In-orbit calibration concept for the SNAP spectrometer Write by: P.Ferruit, A.Ealet, G.Smadja, A.Bonissent Issue Rev. Paragr. Page Date Observations 1 1.0 All 06/04/05 Creation Table of contents 1. Document purpose and scope 6 2. The radiometric calibration of the instrument 6 2.1.Overview 6 2.2.Flat-field calibration of the instrument 6 2.2.1.Description 6 2.2.2.Required observations 7 2.2.3.Constraints on the on-board calibration unit 7 2.2.4.Constraints on the spectrometer 8 2.3.Spectro-photometric calibration of the instrument 8 2.3.1.Description 8 2.3.2.Required observations 9 2.3.3.Constraints on the telescope and on the standard stars 9 3. The calibration of the spectrographic stage distortion of the instrument 9 3.1.Description 9 3.2.Required observations 10 3.3.Constraints on the spectrometer 10 4. The wavelength calibration 10 4.1.Description 10 4.2.Required observations 11 4.3.Constraints on the on-board calibration unit 11 4.4.Constraints on the spectrometer 12 5. The on-board calibration unit 12 5.1.Uniformity and smoothness of the illumination 12 5.2.Continuum sources for the flat-fielding of the spectrometer 13 5.3.Line sources for the wavelength calibration of the spectrometer 13 5.4.Implementation inside the spacecraft 13 6. Calibrating the visible-IR overlap region 14 7. Shutters for the spectrometer 14 8. Validating the calibration concept 15 8.1.Numerical simulations of the instrument 15 8.2.A demonstrator for the spectrometer 16 9.The spectrometer detectors 16 9.1.1.Some key detector properties 16 9.1.2.R&D detector development plan 17 10. References 17 1. Document purpose and scope The SNAP spectrometer is an IFU instrument based on the advanced slicer technology. The calibration scenario follows the standard steps of this type of instrument. Nevertheless, the SNAP mission required a spectrograph calibrated at the one percent level over a large range of magnitude for science in one hand and to measure reference stars for the imager in the other hand. This means than each potential source of error should be well identified and controlled at a level which is, in general, better than 1%. 2. The radiometric calibration of the instrument 2.1.Overview The radiometric calibration of the SNAP spectrometer will be performed following the classical two-step procedure with first a flat-field calibration based on the use of internal sources and second, a spectro-photometric calibration based on the observation of standard stars. 2.2.Flat-field calibration of the instrument 2.2.1.Description The aim of the flat-field calibration of the instrument is to map the relative, spaxel-to-spaxel1 response of the instrument as a function of wavelength. Ideally, this calibration should be performed by illuminating the field of view of the instrument uniformly both spatially (uniform illumination) and spectrally (flat spectrum source). In reality, if one can quite often obtain a reasonably uniform illumination of an instrument field of view, it is not possible to have a flat-spectrum source. The flat-field calibration will therefore provide the relative spaxel-to-spaxel response of the instrument, with respect to a reference spectrum, which is the one of the internal lamp used for the calibration. As it is usually not possible to place the source before the telescope (!), this flat-field calibration will only take into account the response of the instrument (including its detector) and not the response of the telescope itself. As the instrument response will change with time (this is especially true for the detectors), it will be necessary to repeat it regularly during the mission. The exact frequency is still TBD. To meet the high-level requirements on the mission (in particular the scientific requirements), the instrument shall be flat-fielded with an accuracy better than 1.1 % (1), assuming that it will be repeated every TBD months. This includes both accuracy of the calibration itself and the changes in the instrument response between two calibrations (assumed to contribute to a level of roughly 0.2 % over TBD months). Note that this calibration will also make use of the results of the on-ground characterization and calibration campaign of the instrument (including of the detectors), when relevant and possible. 2.2.2.Required observations The flat-field calibration of the instrument will only be performed using the internal sources of the on-board calibration unit (ideally the one used for the calibration of the imager) and will not use “pointed” observations (i.e. observation of the astronomical targets). Flat-fielding calibration sequences will have to be obtained every TBD months. They can be obtained simultaneously for the two channels of the spectrometer if allowed by the internal sources. 