OSU College of Arts and Sciences Department of Astronomy

MODS Instrument Calibration

This page provides links to MODS instrument calibration files and procedures.

This is a work in progress as we transition from commissioning to science observing. More coming soon...

Contents
Overscan Bias and 2D (Zero) Bias
Dark Frames
Flat Fields
Wavelength Calibration
Spectrophotometric Calibration - Updated 2014 May 28
LBT Model Atmospheric Extinction Curve
MODS1 Instrumental Sensitivity - May 2013 (15k PDF)

Overscan Bias and 2D (Zero) Bias

Overscan Bias

write up in progres...

2D Bias (aka "Zero") Images

MODS science CCD biases are very stable and only need to be obtained at once per run in each of the major CCD region-of-interest readout modes. Ten (10) biases provide sufficient signal-to-noise when median combined for most applications.

At the present (2011 Nov 9), 2D Biases are not required for full-frame (8Kx3K) readout as the prescan columns remove most of the bias term without signifcant residual 2D bias structure. Imaging- (3Kx3K) and Prism- (4Kx3K) mode images still require separate bias frames until we get overscan working for subframe readout on the CCDs (a stubborn bug we haven't managed to root out yet), and it remains to be seen if subframe readout introduces significant residual 2D bias structure in the images (we suspect not, but sometimes you fix one problem and create another...).

2D Bias Scripts

Each of these setup and take ten (10) 2D Bias frames in the required format.
Grating Mode (8Kx3K full-frame): bias8K.cal
Imaging Mode (3Kx3K): bias3K.cal
Prism Mode (4Kx3K): prbias.cal

Dark Frames

Measured dark rates on both the red- and blue-channel CCDs are below 0.5 e-/pixel/hour, so we have determined that taking explicit dark frames is not an indicated calibration step with MODS.

No scripts are provided for taking dark frames.


Flat Fields

Updated: 2013 Feb 7

Overview

Flat fields appear to be very stable for a run, so if you have a lot of MOS targets, you don't need to take flats every night. We found that a single set of master pixel and slit flats works for an entire week-long run without problems, and we have used flats between runs separated by a month in a pinch (but we don't recommend it).

All internal flat fields should be taken with the telescope stationary and pointed at the Zenith. They should not be taken while the telescope is moving. The instrument must be dark (closed hatch in "Calibration Mode"), and with all dome enclousure lights turned OFF. We still have some unresolved light leaks in the instrument near the mounting point with the instrument rotator that will be fixed when MODS is next off the telescope (Summer 2012), but our leak mitigation measures may not be suitable in all cases.

We have no indication that flats need to be taken "at position" on a target as the combination of the general stability of the instrument and the image motion compensation system (IMCS) obviates the need for in-place flats. Properly reduced slit flats can be used, with a small shift, to remove fringing at the far red end of the Red Channel range for very red targets, but because the fringe amplitude is at most 2% peak-to-trough, the fringe pattern is only visible in data at very high Signal-to-Noise Ratios.

Pixel-to-Pixel Spectral Flats

The main color-free component of flat-field structure is pixel-to-pixel variation in gain, including gain differences between the quadrants of the device (each CCD is readout through 8 channels: 4 amplifiers with two video channels per amplifier, each with its own bias and gain).

The procedure that works best for grating flats is one we have adapted from the practice with the Keck LRIS instrument (link...) by taking slitless lamp flats with the grating modes. These capture enough light in the UV and far-red ends of the spectral range to work well. The blue flats use a combination of the clear and UG5 filter, the latter to suppress the red-end of the spectra in favor of more blue light (all flat-field lamps are intrinsically very red).

While we have had success taking twilight slitless flats in the red, they are not demonstrably better than internal lamp flats and are not worth the pain of chasing the twilight exposure times (better to use that time for slit illumination correction frames or twilight imaging flats). Blue twilight slitless flats, however, have too much UV in the wrong places (the sky is too blue), and there produce unacceptable pixel-to-pixel artifacts due to the annealing features on the blue CCD that become prominent at UV wavelengths. We therefore do not recommend taking slitless twilight flats in any mode.

