Change page style: 

Known Issues and Warnings

You are here

The following are known problems or unwanted "features" associated with GNIRS

  • Short Red Camera unavailable: During the GNIRS disassembly in sumer 2012 one of the lenses in the short red camera (0.15"/pixel, 3-5 microns) was found to have cracked. It was concluded that it was too risky to leave the lens in the instrument, and was removed. A replacement lens has arrived but has not yet been installed. In many cases programs designed for use with the short red camera camera can be modified to use the long red camera without significant degradation of the science .

  • Degraded resolution of 32 l/mm grating: When used with the 2 pixel wide slits (i.e., the 0.30" slit + short cameras, or the 0.10" slit + long cameras) the 32 l/mm grating produces spectra with somewhat lower resolution than previously. Values of R for this grating and the 0.3" slit should be ~1800, but are currently ~1300-1400 (see footnote h in the GNIRS spectroscopy table). It is thought that this is caused by slight deformation of the grating due to mechanical strain induced by the grating mount. It will be addressed the next time GNIRS undergoes is warmed up for extensive engineering and maintenance). The degradation is much less with wider slits.

  • Radioactive lenses:  Until Semester 2012B long integrations resulted in frames being peppered with spikes, due to the low- refractive index layers of the anti-reflection (radioative)  thorium coatings on most of the lenses in use in GNIRS contain at the time. The effect was by far most severe when using either short camera, because the last lens in the train is close to the array. During the GNIRS upgrade and maintenance work during the summer and fall of 2012 the last lens in the short blue camera was replaced by one which has non-radioactive coatings. The hit rate on the array with this camera has been reduced by a factor of ~40 compared with with the previous camera, and is now ~ 0.01/sec. The result of each hit is a spike plus halo with dimensions as large as 5x5 pixels. (3x3 and 4x4 are more common). Spikes may be removed during data reduction.

  • Array tilt: The science array is rotated by ~1 degree from the ideal, meaning that in the long slit mode the spectrum is not quite vertical (not parallel to columns) and that wavelengths are not quite horizontal (not parallel to rows).

  • "Odd-even effect": Alternate rows in the GNIRS science array have gains that differ by approximately 10 percent, which results in a high frequency "fuzz" in the raw spectra of continuum sources, flat fields, etc. (spectra run vertically on the array). This flat field, taken in imaging mode with the long blue camera, illustrates the effect. As long as the science spectra and the flat fields are not saturated, the fuzz is removed by flat-fielding.

  • Electronic pattern: Like NIRI, GNIRS' GNAAC controller often superimposes vertical striping, horizontal banding, and small numerical offsets on the data. Occasionally more "random" patterns (e.g., stipple) are present. In the cases of striping, banding, and offsets, a python routine under development, cleanir, originally written for NIRI, is available and has proven to be effective at removing the patterns. Common uses of cleanir with GNIRS files are 'cleanir -rq filename.fits', 'cleanir -fq filename.fits', and 'cleanir -rqf filename.fits'. For other usages see the cleanir web page above. The nvnoise task in the GNIRS IRAF package can also be used. Here is an example of a raw file showing some pattern noise, and the same file after cleaning (in this case using nvnoise).

  • Persistence: Signal or background that fills a substantial fraction of the array well capacity is likely to lead to persistence. This is a common phenomenon with InSb detectors (e.g. Campbell & Thompson 2005). For example, the acquisition of a faint target, which may utilize exposures that half-fill the array with background radiation produces a faint image of the imaging keyhole on the array in subsequent frames, often most easily seen in a subtracted pair of images. Likewise the spectrum of a bright calibration star will leave a faint spectrum on the array. The persistence fades away fairly rapidly. Science data are obtained in an ABBA pattern; a linear decrease in the persistence might be removed if data are obtained using that sequence. However, the drop-off is more like 1/time (see below) and thus in practice there is a faint positive residual image in the first subtracted pair (A-B) and a much fainter residual in the following one (B-A), and so they do not cancel. More detailed information follows.

    A. Characterization of Persistence

    • Persistence appears following any single exposure resulting in signal of ~ 1000 ADU or more. It increases linearly with signal above 1000 ADU. E.g., the persistence in a 120 second exposure immediately following a “flat” of 2000 ADU is ~4 ADU, for a “flat” of 3000 ADU the persistence is ~8 ADU, for a flat of 4000 ADU it is ~12 ADU. For a single flat of 1000 ADU it is ~0.
    • Repeated exposures result in a somewhat larger persistence (by ~30 percent) in subsequent exposures. Repeated exposures at or slightly below the above threshold also create persistence.
    • The “persistence current” decreases with time as ~ 1/t whether or not dark exposures (flushes) are taken. The persistence in a frame is then proportional to log (t2/t1), where t1 and t2 are the start and end times of the frame, both times measured from the end of the last exposure to cause persistence.

