Report on ESO calibration meeting

I attended a meeting at ESO on calibrating ESO instruments and friends; LSST counts as a friend, so I gave a talk on our calibration plans.

There were some interesting talks, many on calibrating spectrographs – a topic near to my heart when I wear my PFS hat. There were some more talks that seemed more relevant to LSST, and this conference summary has been reordered to move them to the top of the list.

		________________________

			 NOTES

		 Robert Lupton the Good
		________________________

Table of Contents


1 Wolfgang Hummel: Monitoring the XSHOOTER response function
… 1.1 XSHOOTER and VLT dataflow
… 1.2 Response curves
… 1.3 New pipeline in 2013
2 Pasquale Hibon: HAWK-I in the era of GRAAL
… 2.1 Instrument description
… 2.2 Using GRAAL (GLAO system)
… 2.3 Calibrations in era of GRAAL
3 Susana Deustua: Photometric calibration of WFC3 on HST
… 3.1 Synthetic Photometry
… 3.2 Evolution of filters
… 3.3 Error budget
4 Giuseppe Altavilla: GAIA spectro-photometric standard grid for absolute calibration
… 4.1 “SPSS” Standards
5 C. Jordi: The Gaia astrometric space mission and its deliveries
6 Elena Masciadri: A demonstrator for an automatic operational system for the optical turbulence forecast for ESO sites (Cerro Paranal and Cerro Armazones)
… 6.1 MOdelling Sites ESO: MOSE
7 James Osborn: Characterising atmospheric turbulence with the Stereo-SCIDAR
8 J. Marin: Water vapor forecasting on Chilean sites
9 Florian Kerber: Quantitative characterisation of sky conditions on Paranal with the microwave radiometer LHATPRO – five years and learning
… 9.1 Spatial homogeneity
… 9.2 Photometric sky quality classification
10 Alain Smette: Molecfit
… 10.1 Introduction
… 10.2 Molecfit
11 Carlos Gonzalez talk: Calibrating Wide Field Surveys
12 Robert Lupton: Calibration aspects of the LSST
13 Wolfram Freudling: Data Reduction and archive at ESO
14 Reinhardt Hanuschik: VLT science data processing
… 14.1 Calibrations
… 14.2 Science
15 Andrea Mehner: XShooter
… 15.1 Intro
… 15.2 Issues:
16 Konrad Tristram: Challenges in the midIR with VISIR
… 16.1 VIZIR
… 16.2 Calibration
17 Miwa Goto: IR Spectroscopy
… 17.1 IR spectroscopy in low PWV
… 17.2 Wavelength Calibration
… 17.3 Flux Calibration
… 17.4 Slit filling
… 17.5 Astrometric Calibration
18 Igor Chilingarian: Is there a place for perfectionism in the NIR data reductions?
… 18.1 NIR spectral pipelines
… 18.2 FIRE
… 18.3 Project: Las Companas Stellar Library
… 18.4 FIRE Bright Source Pipeline
… 18.5 MIRS on Magellan (now MMT)
19 Paul Bristow: Metrology
… 19.1 Metrology
… 19.2 Metrology for VLT
… 19.3 CRIRES+ Calibration
… 19.4 CRIRES+ Metrology
20 Holger Drass: Optimisation of MOONS metrology
… 20.1 MOONS
21 Gaspare Lo Curto: HARPS
… 21.1 Improvements
22 Francesco Pepe: High accuracy wavelength calibrations: ESPRESSO
… 22.1 Spectrograph theory
… 22.2 Wavelength calibration
23 A. Szentgyorgyi: GMT spectrograph G-CLEF
… 23.1 GMT
… 23.2 G-CLEF
… 23.3 RV error budget approach
… 23.4 MOS
… 23.5 Lessons from Hectochelle
24 Yuanjie Wu: Frequency combs for astronomical applications
… 24.1 Frequency comb
… 24.2 Astronomical Combs
25 Gaspari Lo Curto: Laser Frequency Combs in Astronomy, from La Silla to Armazones
… 25.1 Astro laser combs
… 25.2 Mode scrambling
26 Jonathan Smoker: CRIRES and CRIRES+
… 26.1 CRIRES
27 Abner(?) Zapata: Study of Radial Velocity Stability in the AIUC Spectrographs: PUCHEROS and FIDEOS
28 C. Janssen: Molecular line parameters
… 28.1 Ozone
29 Gillian Nave: Wavelength references for the calibration of astronomical telescopes
30 S. Bagnulo: Polarimetric calibration and accuracy: lessons learnt from the existing instrumentation
31 A. Cikota: Performance of the VLT/FORS2 spectropolarimetric mode
32 Jonathan Smoker: A very high signal to noise spectrum towards mu Col with UVES
33 F. Selman (presented by E. Johnston and F. Vogt): MUSE Monitoring and Calibration
… 33.1 MUSE
… 33.2 Pipeline
… 33.3 QC
… 33.4 MUSE Challenges
34 Martin Roth: Visible integral field units, such as MUSE
… 34.1 Flat fielding
… 34.2 Line spread function
35 Eleonora Sani: KMOS
… 35.1 Instrument
… 35.2 Pipeline
36 Trevor Mendel: KMOS calibration and data reduction
… 36.1 Science
… 36.2 Pipeline
… 36.3 Possible pipeline improvements
… 36.4 RHL N.B.
… 36.5 Q
37 Myriam Rodrigues: Sky background correction in multi-fiber spectrographs
… 37.1 ESO
… 37.2 OH
… 37.3 Continuum dominated
… 37.4 Strategies
… 37.5 Questions
38 Pasquale Hibon: Astrometry in MCAO imaging
… 38.1 MCAO
39 Posters
40 Michael Sterzik: Calibration and analysis of the telluric O2 bands
41 Suzanne Ramsay: concluding remarks

1 Wolfgang Hummel: Monitoring the XSHOOTER response function

1.1 XSHOOTER and VLT dataflow


  - operated with calibration plan
  - quality control loop
  - certified master calibs [RHL but Andrea Mehner was unhappy about
    them]
  - DRP uses models + polynomials

  Produces science products
  - certified pipeline
  - stable instrument
  - benefit to user (e.g. not simple imaging)


1.2 Response curves
~~~~~~~~~~~~~~~~~~~

  - efficiency (model divided by data with model of the blaze) -- no
    flat field
  - response (counts to physical units) -- need to know lamp spectrum


1.3 New pipeline in 2013
~~~~~~~~~~~~~~~~~~~~~~~~

  - per-night: closer to the data that was actually taken
  - master averaged over many nights (clip non-photometric nights,
    correct extinction)
    - seems naive to RHL -- assume tonight = master*(a + B lambda)
    - monitor the b coefficient with time (observe changes in lamps)
    - lamp replacements not recorded in headers (only in problem report
      system)

