SYMPA, a dedicated instrument for Jovian Seismology. II. Real performance and first results
Patrick Gaulme, F.X. Schmider, Jean Gay, Cedric Jacob, Manuel Alvarez, Mauricio Reyes, Juan Antonio Belmonte, Eric Fossat, Francois Jeanneaux, Jean-Claude Valtier
aa r X i v : . [ a s t r o - ph ] F e b Astronomy & Astrophysics manuscript no. gaulme˙et˙al˙astroph c (cid:13)
ESO 2018November 11, 2018
SYMPA, a dedicated instrument for Jovian Seismology.II. Real performance and first results
Patrick Gaulme , F.X. Schmider , Jean Gay , C´edric Jacob , Manuel Alvarez , Mauricio Reyes , JuanAntonio Belmonte , Eric Fossat , Fran¸cois Jeanneaux , and Jean-Claude Valtier Laboratoire FIZEAU, Universit´e de Nice Sophia-Antipolis, CNRS-Observatoire de la Cˆote d’Azur, F-06108 NiceCedex 2e-mail:
[email protected] Observatorio Astron´omico Nacional, Instituto de Astronom´ıa, Universidad Nacional Aut´onoma de M´exico, Apto.Postal 877, Ensenada, B.C., M´exico Instituto de Astrof´ısica de Canarias, Tenerife, Spain THEMIS Observatory, La Laguna, Tenerife, SpainPreprint online version: November 11, 2018
ABSTRACT
Context.
Due to its great mass and its rapid formation, Jupiter has played a crucial role in shaping the Solar System.The knowledge of its internal structure would strongly constrain the solar system formation mechanism. Seismology isthe most efficient way to probe directly the internal structure of giant planets.
Aims.
SYMPA is the first instrument dedicated to the observations of free oscillations of Jupiter. Principles and theo-retical performance have been presented in paper I. This second paper describes the data processing method, the realinstrumental performance and presents the first results of a Jovian observation run, lead in 2005 at Teide Observatory.
Methods.
SYMPA is a Fourier transform spectrometer which works at fixed optical path difference. It produces Dopplershift maps of the observed object. Velocity amplitude of Jupiter’s oscillations is expected below 60 cm s − . Results.
Despite light technical defects, the instrument demonstrated to work correctly, being limited only by photonnoise, after a careful analysis. A noise level of about 12 cm s − has been reached on a 10-night observation run, with21% duty cycle, which is 5 time better than previous similar observations. However, no signal from Jupiter is clearlyhighlighted. Key words. planets and satellites: formation, Jupiter: oscillations, methods: observational, instrumentation: interferom-eter,techniques: spectroscopic
1. Introduction
Due to its great mass and its rapid formation, Jupiter hasplayed a crucial role in shaping the Solar System. Two sce-narios are generally proposed for the formation of giantplanets: the nucleated instability (Safronov & Ruskol 1982)and the gravitational instability models (Cameron 1978 andMayer et al. 2002). An efficient constraint on the formationscenario would be given by measuring the total amount ofheavy elements inside Jupiter and the size of the planetarycore. Moreover, the knowledge of Jupiter’s internal struc-ture would constrain the high pressure hydrogen equation ofstate, which is still inaccurate, and would particularly solvethe question of the nature of the metallic-molecular phasetransition (e.g. Guillot et al. 2004). Gudkova & Zharkov(1999) showed that the observation of oscillation modes upto degree ℓ = 25 would strongly constrain Jupiter’s inter-nal structure by exploring both the hydrogen plasma-phasetransition and the supposed core level.Attempts to observe Jovian oscillations have beenbrought since the mid 1980’s, thanks to different tech-niques. On one hand, Deming et al. (1989) have lookedfor oscillation signature in thermal infrared. Unfortunately, Send offprint requests to : P. Gaulme their infrared detectors, not enough sensitive, did not detectany signal. On the other hand, oscillations were sought invelocity measurements, obtained by Doppler spectrometry.Schmider et al. (1991, hereafter S91) used sodium cell spec-trometer and Mosser et al. (1993, 2000, hereafter M93 andM00) the Fourier transform spectrometer FTS (at CFHT,Hawaii) at fixed optical path difference. An excess of powerhas been brought out in the spectrum at frequency range[0.8-2] mHz, as well as the large separation of oscillation p -modes around 140 µ Hz. Nevertheless, the oscillation modeshave never been individually identified, hindering any con-straint about the internal structure.SYMPA is an instrument dedicated to Jovian oscilla-tions, which concept and performance have been describedin paper I (Schmider et al. 2007). For the first time, a spe-cific instrument dedicated to Jovian oscillations was devel-oped, including imaging capability. Indeed, full disc obser-vations do not permit to distinguish modes of degree higherthan 3. Moreover, the broadening of the solar lines, dueto the fast rotation of Jupiter, reduces drastically the sen-sitivity of such measurements. The instrument, a Fouriertachometer, is composed of a Mach-Zehnder interferome-ter which produces four images of the planet, in the visi-ble range corresponding to three Mg solar absorption linesat 517 nm. The combination of the four images, in phase
Gaulme et al.: SYMPA: II. Real performance and first results I n t en s i t y ( pho t on / i m age ) Time (day)
Fig. 1.
Total intensity of each image. Nights 2 and 8 andthe end of night 4 present a strongly and rapidly variableamount of photons, because of cloudy conditions.quadrature, allows us to retrieve the phase of the incidentlight, which is related to the Doppler shift generated by theoscillations.Two instruments were built at Laboratoire Fizeau (NiceUniversity). Three campaigns were lead simultaneously ontwo sites: in 2003, at San Pedro Martir (Mexico) andCalern (France) observatories; in 2004 and 2005, at SanPedro Martir and Teide (Canaries) observatories. 2003 cam-paign was mainly dedicated to technical commissioning. InCanaries, bad weather conditions have strongly limited theefficiency of the 2003 and 2004 campaigns, whereas the 2005campaign benefited of better conditions.In this paper we present the data processing, the realperformance and the first results, obtained during the 2005run at Teide observatory. The processing of the San PedroMartir data and the combined analysis of both observing2005 network campaign will be considered in a future work.After a short presentation of the observing conditions (Sect.2), Sect. 3 presents the process of the data analysis. We ex-pose in Sect. 4 all the steps for an accurate calibration ofthe data. Sect. 5 is devoted to the further of the data reduc-tion, for the measurement of velocity maps. The analysis ofthe time series get during the 2005 run at Tenerife is ex-posed in Sect. 6. Section 8 is devoted to conclusions andprospectives.