    B. Effects of persistence on data and possible remedies/workarounds

    Several types of observations result in signal levels that cause persistence and can deleteriously affect subsequent science frames.

    • In the 1-2.5um region observations of bright standard stars (or bright targets) result in their spectra being imprinted on subsequent frames of faint science targets. The remedy for this is to observe the science target and standard star in different locations along the slit. The OT library templates are set up in this manner.
    • In the L and M bands the sky background is bright enough to result in persistence, which is probably unimportant for L and M band science frames but could affect subsequent JHK faint source science spectroscopy frames. This can be mitigated by appropriate scheduling.
    • Emission lines such as strong sky OH lines in the H and K bands and strong arc lamp lines can leave persistence on subsequent frames such as flats. This is difficult to avoid. The observing templates in the OT library are set up so as to obtain flats prior to arcs. The template flats are designed to produce large signals in order that the relative effect of the sky-line persistence is minimized.
    • Observations of spectral flats result in persistence in subsequent frames. Because flats are "flat" in the spatial direction in most cases the persistence will subtract in subsequent A-B subtracted pairs. Problems may arise when a XD program is followed by a long slit program; this is a scheduling issue.
    • The sky background in faint source acquisition frames can leave persistence in the form of an imprint of the keyhole on the first few science frames. Over much of the persistence region the effect goes away when one subtracts the negative spectrum from the positive spectrum (in A-B subtracted frames); but on the curved edges of the keyhole it does not. The acquisition exposure times in the OT library have been shortened to reduce this effect.

    While the GNIRS team is making efforts to understand and avoid persistence as far as possible, it is an intrinsic property of the detector and cannot be avoided in all circumstances. The practical impact of low-level persistence on many science programmes may also be minimal. Users are encouraged to consider the possible effects on their science data and use the above information to plan accordingly.

  • Flexure: As the telescope tracks across the sky the optics within GNIRS flex due to the change in direction of the gravitational force relative to the optics. This results in small wavelength shifts which are not steady but occasionally occur from one frame to the next, and it contributes to motion of the target relative to the slit. The latter, which is also due to flexure between the instrument and the wavefront sensor, requires occasional re-centering of the target in the slit (the frequency of this is discussed elsewhere). The accumulated wavelength shift is typically ~0.5 pixels in an hour, but probably varies depending on telescope orientation. The shift can be measured by comparing the positions of sky lines in successive frames. The amount of degradation of resolution in one's data depends on the slit width (always at least 2 pixels) and on whether the spectra are shifted prior to summing them.

  • Spectral FWHM of GNIRS+AO observations: Typical spectral FWHMs measured with GNIRS+Altair and the 2 pixel-wide slit during commissioning were around 3-4 pixels rather than the optimal 2 pixels. Because of the complexity of the optics in GNIRS we did not expect diffraction-limited performance.

  • Short Camera ghosting: One of the lenses of the short red camera has no anti-reflection coating. In some configurations using that camera (in long slit mode) a ghost spectrum, roughly 5% of the intensity of the true spectrum has been observed horizontally offset from but parallel to the true spectrum, and also offset vertically in wavelength coverage. Similar ghosting, but at the 1% level, has been observed in the short blue camera in long slit mode when used with the 111 l/mm grating at certain wavelengths. In both cases the offset of the ghost spectrum from the true spectrum varies with the location of the true spectrum. For point sources, as long as the ghost does not closely coincide with the A or B spectrum of the target, the ghosting is harmless. Ghosting does have the potential to slightly interfere with L or M band observations of extended sources.

  • ND100X+H2 ghosting: Bright-target acquisitions using the ND100X+H2 filters exhibit a comet-like tail which extends ~20 pixels above the target with a strength ~3% that of the target.

  • Slit length with Long Red camera: The current optical alignment inside GNIRS combined with the current version of the GNIRS control software does not allow simultaneous availability of the full length of the long slit (49") with both the long blue and long red cameras. We have opted to make the full long slit available with the long blue camera. With the long red camera the slit length is then ~45". If it is scientifically important for your program to have the full 49" slit length available for 3-5 micron spectroscopy, contact one of the GNIRS scientists assigned to the instrument.

  • Bad Pixels:The GNIRS detector contains a few small regions of bad pixels. The bad pixels lie between the XD orders in the most heavily-used low-resolution XD modes (0.05"/pix + 10 l/mm and 0.15"/pix + 32 l/mm). However, the exact placement of the orders in the higher-resolution modes depends on the grating central wavelength. If the presence of these bad pixels could affect your science in these modes, please advise your contact scientist as you prepare your phase II observations. The CS will be able to take arc spectra that show exactly where the bad pixels will lie, and help you adjust your central wavelength to avoid them.

Gemini Observatory Participants