The slitless flats are used to create normalized pixel flats: bias-subtracted frames are stacked to remove any cosmic rays, then the color term is divided out to leave just the major pixel-to-pixel variations. We then flat-field any slit flats with the pixel flat, and use that as the basis for any remaining mask-dependent calibrations (e.g., removing fringing, dichroic transmission wiggles, etc). Spectra of standard stars will also serve the same purpose (and eliminate a step).

Prism flats are tricky because of the large spectral pixels. Shows one crack at this. We use the 0.3-arcsec slit to attenuate the calibration light for long-slit mode. For MOS, you can use the stock scripts as templates but must tune the exposure times to avoid saturation. Saturation is a big issue with wider slits and we don't have a good solution yet (maybe a stronger ND filter is a future option).

Spectral Slit Flats

We recommend acquiring a few (5 each) of lamp flats in the individual long-slits in the red for use as fringe corrector frames in the far red end of the spectrum. Blue slit flats, by contrast, do not appear to be useful, but scripts with useful lamp, filter, and exposure time combinations are provided.

The MODS Red-Channel CCD is a thick (40μm) deep-depletion device, so fringing is generally small (2-3% max peak-to-trough amplitude beyond 8500Å) compared to typical CCDs, and should not constitute a major correction for most faint targets, though it will become an issue when the signal to noise is in the high 10s in that part of the spectrum.

Imaging Flats

Imaging flats taken with the variable-intensity flat field ("vflat") lamp work well, but have a 2% top-to-bottom gradient from the calibration illumination system. Twilight sky flats taken as illumination correction frames take out this gradient without problems. Imaging flats are very stable on run timescales because the filters are way out of focus (right in front of the CCD field flattener lenses on the dewars), they are rarely removed or handled, and so are in a protected and thus relatively clean environemnt.

There are no lamp-and-filter combinations that permit simultaneous acquisition of blue and red flats, so we take them serially with the scripts below. The dichroic slightly cuts into the red end of the g filter bandpass, and there is a small change in bandpass shape for the u and r filters, so we take them explicitly for red-only and blue-only modes.

As with other flat fields, lamp flats should be taken at the Zenith with the telescope stationary.

Twilight Sky Flats

Twilight sky (evening or dawn) flats are often used for (1) slit illumination corrections, (2) multi-slit intercalibration corrections, or (3) imaging-mode sky flats.
Slit Illumination Corrections:
Spectroscopic programs that need to use the entire long slit (e.g., rotation curve of a galaxy that fills most of the science FoV), will likely require at least 1 or 2 twilight spectral flats in red and blue to help perform a high-precision illumination correction. There is a slight (2%) gradient from top-to-bottom due to the internal lamp illumination system that such twilight flats will remove. We have achieved good (<0.5%) sky subtractions in the central 1-arcminute segment of the facility long-slit masks for single targets without using twilight sky flats because the slow gradient is free of structure (the laser-cut slits are extremely clean and parallel: <1% width variations on small and large scales).

MOS Mask Intercalibration Corrections:
Because it is impractical (read: total waste of good telescope time) to put a standard star down the slit of every slitlet on an MOS mask, we recommend using 1 or 2 spectra or images taken through the MOS masks to determine the inter-slit calibration corrections. To first order (and perhaps better than that), we expect that slit-to-slit on a given mask the relative photometric calibration is a gray-shift in intensity because of small variations in slit width. The laser-cut masks are of such high quality (recall that 1-arcsec = 0.6mm at the LBT f/15 focal plane, and slit cut quality is measured as ±1-2μm rms), unless there is a problem with slit cleanliness, the slits are very uniform.

Imaging Sky Flats
Imaging programs will need to take at least one set of twilight sky flats during the course of a run, as the 2% gradient due to the internal calibration lamps will produce poor results if used without a proper twilight sky correction.
We are still in the process of developing scripts for taking twilight sky spectral and imaging flats.