2.2.3.Constraints on the on-board calibration unit In order to perform the flat-field calibration of the instrument with the requested accuracy, the on-board calibration unit and its sources shall fulfill a set of basic requirements. In the following we give a brief overview of these requirements, with a short explanation of their origin. Later on, these requirements will be gathered in a specification document for the on-board calibration unit. Spatial uniformity (1) – At any wavelength within the useful spectral range of the spectrograph, the uniformity of the illumination provided by the on-board calibration unit shall be better than 1 % (1, TBC) over the complete field of view of the spectrometer. This requirement aims at minimizing the need and the difficulty of a calibration of the distribution of light provided by the on-board calibration unit. Spatial uniformity (2) – At any wavelength within the useful spectral range of the spectrograph, the uniformity of the illumination of the entrance plane of the spectrometer provided by the on-board calibration unit shall be mappable with a second degree polynomial, with an accuracy better than 0.2 % (1, TBC). This requirement aims at ensuring that no significant high-spatial-frequency variations of the illumination is present (it could jeopardize the flat-field calibration) and at minimizing the number of positions within the field of view where a spectro-photometric star will have to be observed. Spectral smoothness – The spectral gradients in the spectrum provided by the on-board calibration unit at the entrance of the spectrometer for the flat-fielding of the instrument shall be less than TBD percent per TBD nm (1), over the complete useful spectral range of the spectrometer. This requirement aims at avoiding high-frequency spectral gradients in the input spectrum that will, at best, make the flat-fielding of the instrument difficult and, at worst, make it impossible. Photon rate – The number of photons per second provided per spaxel by the on-board calibration unit at the entrance of the instrument, when used for its flat-field calibration, shall be large enough to reach a signal to noise larger than 100 in less than 100 s (TBC) for all wavelengths within the useful wavelength range of the instrument. This shall be possible without saturating the detector. This requirement aims minimizing the time spent for the flat-fielding of the instrument (overall operational efficiency of the spectrograph). Stability (1) – At any wavelength within the useful spectral range of the spectrometer and over its complete field of view, the spatial illumination of the entrance plane of the spectrometer provided by the on-board calibration unit shall be stable to better than TBD % over a duration of TBD months (1). This requirement aims at minimizing the frequency of the observation of the standard spectro-photometric stars. Stability (2) – At any wavelength within the useful spectral range of the spectrometer, the average over the complete field of view of the spectrum of the illumination provided by the on-board calibration unit at the entrance of the spectrometer shall be stable to better than TBD % over a duration of TBD months (1). This requirement aims at controlling the degradation of the illumination over the duration of the mission. 2.2.4.Constraints on the spectrometer We can also infer a set of requirements for the spectrometer itself. As in Sect. 2.2.3, we give in the following a brief overview of these requirements and of their origin. Spectral smoothness – The spectral gradients in the radiometric response of the spectrometer shall be less than TBD percent per TBD nm (1), over the complete useful spectral range of the spectrometer. This requirement aims at avoiding high-frequency spectral gradients in the instrument radiometric response that will, at best, make the flat-fielding of the instrument difficult and, at worst, make it impossible. Stability – At any wavelength within the useful spectral range of the spectrometer and over its complete field of view, the radiometric response of the spectrometer shall be stable to better than TBD % over a duration of TBD months (1). This requirement aims at minimizing the frequency of the flat-field calibrations. 2.3.Spectro-photometric calibration of the instrument 2.3.1.Description The spectro-photometric calibration of the instrument will be performed by observing standard, spectro-photometric star, the absolute spectrum of which is accurately known. It will provide us with the absolute and relative radiometric response of each spaxel of the spectrometer at all wavelengths within the useful spectral range of the instrument. By comparing the absolute (i.e. in physical units) spectrum of reference stars with the observed, flat-fielded spectrum obtained with the instrument, the spectro-photometric calibration of the instrument allows to: remove the residual “signatures” of the on-board calibration unit, namely the signature of its non-flat spectrum and the signature of its “non-uniform” illumination (finishing the relative radiometric calibration), compute the transfer of the spectrum into “physical” units (absolute radiometric calibration). The standard stars will be observed at different positions within the field of view to account for changes in the response of the instrument over its field of view, as well as for the non-uniformity of the illumination of the internal sources used for its flat-fielding. It may also be necessary to use a small dithering procedure to average out changes in the diffraction losses as a function of the centering of the star within a given slice. The accuracy of the spectro-photometric calibration of the instrument shall be better than TBD % (1) over the complete field of view of the instrument and over its complete useful spectral range. 