skolem configuration reaches a maximum spatial frequency of Np/2 by definition whereas the Langford configuration reaches a maximum spatial frequency of (3Np − 2)/4. Because the maximum spatial frequency measured increases from Np/2 to (3Np − 2)/4, the resolution of this hybrid interferometer is therefore improved by about 50% with respect to the interferometer with the Skolem configuration. Maintaining the SPIDER-like interferometer aperture complexity, i.e. Np = 30, an illustration of this Langford aperture configuration is presented in Fig. 5. The represented aperture configuration produces a compact frequency coverage ranging from spatial frequencies 8 to 22, improving the frequency coverage provided by a Skolem configuration, i.e. by considerably stretching the highest spatial frequency. Finally, in Fig. 6b, we have represented the two-dimensional optical transfer function following the merging of the frequency coverages from the monolithic telescope and the SPIDER-like interferometer. REFERENCESKendrick, R. L., Duncan, A., Ogden, C., Wilm, J., and Thurman, S. T.,
“Segmented Planar Imaging Detector for EO Reconnaissance,”
Imaging and Applied Optics, Arlington, Virginia(2013). https://doi.org/10.1364/COSI.2013.CM4C.1 Google Scholar
Su, T., Liu, G., Badham, K. E., Thurman, S. T., Kendrick, R. L., Duncan, A., Wuchenich, D., Ogden, C., Chriqui, G., Feng, S., Chun, J., Lai, W., and Yoo, S. J. B.,
“Interferometric imaging using Si3N4 photonic integrated circuits for a SPIDER imager,”
Optics Express, 26 12801
–12812
(2018). https://doi.org/10.1364/OE.26.012801 Google Scholar
Scott, R. P., Su, T., Ogden, C., Thurman, S. T., Kendrick, R. L., Duncan, A., Yu, R., and Yoo, S. J. B.,
“Demonstration of a photonic integrated circuit for multi-baseline interferometric imaging,”
in 2014 IEEE Photonics Conference],
1
–2
(2014). Google Scholar
Su, T., Scott, R. P., Ogden, C., Thurman, S. T., Kendrick, R. L., Duncan, A., Yu, R., and Yoo, S. J. B.,
“Experimental demonstration of interferometric imaging using photonic integrated circuits,”
Optics Express, 25 12653
(2017). https://doi.org/10.1364/OE.25.012653 Google Scholar
Liu, G., Wen, D.-S., and Song, Z.-X.,
“System design of an optical interferometer based on compressive sensing,”
Monthly Notices of the Royal Astronomical Society, 478 2065
–2073
(2018). https://doi.org/10.1093/mnras/sty1167 Google Scholar
Gao, w., Wang, X., Ma, L., and Guo, D.,
“Quantitative analysis of imaging quality of the segmented planar imaging detector,”
in Tenth International Conference on Information Optics and Photonics,
(2018). https://doi.org/10.1117/12.2505876 Google Scholar
Guo-mian, L., Qi, L., Yue-ting, C., Hua-jun, F., Zhi-hai, X., and Jingjing, M.,
“An improved scheme and numerical simulation of segmented planar imaging detector for electro-optical reconnaissance,”
Optical Review, 26 664
–675
(2019). https://doi.org/10.1007/s10043-019-00548-w Google Scholar
Cao, K., Ye, Z., Jiang, C., Zhu, J., Qiao, Z., and Jiang, Y.,
“Lenslets combination optimal design of the segmented planar image detector for electro-optical reconnaissance,”
Optical Engineering, 59 1
(2020). Google Scholar
Liu, G., Wen, D., Song, Z., and Jiang, T.,
“System design of an optical interferometer based on compressive sensing: an update,”
Optics Express, 28 19349
(2020). https://doi.org/10.1364/OE.394130 Google Scholar
Duncan, A., Kendrick, R., Ogden, C., Wuchenich, D., Thurman, S., Su, T., Lai, W., Chun, J., Li, S., Liu, G., et al.,
“Spider: next generation chip scale imaging sensor update,”
in Advanced Maui Optical and Space Surveillance Technologies Conference,
(2016). Google Scholar
Le Besnerais, G., Lacour, S., Mugnier, L. M., Thiébaut, E., Perrin, G., and Meimon, S.,
“Advanced imaging methods for long-baseline optical interferometry,”
IEEE Journal of Selected Topics in Signal Processing, 2 767
–780
(2008). https://doi.org/10.1109/JSTSP.2008.2005353 Google Scholar
Thiébaut, E. and Young, J.,
“Principles of image reconstruction in optical interferometry: tutorial,”
J. Opt. Soc. Am. A, 34 904
–923
(2017). https://doi.org/10.1364/JOSAA.34.000904 Google Scholar
Dyer, S. D. and Christensen, D. A.,
“Pupil-size effects in fiber optic stellar interferometry,”