  Q: the efficiency drops as the M1 degrades, but cleaning the mirror
  doesn't get back to new.  Is the detector degrading?  A: I don't know


2 Pasquale Hibon: HAWK-I in the era of GRAAL
============================================

2.1 Instrument description
~~~~~~~~~~~~~~~~~~~~~~~~~~

  - IR imager 0.9-2.5microns
  - 0.1" pixels
  - 7.5'x7.5'
  - 4 H2RG detectors
  - daily darks, flats
  - rely on twilight flats
    - good to 3%
  - photometric standards
    - every night
    - colour/extinction once a month


2.2 Using GRAAL (GLAO system)
  • half of turbulence below 1km
  • four lasers
  • expect integration time gain of 1.5 - 2 (20% seeing improvement “the
    vast majority of the time”;
    • 50% EE circle 0.7 -> 0.55 uniform to 7’
  • GRAAL handles everything below 300m (60% of time 60% of turbulence)
  • provisionally accepted mid-2018

2.3 Calibrations in era of GRAAL


  - distortion
    - use a globular cluster
  - flexure
  - photometric calibration
    - use 2MASS fields (but not all sky)
    - also used for FWHM monitoring (?)
  - new pipeline from CASU much better


3 Susana Deustua: Photometric calibration of WFC3 on HST
========================================================

  WFC3
  - 2 4kx4k e2v, 0.04"/pix
  - 1 1kx1k H1RG, 0.13"/pix
  - W and D lamps

  Calibrations
  - 70 science orbits per cycle for calibration (1550 sun-side orbits)
  - UVIS: bias, dark, gain, "bowtie", annealling, CTE, sink pixels,
    traps
  - IR: linearity, dark, gain, persistence (mostly ignored in this talk)
  - flux calibrations using calspec standard (Bohlin)
  - flats (W or D)
  - contamination (repeat observations of a star)
  - new things (e.g. non-linearity tests, new techniques, model testing)

  12 e/pixels enough to deal with CTE -- post-flash if necessary

  - ISR (including CTE)
  - AstroDrizzle combines images from a visit (pointing?)
  - Calibration products (including e.g. drizzle info): FLT FLC DRZ DRC
  - Changed how CCDs calibrated.
    - were treated as one detector
    - now treated separately
      - but QE is different, so bandpasses are different (ACS got lucky,
        two chips from same wafer)
      - normalise the flats separately to median of CCD
      - standard stars in corner of each CCD to normalise things
      - apply CTE corrections to flats
        - improve statistics
      - move stars around the arrays; improved spatial variation


3.1 Synthetic Photometry
~~~~~~~~~~~~~~~~~~~~~~~~

  Define band-weighted flux as "inverse sensitivity" (Bohlin 2014)
  - Propagate all "known" parameters and e.g. measured gain
    - Measured is 20% larger than predicted...
    - 3% photometric errors due to drizzle clipping CRs (RHL: classic
      problem...)


3.2 Evolution of filters
~~~~~~~~~~~~~~~~~~~~~~~~

  - Gosmeyer et al. 2015 (1%) (maybe just throughput)


3.3 Error budget
~~~~~~~~~~~~~~~~

  - 0.7 in Vega
  - WD: <1% in vis, ~2% in IR

  Absolute accuracy:
  - Poisson 0.2%
  - flats 0.4-0.6%
  - repeatibility: 0.2-0.5%
  - Other: 0.5-1%
  - total 1.3% (stat) + 1.22% (systematic)
  Relative accuracy:
  - c. 1%


4 Giuseppe Altavilla: GAIA spectro-photometric standard grid for absolute calibration
=====================================================================================

  Cool overlay of 84k Gaia objects on HST image of Cat's Eye Nebula


4.1 "SPSS" Standards
~~~~~~~~~~~~~~~~~~~~

  photometric standards provided by CU5-DR13 (Bologna)
  - Pancino et al. 2012 MNRAS 426 1767

  Dispersion sufficient to mix wavelengths (given width of LSF) 100-200
  calibrators to model instrument
  - internal few mmag
  - external 1% (?)
  - Catalogue archive is GAIA.SPSS (at ASDC)
    - 94 stars available; total 205
    - V = 9-15 (peak around 13)
    - O-M + dA (+ ?)
    - internal use only
    - HST has 374 R ~ 1000 spectrophotometry stars

  Still working on ground-based spectral reductions

  Absolute photometry
  - shape is correct, but fix ZP
  - tied to calspec standards

  GDR2: late 2017


5 C. Jordi: The Gaia astrometric space mission and its deliveries
=================================================================

  GDR2: T_eff, A_V


6 Elena Masciadri: A demonstrator for an automatic operational system for the optical turbulence forecast for ESO sites (Cerro Paranal and Cerro Armazones)
===========================================================================================================================================================

  Meso-scale models solve the NS equations to predict $C_N^2$


6.1 MOdelling Sites ESO: MOSE
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

  Applied to Paranal and ELT

  Grid models;
  - 3, innermost resolution 500m
  - 5, innermost resolution 100m
    - needed to resolve shape of summit
  - vertical: 62m, delta h = 5m logarithmically growing up to 3500m,
    then 600m
  - 1s to 2min time step
  - take global circulation model ECMWF and match it to the largest
    mesoscale grid
  - need to initialise now to reach equilibrium in meso model for
    tomorrow's forcast

  Predict:
  - surface temp to 1K
    - eliminate dome seeing
  - surface wind speed to 1.45 m/s
  - surface wind direction ?
  - vertical stratification
    - very good
  - optical turbulence

  Lascaux et al. MNRAS 2015

  Applied
  - Gemini (compared to GeMS)
  - ALTA Center PROJECT @ LBT
  - cost includes software license
    - needs to be calibrated using SCIDAR


7 James Osborn: Characterising atmospheric turbulence with the Stereo-SCIDAR
============================================================================

  Correlate the pupil images from two stars C_N^2 profiles through the
  night at Paranal and LaPalma Measure turbulence strength and velocity
  as a function of height

  comparisons with DIMM and SLODAR


8 J. Marin: Water vapor forecasting on Chilean sites
====================================================

  PWV from GOES 6.5 microns.  Built a code


9 Florian Kerber: Quantitative characterisation of sky conditions on Paranal with the microwave radiometer LHATPRO – five years and learning
============================================================================================================================================

  Commercial from Radiometer Physics Gmbh
  - quite long lead time

  PWV c. 0.1mm (?)


9.1 Spatial homogeneity
~~~~~~~~~~~~~~~~~~~~~~~

  - Does zenith tell us what we need to know?