2.3.2.Required observations The spectro-photometric calibration of the instrument will require “pointed” observations of a set of standard spectro-photometric stars. Each star will be observed at different locations within the spectrometer field of view and it may also be necessary to use a small dithering procedure. These observations will have to be repeated every TBD months. 2.3.3.Constraints on the telescope and on the standard stars Knowledge of the absolute spectrum of the standard stars – At any wavelength within the useful spectral range of the spectrometer the absolute spectrum of the spectro-photometric standard stars used for the spectro-photometric calibration of the instrument, shall be know to better than 1 % (1, TBC). This requirement comes from the fact that a poor knowledge of the reference spectrum used in the spectro-photometric calibration will have a direct impact on the accuracy of this latter. Spatial uniformity – At any wavelength within the useful spectral range of the spectrometer and over its complete field of view, the radiometric response of the telescope shall be spatially uniform to better than TBD % (1). This part of the optical path is not included when using the internal sources for the flat-fielding of the instrument. It will not be calibrated and must therefore be controlled. Stability – At any wavelength within the useful spectral range of the spectrograph and over its complete field of view, the radiometric response of the telescope shall be stable to better than TBD % (1) over TBD months. This requirement aims at minimizing the frequency of the observation of spectro-photometric standard stars. 3. The calibration of the spectrographic stage distortion of the instrument 3.1.Description Due to the distortion in the spectrograph, the spectra coming from a given spaxel are not straight lines on the detector. This spectrographic stage distortion must be corrected either prior to or at the same time than the wavelength calibration is performed. This will be done by measuring the lateral shift of the spectrum of a point-like continuum source (i.e. typically a star with a continuum as featureless as possible). These measurements are repeated for different locations of the source within the field of view of the spectrometer and they are used to fit an analytical model of the distortion that will be used to predict the distortion for all possible source locations within the field of view. The accuracy of the spectrographic stage distortion calibration of the instrument shall be better than 0.1 detector pixel (1, TBC) over the complete field of view of the instrument and over its complete useful spectral range. This includes any relevant stability terms. 3.2.Required observations The calibration of the distortion of the spectrographic stage of the instrument will require “pointed” observations of a star with a relatively featureless continuum (a field with several, well-separated stars would be even better). The observation will be repeated for various source locations within the field of view. Note that the spectro-photometric stars used for the radiometric calibration of the instrument are usually hot and have a very featureless continuum. The exposures obtained for the spectro-photometric calibration of the instrument could therefore be used also for the calibration of the distortion. These observations will be repeated every TBD months. 3.3.Constraints on the spectrometer “Smoothness” of the distortion – The lateral (perpendicular to the dispersion direction) distortion of the spectrographic stage of the instrument shall be smooth that, at any given wavelength within the useful spectral range of the instrument, it can be described by a polynomial of degree lower than 5 (function of the position in the field of view) with an accuracy better than TBD detector pixel (1). This requirement aims at minimizing the number of observations necessary to calibrate correctly the distortion (how many source positions have to be explored before obtaining an accurate description of the distortion). Stability – For all source positions within the spectrometer field of view, the lateral (perpendicular to the dispersion direction) distortion generated by the spectrographic stage of the instrument shall not change by more than TBD detector pixel (1) over TBD months. This requirement aims at minimizing the frequency of the distortion calibration observations. 