2. Observations
Observations were conducted at Teide observatory(Canaries islands), with the 1.52-m Carlos Sanchez tele-scope, between march 31st and April 10th 2005. As it hasbeen detailed in paper I, four images of Jupiter come outfrom the instrument. Optical parameters inside SYMPA’sbox are arranged such as the four 1.3-arcmin fields, cover128 pixels on the receptor (a DTA CCD, 1024 ×
256 pixels).During the run, Jupiter at opposition presented a diame-ter of 48 arcsec, corresponding to 69 pixels on the CCDcamera.Seven nights over ten days benefited of good weatherconditions, yielding to a 21% duty cycle. Data quality wasalmost constant from one night to another, excepted fornight 2 and 8, where clouds have reduced the incidentflux. Observation conditions are summarized in Fig. 1 andTable 1. The window function associated to the whole cam-paign was very strong, since we consider only Canaries data.Therefore, in the power spectrum, the amplitude of a singlespike has been divided by a factor 4, since its power gets −50 0 5000.050.10.150.2 W i ndo w po w e r Frequency ( µ Hz)
Fig. 2.
Spectral window of the observation campaign. Theamplitude has been normalized to the spike’s power. Only23% of the total energy remains in the central peak.diluted in high side lobes (Fig. 2). The total flux was ex-pected to be about 2 . photons per 6-s exposure (paperI). However, its mean value along the run is about 2 . photons per exposure (Fig. 1). This discrepancy with theestimated flux introduces a factor 0 .
95 in the ratio to noisevalue. In paper I, a noise level of about 4 cm s − was ex-pected for a 16-nights observation campaign, with 50% dutycycle. Instead of such a performance, by considering onlyCanaries data, the noise level is therefore expected at 10cm s − .
3. Data processing strategy
SYMPA’s instrumental principles are fully explained in pa-per I and are summarized in Fig. (3). The four outputbeams can be described in the detector coordinates ( x, y )by the following approximations : I ( x, y ) = I ( x, y )4 [1 − γ cos φ ( x, y )] (1) I ( x, y ) = I ( x, y )4 [1 − γ sin φ ( x, y )] (2) I ( x, y ) = I ( x, y )4 [1 + γ cos φ ( x, y )] (3) I ( x, y ) = I ( x, y )4 [1 + γ sin φ ( x, y )] (4)where I is the continuum component of the incident light,that is to say the Jovian figure, γ the fringe contrast and φ ( x, y ) the incident wave phase map : φ ( x, y ) = 2 πσ ∆( x, y ) (cid:16) v D c (cid:17) (5)where σ is the central wavenumber of the input filter,∆( x, y ) is the optical path difference (OPD) and v D and c are the Doppler and the light velocities. The Doppler shiftof the solar Mg lines comes from the combination of the rel-ative motion of Jupiter to the Sun v J / S , relative motion ofthe observer to Jupiter v E / J + v E , rot (distance between thetwo planets and Earth’s rotation), Jupiter’s rotation v J , rot and, finally, the oscillations v osc . In the following, we writethe Doppler velocity as the sum of : v D = 2 (cid:0) v J / S + v E / J + v E , rot + v J , rot + v osc (cid:1) (6)The factor 2 is due to the fact that the Doppler effect getsdoubled after reflection on Jupiter’s atmosphere. The orders aulme et al.: SYMPA: II. Real performance and first results 3 Table 1.
Starting date Ending date Duration Mean sampling Number of Mean Intensity Standard deviationm d h m s (UT) m d h m s (UT) h m s s acquisitions photons image − photons image − Apr-02, 23:48:14 Apr-03, 07:03:40 7:09:35 5.45 4727 1.89 10 Apr-03, 23:30:14 Apr-04, 07:02:29 7:31:59 7.13 3792 1.79 10 Apr-04, 23:16:12 Apr-05, 06:58:17 7:39:30 6.75 4080 2.50 10 Apr-05, 23:25:29 Apr-06, 06:50:23 7:22:12 6.68 3971 2.44 10 Apr-09, 23:20:17 Apr-10, 06:31:59 6:37:48 6.28 3799 2.06 10 Apr-10, 22:51:11 Apr-11, 06:33:11 7:41:48 6.33 4376 2.38 10 Apr-11, 21:09:35 Apr-12, 06:30:17 7:20:42 6.83 3872 2.36 10 Fig. 3.
Schematic view of the SYMPA instrument. Theincident light coming from the 1.5-m telescope, passesthrough a 120-mm collimator and the 5 nm bandwidth in-terference filter. The optical path difference ∆ occurs insidethe Mach-Zehnder prism; it is a function of the heights H and h , refraction index N and n and incidence angles intothe prism I and i : ∆ = 2( HN cos I − hn cos i . The wollas-ton polarizer separates each output from the interferometricdevice into two separated beams. In total, the instrumentproduces on the camera four images of the same field, sep-arated by π/ πσ ∆( x, y ), anda velocity term, 4 πσ ∆( x, y ) v D /c . The art of extracting theoscillation signal resides in the ability of eliminating stepby step the motionless fringes and the “spurious” veloc-ity fields v J / S , v E / J , v E , rot and v J , rot . All the steps of theschematic process presented below are detailed in Sect. 4and 5. Table 2.
Orders of magnitude of different Doppler shiftsduring the 2005 run at Teide Observatory.
Velocity (m s − )Jupiter-Sun 0.4Jupiter-Earth 3142Earth rotation [ − , − , < . Let us consider a quadruplet of interfering images on thedetector’s field. The differences between the two couples ofimages, which are in phase opposition, allow to cancel thecontinuous component of the interferograms I , in orderto keep the interfering patterns. We write U and V thenormalized interfering patterns : U = I − I I + I ∝ γ cos φ (7) V = I − I I + I ∝ γ sin φ (8)where we let go of ( x, y ) dependence in order to simplifynotations. The incident wave phase is retrieved by takingthe argument of the complex interferogram: Z = U + ıV ∝ γ e ıφ (9)The data processing expands in three main steps: correctionof the motionless fringes, elimination of Jupiter rotationand elimination of the relative motion of the observer to thetarget and of the target to the Sun. Mathematically, it liesin creating successively four complex interferograms, asso-ciated to each step of the data processing: Z for motionlessfringes, Z J , rot and Z E , rot for Jupiter’s and Earth’s rotation, Z E / J for the relative motion of the Earth to Jupiter and Z J / S for Jupiter’s motion with respect to the Sun. Then,the rough Jupiter complex interferogram Z jup comes decon-voluted from additive signals: Z J , flat = Z jup × Z ∗ × Z ∗ J , rot Z ∗ E , rot × Z ∗ E / J × Z ∗ J / S (10) ∝ exp (cid:16) πσ ∆ v osc c (cid:17) (11)where the asterisk indicates the complex conjugation. Theresulting interferogram is so-called Z J , flat because of theappearance of its argument, i.e. the velocity map. Indeed,since oscillation amplitude is expected to be lower than 0.6m s − , it is absolutely impossible to see directly oscilla-tion modes in a single phase map, where mean noise levelis expected to be about 900 m s − per pixel (Paper I).Oscillations might be picked out only in the spectrum oflong time series.Note that Doppler shifts due both to Earth relativemotion to Jupiter and Jupiter relative motion to the Sunare uniform across the field. On the contrary, the JovianDoppler component is very sensitive to Jovian rotationsince its value varie from − .