  IR channel
  - locked onto LHATPRO
  - separate camera bolted to the side
  - sky brightness at 10.5 microns
  - down to -120K
  - can detect cold high thin clouds

  Pointing
  - Mostly operates at zenith
  - 6 minute all sky scan (12 degree steps)
  - azimuthal scan

  structure
  - PWV is homogeneous to a few % all sky (above 30 degrees)


9.2 Photometric sky quality classification

Conditions are classified as:

  • PHOT CLR THN THK
    can we do better using the IR camera?

Detrended Fluctation Analysis (DFA)

  • Cut brightness temperature data into 2 hour segments
  • detrend separately
  • calculate autocorrelation, characterise as slope and amplitude
    (alpha, <F(t)^2>^{1/2})
    • photometric: flat
    • clouds: slope
      Plot alpha v. amplitude maps cleanly to nightlog conditions

10 Alain Smette: Molecfit

10.1 Introduction


  Usual method:
  - observe rapidly rotating B star
  - higher S/N than target (so as to use point-by-point correction)
  - ESO has delta airmass < 0.2 -- not nearly good enough
  - seeing can change (and hence LSF).  RHL this shouldn't be right if
    the slit is well chosen


10.2 Molecfit
~~~~~~~~~~~~~

  - instrumental LSF
    - fitted or model
  - LBLRTM (previously RFM)
    - HITRAN molecular parameters
  - Atmospheric profiles
    - for T, P, water pull from GDAS
    - use onsight T, P up to 5km
  - adjusts atmospheric composition as a function of height to fit data
    - using sensitive wavelength regions
  - vertical water profile matters for line shapes (and EW?)
  - 2K or 1mbar changes matter
  - parameters for continuum (polynomial per region?) and errors in
    wavelength calibration
  - GUI to fiddle with things
  - predicts emission lines (except OH)


11 Carlos Gonzalez talk: Calibrating Wide Field Surveys
=======================================================

  CASU

  Vista
  - 16 2048x2048 HgCdTe
  - 0.34" pixels
  - Z-K_S
  - VHS is most of the southern hemisphere

  Calibration:
  - fluxes are consistent internally and externally within the errors
  - sky flats
  - correct for distortion
  - measure flux (no details)
  - zero point
    - atmosphere ("easy")
    - system
    - global offset to e.g. AB
      - Use global calibrators.  Ideally
        - proper magnitude
        - measured in instrumental system
        - primary calibrators
      - in practice, 2MASS
        - use classic linear colour terms
        - extrapolation is bad
      - correct using reddening from SFD (why??)
    - Onto a standard system
      - assume linear correction to vega
      - use A0V stars and synthetic spectra
      - compare to 2MASS

  Vista
  - One ZP per image (with per-chip corrections)
  - 2% where there are enough photons
  - Then ubercal (slightly simplified)
    - find 5% structure near the readouts
    - post-ubercal quality??
  - Downsides
    - computationally demanding
    - wait until there's enough data
    - need overlaps -- PIs may not be happy
  - Conclusions
    - need classical for real time
      - hard below 5%, can't beat 3%
    - ubercal needed for < 1%


12 Robert Lupton: Calibration aspects of the LSST
=================================================

  Standard summary of LSST calibration plans.  You've heard it before.


13 Wolfram Freudling: Data Reduction and archive at ESO
=======================================================

  Observatory
  - QC
  - preimaging data
  - master calibs
  - science data products
  Community
  - custom setting
  - human checking
  - custom steps

  CPL is used to tie algorithms together workflow
  - ?? at Paranal
  - Reflex at ESO
  both use ??

  Data layout defined by "OCA rules"
  - classification "this is a dark"
  - organisation "define a set of darks"
  - association (a SQL-like statement that makes a dark)

  Consortia
  - deliver pipelines in CPL to run as plugins (P. Ballester
    VLT-SPE-ESO-19000-1618)
  - reflex workflow + docs
  - ESO Pipeline systems group takes (PSG) over after start of operation
    (9 developers)
  - Scientific oversight by Garching scientists during development; PSG
    + ? thereafter

  C library with primitives such as CR removal called HDRL ?how does
  this relate to CPL?)

  Workflows "reflex" built on "kepler"; it's a set of actors to execure
  ESO recipes, GUIs, and python scripts
  - lazy evaluation
  - organizes the data
  - book-keeping, provenance
  - GUI to display intermediates, modify and retry
  - monitor progress
  - can insert command line scripts
  - one click activation
  - test data and tutorial

  Science Archive Facility(?) SAF
  - internal data from "scientifically validated pipelines".  RHL:
    apparently not "high level products"
    - Science Data Products Group (SDPG)
  - external PIs (e.g. post-CASU) provide high level products (mosaics,
    source catalogues)

  18 pipelines; 9 SAF developers

  ESO questionnaire ESO2020
  - 1439 answers
  - 80% want raw in archive

  Romaniello (The Messenger 2016)
  - all types of reduced data are desired
  - both types of SAF data are in demand
  - didn't reduce demand for raw data
  - archive is c. 25% of papers


14 Reinhardt Hanuschik: VLT science data processing
===================================================

14.1 Calibrations

Run pipelines on calibration data, run QC, and generate trending
plots.

  • Both on the mountain and in Garching
  • cronjob runs every hour looking for new data
  • Garching certifies master calibs (human intervention)

Observation based on user-defined OBs

  • data taken
  • raw data moved to archive
  • add calibs

Calibrations II

  • compare to old values (?why didn’t QC do this?)

14.2 Science


  traditionally:
  - raw data
  - astronomers use any tools they like to process raw data
  - do science

  More efficient
  - raw data
  - ESO generates `Internal Data Products' (IDPs)
    - aim for "science level"
      - ISR + atm removed
      - physical units
      - errors
      - ready for analysis
  - do science

  N.b.
  - standard reduction
    - may not be optimal
  - QC
  - at scale

  IDP requires:
  - reviewed pipeline (SDPG)
  - master calibs
  - QC scheme
  - preview plots
  - data product standard

  IDPs for
  - UVES
  - XSHOOTER
  - GIRAFFE
  - MUSE
    - soon MUSE_DEEP
  - HARPS, FEROS
  - covers complete history of instruments (16 years for UVES)
  - ingest new data every month or two
  - Use Phoenix workflow
  - Trust master calibrations
  - Do statistical quality checks for *process* QC (not data QC)
  - flag saturation, CRs
  - Our most popular data products

  MUSE
  - Projects #4 and #5
  - Big; 1GB files, 9e4 spectra per exposure (400 Mpixels)
  - IDP has 5 recipes and 6 steps
  - 24 cores (one per spectrograph); 100 GB
  - 2 hours
  - ~ 3GB datacube
  - flux calib, sky subtracted
  - QC
    - QC number
    - visual checks
  - going deep (but a single OB is <= 1 hour)
  - now combing all OBs for each source
    - e.g. 15 OBs (11.2hr)
    - 16 hrs on 48-core blade (1TB)


15 Andrea Mehner: XShooter
==========================

15.1 Intro
~~~~~~~~~~

  - 300nm -- 2400nm; 4 arms (photometry; uv, opt, ir)
  low resolution

  0.4x5" slit (popular) 4x1.8" IFU (unpopular)

  stare, nodding, mapping

  standards:
  - telluric standards (but moving to OH lines - 14% -> 5%)
  - can accept user standards (velocity); rare
  - QA once a month

  Instrument health: dark arc linear ADCs+??


15.2 Issues:
  • mechanical failure of ADC (disabled in 2012; fixed in May 2017)
  • centering poor due to incorrect reference pixel position (0.2") –
    found after 4 years. Check every day!
  • humidity
    • Condensation
    • absorption on dichroics (2 week timescale)
      • manual airflow system
      • correlate humidity with outputs
  • high readnoise in opt arm some of the time. Cabling and grounding?
    • unsolved. Tell the users about it
  • unsuitable telluric standards (19% of 1079 B-type stars were Be,
    binaries, SGs, LPV, strange line profiles)

pipelines

  • IFU
    • no science ready data (15% of 35 programs are published)
    • Pipelines are important!
  • Sky subtraction
    • solution in 2017, CRIRES + consortium?
  • Master calibs
    • bad due to two lamps in the UV which vary (apparently a problem in
      ESO calibration choice – see Hummel’s talk)
    • RHL: Why not use the telluric standards?

16 Konrad Tristram: Challenges in the midIR with VISIR

16.1 VIZIR


  - Imaging spectroscopy 5-20 microns
  - Built in 2004; upgraded by ESO in 2014
  - UT3 Cassegrain
    - imager
      - M1, M2 image onto cold stop; M3-M5 focus onto detector.  0.045
        pixels
      - broadest filters in N reach a few mJy in an hour
    - spectrometer
      - two arms simultaneusly
        - LR (R ~ 300) & MR (R ~ 3200; not yet available) prism/grating
        - HR echelle R ~ 25000
    - two Aquarius 1024x1024 InSb detectors (110Hz) (one imaging, one
      spectroscopy.  How do the two arms work?)
      - "excess low frequency noise" mitigated by faster chopping (M2 @
        4Hz; telescope at 0.01Hz)
      - good cosmetics
      - bleeding if bright stars fall on splits between readouts.  Not
        flux
    - calibration unit


16.2 Calibration
  • Transmission and emission strongly dependent on humidity
  • sky and telescope emission (instrument is at 30K)
    • water vapour, dust, air temperature, airmass
  • need to chop M2 (leaves residuals as M2 moved); remove by chopping
    telescope in same or perpendicular direction (perp. not an option
    for slit spectroscopy)
    • careful to guide independentally on two chop positions

Health:

  • monitored by “HealthChecker”

17 Miwa Goto: IR Spectroscopy

17.1 IR spectroscopy in low PWV


  - 2.4mm -> 0.1mm for 1/2 a night (Sofia: 0.005 - 0.020 mm)
  - can observe DH vibrational lines


17.2 Wavelength Calibration
~~~~~~~~~~~~~~~~~~~~~~~~~~~

  - options:
    - Fit a linear or higher function to arc lines
    - Fit to the atmospheric absorption (uses all pixels -- I don't buy
      it)
  - irregular sampling
    - don't use linear sampling.  cubic can be worse.  Resample once


17.3 Flux Calibration
~~~~~~~~~~~~~~~~~~~~~

  - linearity corrections
    - N.b. within each reset cycle
    - RHL deduces that either IR spectroscopists are naive or the speaker
      is


17.4 Slit filling
~~~~~~~~~~~~~~~~~

  - features dividing target by standard.  Features are anti-correlated
    in object and standard
    - Unresolved lines in spectra are aliased in the point source but not
      in the extended source (i.e. resolution wasn't set by slit)


17.5 Astrometric Calibration
~~~~~~~~~~~~~~~~~~~~~~~~~~~~

  - astrometric problems from wavelength solution (I didn't understand
    -- trace curvature)


18 Igor Chilingarian: Is there a place for perfectionism in the NIR data reductions?
====================================================================================

18.1 NIR spectral pipelines
~~~~~~~~~~~~~~~~~~~~~~~~~~~

  Two limiting cases:
  - high S/N echelle of bright stars
  - low S/N

  - H2RGs compare well to 20year old CCDs, but
    - persistence
    - pix-to-pix variations
    - non-linearity
    - ITAR
  - Significant thermal background beyond 2 microns
  - Sky emission and absorption
    - airglow and absorption highly variable in space and time
    - use telluric standards


18.2 FIRE
~~~~~~~~~

  - R ~ 4000-6500 echelle
  - 830nm -- 2500nm
  - optical design very sensitive to flexure (double pass through some
    prisms)


18.3 Project: Las Companas Stellar Library
  • Goal: 1200+ stars, R = 6500, S/N 150-400
  • scanning across the slit to avoid slit losses, psf variation, bright
    stars (RHL: if constant)

18.4 FIRE Bright Source Pipeline


  - flux to 3% z-K; 1% in 200AA (?)
  - Nasty H2RG:
    - 4 readouts slightly different
    - strong pixel-to-pixel
    - 10% non-linear near saturation
    - many bad pixels
  - Flexure gives 3 pixel offsets (and slit isn't repeatable)
  - Create superflat to correct pixel-to-pixel effects
  - Wavelength calibration hard
    - Using ThAr (too few lines in the red)
    - Can't rely on OH as some exposures are very short
    - fit ThAr first, then fit simultaneously with OH.  10km/s
  - PSF across slit isn't Gaussian (see speckles)
    - 30% z to K [RHL: consistent with Kolmogorov]
    - fit PSF, do optimal extraction
  - Telluric correction
    - Use 2MASS if possible
    - fit models (inc. telluric lines) to get to 2 km/s


18.5 MIRS on Magellan (now MMT)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

  - 7' slit
  - Multiobject over 4x7'
  - R ~ 3000

  - Standard processing (Published 2015)
  - sky subtraction
    - use all skies simultaneously to recover sampling (using Kelson
      2003)
    - Claims Poisson limit (but maybe for bright sources)
    - reached continuum 24 in J at Magellan in 20 hours


19 Paul Bristow: Metrology
==========================

19.1 Metrology
~~~~~~~~~~~~~~

  Metrology:
  - "Instrumenting anything that can change, and providing feedback to
    fix problems"
  Calibration:
  - postprocessing (can't get the instrument aligned after the fact)
  Calibration reference data
  - traceable to alboratory standards (with metadata)
  - errors
  - documented (preferably published)


19.2 Metrology for VLT
~~~~~~~~~~~~~~~~~~~~~~

  - XSHOOTER
    - keeping target on three slits
      - take calibration arc spectrum with pinhole at input slit and at
        cass.  Realign
    - correct dispersion for flexure (in pipeline if you have the
      metrology)
      - make physical model based on those calib arcs

  - VIMOS
    - Similar flexure problem (due to rotation of instrument)
      - solution also similar to XShooter
    - inability to put targets on masks
      - fixed by a better sky to mask calibration

  - MOONS
    - fibre positioning more predictable, but still allow for closed loop
      - crosshairs as back-illumination isn't possible
      - see next speaker

  - MUSE
    - rotator/ADC violate 10microns per hour specification
      - guide on outer parts of field

  - AOF (AO)
    - slow drift between wavefront sensors and what the DM thinks it
      looks like
      - update control matrix on minute timescales

  - 4MOST on Vista
    - 2436 fibres
      - measurement and repositioning using back illumination (very
        similar to PFS's MCS)


19.3 CRIRES+ Calibration
~~~~~~~~~~~~~~~~~~~~~~~~

  Upgrade from CRIRES AO-assisted R ~ 1e5 spectrograph, now grating
  cross-dispersed for spectral range (was pre-dispersed)

  Calibration sources feeding integrating sphere
  - halogen, Ne, Kr, heNe
  - replace ThAr with UNe (may not be able to get ThAr?)
  - Fabry-Perot, fibre feed to integrating sphere
  - absorption cells (NH3, CH4-13, C2H2, ?)

  Fabry-Perot because arcs are irregular spacing and have varying
  intensities
  - reach 10m/s
  - zeropoint from emission lamps
  - not simultaneous with science data (unlike absorption cell)

  - Location of sources important -- pick-off from integrating sphere.


19.4 CRIRES+ Metrology
~~~~~~~~~~~~~~~~~~~~~~

  - dispersion stability 0.1 (0.05) pix/4 hours; repeatability 0.5 (0.2)
    pix/24 hours
  - oCRIRES tall poll was echelle (0.15 pix in 4 hours) -- will be in
    CHIRES+ too
    - fiddled ArTh injection cleverly, but caused problems
    - arcs appeared above science spectrum
  - Solution: figure out closed-loop mechanism (e.g. piezo activate fold
    mirror and gratings) to setup instrument (took 80s).  I missed
    details
    - never tested as oCRIRES was removed before it could be tried in
      anger

  CRIRES+ has two bright Ar lines injected near field stop to only
  sample grating (could use during science exposures) Must move fold
  mirror last as it's `blind' (doesn't affect metrology fibres)


20 Holger Drass: Optimisation of MOONS metrology
================================================

  Member of "MOONS Metrology Team" (scary for PFS)


20.1 MOONS
~~~~~~~~~~

  - 1024 fibres (or in pairs)
  - 0.64-1.8 microns
  - 1000 fibre positioners mounted on curved plate at Naysmyth
    - 2-arm; motors with encoders and anti-backlash
    - can reach center of neighbouring positioner
    - 10 micron tolerance
    - 12 cameras around support plate (overlapping).  Canon 18Mpix; 8.3
      pix/mm (need 1/10 pixel)
  - Status: final design (passed PDR)
  - 30s cycle time
  - Use at beginning of night, and maybe for each observing block

  100 images and detect fibres.  Given an externally-measured set of 3-d
  positions solve for the mapping.  Failed:
  - temperature stabilisation
  - Got to the point where there's structure in the residuals
    - Chromatic aberration in cameras?
      - switch to filtered/monochromatic light?


21 Gaspare Lo Curto: HARPS
==========================

  80k-115k 1mK/day

  fibre fed, one arm has "double scrambler" -- extra 1m within
  instrument "Low" res arm has sky fibre (EGGS)

  pipeline running on/offline mode "delivering science data products"

  All data re-released when a new pipeline is released

  Daily:
  - bias, flat, etc.
  - fibre cross cal
  - wavelength (ThAr)
  - drift (ThAr + FP)

  20 minutes

  EGGS takes better flats (S/N 700)

  Na-saturated stars to monitor scattered light within instrument

  0.6 m/s RV demonstrated


21.1 Improvements
~~~~~~~~~~~~~~~~~

  - Hollow-cathode ThAr.  EU regulation change in 2006 led to much lower
    purity
    - some lines are buried in "continuuum"
    - buy many old lamps
    - switch to using an FP for drift measurements
  - FP added in 2011
    - < 3cm/s photon noise per frame
    - < 10cm/s drift/night
    - no absolute wavelength scale
  - guiding
    - only azimuthal scrambling is efficient
    - 3 m/s from a 0.5" offset (10% variation in intensity)
    - RHL N.b. non-uniform output intensity
    - RHL: do I care?  GA?
    - solution: use tip-tilt table to guide
  - octagonal fibres
    - excellent scrambling!
    - RV insensitive to guiding and focus
    - increased throughput (just newer technology)
    - small drift in zero-point in velocity (15 m/s), dependent on
      spectral type and linewidths
  - Laser frequency comb
    - stable/accurate at 2e-11


22 Francesco Pepe: High accuracy wavelength calibrations: ESPRESSO
==================================================================

22.1 Spectrograph theory
~~~~~~~~~~~~~~~~~~~~~~~~

  - We want omega, but are obliged to measure wavelength
  - encode wavelength in angular direction (collimate first to remove
    degeneracy)
  - produce monochromatic image of slit
    - convolved with the spectroscopic PSF


22.2 Wavelength calibration
~~~~~~~~~~~~~~~~~~~~~~~~~~~

  Wavelength calibration: pixel to wavelength
  - errors in both
    - arc: photon noise, line variation, blending, algorithms
    - science: photon noise, spectrum. algorithms
  - Two approaches:
    - HARPS-like: simultaneous imaging of standard spectra (doesn't work
      for slits)
    - HIRES-like: I_2 cell superimposed on spectrum
      - requires deconvolution

  - Bad radial scrambling
    - polygonal fibres
  - finite number of modes and coherence 2(pi r /lambda NA)^2
    - scramble modes

  PRNU effects.

  flats
  - laser annealing
  - fringing

  Pixel size variation
  - stitching errors
  - see in low-res model (ArTh) model applied to high (laser comb)
    wavelength calibrations
  - Use floating-point pixel positions measured using superflats (RHL:
    why not laser?)