4. The wavelength calibration 4.1.Description The wavelength calibration of the spectrometer is similar to the classical one used for long-slit observations (each pseudo-slit of the sliced can be considered as a small “long-slit”). We plan to use dedicated sources in the on-board calibration unit. These sources will be “line sources”, with a set of spectrally well-separated lines distributed over the complete wavelength range of the instrument. The intrinsic wavelength of all these lines will have to be known very accurately. Note that it is not necessary for the “line” to be spectrally unresolved (it might even turn out to be advantageous to have slightly resolved lines to increase the accuracy of the determination of their centroid). The accuracy of the wavelength calibration of the instrument shall be better than 0.2 detector pixel (1, TBC) over the complete field of view of the instrument and over its complete useful spectral range. This includes any relevant stability terms. 4.2.Required observations The wavelength calibration of the instrument will not require “pointed” observations, it will use dedicated line sources that will (hopefully) be present in the on-board calibration unit. They will have to be repeated every TBD months. Depending on the availability of line sources suitable for the complete wavelength range, the wavelength calibration of the two channels of the spectrograph could be performed simultaneously. 4.3.Constraints on the on-board calibration unit In order to perform the wavelength calibration of the instrument with the requested accuracy, the on-board calibration unit and its sources shall fulfill a set of basic requirements, which are different from those inferred for the flat-field calibration. Spectral lines (1) – The line source(s) of the on-board calibration unit shall provide a minimum of 20 (TBC) well-separated, high-contrasts line over the complete useful wavelength range of the instrument. This requirement aims at ensuring that the number of line that will be used as references for the wavelength calibration of the instrument is large enough to ensure an accurate determination of the calibration relation over the complete useful wavelength range of the instrument. Spectral lines (2) – The line source(s) of the on-board calibration unit shall have FWHM (full width at half maximum) smaller than 15 (TBC) detector pixels. This requirement aims at making sure that the lines (that can be spectrally resolved) are not too large. Spectral lines (3) – The line source(s) of the on-board calibration unit shall have peak intensities (inferred from their integrated flux and their FWHM assuming a simple Gaussian profile) differing by no more than a factor 5 (TBC). This requirement aims at making sure that if in order to reach a minimum signal to noise of 100 on the peak of a given line, we do not saturate any other line (i.e. all lines can be used for the wavelength calibration). Spatial uniformity – At any wavelength within the useful spectral range of the spectrograph, the uniformity of the illumination provided by the on-board calibration unit when using “line” sources shall be better than 10 % (1, TBC) over the complete field of view of the spectrometer. This requirement aims at ensuring that, for a given integration time, there are no significant differences in signal to noise on within the field of view due to the non-uniformity of the illumination provided by the on-board calibration unit. Note that this requirement is automatically fulfilled if the on-board calibration unit fulfills the equivalent requirement for the flat-field calibration of the instrument. Photon rate – The number of photons per second provided per spaxel by the on-board calibration unit at the entrance of the instrument, when used for its wavelength calibration, shall be large enough to reach a peak signal to noise larger than 100 in less than 20 s (TBC). This shall be possible without saturating the detector at any wavelength within the useful spectral range of the instrument. This requirement aims at ensuring that the signal to noise on the lines is high enough to accurately measure their centroid and also that it is obtained quickly enough (the need for long calibration exposures would reduce the overall operational efficiency of the spectrometer). Stability and knowledge of the position of the lines – At any time during the mission, the intrinsic central wavelength of a given line used for the wavelength calibration of the instrument shall be know to better than 1/40th of a detector pixel (1, TBC). This requirement is easily and automatically fulfilled for an emission-line source (classical Neon lamp as an example) but this may not be the case for other types of “line” sources. 