57 km s − to 12 .
57 km s − from east to west on Jovian equator. In terms of velocity, itimplies that the error in positioning the phaser Z J , rot mustbe smaller than the photon noise. Therefore, it supposes Gaulme et al.: SYMPA: II. Real performance and first results
20 40 60 80 100 12020406080100120 20 40 60 80 100 1202040608010012020 40 60 80 100 12020406080100120 20 40 60 80 100 1202040608010012020 40 60 80 100 12020406080100120 20 40 60 80 100 12020406080100120
Fig. 4.
Simulation of interferograms along the data process-ing chain. Top: motionless interferograms ( U , V ). Middle:Jovian interferogram ( U jup , V jup ) when Jupiter is inclined of − ◦ with respect to vertical. The interference pattern ismainly due to the coupling between motionless fringes andJovian rotation (Eqt. 6). Bottom: Jovian interferograms de-convoluted from motionless contribution, ( U rot , V rot ), i.e.fringe pattern associated to Jovian rotation. Note thatfringes associated to solid rotation present velocity iso-values parallel to the rotation axis. Differential wind profilehas not been introduced.to know Jupiter’s position better than 1/20 th of pixel.Figure 4 presents simulations of the motionless interfero-grams, Jovian interferograms and Jovian rotation interfer-ing pattern.
4. Data calibration
Pre-processing consists in cleaning each quadruplet ofJovian image ( I , I , I , I ) in order to create couples ofinterferograms ( U, V ). This implies three main operations.First, the camera dark current contribution has to besubtracted, by using offset images. Second, the inhomo-geneities of the single pixel responses to light intensity (pho-ton/electron gain) have to be compensated by dividing eachimage by a flat field image. Third, the construction of theJupiter phase map, using the argument of the complex im-age created by the difference of two couples of images, re-quires that all the four images overlap one to each other. x (pixel) y ( p i x e l )
20 40 60 80 100 12020406080100120 x (pixel) y ( p i x e l )
20 40 60 80 100 12020406080100120
Fig. 5.
Overlapping of grids 1 and 3, before then after dis-tortion rectification. The position of the horizontal and ver-tical lines are determined by fitting the zeros of the gridimage derivatives, along both directions x and y . Secondorder polynomials are sufficient, since rotation and barreldistortion do not require higher orders. Then, each intersec-tion point coordinates are calculated by solving numericallythe four order equation coming from the combination of thevertical and the horizontal line fits. The mean distance be-tween intersections points is equal to 5.32 % pixel − If the two first points are easy to realize, the last one ispretty delicate because of the required accuracy of about1/20 pixel.As for every optical system including lenses and prisms,field distortion is unavoidable. Although Jupiter was po-sitioned as close as possible to the optical axis, its largediameter involves differential distortions between the fourimages, making the overlapping impossible. The whole dis-tortion effect is supposed to be composed of only transla-tion, rotation and barrel distortion. This problem has beenanticipated by putting a regular grid, engraved on a glassslice, at the instrument focus. The grid intersection posi-tions are used to characterize the distortion (Fig. 5).Let us consider one image among the quadruplet I i ( x, y ), i ∈ [1 , x, y ) are the detector coordinates(CCD pixels). Because of optical distortion, the image valueon the ( x k , y k ) point, k ∈ [1 , x ′ k , y ′ k ) point. The repositioning algorithm must pro-duce for each image I i an image I ′ i , defined in the detector’scoordinates, as: I ′ i ( x, y ) = I i ( x ′ , y ′ ) (12)The distorted coordinates ( x ′ , y ′ ) are related to the regulardetector coordinates by the relations: x ′ = x + f ( x, y ) and y ′ = y + g ( x, y ) (13)where f and g are polynomials expressed as P k C k x j y k − j ,where k is the polynomial order and j ≤ k . The polyno-mial coefficients C k are obtained by minimizing the differ-ence between the intersection point coordinates of the gridassociated to the considered image I i and the intersectionpoint coordinates of the regular grid. The knowledge of theset of distorted coordinates ( x ′ , y ′ ) is used to build the newrectified image I ′ i ( x, y ), by interpolation of I i ( x ′ , y ′ ) uponthe regular detector coordinates.Interpolation is realized with cubic method. In Fig. 5 wepresent a couple of 2-grid images before and after reposi-tioning. Note that the translation repositioning representsthe zero order of the transformation. A way to evaluatethe accuracy of the repositioning process is to apply it tothe previously repositioned grids; they should overlap (Fig. aulme et al.: SYMPA: II. Real performance and first results 5
5, right). The mean distance between the repositioned in-tersection points from one grid to another belongs to therange [1 / , /
15] pixel, which almost fulfills the requiredaccuracy.In order to respect the conservation of the flux, correct-ing operations have to be processed in the following order:first, subtraction of the offset to the image and to to theflat-field, then rectification of the offset-corrected image andflat-field. I pre − processed = [ I ( x, y, T ) − O ( x, y, T )] rectified [ F ( x, y, T ) − O ( x, y, T )] rectified (14)where I indicates the considered image (e.g. Jupiter), O theoffset image and F the flat-field image. T stands for thetemperature of the camera. Note that assessing the fouroutput intensities are not strictly equal, no photometricbalance has to be performed since it is implicitly done bydividing each image by the flat field. The next step of data processing is the deconvolution ofJovian complex interferograms Z jup from motionless inter-ferogram Z . Therefore, we have to characterize the mo-tionless phase term 2 πσ ∆( x, y ) (Eqt. 5). Furthermore, asit has been reported in paper I, the four output beams arenot in perfect phase quadrature: the discrepancies betweenactual measurements and theoretical expectation are about ε = 28 ◦ . This shift compared to quadrature has to be quan-tified precisely across the whole field ( x, y ≃ U, V ) and( U ′ , V ′ ), taken at two different dates. The phase difference δφ between them is due to terrestrial motion and temper-ature variation. Their expressions are given by : U = U + γ U cos( φ ) U ′ = U + γ U cos( φ + δφ ) (15)
50 100 150 200 250204060801001200 50 100−10−505 x 10 −3 x (pixel) A m p li t ude −3 x (pixel) Fig. 6.
