79 GR/MM GRATING, ORDER #60 relatively coarse resolution detector. If, 10000 in the future, the full MAMA format be- comes available, the CASPEC Long Camera could be used and the detector pixel size would be better matched to the spectrograph optical resolution. 7500 The present data taking software used with the MAMA does not allow access to IHAP during integrations. The computer zb is therefore not then available to perform 2 other tasks, such as the precise estima- u 5000 tion of the signalhoise ratio of an expo- -4 sure as the integration proceeds, arith- E- 0 metic on previous exposures, computing b the continually changing atmospheric dispersion parameters, etc. The atmo- 2500 spheric dispersion parameters are needed to accurately centre the UV im- age of an object in the slit since the ESO TV cameras detect only the visual image of the object. Because the differential 0 atmospheric dispersion is so large in the 100 200 300 400 500 600 PIXEL UV, the ability to calculate the position angle and amplitude of the dispersion is Figure 5: The Thorium-Argon spectrum. It may be compared with that given in the ESO Atlas of particularly important for a UV sensitive the Thorium-Argon Spectrum (ESO Scientific Report No. 6 - July 1987). detector such as MAMA. Future minor revisions to the data taking software will allow the MAMA detector to be used On-line monitoring of centring errors, resolution, the rotation of the Earth more efficiently. clouds, etc. is possible when the smears long exposures. With the In conclusion, the ESO MAMA is a ratemeter is attached to the detector. MAMA, long exposures of faint objects working detector with particular advan- Because the astronomer can see the can be broken in order to obtain velocity tages for UV observations. A few minor spectrum building up, the object iden- calibrations. The MAMA has a very low modifications presently being underta- tification is obtained early in the obser- dark background (about 0.1 event/pix- ken will make the detector even easier vation, and integration can be halted el-' hour-' at 5C), and it also dis- to use. It thus fills a needed capability in when it is judged that a sufficiently high criminates against the cosmic rays that the instrument compliment of ESO. signal/noise ratio has been obtained. It plague long integrations with CCD de- If there is sufficient user interest in the is therefore very well suited to those tectors. MAMA detector, ESO could purchase problems in which very long exposures The ESO MAMA detector has a some- additional tubes with improved charac- of faint objects are needed. Because what smaller (- 1000 x 250 pixels) for- teristics. We suppose that this will de- there is no readout noise, long ex- mat than the CCDs and the loss of a pend to some extent in the future availa- posures can be broken into short seg- number of pixels leaves one with only bility of improved CCD detectors that ments that are interspersed with calibra- about 670 x 250 contiguous pixels. To could become competitive with the tion spectra. This is a significant advan- achieve overlap of the spectral orders MAMA for the UV applications, as well tage for instruments, such as CASPEC, the short Camera must then be used. as the interest in applications where the that suffer from flexure problems. For This results in a less than optimal map- fast temporal response time of the spectra taken with very high spectral ping of a high resolution image onto the MAMA can be utilized.

Deconvolution of NTT Images of EISO F. ZAVA~~I',0. BEND~NELL~~,G. PARMEGGIANI~, G. G.C. PALUMBO',~~~~G. F. B1GNAMl3 ' Dipartimento di Astronomia, Universita di Bologna, Italy. Osservatorio Astronomico di Bologna, Italy. 1. F. C. T. R./C. N. R., Milano, Italy. 1. TE. S. RE./C. N. R., Bologna, Italy.

Introduction The first glimpses of the high quality On the other hand, in recent years a The New Technology Telescope (NT) images obtainable by virtue of the considerable effort has been put into the has recently become fully operational happy combination of advanced tech- development of highly sophisticated and subsequently made available by nology and excellent seeing conditions mathematical techniques to deconvolve ESO to European astronomers in the at La Silla have already been published images in one and two dimensions. standard scheduling procedure. (The Messenger, issues 56 and 58). Although the effects of atmospheric posures are shown, and from Figure Ib where the adopted average PSF from NTT PSF the 5-minute frame is compared with the - S4 1 rnxn most luminous in the I-minute ex- -1 - average posure image as well as with a star in the field of the ESO 440 G 37. The average PSF (HWHM -0.34 arc- -2 sec) was fitted with a sum of 3 Gauss- ians (e.g. Bendinelli et al., 1987, 1989) -3 as well as a Moffat (1969) function; the fit parameters are listed in Table I. -4 This comparison brings the conclu- sion that the NIT and the seeing condi-

-5 tions were considerably stable both within each single frame and from one frame to another. -6