4.4.Constraints on the spectrometer We can also infer a set of requirements for the spectrometer itself. Smoothness of the calibration relation – The calibration relation (relation providing the wavelength as a function of position in a spectrum) of the spectrometer shall (spectrally) smooth enough so that, for any spaxel and over the complete useful spectral range of the instrument, it can be described by a polynomial of degree lower than 5 (TBC) with an accuracy better than 0.02 detector pixels (1, TBC). This requirement aims at making sure that the requested number of lines for the sources is high enough to accurately constrain the fit of the calibration relation for all spaxels. Stability – At any wavelength within the useful spectral range of the spectrometer and over its complete field of view, the calibration relation (relation providing the wavelength as a function of position in a spectrum) of the spectrometer shall be stable to better than TBD % over a duration of TBD months (1). This requirement aims at minimizing the frequency of the wavelength calibrations. 5. The on-board calibration unit It is currently foreseen that the SNAP spectrometer will use the same on-board calibration unit than the imager. However, the spectrometer has needs that can be very different from the ones of the imager. It is therefore necessary to make sure the on-board calibration unit can be used both for the imager and the spectrometer. In the previous sections, we identified requirements on the on-board calibration unit originating from various calibrations. In the following we discuss the impact of some of them on various characteristics of the on-board calibration unit (in particular its sources). 5.1.Uniformity and smoothness of the illumination When discussing the flat-field calibration of the spectrometer, we have identified a very stringent requirement on the uniformity (less than 1 %) and smoothness of the illumination provided by the on-board calibration unit at the entrance of the spectrometer. It is not yet clear if the current design of the on-board calibration unit allows meeting this requirement (it is a stringent requirement but it applies to a very small field of view compared to that of the imager). At the end, this may push toward the use of an integrating sphere if possible. 5.2.Continuum sources for the flat-fielding of the spectrometer We have also identified a set of requirements for the sources that will be used for the flat-field calibration of the instrument. One major difficulty is probably the need to have a fairly uniform photon rate over such a large wavelength domain (we want to be able to reach a good signal to noise in the dimmest parts of the spectral range without saturating the brightest ones). It is not yet clear if this is possible with a classical tungsten-filament lamp or if it will be necessary to use LED. In particular, to be able to use tungsten filament lamps in the blue, it is necessary to use high filament temperatures and this may strongly degrade their stability and decrease their lifetime. The final solution may be to use a combination of different lamps and LED. It is not clear yet if it will be possible to use the same sources than the imager (this would of course be the ideal scenario). 5.3.Line sources for the wavelength calibration of the spectrometer The need of “line” sources is clearly specific to the spectrometer and is not present for the imager. It will therefore be necessary to have dedicated sources for the wavelength calibration. We have identified a set of different possible solutions for these line sources: Actual “emission-line” lamps – These lamps are classically used on the ground for the calibration of spectrometers. The difficulty is usually to have strong, unblended lines over the complete spectral range of the instrument. In particular, this is an issue at the blue and red wavelength ends. “Pérot-Fabry” sources – An second possible solution is to use a classical continuum source and combine it with a Pérot-Fabry type of filter (i.e. displaying regularly spaced transmission peaks). The difficulty in this case is to design such a Fabry-Perot filter over such a large wavelength range and to find a suitable continuum source (to have enough photons for every line, without saturating others). Note also that in this case the wavelength of the peaks will depend (slightly) on the temperature of the filter that will therefore have to be controlled and/or known. LED – It is not yet clear if a set of LED could be used as single line sources. 