Top: interference pattern (
U, V ) obtained with solarlight scattered by the telescope dome, over one minute (sumof 3 images). Bottom: cut along x -axis of U and V , at y =55 pixel. Note that both interferograms present a varyingcontrast between 0.3 and 0.5 % and a significant unflatness.Moreover, they are not centered around 0, that makes thephase φ = arg( U + iV ) impossible to recover.
20 40 60 80 100 120−6−4−20246 x 10 −3 x (pixel) A m p li t ude
20 40 60 80 100 120−6−4−20246 x 10 −3 x (pixel) Fig. 7.
Cut of interference patterns (∆ U, ∆ V ) along x -axis.After correction of the low spatial frequency term, fringecontrast still varies strongly, from 0.2 % to 0.6%. V = V + γ V sin( φ + ε ) V ′ = V + γ V sin( φ + δφ + ε )(16)where ( U , V ) describe the unflatness, and ( γ U , γ V ) thefringe contrasts. The subtraction of these signals eliminatesthe unflatness terms and brings to the two interfering pat-terns :∆ U = U ′ − U = − γ U (cid:20) (cid:18) δφ (cid:19)(cid:21) sin( ϕ ) (17)∆ V = V ′ − V = γ V (cid:20) (cid:18) δφ (cid:19)(cid:21) cos( ϕ + ε ) (18)where ϕ is the mean phase ϕ = φ + δφ/
2. Figure 7 showscuts of (∆ U, ∆ V ) along the x -axis; the unflatness problemhas been corrected. The fringe contrast presents optimizedvariations across both fields, from 0.2% to 0.6%.The motionless complex interferograms Z = U + ıV is obtained by fitting both quadrature shift and phase on Gaulme et al.: SYMPA: II. Real performance and first results
Fig. 8.
Left: plot of the interference pattern ∆ U as a func-tion of the other ∆ V . The spiral structure comes from thefact that both fringe contrast and ellipticity vary across thefield. Right: plot of the interference pattern ∆ U ′ as a func-tion of the other ∆ V ′ . Points are distributed along a circle(solid line, white), which indicate that the two interfero-grams are in phase quadrature. In terms of phase accuracy,the standard deviation of these points is 5 . ◦ .the couple (∆ U, ∆ V ). First, the plot of one of the inter-ference pattern as a function of the other highlights thephase quadrature imperfections (Fig. 8). In case of perfectquadrature and uniform contrast in both interferograms,points would be distributed around a circle centered on 0.As a result, points are distributed along an ellipse, whoseradius varies strongly on detector field. The phase shift ε with respect to quadrature is related to the ellipse param-eters ( A, B ) by: ε = arcsin (cid:18) A − B A + B (cid:19) (19)Since ellipticity and amplitude vary across the field (Fig. 8,left), a further correction has to be applied. Therefore, thefield ( x, y ) is divided in 10-pixel large squares, in which allthese parameters are considered as uniform. Ellipse axis areestimated by least square fitting. The resulting parametersestimate ( A, B ), obtained for each sub-region, are interpo-lated for each pixel, by fitting their values by a 4th orderpolynomial. Thereafter, thanks to A , B and ε , the ampli-tude of both interference patterns (∆ U, ∆ V ) is normalizedto 1 and the phase shift is set to 90 ◦ , by the operation:∆ U ′ = ∆ U norm cos( ε/ − ∆ V norm sin( ε/ ε/
2) (20)∆ V ′ = ∆ V norm cos( ε/ − ∆ U norm sin( ε/ ε/
2) (21)where the subscript “norm” indicates that amplitudes havebeen normalized to 1. These new variables are now in therequired quadrature. Figure 8 shows the plot of the in-terference pattern U as a function of the other V ; phasequadrature is reached. At last, the motionless phase φ instru is obtained by fitting with a 4th order polynomial the ar-gument of the complex interferogram Z sky = ∆ U ′ + ı ∆ V ′ .The resulting phase standard deviation is about 5 . ◦ .The motionless interferogram used in the following toprocess Jovian data is simply built as: Z = eıφ instru (22)Indeed, since only the phase term matters, the fringe con-trast is set equal to 1 in the whole field. y ( p i x e l )
50 100 150 200 250 300 350 400 450 50020406080100120 50 100 150 200 250 300 350 400 450 50001234 x 10 x (pixel) I n t en s i t y ( A DU ) Fig. 9.
Top: quadruplet of Jupiter images ( I , I , I , I ),taken during night of April 2nd 2005 at Teide observatory.Bottom: cut of the top images. The instrument exhibits aglobal polarization effect, resulting in a better transmissionof one channel with respect to the other.
5. From four Jupiter images to a velocity map
Processing the data lies in turning each quadruplet ofJovian images into a calibrated radial velocity map. Thefirst step, developed in the previous section, gives cleanfringes on both interference patterns ( U jup , V jup ), in orderto create the complex Jovian interferogram Z jup . Second,motionless phase is deconvoluted to Z jup with the help of Z (Eqt. 22). Third, Jovian rotation and other uniform ve-locity drifts have to be deconvoluted, in order to extractthe velocity map. Let us consider a quadruplet of 4 pre-processed Jovian im-ages. Since they have been corrected from flat field and op-tical distortions, each couple of images ( I , I ) and ( I , I )present the same intensity level and shall overlap (Fig. 9).Then, the two interfering patterns ( U jup , V jup ) are createdfollowing Eqts. 7 and 8. Moreover, they are set in phasequadrature following relation (20) and (21). As it can beseen in Figs. (10) interference fringes appear, but, as for the“sky” interferograms, the fringes do not oscillate around 0,but around a distorted surface. These features are photo-metric residues, which have not disappeared with opera-tions (7) and (8). However, fringes have to be repositionedaround a flat surface, centered around 0, in order to followthe phase of the Doppler signal.The best way to separate the photometric noise andspurious signal resides in filtering in the spatial frequencydomain. The two dimension fast Fourier transform (FFT) isapplied to the complex interferogram Z jup . In order to avoidspectral leakage due to the finite size of the image, we ap-ply an oversampling on the data Z jup , by a factor 2, beforepassing to the Fourier domain. Besides, Jupiter’s bound-ary is apodized with a cos function, to avoid Airy-like re-bounds due to Fourier transform. In Fig. 11, we present theFourier transform modulus. The central region contains thelow frequency information (mean value, slow distortions).The horizontal line corresponds to the interference pattern;the horizontal structure comes from the fact that opticalpath difference varies essentially with x -coordinates. Theinclined alignment of stains constitutes the photometric ef- aulme et al.: SYMPA: II. Real performance and first results 7 x (pixel) y ( p i x e l )
50 100 150 200 25020406080100120 −8−6−4−20246x 10 −3 Fig. 10.