-7 The Deconvolution 0 5 1 1.5 2 0 5 1 1.5 2 R (arcsec) R (arcsec) Surface brightness profiles were ex- Figure 1: (a) Brightness profiles of some near PKS 1209-52 from 5 and 1 minute tracted using the VISTA code. integration time. (b) Brightness profile of the adopted (average) PSF (solid circles and line) All galaxies in our sample were de- compared with the one of the most luminous star in the I-minute frame (solid line). The dashed convolved by means of three quite diffe- line is the PSF derived from a star in the ESO 440 G 37 frame. rent techniques: - Regularized numerical inversion of the convolution integral equation (RLS), applied to brightness profiles. Such turbulence which produces image blur- era (515 x 31 3 pixels; Nasmyth focus) at method was previously used in the de- ring cannot entirely be abolished in any the ESO NIT telescope in the V band. convolution of a variety of very different image taken with a ground-based tele- The images were acquired on the night of objects (see e.g. Bendinelli et al., 1986 scope, one can aim, deconvolving with January 24, 1990. The typical seeing and references therein). With a PSF ex- the Point Spread Function (PSF), at FWHM was 0.68 arcsec for the duration pressed in terms of Moffat function. "cleaning" the observed image in order of these observations, and the sampling - Regularized Multi-Gaussian (RMG), to approximate the "real" image, as it was 0.246 arcseclpixel (rebinned im- based on the analytical solution of the would be obtained in the absence of the ages). A total of eight CCD frames were convolution integral equation when both atmosphere. This process has been ap- obtained with integration times ranging object and PSF are expressed as a sum plied on various occasions using a vari- from 25 to 21 0 sec. Standard processing of Gaussian functions. ety of deconvolution techniques and of .framesfor flat field and bias was used. This method has two major advan- has been proven to be quite powerful. In this note we discuss the deconvolution tages: There still is a debate, however, on of these elliptical galaxies. (a) it is insensitive to undersampling which mathematical technique works (b). . it runs a whole deconvolution cvcle best. Combining high quality images in a few seconds on a Personal Com- The Point Spread Function with robust deconvolution Drocesses, Duter. one can hope to reach a better under- Star images suitable for PSF extrac- The major limitation of the RMG standing of both the relative power of tion were not available in all frames. Two method, however, is that it is applicable the algorithms used and the goodness frames of the stellar field near PKS only to round or elliptical images and of the images obtained, or, in turn, to 1209-52 (1 and 5 minutes exposure does not work on images lacking test the telescope characteristics. time) were therefore used in order to symmetry. During the first general observer NlT derive an average PSF. Such Point - Deconvolution adopting a method run, a few images of well-known bright Spread Function was used for decon- devised by Hunt (1973) (HUNT). This is a galaxies were obtained in good seeing volving all extended objects in the pres- bidimensional deconvolution and makes conditions by one of us (G. F. 6.). ent sample. The seeing conditions for use of FFT, and basically is a generaliza- The present paper describes the ob- this reference star field were monitored tion of Wiener's filter. As far as we are servation of those objects and the three to be identical to those of the galaxy aware the HUNT method has only been mathematical algorithms used to de- images. applied to astronomical images once by convolve their images. From the results The stability of the star profiles can be Heap and Linder (1987) for deconvolv- obtained, conclusions can be drawn on deduced from Figure 1 a, where some ing the image of SN 1987A. the relative power of the deconvolution stars from both 1 and 5 minutes ex- The three deconvolution methods techniques used, the aualitv. . of the im- ages the NIT is able to acquire and some interesting astrophysical aspects TABLE I: Parameters of the fitted average PSF of the objects studied. FWHM 3-Gaussian fit Moffat fit (arcsec) L,, a,, i = 1, 2, 3 a, B Observations Five EISO galaxies: ESO 440 G 37, 0.306 0.51 8 NGC 31 15, NGC 3585, NGC 3904, NGC 0.115 1.121 3923 were observed. These images were Note: o,,rr in arcsec. obtained using the RCA # 15 CCD cam- . OBS RLS BSO 440 G37 -RUG NCC 3585 - RMC NCC 3904 -RUG -- HUNT - -HUNT - - - PSF - PSF - PSF

- -

-

- -3 - -3 \\ - t -3

-4 ' '1'"1'"1'1'1"' -4 '"l"'J'lll'l'l''l 0 2 4 10 0 2 10 R (arcrec) 4~ (arcrec)

- OBS - RLS NCC 3115-RMC NCC 3923 - RMC - -.HUNT - PSF - PSF

-4-'h'~''f~'''~~~@111~ 0 2 I 8 8 I0 R (arersc)

Figure 2: Observed (solid dots) and deconvolved (lines) profiles of the five galaxies. Observed PSFprofiles, normalized to the central brightness of the galaxies, are also shown as solid lines and dots.

