From erik Wed Dec 23 11:58:40 1998 Subject: CUO_51.txt sampling in photometry To: makarov (Valeri Makarov), Frederic.Arenou@obspm.fr (Arenou), cf (Claus Fabricius), indus (Jens Knude), gaia-sag@astro.estec.esa.nl (GAIA SAG) Date: Wed, 23 Dec 1998 11:58:40 +0100 (MET) The sampling in GAIA photometry =============================== E. Hoeg 23 Dec 1998 SAG_CUO_51 ABSTRACT: It appears from Gerry's note (GG-011, 16-12-98) that he is very concerned about the use of long rectangular samples for GAIA photometry. He does not mention astrometry. I will discuss some of Gerry's propositions and concerns and I find no reasons to be worried. A new NBP design is briefly described. Discussion related to long samples ---------------------------------- The first priority in the SAG baseline design (the October'98 baseline, described in CUO_50) is that the photometry must be optimized for stars and other point-like sources on a uniform and a non-uniform background. Photometry of point sources must have priority because only these can obtain highly accurate astrometry. We will of course - as long planned - make simulations of photometry in star forming regions and stars in galaxies. We intend to do the latter at CUO. The SAG baseline (which is not just a 'suggestion' as Gerry calls it) should have priority since it is the only consistent design we have. Photometry of extended sources is also obtained, but with a lower priority than stars. Could anybody explain the meaning of 2) on p.1 ? Does it mean: Do the proposed samples include a *too* large large area of sky *so* that... ? Note that Gerry wrote 'sufficiently' instead of my 'too', and he does not write 'so'. My English is surely not perfect, but perhaps easy to understand ? I believe that Gerry's concern about 1% disturbance from galaxies, and the disturbance from unknown minor planets are unwarranted. Please check and let me know if you agree. Could anybody explain to me (GG, p.5, 2.3) why the spectrometer option may reach fainter magnitudes than the NBP in areas of high star density, according to Gerry. I believe the opposite is true. Please note the simplicity of the SAG baseline. Positions are estimated e.g. by MCC (see Sect.7.3.1). The flux is subsequently estimated by aperture photometry or PSF photometry. Flux maps are generated by stacking of patches. In a very general sense all this could be called 'deconcolution', but I propose to call it parameter estimation. The term deconvolution should be reserved for more complicated operations where a whole image is transformed, i.e. a large set of output numbers is produced. So, no deconvolution is proposed anywhere in CUO_50 (see Sect.7.3.2). After a parameter estimation for a stack of patches a chi-square should be calculated in order to see if the image model was adequate, i.e. whether the number of point sources was correct and/or the point-source assumption was correct. If the chi-square is too large the image model should be modified. A method suited to analyse the PSM patches in order to find stars in the 3 arcsec^2 area has been described in SP-1200, Volume 3, Section 5. This method called Tycho Recognition is computational very efficient and was applied to 3 million small areas on the sky of about 100 arcsec^2 in 1991. It could well be adapted to cope with a billion stars with the computers available in 2011, twenty years after Tycho. Two other method for similar purposes were used during the Tycho-2 reduction in 1998. The first method was applied to 4.3 million areas in order to estimate the background, positions and magnitudes of single stars. This took 5 weeks with 3 CPUs running parallel. The reduction also gave quality and duplicity indicators for each star. Subsequently a special program was applied to 120 000 suspected double stars for finding and estimating the components, resulting in about 8000 resolved pairs. The computation took 4 weeks with 3 CPUs. The CPUs are 250 MHz Ultrasparc, thus not the fastest you can get, and the speed of available CPUs appears to increase by a factor ten in 5 years. Our impression is that a proper treatment of a billion stars with respect to duplicity and multiplicity will be possible ten years from now, but it is of course a large task. Please note that Gerry's conclusions from Leinert etal could not be confirmed: that "the majority of the sky below Galactic Latitude about 10 deg will require deconvolutions which allow for more complex brightness distributions than just point sources." But Leinert does not give any values on the non-uniformity of the sky background on the scales relevant for GAIA, i.e. less than about one arcsec. Averages are given over much larger areas. The brightest area in Leinert's Fig.66 is 2600 S_10 units = 19.2 mag/arcsec^2, only 1.8 mag brighter than the average sky background assumed in CUO_50, Table 2. This means e.g. that the sky and the readnoise will contribute about equally in the PSM in the brightest areas. In the BBP the readnoise will still dominate. This is a good reason for keeping the length of the samples in the SAG baseline of October'98. In conclusion, there are no reasons to worry, but a lot of careful detailed analysis and hard work lies ahead. New NBP design -------------- I have completed a new design of the NBP and described it in two reports to be distributed in January. We can double the integration time for the present Stromvil bands, and also include infrared bands. The total available integration time for NBP will be increased by a factor 3 by introduction of a dichroic mirror in front of the focal plane, as proposed in Section 7.1.2 of CUO_50. The separation wavelength should probably be 700 nm. This will result in better precision at a given magnitude, e.g. 0.010 mag in the y band at V=18.0 for G2V, and Stromvil photometry is still obtained at V=19.0 containing more astrophysical information than the Sloan photometry. This can be achieved in more than 99 per cent of the sky, excepting areas with the highest star density, which can mostly be observed from the ground. I am looking forward to get proposals for infrared bands from V. Straizys and from the Padova group. - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - *