Brazilian Tunable Filter ImagerBrazilian Tunable Filter Imager(BTFI)(BTFI)

Preliminary Design Review (PDR)Preliminary Design Review (PDR)

USP-IAG Universidade de São USP-IAG Universidade de São PauloPaulo

18-19th June 200818-19th June 2008

BTFI Data ReductionBTFI Data Reduction(Keith Taylor)(Keith Taylor)

Design Philosophy(Optimize Science Impact)

FPs notoriously difficult to calibrate, regularize and interpret: iBTFs are unheard of – doesn’t help; How do we tackle “fear of the unknown”?

Develop intuitive approach to problem Lots of graphical feed-back – “Picture = 103 words” Lead user through process “by the nose” Heavy use of procedural macros Minimize trips to the Nasmyth platform Make user feel they are in familiar Spectrograph-Land

Propaganda (catch-phrase is Think Slick)

How do we achieve “Slickness”? Inherit code without fear or favour

French, Canadian, French-Canadian, Australian (TAURUS c1980 not available!)

Deploy lots and lots of post-doc time (it’s free) in Brazil ; at SOAR

Over-arching Goals(Calibration)

Efficient calibration of FPs require:1. Calibration sources (line & continuum ; wide-field & point)

2. Slick parallelism determination (daytime only?) FPI & FPP (different procedures & algorithms) Stability of SESO FPs unknown

3. Slick gap (FP order) determination (daytime only?) FPI & FPP (different procedures & algorithms) Need to iterate on gap determination?

Slick -calibration (how often?) -calibrated data-cube – for -correction (daytime, only) 3D flat-fields

Efficient calibration of iBTF requires: 1, 4, 5 & 6 – ie: much simpler than FPs (hopefully!)

Over-arching Goals(Observing)

Efficient observing requires: Slick ITC

– on/off-line Slick acquisition

- Field position + 0, , & STEP

On-line data assessment (continuously updated during acquisition) SNR(,time) in multi-RoIs

On-the-Fly data reduction (next integration time-scale) for acquisition assessment and immediate feedback

Pipe-line data reduction (next night) State-of-the-Art

Off-line data reduction (for the anally retentive) Toilet paper not supplied

Archiving See comments by Chris Smith (NOAO)

Strategies to achieve goals(FP Parallelism)

FPI parallelism determination (low-R) SOAR’s ISB calibration unit (with fibre-feed modification)

NB: Small gap – no by eye method Use 4 corner-mounted (flexure) fibres Scan rapidly over line ; determine 4 -centroids ; iterate controller

(x,y) offsets until 4 -centroids are equal parallel

FPP parallelism determination (High-R) SOAR’s ISB calibration unit (wide-field) Mexican (hat)trick?

4 prisms in pupil filter wheel

or Pupil re-imager (practicality?)

Mexican (hat)trick

Left the FP is parallel. Right is not

PUMA

Further strategies to achieve goals

FP gap/order determination Calibration sources (line): SOAR’s ISB calibration unit (full-field or fibre-feed)

Need at least 2 lines (1, 2 …) widely spaced Scan FPI or FPP over FSR

Robust algorithm to determine FP gap Determine -calibration (locally)

PhotonEtc’s tunable source? ($50k?)

FP -stability (periodically through night if nec.) FPI – repeat parallel determination FPP – determine Ring Radius (classic technique)

FPI + FPP - tandem (coordinated) scans Needs -calibration for both I P but I = P zSTEP

I zSTEPP

Yet further strategies to achieve goals

iBTF calibrations Determine -calibration (single order – should be simple) Determine -map (~1D)

Slick ITC – on/off-line Includes models of all instrument & detector modes Includes all sky params (sky brightness, moon, extinction etc) Includes magnitude or surface-brightness switch

On-line data assessment (during acquisition) EMCCD (amp. mode) – view SNR increase on-line

Don’t have to wait for next CCD read-out

On-line z-compressed image (cf: TAURUS + IPCS: c1980) Select RoIs (~4-6) interactively Assess SNR of -profiles (not -corrected) on-line

Data Reduction (at last)

Preliminaries (during afternoon, for all configs.) -calibration (FPs and/or iBTFs) -map (FPs and/or iBTF) 3D flat-fields (FPs and/or iBTF) - -corrected

Acquisition protocols 3D data-cubes recorded as individual 2D frames

OtF data-cubes are not necessarily archived

On-the-Fly (during & immediately after acquisition) CR rejection (sigma clipping) on each individual z-frame Shift&Add into data-cube taking into account

Flexure determination (at least once per pass through data cube) Guiding corrections (gaussian fits to bright stars)

Interactive 2D gaussian fitting to field stars Interactive -profile fitting and SNR determination

Applying -correction on each RoI

The Pipe

Post-acquisition processing (next day) Flat-fielding (2D) individual frames – from Twilight? CR rejection (median rejection over 3 frames) Shift & Add (x,y,) into data-cube -correction 3D flat-field correction -calibration Sky-subtraction (in 3D - mainly for iBTF and FPI modes) Flux calibration (assuming calibration sources have been observed) -profile function fitting

Adaptive binning Wavelet analysis (?)

2D maps of flux, velocity, line-width etc. Rotation curve fitting (in your dreams!)

The Archive(email between: Steve Heathcote / Chris

Smith) Steve: “Needs to be able to swallow

40GBytes/night of data and transport it back to the archive?”

Chris: “Yes.  We are committed to producing a "Data Transport System" (DTS) interface that can swallow at least the 300-400GB/nights by 2010”

Steve: “What software/hardware they will need for running this kind of pipeline.”

Chris: Hardware: “Basic Linux cluster, being a rack of 1U

units, each sporting 1 or 2 quad-core CPUs” Software architecture options:

1. “MOSAIC/NEWFIRM pipeline infrastructure” or 2. “LSST software”