Rough interferograms ( U jup , V jup ) obtained follow-ing Eqts. 7 and 8. The interference pattern is clearly visible,but some photometric residuals of same order of magnitudeas fringe’s contrast, related to Jovian bands and zones, stillremain.fect. Indeed, it corresponds to the spatial frequencies alongthe rotation axis of Jupiter, which is inclined of about − ◦ in the detector’s field. These features are actually remnantsof cloud zones and belts.Wiener filtering stipulates that signal can be filteredout from noise in the Fourier domain if they were clearlydifferentiable. As it can be seen in Fig. 11, fringes occupydistinct positions from photometric residues. Actually, forhigh degree modes, a coupling between photometric noiseand oscillation signal still exists, since high degree sphericalharmonics extend largely in the Fourier domain. Therefore,part of the energy in the largest spatial frequencies maybe filtered during this operation. Nevertheless, higher fre-quency mode information will be still available, since mostof its signal is not cancelled by filtering operations, but maysuffer from amplitude estimate uncertainty.In order to limit the damage generated by filtering thenoise, the spatial filter is as smooth as possible. It consistsof an ellipse, which major axis inclined of − ◦ with respectto vertical, which includes only the photometric stains andthe central region. Besides, as for Jupiter, filter’s bound-ary is apodized to avoid rebounds by applying the inverseFourier transform, when getting back to the image plane.The resulting couple of interference patterns, after spatialfiltering is presented in Fig (10). Now, fringes are centeredaround the 0 value. Hence, requirements for deconvolutingthe motionless phase are achieved. The motionless fringe deconvolution is realized by the fol-lowing operation: Z j , rot = Z jup × Z ∗ (23)The resulting interference pattern presents fringes parallelto the planetary rotation axis (Fig. 13). Indeed, the projec-tion of Jupiter’s velocity field v = Ω R ( L ), where Ω is theangular velocity and L the latitude, towards the observerreduces to v = Ω x , where x is the abscissa along Jupiter’sequator. At zero order, Jovian rotation can be consideredas solid rotation since differential rotation with respect tosolid-body rotation is about 1% on the equator. Thus, Ωis almost uniform on the Jovian disk, and the phase of thecomplex interferogram presents iso-values along the rota-tion axis. Moreover, note that since noise level is about 900ms − per pixel, differential rotation is definitely invisibleon a single image. k x (pixel −1 ) k y ( p i x e l − ) −0.4 −0.2 0 0.2 0.4−0.5−0.4−0.3−0.2−0.100.10.20.30.4 photometric residuesfringes Fig. 11.
Modulus of Fourier transform of the complex in-terferogram Z jup = U jup + ıV jup . Photometric residuals arelocated along the direction indicated by the two upper ar-rows. This direction fits with Jupiter’s inclination in thedetector’s field. Main fringe information relies in the hori-zontal line, under the central region. x (pixel) y ( p i x e l )
50 100 150 200 25020406080100120 −505x 10 −3 Fig. 12.
The same couple of interferograms after filteringof the photometric residuals. Note that fringes oscillatearound 0, and that contrast is about 0.6%, which is veryclose to expected values. x (pixel) y ( p i x e l )
50 100 150 200 25020406080100120 −5−4−3−2−1012345x 10 −3 Fig. 13.
Interference pattern after deconvolution of the mo-tionless component. Fringes present iso-values along the ro-tation axis of Jupiter. The white circle indicates Jupiter’ssize before resizing.Hence, a complex phaser which reproduces the solid ro-tation is applied to each Jovian complex interferogram; it isdefined by Z solidrot = exp(4 πσ ∆ v rot /c ), where v rot = 2 π/T and where T = 9 h 55 m 30 s (system III) is the mean rota-tion period. The main difficulty of such a process lays on theaccuracy of the estimate of Jupiter’s position on the detec-tor, because 1 pixel corresponds to 350 ms − . Two meth-ods have been envisioned to make the center of the solidrotation phaser Z solidrot overlap on the center of Jupiter’s Gaulme et al.: SYMPA: II. Real performance and first results interferogram Z j , rot . In both cases, a threshold is appliedto each image, in order to get rid of spurious photometricsignals, such as Jovian satellites of terrestrial atmosphericlight scattering. The “barycenter” method consists in tak-ing the coordinates of the barycenter of the total photo-metric image I = P i =1 I i . The “interspectrum” methodconsists in determining the relative distance between twoimages, taken at different dates, by measuring the phase Φof the interspectrum of the two images. Indeed, the phase ofthe interspectrum of the couple of images ( I , I ) is definedas:Φ = arg {F ( I ) × F ( I ) ∗ } (24) ∝ ( x − x ) + ( y − y ) (25)where ( x , y ) and ( x , y ) are the coordinates of the cen-ter of Jupiter and F and F ∗ indicate the Fourier and in-verse Fourier transformations. The second method has beenpreferred because the barycenter estimate is too sensitiveto high spatial frequency photometric details, which varyalong the night, as cloud features or satellite transits. Onthe contrary, the interspectrum method is sensitive only tolow spatial frequency.After deconvolution of Jovian mean rotation (i.e. solidbody approximation), interferograms become flat since theremaining phase φ = 4 πσ (cid:0) v J / S + v E / J + v E , rot + v osc (cid:1) isuniform across the field (Fig. 14). The subtraction of v J / S and v E / J is exposed in the next section, since they do notchange the noise to signal ratio of the phase map. Therefore,the velocity map is retrieved thank to Eqt. 6 in paper I: v = v arg { Z flat } (26)with v ≃ − . Note that because of apodization cre-ated by spatial filtering of photometric residuals, the entirephase map is not exploitable. A part of the external regionis cut down, whose proportion is a function of the apodiza-tion strength. Here, an external ring representing 1 / − . However,the performance decrease is limited by the low weight ofexternal regions in the Doppler signal.The standard deviation of velocity across the Joviandisk is about 890 ms − per pixel, that is to say 18.9 ms − when integrating the 2200 pixel of the resized Jovian disk.This performance matches with expectations. If photonnoise is reached, a 7-hour night integration with 6-s sam-pling yields a noise level as small as 29 cm s − .
6. Temporal analysis over one night
In the previous section, we have exposed the data process-ing method for a single Jupiter quadruplet, for the extrac-tion of the Doppler signal. In this section, we analyse thetime series of the median phase extracted for each Jovianquadruplet. Then, we identify the different sources of noiseand present correction methods.