Spectral Flat-Field Calibration Scripts

These are copies of the standard facility calibration scripts that we recommend observers use at the LBT for routine long-slit mode flat-field calibrations in grating and prism modes. Comments at the top of each file explain what each one does.
Grating Slitless ("Pixel") Flats:
Dual Mode: grpixflats.cal
Red-Only: grpixflats_r.cal
Blue-Only: grpixflats_b.cal

Grating Slit Flats:
Dual Long-Slit: slitflats.cal
Red Long-Slit: slitflats_r.cal
Blue Long-Slit: slitflats_b.cal
MOS Example: grmosflats.cal

Prism Slit Flats:
Dual Long-Slit: prflats.cal
Red Long-Slit: prflats_r.cal
Blue Long-Slit: prflats_b.cal
MOS Example: prmosflats.cal - Example

Imaging Internal Flats:
Dual Mode: imflats.cal
Red-Only: imflats_r.cal
Blue-Only: imflats_b.cal
Special SDSS-u Flat Script: imflat_u.cal

All Flat-Field Scripts:
Tar GZip File: flatScripts.tgz
These can be used as templates for creating multi-object slit (MOS) mask flat-field scripts. The example MOS mask script uses a dummy slitmask ID, so do not execute this script at the LBT without modification.

Wavelength Calibration

Updated: 2013 July 3

We recommend taking wavelength calibration lamps through the 0.6-arcsec longslit mask for grating wavelength calibrations, and the 0.3-arcsec longslit mask for the prism flats. For the grating, 0.6-arcsec is the design-reference slit and you will have 80-100 reasonably bright and unblended lines to work with in the red, maybe 50 in the blue, which covers the range of interest.

Wavelength calibrations are very stable, modulo a small flexure shift of order a few pixels at most which is readily measured and removed using night sky lines (another advantage of having an 8.4m telescope). Note that because the red and blue channels are mechanically different relative to gravity below the common focal plane, any amount of shift in wavelength due to absolute instrument flexure that is not corrected for by the IMCS is expected to be different between the red and blue channels.

Only one (1) exposure per lamp or lamp combination is needed: the exposures are very short, and there are many 10s of lines, so even single CR hits are not a problem. Short exposures plus long readout times in full-frame mode mean you can waste a lot of time taking superfluous data. The MODS calibration lamp system's integrating sphere and projection optics are very efficient, and we generally have to observe the comparison lamps through strong neutral density filters (ND1.5) to avoid badly saturating the spectra.

Unlike flat fields and bias frames, it is possible to take comparison lamp spectra during the afternoon with instrument dark and full dome lights on(!), but the telescope must be at the Zenith.

Dispersion Solutions

For the grating modes, using the IRAF identify task as a prototype, we recommend performing a 4th-degree polynomial fit (terms to x4) to the data. The best fit polynomials, linear dispersions, and typical RMS fit residuals are summarized as follows:

Blue Grating
Wavelength Solution Plot
Linear Dispersion: 0.515Å/pixel with typical rms residuals of 0.03Å
Non-Linear Departure: +1 to -2Å

Red Grating
Wavelength Solution Plot
Linear Dispersion: 0.845Å/pixel with typical rms residuals of 0.01Å
Non-Linear Departure: +6 to -3Å
In the wavelength solution plots we show the polynomial term after removal of the nominal linear dispersion in the top panel with the best-fit residulas for the full 4th-degree polynomial fit in the lower panel. Both were measured using the 0.6-arcsec long slit and unbinned pixels.

The best fit RMS deviations for the grating wavelength solutions depend on the width of the slit. We found that a 0.6-arcsec long slit mask gives the best solutions. This is summarized in the table below:

Wavelength Fit RMS and Slit Width
Slit
Width
Blue
RMS
Red
RMS
0.3 0.059Å 0.010Å
0.6 0.024Å 0.008Å
1.0 0.045Å 0.018Å
1.2 0.117Å 0.026Å
The reason is that because MODS has such a sharp line spread function, as you use progressively wider slits the emission lines become more flat-topped (each line is a monochromatic image of the wide slit), degrading the precision with which IRAF and other standard packages measure the line centroids. Similarly, when using the narrowest facility slit mask (0.3-arcsec), the lines start to become marginally sampled (FWHM becomes less than about 3 pixels), and the centroid precision is decreased. The peak is at about 0.6-arcsec slit width.