listed above are compared on the same Discussion of the Results ment that with these images (i.e. pixel image here for the first time although the size, object size, PSF) all three methods first two (RLS and RMG) were discussed Table II should be regarded as the work equally well. in Bendinelli et al. (1990). main result of this work which goes to- This means that the unambiguous Table II gives some of the relevant ward the analysis of core structure of identification of a "point" mass in the parameters of the deconvolutions: ob- early type galaxies, as outlined in Ben- centre of nearby galaxies, if present, is served and deconvolved HWHM, in arc- dinelli et al. (1990). possible by a variety of methods, if well- sec, for the different methods and the Differences among the deconvolution sampled data are available. This central difference in observed and deconvolved methods are of the order of some per mass concentration with higher mass- central surface brightness. cent, small enough to justify the state- to-light ratio than the stars that appear If a gain in resolution is defined as the ratio between the HWHM of the ob- served and deconvolved profiles, then TABLE II: Parameters of the deconvolved profiles from this table values can be derived ranging from 1.3 to 3.7, depending on Galaxy OBS RLS RMG HUNT the relative dimension of galaxy and HWHM HWHM Ah HWHM ALL HWHM Ay PSF. In Figure 2 observed and decon- E 440 G 37 0.75 0.42 0.73 0.38 0.80 0.40 0.86 volved profiles of the five galaxies are NGC 3585 0.88 0.40 0.71 0.24 0.82 0.49 0.70 presented. NGC 3904 0.90 0.52 0.59 0.50 0.59 0.57 0.58 The results do not depend on the NGC 3115 1.07 0.48 0.60 0.43 0.66 0.58 0.60 functional form choosen for the PSF NGC 3923 2.27 1.80 0.18 1.80 0.1 8 1.74 0.22 representation, confirming previous Note: HWHM in arcsec; Ap = (observed-deconvolved) in mag/arcsec2. tests (Bendinelli et al., 1988). Figure 3: NGC 31 75-50 galaxy; BT = 9.75. Figure 4: NGC 3585, also ESO 502 G 25-E Figure 5: ESO 440 G 37-E-SO galaxy; Bo = Left panel: original CCD frame (1 .'7 x O,'9). galaxy; Bo = 11.40. Left panel: original CCD 14.30. Left panel: original CCD frame Right panel: Hunt deconvolution. frame (1 .'7 x 0.'9). Right panel: Hunt decon- (1 .I7 x O.'9). Right panel: Hunt deconvolution. volution. to dominate in any tested object the be regarded as a step forward towards Bendinelli, O., Parmeggiani, G., and Zavatti, galaxy mass distribution on scales of the understanding of galaxies them- F.: 1988, J. Astrophys. Astron. 9, 17. tens to hundreds of has tenta- selves and the existence of black holes Bendinelli, O., Parmeggiani, G., and Zavatti, tively been identified with a black hole in their centres. F.: 1989, in Highlights of Astronomy, proc. by many authors. This work was supported in part by 20th IAU General Assembly, in press. Dor- drecht: Kluwer. the Minister0 della Pubblica lstruzione The debate on the presence of black Bendinelli, O., Parmeggiani, G., Zavatti, F., holes in the centre of galaxies is still and Osservatorio Astronomico di and Djorgovski, S.: 1990, Astron. J. 99, quite open since the data seem only to Bologna. 774. imply that some additional mass is re- Gouiffes, G., Wampler, E.J., Baade D., and quired on rather small scales in the References Wang, L.-F.: The Messenger, 58, 11. centre of galaxies. Heap, S.R., and Linder, D.J.: 1987, Astron. Bendinelli, O., Parmeggiani, G., and Zavatti, Astrophys. 185, L 10. Hence any deconvolution method F.: 1986, Ap. J. 308, 61 1. Hunt, B.R.: 1973, IEEE Trans. Comp., C-22, which will permit to constrain the mass Bendinelli, O., Parmeggiani, G., Piccioni, A., 805. distribution in the core of galaxies may and Zavatti, F.: 1987, Astron. J. 94, 1095. Moffat, A.F.J.: 1969, Astr. Ap., 3, 455.

MlDAS Memo ESO Image Processing Group

ments and developments of new appli- states that MlDAS is the image process- 1. Application Developments cation programmes in MIDAS. Not only ing system of ESO to be used both for The main efforts were placed on vali- will this make it possible to have new off-line data reductions and for on-line dation of basic MlDAS commands and algorithms and applications included evaluation of data from ESO telescopes the subsequent correction of problems into MlDAS after a period of limited im- including the VLT. MlDAS is available to reported. Since the next major release provements in this area, but will spread all non-profit research organizations. 90 NOV will be frozen in September al- long-term, detailed knowledge of Such organizations must sign a User ready, only a limited number of these MlDAS in the community when people Agreement with ESO before obtaining improvements will be available in that in these positions return to their home the package. This agreement will reg- release. institutes. ulate the usage of MlDAS and ensure The COPY/DISPLAY command which In addition to these positions, it will be that it is not exploited commercially. can produce hard copies of the image possible to invite people who have This policy will be implemented as of the display has been upgraded to be able to made interesting algorithms and pro- 90 NOV release of MIDAS. provide output for colour Postscript grammes, to ESO for an implementation printers. This makes it possible to get of them into the MlDAS environment. 4. MlDAS on New Systems good working hard copies in both BAN People interested in contributing and/or and colour directly from the image dis- making new applications to MlDAS may The 90 MAY release of MlDAS was play in MIDAS. contact the IPG with detailed descrip- installed on an IBM System 6000 Model tions. 540 (the new RlSC CPU) made available by IBM. Only very mino; problems were 2. New Positions detected, all relating to the operating 3. Distribution Policy Two additional short-term positions system AIX 3.1 which was a preliminary (with durations of up to two years) have The ESO Council, during its last meet- version during the tests. Those prob- been allocated to the MlDAS group. ing in June 1990, defined the policy for lems have been resolved in the official They will be used mainly for improve- usage and distribution of MIDAS. It release of AIX 3.1. The system had an