The study of median phase along a single night allows usto evaluate the mean noise level and to identify spurious x (pixel) y ( p i x e l )
20 40 60 80 100 12020406080100120 0.8 0.6 0.4 0.200.20.40.60.8 U V Fig. 14.
Left: Jupiter phase map, after deconvolution ofmotionless phase and Jovian rotation. Terrestrial motionand temperature variations have not been taken into ac-count, since they only shift the whole phase by a uniformvalue. Right: complex diagram of the same interferogram; V flat is plotted as a function of U flat . The white circle in-dicates the standard deviation of the phase of the interfer-ogram, which value is 8 . − rad. In terms of velocity, itcorresponds 890 ms − per pixel, that is to say 18.9 ms − on the whole disk (2200 pixels are still available on Jupiterafter resizing).signals. Median phases are extracted from Jupiter’s flat in-terferograms Z flat as follow: φ med = arctan (cid:18) V med U med (cid:19) (27)where the subscript “med” indicates the median value ofthe considered variable. This estimate of Jupiter medianphase has been preferred to the direct median of the phasemap, in order to avoid noise coming from 2 π jumps (seeFig. 14). In Fig. 15, median phase along night 9 and itspower spectrum have been plotted.First, according to data processing chain (Eqts. 5, 10and 11), phase measurement is proportional to the red-shift of spectral lines, it appears that the signal is over-whelmed by a strong low frequency noise. Indeed, alongthe night, the velocity variation should be dominated bythe Earth rotation (409 m/s amplitude at Teide observa-tory), so the phase shall increase of about 0.41 rad insteadof decreasing of about 1.12 rad. This implies that an unex-pected stronger low frequency drift dominates. Besides thislow frequency noise, the power spectrum highlights a rapidoscillation noise source around 6-mHz frequency.Anyway, the standard deviation of median phase isaround 27 m s − per image; images were taken every 6s. By supposing an only photon noise origin, the noise levelreduces to 15 cm s − for 7 nights of 7 hours, which is 1.3 timeworse than expected from previous section. In the followingsubsections, we interpret the noise origins and describe themethod which has been used to get rid of their main effects. Paper I has expose that the interferometer is made of twopieces of two different glasses, specially chosen as to com-pensate the index variations and the dilatation, in orderto have a stable OPD. In particular, a temperature varia-tion of about 1 ◦ C should have no effect on OPD in 14 ◦ Cenvironment.The median of Jupiter’s median phase over each nighthas been plotted as a function of the corresponding me- aulme et al.: SYMPA: II. Real performance and first results 9 P ha s e (r ad ) Measured phaseTheoretical phase0 10 20 30 40 50 60 70 80012345 Frequency (mHz) A m p li t ude s pe c t r u m ( m s − ) Fig. 15.
Top. Median phase along night 9: measured phase(full line) versus theoretical phase drift due to Earth rota-tion and relative motion of Jupiter toward the Sun and theEarth (dashed line). Bottom. Amplitude spectrum of themeasured phase as a function of the time frequency. Twospike forests blow out the mean noise level at less than 0.5mHz and at 6 mHz. Note that mean noise becomes totallyflat in the frequency range above 20 mHz, which means thatphoton noise is achieved, around 41 cm s − .dian temperature (Fig. 16). Phase and temperature appearclearly correlated. A variation of 1 ◦ C introduces a phaseshift of 0.33 rad (i.e. 330 m s − ), which is much abovethan expected (between −
60 m s − and 30 m s − for tem-perature between 0 and 20 ◦ C; see paper I). In fact, sucha thermal effect is retrieved when taking into account theerror bars upon dilatation coefficient and refractive indexthermal dependance ( ±
10 %).Unfortunately, the correlation coefficient reported inFig. 16 does not allow us to correct Jovian phase map fromits thermal dependence. Indeed, the comparison of phaseand temperature over one night does not match correctly: aseveral hour delay is still present (Fig. 17). This discrepancyis due to the fact that temperature measurements do notcorrespond to the Mach-Zehnder prism, but to the metallicbox which holds it on. Because of a greater thermal iner-tia, glass actual temperature is time-shifted with respect tometal temperature. Therefore, a low frequency filter is ap-plied to data in order to reduce the noise in the frequencyrange below 0.2 mHz.Beyond a mean OPD drift, temperature variations gen-erate differential OPD variations with respect to light in-cidence angle onto the prism. Since Jovian interferogramsget deconvoluted from motionless fringes with the help of aconstant interferogram (Eqt. 23), the differential OPD vari-ations introduce a slowly varying inclined surface in Jovianflat phase maps (Fig. 18). This surface is bent along the x -axis because OPD is only x -dependent. Such an inclinedsurface yields a high frequency perturbation correlated toJupiter’s position in the observed field. This noise is re-duced by fitting the spurious surface on each phase map,by a plane inclined only with respect to x -axis. Thereafter,Jupiter’s phase map are set flat by using a smoothed esti-mate of the fitted parameters. M ed i an pha s e (r ad i an ) Median measured temperature (°C) φ = 0.33 T − 1.77 Fig. 16.
Phase-temperature correlation. For each night, wehave plotted the median of the median phase as a func-tion of the median temperature over the entire night. Thestraight line represents the least square estimate of the cor-relation between phase and temperature. P ha s e (r ad ) Fig. 17.
Jovian median phase, corrected of Earth’s rotationcomponent, over night 10 (full noisy line) and temperatureconverted to phase with the previously fitted correlationcoefficient (full smooth line). Note that measured phase fol-lows the temperature with a several hour delay. M ean pha s e (r ad i an ) M ean s l ope (r ad i an p i x e l − ) Fig. 18.
Top: phase map averaged along y-axis as a func-tion of the x-axis. A negative slope is clearly visible.Bottom: root mean square estimate of the slope of the in-clined plane along night 10.
The so-called “guiding noise” is the perturbation relatedto Jupiter’s position, which should not occur if Jupiterwere at fixed position. Theoretically, SYMPA’s Dopplervelocity measurements are not sensitive to Jupiter’s po-sition onto the field, because after motionless fringe de-convolution (Eqt. 23) nothing should depend on the co-ordinates. However, the comparison between spectra of J up i t e r c oo r d i na t e s ( p i x e l ) Abscissa xOrdinate y0 10 20 30 40 50 60 70 8000.20.40.60.8 Frequency (mHz) A m p li t ude s pe c t r u m ( p i x e l s − ) Ordinate yAbscissa x
Fig. 19.