For the prism modes, the wavelength solutions are dominated by the strong wavelength dependence of the index of refraction of the prism glasses. The blue prism is best fit by a 4th-degree polynomial (terms to x4) whereas the red prism is best fit by a 6th-degree polynomial (terms to x6). Unlike with gratings, the prism dispersion is so nonlinear we cannot characterize the dispersion as a linear term plus polynomial departures from linearity. Thus instead of giving a nominal linear dispersion in Å/pixel, we instead plot the spectral pixel size in Å/pixel as a function of pixel and wavelength for each channel. The results for the MODS prism modes are summarized below:

Blue Prism
Wavelength Solution Plot
Spectral Pixel Size Plot

Red Prism
Wavelength Solution Plot
Spectral Pixel Size Plot
Both were measured using the 0.3-arcsec long slit and unbinned pixels.

Calibration Lamp Files

Links following provide annotated plots of MODS 1D comparison lamp spectra taken in the grating and prism modes. Lines are identified with wavelengths in units of Ångstroms.

Line lists are provided in 1-column ASCII text format (the default format for IRAF). Only lines positively identified in MODS comparison lamp spectra are listed. The wavelengths are taken from the NIST Handbook of Basic Atomic Spectroscopic Data.

Grating Mode

Blue Channel: (PDFs)
Hg(Ar): [ 3500-6000Å ]
Argon: [ 3500-4500Å | 4500-6000Å | All ]
Xe+Kr: [ 3200-4450Å | 4000-5600Å | 4500-6000Å | All ]
All Plots: [ blueLamps.tgz ]

Red Channel: (PDFs)
Argon: [ 5000-7000Å | 7000-8500Å | 8500-10500Å | 6500-10000Å | All
Neon: [ 5000-6100Å | 5600-7800Å | 6800-9600Å | 7800-9800Å | All
Xe+Kr: [ 5000-6500Å | 6500-7500Å | 7500-10000Å | 7500-1000Å Zoom | All
All Plots: [ redLamps.tgz ]

Line Lists: (ASCII)
Blue Channel: [ Hg(Ar) | Argon | Xenon | Krypton ]
Red Channel: [ Neon | Argon | Xenon | Krypton ]
All Lists: [ lineLists.tgz ]

Prism Mode

For prism mode the lower resolutions (R=100-500) with subsequent greater line blending and strongly second-order dispersion solutions make wavelength calibration very challenging. We recommend using the 0.3-arcsec slit for long-slit prism comparison lamps, and not mixing lamps together.

A second effect is that with the larger spectral pixels from the prisms and the finite (if small) leakage through the dichoric in Dual-Channel mode, so you will see significant artifacts beyond the nominal blue- and red- limits of the red- and blue-channels, respectively. This cannot be avoided: prisms have lower dispersion at longer wavelengths, and even with the strong dichroic rejection in the out-of-band regions you still get some leakage from strong sources. Beware especially if trying to horizontally "stack" multi-object slits in prism mode.

Plots of MODS 1D comparison lamp spectra with lines identified are in PDF format are below. These are plotted in two forms: as a function of pixels to help identify the lines in uncalibrated spectra, and as a function of wavelength (lambda).

Blue Channel: (PDFs)
HgAr: [ pixels | lambda ]
Kr: [ pixels | lambda ]
Xe: [ pixels | lambda ]
All Plots: [ bluePrism.tgz ]

Red Channel: (PDFs)
Ar(+HgAr): [ pixels | lambda ]
Ne(+HgAr): [ pixels | lambda ]
Kr: [ pixels | lambda ]
Xe: [ pixels | lambda ]
All Plots: [ redPrism.tgz ]

Line Lists: (ASCII)
Blue Channel: [ Hg(Ar) | Xenon | Krypton ]
Red Channel: [ Neon | Argon | Xenon | Krypton ]
All Lists: [ prismLists.tgz ]
Note that the line lists above are abbreviated lists that only include the well-identified (and unblended) lines for each species shown in the prism comparison lamp spectral plots. While you can use the full line identification tables for the Grating mode, this reduced list will usually give quicker identifications (e.g., in IRAF or IDL) without a lot of initial mis-identifications.

Wavelength Calibration Acquisition Scripts

Updated: 2011 Nov 9

These are copies of the standard facility calibration scripts that we recommend observers use at the LBT for routine long-slit mode wavelength calibrations in grating and prism modes. Comments at the top of each file explain what each one does.

The scripts below have our best determination of good combinations of lamps, exposure times, and blocking filters, and choice of long-slit masks for efficient acquisition of wavelength calibration data in the long-slit modes. These also serve as good templates for multi-object comparison lamp scripts.