5.4.Implementation inside the spacecraft The spectrograph is placed on the back side of the focal plan and the entrance pupil is in the middle of the plan (see picture …). The baseline for the calibration is to use the same on-board calibration unit as the imager. However, we have identified some caveats to be solved. Ensure first an illumination of the entrance of the spectrograph (which is not the case actually) and secondly ensure than the imager lamps fulfill the requirements of section 5.1 and 5.2. If this solution is implemented, we can also study the implementation of the lines sources in front of the illumination system. If it is not possible, we propose to have an internal calibration unit for the spectrometer, using a screen shutter to project the light in the spectrometer. In this case, the spectrometer will have an independent calibration unit for both flat fielfing and wavelength calibration. 6. Calibrating the visible-IR overlap region In order to cover the complete spectral range running from 0.3 to 1.7 µm, the SNAP spectrometer will have two different arms and will use a dichroic to divert light to each channel. There will be a small overlap region between the blue and red channels of the instrument. The exact location of this region that is foreseen to be somewhere between 0.8 and 1.1 µm is still TBD. It will mainly depend on the properties of the detectors, namely the fringing and efficiency toward the red for the visible detectors and the efficiency toward the blue for the infrared detectors. It has however already identified that a good radiometric calibration of this overlap region will be needed to make sure we have consistent radiometric measurements between the blue and red channels. Indeed, the transmission curve of the dichroic can display oscillations in the overlap region. These oscillations will typically get stronger if the transition is sharp and they could harm the radiometric calibration of the two channels in this region. 7. Shutters for the spectrometer When acquiring scientific exposures and for most calibrations, the spectrometer will not require any shutter and will make use of a so-called “electronic shutter”. This is implemented in two very different ways in the blue (visible) and red (infrared) channels of the spectrometer. In the visible (CCD detector) we will make use frame-transfer technics to move the charges into a non-illuminated region of the detector at the end of the exposure. In the near-infrared (IR-type detector), the detector will actually be reading permanently and as there is a direct addressing of each “pixel” there is not worry for differential exposure time effects during the read-out of the detector. These types of “electronic shutters” are appropriate for the long exposure times foreseen for the scientific operation of the instrument. However two others specific needs have been identified that cannot be fulfilled with the current “electronic shutter” baseline. Measurement of the dark current – we need to guarantee that no photon reach the detectors during this measurement. This requires an external shutter. No special constraints are required neither on its speed (it can be very slow) nor on the reproducibility of exposure time. A possible extension of the shutters implemented at the Cassegrain focus of the telescope for the imager can be studied and would probably be adequate. Another solution is an internal shutter for the spectrograph but it would add a mechanism. Radiometric calibration of the fundamental spectro-photometric standard stars -- In the initial phase of the mission, the SNAP spectrometer can be used to transfer the calibration from Vega (magnitude 0) to fundamental standards (magnitude of order 12-15) and then to primary stars (magnitude 16-19), to establish a catalog of calibrators for the imager. It increases the dynamic range the spectrometer needs to accommodate and will require very short exposure time at the level of the ms. Dedicated electronic shutters, running under special modes can be implemented and are described in RD01 (put reference). Preliminary studies show that the required precision can be achieved. This need to be tested more precisely by conducting specific studies using lab prototypes (as planned during the R&D period). 8. Validating the calibration concept To validate the procedure and ensure the level of accuracy, we have undertaken a R&D program based on both an end-to-end simulation of the instrument at pixel level and on the construction of a prototype spectrograph on an optical bench to verify the optical performances both in visible and infrared. The combination of the output of the simulations and of the test results will allow us to validate not only the optical design of the spectrometer but also its calibration concept. It will also allow us to refine the procedures and refine its specifications. 