  CTE
  - line shape a function of line flux (due to finite capacity of traps)
  - recursively correct raw (?2-d?) frames
  - RHL: do you allow for noise when doing this?

  ThAr
  - Th moves relative to Ar lines by 15 m/s

  Ideal calibrator
  - Laser comb?
  - FP?
    - stable to 10 cm/s in 10 hours
    - group dispersion delay (due to coatings) -- effective spacing a
      function of wavelength. Very stable

  FP
  - determine structure of solution from FP
  - get low-order from ThAr (or limited lambda range laser comb?)


23 A. Szentgyorgyi: GMT spectrograph G-CLEF
===========================================

23.1 GMT
~~~~~~~~

  - f/8
  - adaptive optics for all instruments
  - 4-mirror first light 2023


23.2 G-CLEF
~~~~~~~~~~~

  - R ~ 20k -- 110k
  - 350(?) - 950 nm
  - fibre fed echelle
  - all optics in vacuum within precision thermal control.  Gravity
    neutral
  - Will have MOS frontend
  - 2 cameras, red and blue
  - carbon fibre optical bench

  Largest aperture of all precision RV instruments for a while...


23.3 RV error budget approach
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

  - spec 40 cm/s (goal 20 cm/s)
  - used for cost/benefit analysis (e.g. carbon fibre bench)


23.4 MOS
~~~~~~~~

  - 40-50 multiplex
  - 20 arcmin fiels
  - non PRV; maybe globular cluster dynamics
  - uses MANIFEST with "starbugs" to move the fibres (walk using piezo)


23.5 Lessons from Hectochelle
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

  - ?
  - fibres should be telecentric
  - use tunable laser + "wavemeter" (25k$ for 2ppm)
    - 5 times better than ThAr (80 m/s per feature -- maybe 15m/s
      achieved with sqrt(N))
    - 350nm - NIR
    - very bright
    - RHL line width?


24 Yuanjie Wu: Frequency combs for astronomical applications
============================================================

24.1 Frequency comb
~~~~~~~~~~~~~~~~~~~

  - mode-locked laser with short pulses somehow locked to atomic clock
    (GPS)
  - spectral envelope is tunable


24.2 Astronomical Combs
~~~~~~~~~~~~~~~~~~~~~~~

  - mode spacing: 3*resolution
  - HARPS/ESPRESSO: 18 GHz
  - standard comb -> cavity -> non-linear "broadening" fibre ->
    "spectral flattener" (throws "excess" light away?)
    - cavity filters 250GHz -> 18 GHz (finesse 2300; > 46dB suppression
      of sidelobes)
    - broadening fibres change in time (colour centre formation?  grating
      formation?)
      - need to change them out
  - May be able to reach 380nm by second harmonic generation by rotating
    non-linear crystal (?)
  - If fibre illumination constant 3cm/s drifts, close to photon
    - bins down to 1 cm/s
  - need to agitate fibres to scramble modes
  - Compare 18-18GHz and 18-25GHz -- 0.5 m/s shift
  - Intensity changes give RV shifts
    - CTE and pipeline problems
  - continuum of 3-6% due to "noise in fibre amplifier"

  Winds move atmospheric lines by 5-10 m/s -- can correct using a
  radiosonde


25 Gaspari Lo Curto:  Laser Frequency Combs in Astronomy, from La Silla to Armazones
====================================================================================

  2cm/s : 3.2AA (Si lattice constant is 5.4AA)

  omega_n = n omega_r + omega_{CE}

  omega_r atomic clock frequency (Rb not good enough; need Cs but too
  expensive; use GPS) n: integer ~ 1e6


25.1 Astro laser combs
~~~~~~~~~~~~~~~~~~~~~~

  Commercial products (orange box, 250MHz, pulse width 100fs -> 40nm
  bandwidth, 1060nm Yb laser)
  - FP filtering down to c. 20GHz spacing
  - photonic crystal fibres for non-linear mixing
    - 4-wave mixing (3 photons -> 1; omega = omega_1 +- omega_2 +-
      omega_3)
    - honeycomb structures
    - finite lifetime (high energy pulses)
    - non deterministic
    - amplify unwanted photons
    - can spread 1060nm to 400nm -> 1500nm (but died after 30 minutes)
      - 500-800 lives 100s hours
  - flattening coupled to single-mode fibre


25.2 Mode scrambling
~~~~~~~~~~~~~~~~~~~~

  - Laser, so see all modes
  - dynamic scrambling (shake ("dynamic") and squeeze fibre); dynamic is
    essential
  - couple single-mode to multimode fibre -- fill as many modes as
    possible

  Studied PSF as a function of position (Fei Zaho poster XXVIII IAU).
  CTE visible

  Can "scan" the detector by varying mode spacing
  - characterise each pixel
  - DB of PSF for each pixel (RHL: what about FRD?)

  Photon noise:
  - 1 m/s -> SN = 100 (La Silla 3.5m)
  - 1 cm/s -> SN = 10000 (Armazones == ELT)


26 Jonathan Smoker: CRIRES and CRIRES+
======================================

26.1 CRIRES
~~~~~~~~~~~

  - CRyogenic IR Echelle Spectrograph
  - 2007-2016
  - now being upgraded in Garching -> CRIRES+
  - not very efficient (3 mags worse than XShooter)
  - Behind a curvature-based AO system (shared with e.g. SINFONI)

  Telluric lines
  - Remove telluric lines using a radiometer
  - use molecfit not standards for CRIRES+ by default

  Water vapour seen in flat fields (!)


27 Abner(?) Zapata: Study of Radial Velocity Stability in the AIUC Spectrographs: PUCHEROS and FIDEOS
=====================================================================================================

  AIUC: Astro Ingeneria University Catolica

  PUCHEROS RV shift in ThAr correlates with temperature (10 km/s/K)


28 C. Janssen: Molecular line parameters
========================================

28.1 Ozone
~~~~~~~~~~

  Cross section uncertain at 1.5% level at 325nm


29 Gillian Nave: Wavelength references for the calibration of astronomical telescopes
=====================================================================================

  NIST databases

  Lots of muttering about Th lines
  - agree well (2e-7) with Engleman et al. 2003
  - Ar doesn't 2e-3 cm^{-1} for transitions involving high energy levels

  N.b.  U Ne lamps


30 S. Bagnulo: Polarimetric calibration and accuracy: lessons learnt from the existing instrumentation
======================================================================================================


31 A. Cikota: Performance of the VLT/FORS2 spectropolarimetric mode
===================================================================


32 Jonathan Smoker: A very high signal to noise spectrum towards mu Col with UVES
=================================================================================


33 F. Selman (presented by E. Johnston and F. Vogt): MUSE Monitoring and Calibration
====================================================================================

33.1 MUSE
~~~~~~~~~