Grating Mode:
Dual Long-Slit: grlamps.cal
Red Long-Slit: grlamps_r.cal
Blue Long-Slit: grlamps_b.cal
MOS Example: grmoslamps.cal

Prism Mode:
Dual Long-Slit: prlamps.cal
Red Long-Slit: prlamps_r.cal
Blue Long-Slit: prlamps_b.cal
MOS Example: prmoslamps.cal

All Comp Scripts:
Tar GZip File: compScripts.tgz
These can be used as templates for creating multi-object slit (MOS) mask calibration scripts. The example MOS mask script uses a dummy slitmask ID, so do not execute this script at the LBT without modification.

Spectrophotometric Calibration

Updated: 2014 May 28

Because MODS works from 3200Å to 10000Å, we need to use standard stars with well-determined fluxes from UV to Near-IR. This section describes our preferred set of standard stars, an approximate model atmospheric extinction curve, standard star observing scripts, and our recommended spectrophotometric calibration strategy.

Primary Spectrophotometric Standards

The recommended MODS Primary Spectrophotometric Standard Stars are derived from the HST CALSPEC database. These are a combination of the well-observed northern-hemisphere standards from the list of Oke (1990 AJ, 99, 1621), and the four HST White Dwarf Primary spectrophotometric standards of Bohlin, Colina, & Finlay (1995 AJ, 100, 1316). These latter 4 stars are listed in boldface in the table below.

For each primary standard star we provide flux tables with 10Å sampling. These tables are in IRAF-style ASCII 3-column format ready to be used with IRAF. Where necessary, we used HST stellar models or other HST calibration data to extend the near-IR wavelength coverage to 10500Å. These are suitable for grating-mode calibrations. Wavelengths are vacuum wavelengths as in the original CALSPEC source tables.

We also provide 50Å-sampling flux tables suitable for prism-mode calibrations based on CALSPEC and the Spec50 flux tables (Massey et al. 1988 ApJ, 328, 315) with the Massey & Gronwall (1990 ApJ, 358, 344; M&G90). 50Å tables are not yet available for all primary standards, but will be added as we find suitable data sources.

These tables have the major telluric absorption complexes censored, and we have cut out the brighter stellar absorption lines in most (but not all) tables. This regrettably crowded plot graphically summarizes the primary standard flux table contents. All of these stars are roughly the same color except the redder BD+33°2642, and span about 3 magnitudes in apparent brightness.

The original, uncensored flux tables are available from the HST CALSPEC database.

Note that the modsIDL Reduction Pipeline uses the full CALSPEC flux tables which are included in the software distribution.

MODS Primary Spectrophotometric Standard Stars
StarRADecSp Type m(5556)pmRApmDec t(Grating)t(Prism) Downloads
G191-B2B [1] 05:05:30.613+52:49:51.96 DA011.85 +7.45 -89.54 9010 Finder Chart
Acq Script | Obs Script
10Å Flux Table
50Å Flux Table
GD 71 [2] 05:52:27.614+15:53:13.75 DA113.03 +85 -174 18030 Finder Chart
Acq Script | Obs Script
10Å Flux Table
50Å Flux Table
GD 153 [3] 12:57:02.337+22:01:52:68 DA1 13.35 -46-204 18030 Finder Chart
Acq Script | Obs Script
10Å Flux Table
Feige 34 10:39:36.740+43:06:09.26 sdO 11.25 +14.09-25.01 12015 Finder Chart
Acq Script | Obs Script
10Å Flux Table
50Å Flux Table
Feige 110 23:19:58.398-05:09:56.16 sdO 11.88 -10.68 +0.31 9010 Finder Chart
Acq Script | Obs Script
10Å Flux Table
50Å Flux Table
 J2000J2000  mag mas/yrmas/yr secsec  

