
Discussion of the telescope + instrument + site (reference Gunn paper)
  Telescope
  Instrument: spectrographs + cartridges
  Site, transparency, seeing
  Plate drilling + Fiber mapping
  Observing strategy, one vs. multiple-night plates
  How we take flats using the flat-field screens
  Collimation procedure
  Guiding
  On-the-mountain reductions + S/N limited-integrations

2D pre-processing (DJS)
  Overscan + bias
  Pixel-to-pixel flattening
  Question: Do we introduce the ADDERR term here?

Spatial tracing from flatfield (SB)
  never mis-identify fibers
  fit traces to smooth functions

Optimal extraction (SB)
  Extract flat, arc + objects the same
  Characterize PSF in spatial direction
  Reject CRs
  Remove cross-talk impliticitly
    * Test: Measure cross-talk next to bright objects
  Remove scattered light

Wavelength calibration
  Characterize PSF in wavelength direction
    (save the dispersion mapping in final outputs)
  Fit to arc lines + sky -> remove flexure
  Robust against bad fibers, CRs,...
  Achieve residuals of 3 km/sec
    * Test: Sky line centroid residuals
    * Test: M67 velocities
    * Test: 5007/3727 centroids vs. redshifts for em. line galaxies
    * Test: Compare redshifts from repeat observations, as fn of S/N?
      (This will be an upper bound to these errors)

Flat-fielding
  Remove fiber-to-fiber variations
  Remove dichroic effects
  Model relative wavelength-response between fibers

Sky-subtraction (SB)
  Construct "supersky"
  Re-scalining the errors using the chi^2 from sky fibers
  * Test: Chi distribution of sky fibers vs. wavelength
  * Test: Chi distribution of LRGs minus best-fit-models
    Reference: Bolton, Ly-alpha forest papers

Flux-calibration and flux-correction vectors (DJS)
  Selection of F star standards
    * Plot: F stars in color-color space
  Fit type of star using Kurucz models
  Includes telluric absorption
  System throughput
    * Plot throughput vs. wavelength
    * Plot throughput vs. time at mean wavelength of CCD
  * Test: Spectro-photometry of synthetic mags vs. photo mags
  * Plot: Mean spectra vs. mean best-fit models in observed wavelength
  * Plot: Test of spectro-photometry using WD models

Co-adding spectra (DJS + SB)
  Re-sampling with B-splines, retaining variance maps too
  Deal w/variable PSF 
     (should cite Kelson's Gauss-Hermite paper, that's for spatial profiles)
  More outlier rejection
  Demonstrate no residual from dichroic

Application of these methods to other pipelines: DEEP-2, etc.

Appendix: B-splines (SB)

Appendix: Data Model (DJS)

Appendix: Remaining problems in the data?

Appendix: How to get the code + the data