8.1.Numerical simulations of the instrument A simulation has been implemented in the SNAP JAVA framework with the simulation group. It is a full simulation of the optical system based on Fourier optics. It is coupled to the detailed design of the instrument via Zernike coefficients produced by the Zemax optical design program. A parameterization based on shapelets decomposition of the PSF has been developed. Shapelets coefficients are interpolated on a continuous grid using a neural network. The output is a discrete PSF at the detector level, for a monochromatic point source at a given position within the field of view. It can be used to simulate a complete SN spectrum in the 0.35-1.7 um wavelength range (see Figure 8.1-8.1.1), and to verify that the expected performances can be reached. In particular all optical effect, losses or default can be implemented and their effects evaluated in a complete way. Figure 8.1‑8.1.1: Simulated detector exposure for the observation of a SN and its host galaxy. Simulation performed using the end-to-end simulator. 8.2.A demonstrator for the spectrometer We are preparing a demonstrator for 2006, focusing on the optical performances of the image slicer and on the validation of the calibration concept for this type of IFU. We plan to test it not only in the visible, but also in the NIR (using a Rockwell detector). The demonstrator will allow us to: reconstruct an accurate PSF, spatial and spectral, over the full wavelength range (visible and IR) control the optical calibration control distortions control diffraction losses control stray light test dithering procedures, validate the simulation and help to correctly model the instrument response. The demonstrator is currently being designed and we will start manufacturing at the end of the year. 9.The spectrometer detectors In addition there is on-going work on the detectors that may have a significant impact on the calibration concept (the actual detector properties will have a significant impact on the instrumental parameters and the removal of the detector signatures will also depend on the actual detector properties). The base line is to use a LBNL CCD detector for the visible arm and the same Rockwell HgCdTe detector than the imager for the infrared arm. We will face the same problems for calibration but we remain here the properties which are crucial for the spectrometer and the plan test for specific studies. 9.1.1.Some key detector properties We describe here briefly several physical effects (which are actually common to the imager and the spectrometer) that will have to be investigated and monitored thoroughly before and during flight (some of them will be associated with key steps of the removal of the detector signature in the data). Remanence -- The choice of the operating temperature will imply a trade-off between the dark current level and the remanence effects after an intense illumination. The forthcoming tests will evaluate quantitatively the temperature dependence of the remanence. As the fluxes are lower in the spectrograph, the nominal choice of operating temperature might be different than for the imager. Non linearity -- We do not anticipate any specific problem in the evaluation of the non linearity of the detector and readout electronic responses. The possibility of distinguishing both contributions would imply the inclusion of a test signal on some of the reference unconnected pixels. These pixels may be less suitable for the common mode noise subtraction if such a configuration is implemented. Light signals of varying duration seem the easiest way of ensuring a linear variation, but monitoring photodiodes would help control secondary effects, such as an increased temperature for longer signals. Fringing -- It is assumed at present that the CCD thickness is limited to 100 µm in the spectrograph, in order to lower the number of pixels affected by a cosmic ray impacts.  Some residual fringing oscillation might still be observed beyond 0.8 µm. This will have to be measured during the ground tests of the spectrometer. The fringing corrections may require a detailed understanding of the detector, as shown in the HST analysis. Given the complexity of implementing fringing corrections, the level of fringing of a given detector will be an important consideration in the choice of the location (in wavelength) of the overlap region between the blue and red channels of the spectrograph. 9.1.2.R&D detector development plan Some specific programs for testing detectors both in the visible and the NIR will be conducted, (mainly in US, by the GSFC group). They will do some specific trade off studies: -  test for QE, noise HgCdTe detectors  and test single pixel and single row readout for bright star in the spectrometer Test the photometric precision of fast electronic shuttering of  CCD Test cascade CCDs slow and cold for photon counting for faint supernovae with visible spectrograph. Radiation testing of  p-channel CCDs Trade off between higher red QE of p-channel CCDs and lower cosmic ray effects of n-channel CCDs. 10. References Spectrograph calibration short exposure times A.Bonissent, A.Ealet,G.Smadja internal SNQP note Decembre 2004 Calibration of the SNAP IFU spectrometer, error budget P.Ferruit, internal SNAP note April 2003 Spectrograph calibration A.Bonissent, A.Ealet,G.Smadja internal SNAP note Novembre 2002