Notes on the Standards

Table Notes:
RA & Dec are FK5 coordinates, Equinox J2000, Epoch 2000 from Simbad
pmRA & pmDec are proper motions in mas/yr, FK5 Epoch 2000 from Simbad
t(disp) is the recommended exposure time for disp = Grating and disp = Prism, respectively
Finder Chart is a modsView finder chart with guide star indicated in PNG format
Acq Script is a template MODS target acquisition script. These have the target coordinates, guide star, and recommended PA selected. Guide star names and coordinates are from the USNO-B1 catalog.
Obs Script is a template MODS observing script, including configuration and exposure times for each spectrograph mode. Edit this template to suit.
10Å Flux Table is the HST CALSPEC data binned to 10Å intervals and presented in IRAF-style ASCII format. Where needed additional near-IR HST points have been added using model atmospheres. Table are 3-column: wavelength in Angstroms, AB magnitudes, and the bin size in Angstroms.
50Å Flux Table is the coarse-grained 50Å flux table in IRAF-style ASCII format, with strong absorption features censored, in the same format as the 10Å tables. These tables are good for performing prism-mode flux calibrations. 50Å tables are not yet available for all stars.

Notes on Individual Stars:
[1] - G191-B2B is an HST primary white dwarf star. On the finder chart it is the northernmost of the two bright stars in the field.
[2] - GD71 is an HST primary white dwarf standard stars. The NIR extension is based on HST model spectrum.
[3] - GD153 is an HST primary white dwarf star. The NIR extension is based on the HST fluxes shifted by -0.02mag.

We are in the process of compiling a library of MODS spectra taken in all modes of the instrument which will be available online with the recommended flux tables as we progress with reducing the data and verifying the flux tables.

Secondary Standard Stars

After review of data taken since 2011, we have decided that the four secondary spectrophotometric standard stars (GD140, PG0823, Hiltner 600, and Wolf 1346) are not useful for MODS flux calibration. Their wavelength coverage is too coarse and not blue or red enough for MODS, and the data quality is low compared to the CALSPEC standards. In particular, these standards perform poorly in taking out the "wiggles" in the dichroic transmission function, and perform only marginally better in blue- or red-only spectra with the dichroic out of the beam.

Therefore, as of May 2014 we are withdrawing these standard stars from our recommended list, and removing them from the default calibration program at LBTO.

Legacy copies of the flux tables are available

modsSecondary.tgz (2.5Mb)
to help reduce archival spectra of these stars taken before May 2014.

Model Atmospheric Extinction Curve

At present there is no empirical atmospheric extinction curve for the Mt. Graham site. At an elevation of 3200-meters, Mt. Graham is above Kitt Peak (2100-m) and Paranal (2600-m), below Mauna Kea (4100-m), so simply adopting a published standard extinction curves from these well-characterized sites would result in significant over-/under-estimation of the approximate atmospheric extinction, especially towards bluer wavelengths.

As an interim measure until we can either obtain a good empirical extinction curve for Mt. Graham or generate an atmospheric model for the site, we have created an approximate extinction curve for Mt. Graham using the well-measured KPNO and Paranal extinction curves scaled to an elevation of 3200-m assuming an atmospheric scale-height of 7000-m and combined. This curve is available for download as an IRAF-style ASCII table:

lbtextinct.dat
The scaling ignores elevation-independent components of extinction like high-altitude (>5-10km) suspended aerosols, but that is a minority component and highly variable in any event (e.g., injection of particulates from volcanic eruptions, powerful sand storms, etc.). The table above also does not include time- and position-variable molecular absorption features of water-vapor, O2, and ozone.

Standard Star Observing Scripts

You can download all of the MODS Spectrophotometric standard star observing scripts, finding charts, and flux tables as one tarball:
modsSpecPhot.tgz (6.0Mb)
This includes a copy of the current approximate LBT extinction curve.

Standard Star Observation Strategy

For applications that only require a good response curve, we recommend using standard stars observed with the 5-arcsec spectrophotometric slit mask (LS60x5). This will provide spectra across the full wavelength coverage with minimal losses due to atmospheric dispersion and seeing. Note that resolution in the 5-arcsec slit will be seeing-dependent, but given that the flux calibrations are typically in 10-50Å bins, this will have little impact on the derived response curves.

Standards taken through the long-slit masks are recommended only if you also want to use the standard-star spectra to correct for telluric absorption and are willing to accept some atmospheric dispersion losses at the far blue end. Many of the standard stars are in sparse fields that do not always have convenient guide stars for all position angles. This is not as critical as at near-IR wavelengths (e.g., with LUCI), but is an option.


Updated: 2019 Aug 05 [rwp/osu]