  - image slicer
  - 24 dewars


33.2 Pipeline
~~~~~~~~~~~~~

  Per CCD:
  - ISR
    - bias
    - dark
    - flat
  - flux calib, lambda-cal, sky subtraction


33.3 QC
~~~~~~~

  - monitor bias level, readout, trends
  - drill down to thumbnails of first/last lamp/bias frames
  - daily bias/flat/arc/standard star/illumination corrections
  - weekly twilight
  - monthly astrometry + ?
  - will compare standards as a function of time (e.g. extra flux after
    installing new DM) -- add more diagnostics


33.4 MUSE Challenges
~~~~~~~~~~~~~~~~~~~~

  - 24 spectrographs; 24 CCDs
    - make them behave as one
  - sensitive to temperature (image slicer moves beams?)
  - everything else

  - Old problems
    - bad electronics
  - Current problems
    - temperature dependent flat fields
      - range in pixel values (delta) evolves with temperature and time
      - locally causes failure to follow sqrt(N)
      - large scale bad flats cause stiching failures
    - weird structures in science frames "Ferris wheel"
      - optical artefact in some science frames; rings + radial structure
      - due to bright point source at edge of VLT field of view
      - mostly repeatable when revisiting the field; sometimes disappears
    - residuals
  - Future
    - Raman scattering making Na beacon
    - Mostly scattering of bright Na doublet
    - Raman: laser photon hits molecule, and loses energy by exciting a
      V-R state, losing a quantum of energy
    - So we get N_2 and O_2 bands
    - Shouldn't be a problem, but scattered light (cf. Ferris wheel).
      Should be subtractable with the sky


34 Martin Roth: Visible integral field units, such as MUSE
==========================================================

34.1 Flat fielding
~~~~~~~~~~~~~~~~~~

  Model PSF as a Moffat works in a crowded field in a data cube doing
  crowded field photometry

  Advantages:
  - no slit losses
  - insensitive to pointing/differential chromatic refraction
  - advantages for sky subtraction

  - few photons
  - lots of resampling
  - complex light path
    - scattering, vignetting, edges, ...

  - sky flat affected by absorption lines
  - superflats from data inapplicable for extended sources
  dome flats have incorrect illumination

  PMAS (Potsdam) built a calibration unit
  - integrating sphere
  - relay lense
  - pupil mask

  VMOS (fibre) IFU has problems with second order light

  MUSE uses sky or internal flats

  lots of resampling
  - variance isn't enough (factor of 6 variation) RHL: carry covariance

  flats to +- 2% of sky


34.2 Line spread function
~~~~~~~~~~~~~~~~~~~~~~~~~

  - Science
  - lamda cal
  - OH!

  Fit Gaussian to 19 groups of lines to see smooth variation with lambda
  different methods give different results

  Tried using model
  - line intensities
  - Hermite expansion of line profiles

  Fibre-fed systems
  - bending tests did NOT just look like Gaussian FRD Schmoll et
    al. 2003
  - studying for (some new instrument)


35 Eleonora Sani: KMOS
======================

35.1 Instrument
~~~~~~~~~~~~~~~

  - 24 pickoff mirrors in patrol region (7.2' wide); each 2.8x2.8"
    (14x14 pixels)
  - arms have cryogenic stepping motors with LVDT encoders (used to
    check positions)
  - arm collision avoidance in software, hardware collision detection
  - filter per arm
  - 3 independent groups of 8 arms; so each spectrograph has 8*14 pixels
  - cryostat at 125K; detectors at 40K

  Visualise position of arms and image cubes during acquisition
  - nod to sky (each IFU has associated sky IFU)
  - stare (some sky IFUs)


35.2 Pipeline
~~~~~~~~~~~~~

  - Richard Davies sky subtraction
  - ESO-style codes, some standard workflow
  - arcs to determine wavelength solution

  Calibration plan
  - day
    - dark
    - arcs
    - quartz (arcs and quartz are taken at 6 rotation angles)
  -nightly
  - standard star
    - telluric, flux calibration within 2 hours; delta airmass < 0.2.  1
      arm per spectrograph
  - fortnightly
    - relative astrometry
      - check in 45 minutes; redo in 0.5 night
  - monthly
    - sky flats for illumination correction


36 Trevor Mendel: KMOS calibration and data reduction
=====================================================

36.1 Science
~~~~~~~~~~~~

  - KMOS^{3D} E.g. Genzel
    - gas dynamics as 1 < z < 3
  - KMOS cluster survey
    - stellar pops at z > 1.5; 20+ hours

  Problems
  - registration to preserve spatial information
  - sky/telluric correction


36.2 Pipeline
~~~~~~~~~~~~~

  - Nightly arcs/flats
    - illumination correction
    - telluric correction/zero points
  - Flexure compensation loop (update lambda table)
  - OH using offset sky frames
    - adaptive stretch-and-shift
    - split sky into OH transitions and scale each separately (Richard
      Davies)
  - 1 star per arm for PSF, dither, tracking, throughput, telluric
    - don't acquire guide stars -- 15 minutes
  - correct for offsets, white-light stack, then cross-correlate to
    align visits
  - correcting for IFU illumination as a function of rotation angle
    - 2-5% across IFUs
    - 10-20% between IFUs
    - may be discontinuous...
  - telluric absorption at 1100 - 1150nm.  Uses molecfit (rescale water
    vapour column)
    - does a pretty good job
  - bias level drift
    - using reference pixels on H2RGs helps
    - output channel jump (in read/control electronics?)
  - cross talk (claimed to be capacitative)
  - persistence.  Maybe global reset?


36.3 Possible pipeline improvements
  • move as much as possible into detector frame (avoid resampling)
    • sky (Kelson et al. approach)
    • IFUs are too small to define clean sky pixels
      • subtract sky
      • model residuals at constant lambda/wavelength (on some other
        scale?)
    • frame-specific illumination using sky lines
  • optimised 1-D extraction
    • can do per image using absolute positions
      • permits bootstrap resampling
        Comparison with MOSFIRE encouraging
  • comparable sky subtraction
  • flux calibration off by 20%

36.4 RHL N.B.


  - Mendel seems pretty good
  - persistence builds up during night even from science frames

  - spatial variability
  - cross talk coefficients from CRs


36.5 Q
~~~~~~

  - how does telluric correlate with e.g. GPS or radiometer
    measurements?