Top. Jupiter’s coordinates ( x, y ) in the CCD fieldover night 9. Note that y varies much more than x coordi-nate because Jupiter is rotated about − ◦ on the imageand because most of the telescope guiding problems occurwith the right ascension. Bottom. Amplitude spectrum ofboth coordinates. They present strong components below 6mHz and significative spikes up to 18 mHz.Jupiter’s phase and coordinates ( x, y ) (Figs. 15 and 19)shows a strong correlation at 6 mHz. In fact, many othercorrelated signatures occur at many frequencies.This guiding noise has two main origins. First, thetemperature high frequency noise cannot be totally can-celled because of the inclined surface fit inaccuracy. Second,Jupiter’s position is estimated around 1/15 of pixel.Therefore, the deconvolution of Jovian rotation introducesa spurious signal correlated to Jovian position. Thus, theguiding noise is a combination of these two effects, for whichit is hard to determine which of them predominates.The guiding noise is strongly reduced by removingthe intercorrelation of Jupiter’s phase and coordinates.The decorrelated phase with respect to both coordinated φ decorr / x / y is obtained as follow: φ decorr / x = F − { (1 − I x ) F{ φ }} (28) φ decorr / x / y = F − { (1 − I y ) F{ φ decorr / x }} (29)where φ decorr / x indicates the phase decorrelated with re-spect to x , and I x and I y the interspectra relative to x and y , defined by: I x = 2 π arcsin ( ℜ{F ( φ ) F ∗ ( x ) } ) |F ( φ ) ||F ∗ ( x ) | (30) I y = 2 π arcsin (cid:0) ℜ{F ( φ decorr / x ) F ∗ ( x ) } (cid:1) |F ( φ decorr / x ) ||F ∗ ( x ) | (31)Note that we apply a 75% threshold upon I x and I y in orderto decorrelate only significant guiding noise spikes.Jupiter’s phase standard deviation drops down from 27m s − per image to 21 m s − , which is very close to theexpected photon noise level (19 m s − per image). Hence,by supposing the remaining noise is only due to photons,the noise level reduces to 32-cm s − on a 7-hour integrationand to 12 cm s − after 7 nights. The remaining not-photonnoise is ignored.
7. The search for oscillations: no evidence for aJovian signal
A stationary oscillation mode can be described as the sumof two spherical harmonics of degree ℓ and order ± m .Jupiter velocity field related to p -modes follows such a de-scription. Therefore, the Doppler signature of the radialvelocity field expands in a base made of the projected com-plex spherical harmonics towards the observer.The coefficients c mℓ associated to each spherical harmon-ics Y mℓ are obtained as follow: c mℓ = P pixels ℜ{ Y mℓ } × v P pixels ℜ{ Y mℓ } + ı P pixels ℑ{ Y mℓ } × v P pixels ℑ{ Y mℓ } (32)where v indicates the velocity map (Eqt. 26). The normali-sation coefficient is done with respect to the actual numberof pixel after resizing. The c mℓ coefficient modulus are ex-pressed in m s − . Oscillation search is performed in thespectra of all spherical harmonics up to the degree ℓ = 25. Velocity maps are extracted from mean flat interferograms,averaged within 30-s intervals, but calculated every 15 s.Such a process has two reasons. On one hand, the meannoise level inside velocity maps drops down of a factor √ π jumps which appear when applying Eqt. 26. On the otherhand, it allows to use the fast Fourier transform (FFT) tocalculate the power spectra. The procedure is of great inter-est because the search for modes up to 25 means 625 spectraof 27000 point in the time series. The spacing of data ev-ery 15 s imposes a cut-off frequency at 16.3 mHz, whichis well beyond the expected p-modes (less than 3.5 mHzfrom Mosser 1995). The averaging within 30-s intervals isdone in order to avoid spectrum leakage when calculatingthe Fourier transform. ( ℓ = 0 , m = 0) and ( ℓ = 1 , m = 0)We present the power spectrum of the time series corre-sponding to spherical harmonics Y and Y , for all the dataof Canaries observation campaign. The choice of these twomodes among 625, permits to present the two main typesof spectra. The first is sensitive to the remaining guidingnoise, whereas the second is much less sensitive. Indeed,since guiding noise is mainly due to right ascension controldefects, the mode ( ℓ = 1 , m = 0) (hereafter (1 , ,
8] mHz, excepted around 6 mHz, where a guidingsignal subsists. Beyond 8 mHz, the averaging over 30 s cuts-off the signal. Guiding signature is reduced from 4 m s − to 1.3 m s − after decorrelation processes. The mean noiselevel is about 12.6 cm s − for mode (0,0) and 11 cm s − formode (1,0), which squares with the last-estimated photonnoise level. Note that such a performance has never beenreached on Jupiter and proves that the instrument and thedata processing chain works efficiently. aulme et al.: SYMPA: II. Real performance and first results 11 Fig. 20.
Power spectrum of time series related to modes (0,0), top, and (1,0), bottom, after concatenation of data alongthe whole run. The mean noise level are respectively of about 12.6 cm s − and 11 cm s − , what matches with the lastestimate of the photon noise. The 6-mHz guiding spike still is about 1.3 m s − for the (0,0) mode, whereas it is onlyabout 0.5 m s − for the (1,0) mode.As regards the comparison to previous observations ofS91 and M93 and M00, no excess of power is present be-tween in the [1 ,
2] mHz frequency range. Moreover, no largespacing ν is highlighted, whose value is estimated around150 µ Hz, and which was detected around 136 and 143 µ Hz,respectively, by S91 and M00. Such a dissension with pre-vious observations will be analyzed in detail in a futurework.As regards the excess of power in the frequency range[0 . , .
6] mHz, no indication for a Jovian origin can be fur-nished at this step of the data analysis. It could be a re-maining low frequency noise, related to temperature andposition. A global analysis over all the modes up to degree ℓ = 25 is required to determine its origin and to high-light global signature as Jovian rotation frequency or ν frequency. As for helioseismology, simultaneous temporal and spatialfrequency analysis may reveal the presence of significant in-formation lost among noisy spectra. In Fig. 21 we presentthe ( ℓ, ν ) and ( m, ν ) diagrams. In the ( ℓ, ν ) diagram, the lowfrequency excess of power, picked out in the (0 ,
0) and (1 , ℓ in thefrequency range [0 . ,
1] mHz. It does not exhibit an orga-nized structure. Moreover, in comparison with past obser-vations, no signal is distinguishable in the [1 ,
2] mHz range.On the other hand, in the ( m, ν ) diagram, the same excessof power appears to be strongly structured. The energy isdistributed along 2 main lines, symmetric with respect toabscissa, beginning at 0.5 mHz and ending at 1.2 mHz for m = ±
25. Moreover, a second couple of lines, almost par-allel to the first ones, is still visible between frequency 0.7
20 40 60 80102030405060708090
Fig. 22.