JOSA A 16, 2275
–2280 Publisher: Optical Society of America.1999). https://doi.org/10.1364/JOSAA.16.002275 Google Scholar
Mege, P., Malbet, F., and Chelli, A., International Society for Optics and Photonics,
“Interferometry with singlemode waveguide,”
Interferometry for Optical Astronomy II, 4838 329
–337
(2003). https://doi.org/10.1117/12.458917 Google Scholar
Mugnier, L. M., Michau, V., Debary, H., and Cassaing, F.,
“Analysis of a compact interferometric imager,”
in Proc. Soc. Photo-Opt. Instrum. Eng. (2022),
(2022). Google Scholar
Debary, H., Mugnier, L. M., Michau, V., and Lopez, S.,
“Optical interferometry imaging from space: optimization of the aperture configuration of a dense array,”
in Proc. Soc. Photo-Opt. Instrum. Eng. (2022),
(2022). Google Scholar
Cassaing, F. and Mugnier, L. M.,
“Optimal sparse apertures for phased-array imaging,”
Opt. Lett., 43 4655
–4658
(2018). https://doi.org/10.1364/OL.43.004655 Google Scholar
Badham, K., Kendrick, R. L., Wuchenich, D., Ogden, C., Chriqui, G., Duncan, A., Thurman, S. T., Yoo, S. J. B., Su, T., Lai, W., Chun, J., Li, S., and Liu, G.,
“Photonic integrated circuit-based imaging system for spider,”
in 2017 Conference on Lasers and Electro-Optics Pacific Rim (CLEO-PR),
1
–5
(2017). Google Scholar
Damé, L. and Martic, M.,
“Study of an optimized configuration for interferometric imaging of complex and extended solar structures,”
Targets for Space-Based Interferometry, 354 201
–208
(1992). Google Scholar
Mugnier, L. M., Rousset, G., and Cassaing, F.,
“Aperture configuration optimality criterion for phased arrays of optical telescopes,”
J. Opt. Soc. Am. A, 13 2367
–2374
(1996). https://doi.org/10.1364/JOSAA.13.002367 Google Scholar
Harvey, J. E. and Rockwell, R. A.,
“Performance characteristics of phased array and thinned aperture optical telescopes,”
Optical Engineering, 27
(9), 279762
(1988). https://doi.org/10.1117/12.7976758 Google Scholar
Nordh, G.,
“Perfect skolem sets,”
Discrete mathematics, 308
(9), 1653
–1664
(2008). https://doi.org/10.1016/j.disc.2006.12.003 Google Scholar
Skolem, T.,
“Some remarks on the triple systems of steiner,”
Mathematica Scandinavica, 6
(2), 273
–280
(1958). https://doi.org/10.7146/math.scand.a-10551 Google Scholar
Davies, R. O.,
“On langford’s problem (ii),”
The Mathematical Gazette, 43
(346), 253
–255
(1959). https://doi.org/10.1017/S0025557200041693 Google Scholar
Bermond, J.-C., Brouwer, A. E., and Germa, A.,
“Systemes de triplets et differences associees,”
Probèmes combinatoires et théorie des graphes, Colloque international CNRS 260, 260 35
–38
(1978). Google Scholar
Simpson, J. E.,
“Langford sequences: perfect and hooked,”
Discrete Mathematics, 44
(1), 97
–104
(1983). https://doi.org/10.1016/0012-365X(83)90008-0 Google Scholar
Debary, H., Mugnier, L. M., and Michau, V.,
“Aperture configuration optimization for extended scene observation by an interferometric telescope,”
Optics Letters, 47
(16), 4056
–4059
(2022). https://doi.org/10.1364/OL.462561 Google Scholar
Langford, C. D.,
“Langford problem,”
Math. Gazette, 42 228 https://doi.org/10.2307/3610395 Google Scholar
Notes[2] For each spatial frequency (u, v), the demodulation consists in estimating y(u, v) as: ℜ(y(u, v)) = (i(u, v, ϕ1) – i(u, v, ϕ3))/2 and . [3] This condition can also be obtained by specifying that the optical path difference between two apertures separated by B for a direction θ = λ/(2D) at the edge of the field of view, which is B.θ = Bλ/(2D), must be less than half the coherence length Lc = λ2/(dλ). [4] The variance of the complex variable y can be defined as , and it is easy to show that Var(y) = E(|y|2) – |E(y)|2 = Var(ℜ(y)) + Var(ℑ(y)). If moreover, as we can reasonably assume in this study, ℜ(y) and ℑ(y) are two independent Gaussian variables of the same variance, then y is a complex circular Gaussian variable (distribution invariant by any rotation). [5] The set of spatial frequencies to be measured to cover the same frequency support as a conventional imaging instrument of the same resolution, i.e., of diameter Bmax, is contained in a half-disk of radius Bmax/λ, that is to say of the order of π(Bmax/D)2/2 distinct frequencies, if the frequency sampling step is D/λ (which requires joint apertures). |