37 Myriam Rodrigues: Sky background correction in multi-fiber spectrographs
===========================================================================

37.1 ESO
~~~~~~~~

  - FLAMES/GIRAFFE
  - MOONS (VLT 2019)
  - MOSAIC (proposed for ELT)

  FLAMES is OH dominated MOONS is background dominated (not OH) -- 25th
  mag in 16 hours


37.2 OH
~~~~~~~

  - vary by 20% on minutes to decades
  - diurnal
  - gravity waves (minutes)
  - spatial variation

  Strategies:
  - mean sky
  - closest sky fibre
  - clean up by PCA (Wild & Hewett; Sharp; sparse Zhang and Zhang 2016)
  - Physical modelling by V and R groups (Richard Davies) (lines from
    Noll et al 2014)

  - surface recomposition model (Rodrigues 2008) for continuum after sky
    subtraction


37.3 Continuum dominated
  • moon
  • zodi
  • stellar background
  • airglow
    • nitric oxide?
    • lorentzian wings of OH lines?
    • scattered light

Studies at 900nm

  • using FORS2 Puech et al 2012, Yang et al. 2012
  • spatial variation 1", 30" to 150" with amplitude 0.5% of mean sjy
  • timescales below 30minutes

37.4 Strategies


  - two fibres per object (< 20")
    - lose half your fibres
  - beam switching
    - only 50% duty cycle (or 67%)
    - non-simultaneous
  - cross-beam switching (2 fibres per object, nod between them)
    "XSwitch"
    - lose half your fibres
    - subtract 2-D images

  Tests with FLAMES:
  - R ~ 6500
  - I_AB ~ 21st mag
  - 28 degrees from the moon
  - targets 15% of sky

  cross beam gives <~ 0.6% sky -- 10* better than other methods.

  Hard to design cross-beam fibre assignment for MOONS
  - enter catalog of all sources
  - priority for your sources
  - fibre assignment maximises number of pairs

  Simulations (naive?) say that XSwitch will work well

  sky variation (Flores et al. 2016)
  - Bright lines vary by 5-15% on 10 minutes
  - faint lines (1/100 of bright) vary by 100%
  - looked at APOGEE, lines that aren't in models
    - O_2?


37.5 Questions
~~~~~~~~~~~~~~

  - continuum from Lorentzian wings would scale with OH flux (modulo V/R
    excitation)


38 Pasquale Hibon: Astrometry in MCAO imaging
=============================================

  Classic adaptive optics
  - Natural guide stars 15 -> 5% of sky
  - cone effect (Strehl loss of 50% for 8m telescope at 1 micron)
    - worse for bigger telescopes


38.1 MCAO
~~~~~~~~~

  - removes cone effect
  - average sky coverage 50% in H
  - One DM for each layer, e.g.
    - ground layer
    - high altitude

  - MAD at VLT in 2012 K-band over 2' (natural guide stars)
  - LINC-NIRVANA (newest system on VLT)
  - GeMS on Gemini-S
    - GEmini Multiconjugate ao System
    - 85"x85"
      - very uniform in Strehl and PSF width -- close to diffraction
        limited
    - Need 3 natural guide stars
    - GSAOI is the imager
      - 0.02"/pixel

  MCAO good for astrometry
  - large FoV
  - good knowledge of PSF
  - can control scale

  Calibrations:
  - non-common-path aberrations (NCPA)
    - science path v. reference path
    - correct if we can measure it
  - darks
    - library with varying exposure times
    - number of hot pixels depends on the exposure time (?)
  - flats
    - use dome flats
    - can use twilight flats
  - flux
    - standard stars
    - 2-3 standards per night at different airmasses
    - photometric accuracy 2-3%
      - flat fielding
      - ...

  Challenges:
  - PSF depends on the position in the field and location of guide stars
    - 50-90 mas, better in centre of field
  - Distortion varies with time
    - night2 - night1 ~ 0.2mas
    - can use pinhole mask to measure

  MCAO in the visible
  - GMOS
    - 5.5'x5.5'
    - 73 mas/pixel
  - GeMS + GMOS
    - 2.5'x2.5'
    - 36 mas/pixel
    - only i, Ca triplet, z
    - It worked!  14 targets in 2012 (mostly globulars)
    - 0.3" FWHM in 10s on at least one night
    - many more mirrors -- lose factor of 2.9 in i (narrow passband?);
      gain 1.5 in z
    - 3ish mas positions in globulars
    - Pretty picture with 80mas FWHM with 2' field in z and i


39 Posters
==========

  - Modigliani
    - improvement over Kelson


40 Michael Sterzik: Calibration and analysis of the telluric O2 bands
=====================================================================

  Mostly medium-resolution polarized looking at earthshine.


41 Suzanne Ramsay: concluding remarks
=====================================
1 Like

I’d like to know more about flat field calibration. My understanding is very basic. A popular method seems to be taking flat field images at twilight. At that point the sky is dark enough not to saturate pixels but there are no stars or especially galactic star clusters to make the field uneven.

The reason I decided to learn about this was an article in the NY Times quoting a scientist affiliated with LSST: “His preliminary results suggest that avoiding the satellites would be difficult during twilight — a serious problem given that potentially hazardous asteroids and many objects in the solar system are best seen during this time. The satellites thus limit the ability of astronomers to observe them.”

That appears to be incorrect. Objects like asteroids that are lit by the sun are not “best seen” at twilight; they are best seen at the darkest time of night. The satellites that are being discussed in the NY Times article are indeed brightly lit at twilight like the Space Station which I have observed many times across a wide angle of sky. But once in the earth’s shadow those satellites will not be lit.

What appears to be the case, in my limited understanding, is that flat field calibration will be affected by the satellites. Each pixel in a frame impacted by a satellite will have to be tossed as unusable for calibration. Once observed, the satellite-lit pixels will be predictable and may even be predictable from orbital data. But a large number of satellites could affect enough pixels for enough frames to make it impossible to calibrate those pixels.

Link to NY Times: https://www.nytimes.com/2019/11/11/science/spacex-starlink-satellites.html

You’re quite right about asteroids and low-earth objects. For a number of reasons it’s not a good idea to use twilight flats (for example, they have a very different colour and the illumination pattern is not well-defined).
Flat fielding is usually done using a flat-field screen illumination in conjunction with a set of observations of stars to correct for illumination, ghosting, and such like problems. In the case of LSST we will also have a “collimated beam projector” which provides a set of monochromatic “stars” allowing us to perform the star corrections based on in-dome data; we have not yet demonstrated that this works.

Note that objects like asteroids that are inside the Earth’s orbit cannot be seen at the “darkest time of night” (as the Earth itself is in the way).

Aha, that makes sense. Anything in or inside our orbit would need to be seen in that side view.