Velocity maps averaged over 5 minutes. A fringestructure, inclined of about Jupiter’s inclination on the de-tector is still present. The coupling of these feature withthe spherical harmonic masks may explain the observedfeatures in the ( m, ν ) diagram.and 1.6 mHz. The mean slope of the principal lines is about28 µ Hz, which corresponds to Jovian rotation frequency.A quick analysis of the ( m, ν ) diagram gives some indi-cations. A simple guiding noise origin is excluded, becausesuch a spurious signal contaminates all the eigenmodes atthe same frequency, as for the 6 mHz spike (see Fig. 21).Furthermore, guiding noise becomes negligible beyond or-der m = 15, since its effect compensates when applyinghigh order spherical harmonic filtering to velocity maps.At last, if the slope of about 28 µ Hz could be casual, itcould indicate a Jovian origin to the observed signal. D eg r ee l O r de r m Frequency (mHz)0 1 2 3 4 5 6−20−1001020
Fig. 21.
Time-spatial frequency analysis. Top, the ( l, ν ) diagram represents the mean of all the power spectra at a givendegree ℓ , as a function of all the explored degree. Bottom, the ( m, ν ) diagram represents the mean of all power spectraat a given order m , as a function of all explored orders. Note the guiding noise is still present at 6 mHz, within a verticalline in both graphics. The maximum amplitude is around 1.2 m s − .However, the Jovian signal hypothesis may vanish bysupposing a coupling between two spurious signals. First,it can be noticed, after summation of velocity maps over10-min, that a fringe structure still remains, which peri-odicity fits almost with m = 16 (Fig. 22). Second, it canreasonably supposed that photometric signal has not beentotally removed after the Fourier filtering step (see Sect.5.1). Thus, a signal modulated by Jupiter’s rotation prob-ably underlies inside velocity maps. The coupling of thesetwo spurious signals introduces a modulation of the signalby a cos( m Φ) factor, where Φ = Ω r t comes from the pho-tometric remnants; Ω r = 2 πν r indicates the Jovian rotationfrequency. Consequently, when applying the spherical har-monic search algorithm (Eqt. 32), a coupling appears be-tween the cos( m Φ) modulation of velocity maps and thecos( m Φ) associated to Y mℓ . With this assessment, velocitymaps are modulated in the following way: v = v D cos(Ω r m t ) cos(Ω r mt ) (33)= v D { cos [Ω rot ( m + m ) t ] + cos [Ω rot ( m − m ) t ] } (34)Therefore, a linear dependence ν = m ν r + C appears,where the constant term C = m ν r is about 450 µ Hz, whichmatches with the observed origin of the two lines.
8. Conclusion and prospects
The aim of SYMPA instrument was the detection and mea-surements of acoustic modes on the giant planets of the so-lar system, with a previously unequalled sensitivity around4 cm s − . Such a performance was estimated for a 16-day observation campaign with 50% duty cycle (see paper I).By choosing to process only the Canaries data, since theinstrument used in San Pedro Martir Observatory presentssome defects which make the data more difficult to process,the duty cycle is only 21% over 10 nights. In this case, thenoise level is reevaluated at 10 cm s − . By taking into ac-count the lack of photons by a factor 2, underlined in Sect.2, the 1 − σ sensitivity decreases to 12 cm s − . After decor-relation of time series with respect to Jupiter’s position onthe detector, outside the low frequency range ( ≤ . − . Such a noise level is 5time better than previous observations which were limitedaround 60 cm s − (M00).However, the data processing has highlighted some de-fects, as the too strong optical path difference dependenceto temperature and some difficulties, as the separation be-tween photometric and spectrometric information. The lat-ter make the extraction of Doppler velocity hard to practise.This is mainly due to an insufficient accuracy in rectifyingthe distortion, particularly about the photometric effect ofthe distortion (variable PSF upon the detector). This pointsuggests some instrumental modifications, as the time mod-ulation of OPD in order to modulate the phase of the fringepattern of about π . It would allow us to replace the spatialsubtraction between interferograms by a time subtraction.Moreover, the spectral information would be emphasizedmore easily by narrowing the entrance filter, which wouldaffect the global sensitivity. This would increase the fringecontrast. aulme et al.: SYMPA: II. Real performance and first results 13 Jupiter’s seismological observation of S91, M93 and M00have all exhibited an excess of power in the frequency range[1 ,
2] mHz, which matches with the theoretical expectations(e. g. Mosser et al. 1996). Moreover, they did not provideany spatial resolution. If both observation methods (sodiumcell and Fourier transform spectrometry at fixed OPD) havepresented an excess of power in the same frequency range,the large spacing ν estimate was very noisy and differedquite sensitively with respect to the theoretical value of 153 µ Hz (Gudkova et al. 1995).Our observations do not present features close to p -modes signature: the absence of the large spacing ν inpower spectra and ( ℓ, ν ) diagram is significative. However,it is worth to notice that because of the resizing of Jovianvelocity maps, the projected spherical harmonic base be-comes a quite degenerated base. Therefore, informationfrom a given mode goes diluted into other mode spectra,which may diminish strongly the p -mode signature and theidentification of specific related structures. A way of leavingdegeneracy has to be developed.Moreover, Bercovici and Schubert (1987) roughly esti-mated Jupiter’s oscillation velocity amplitude between afew cm s − and 1 m s − . Therefore it is not abnormal thatoscillations are not enlightened with a 12-cm s − noise level,with a 21% duty cycle. This reduces the observed amplitudeof any oscillations by a factor (0 . . . In these conditions,only a 25-cm s − signal could be detected at 1 − σ level. SYMPA has demonstrated to work properly, after takinginto account its technical defects (temperature dependance,field distortions, difference of intensity between the two po-larized outputs, lower sensitivity of the CCD). It has per-mitted to reach the best remote sensing velocity measure-ments upon giant planets. Some improvements are to beperformed on the existing instrument, as a better thermalinsulation and a slightly modified optical design in orderto reduce the geometrical distortions. An alternative andmore efficient solution is to rebuilt a prism, equipped withan OPD time modulation.From a most general point of view, this seismometeris a tachometer, which furnishes velocity maps, insteadof point to point measures, which is the case of echelle-spectrometers. Thus, it can be used to other kind of ob-servations, such as wind velocity measurements. SYMPAhas tested an extra seismological application in November2007, by participating to a ground-based observation cam-paign organised in sustain to ESA Venus-Express probe, inorder to characterize the lower mesosphere wind velocity.