Where to Start
News and Updates
Data Products
Data Access
Sky Coverage
Data Flow
Help and Feedback


Up: Data Products Sections: Images - Object lists - Spectra - Tiling

About spectroscopic samples

The spectroscopic survey is predominantly a survey for

There is also a large number of spectra of stars. Brown dwarfs, ROSAT and FIRST counterparts are targetted as well. Refer to the target selection quality or target selection algorithm for details about the spectroscopic survey targets.

Getting and using spectra

The spectra distributed by the SDSS have been sky subtracted, corrected for telluric absorption spectrophotometrically calibrated by the Spectro2d pipeline. Spectral classification, redshift determination, and emission and absorption line measurements are carried out by the Spectro1d pipeline. The algorithms page contains details about spectroscopic data processing.

The Data Archive Server provides the same spectrum in two files:

  • Spectro2d spPlate*.fits: all calibrated spectra from a single plate, without any parameters or continuum fits. This file also contains the resolution at each pixel as measured from arc spectra (see Data model for spPlate*.fits resolution HDU #4).
  • Spectro1d spSpec*.fits: multiple fits extensions (images and tables) with the calibrated spectrum, a continuum-subtracted spectrum, and all measured parameters (redshift, line fits, derived quantities). See how to read an spSpec file.

The data access page contains various query forms to get spectra by coordinates, or to search for spectra by redshift, object magnitude, color etc., and to retrieve them from the archive. In particular, the Catalog Archive Server provides a fast search capability for object lists and spectroscopic parameters as well as pointers to the files in the Data Archive Server. The Spectro Query Server query form is dedicated to the search of the spectroscopic database.

Please note the caveats below, which are essential to obtain meaningful scientific results from analysing SDSS spectra, in particular the redshift status caveat.

Imaging information associated with each spectrum

Very often one wishes to obtain all the SDSS photometric (imaging camera) data associated with each spectrum. One can obtain this information most easily for nearly all plates through the SQS interface by selecting imaging parameters to return with your selected spectra.

For those who prefer working with the full fits files of spectra and imaging catalogs, we will make available a special set of files (spObj-*.fit) containing a bundled version of the spectroscopic data and all the corresponding imaging data (as for DR1) in the near future

These files are available via the DAS by requesting the data product tsObjFromMap (or spObj) for the plates you wish. To get these files in bulk for the entire survey, see getting SDSS spectra and corresponding imaging data.

Quality of spectra

About the spectra

For details about the spectrographs, see the spectrograph page.

Plate diameter3 degrees
Fiber diameter3 arcsec
Wavelength coverage3800 - 9200 Å
Wavelength calibrationbetter than 5 km/s
Wavelength referenceheliocentric vacuum wavelengths
Binninglog-lambda, 69 km/s per pixel
Resolution1850 to 2200; value for each spectrum in spPlate*.fits
Flux Units 10-17 erg/s/cm2

Further details

  • Error and mask arrays are available (see the data model.) The sky spectrum subtracted from each object spectrum is not currently available.
  • Spectroscopic observations are undertaken in non-photometric conditions when the imaging camera is not in use. At least three fifteen-minute exposures are taken until the cumulative mean S/N per pixel exceeds 4 for a fiducial fiber magnitude of g = 20.2 and i = 19.9.
  • We provide the cross-correlation templates used to obtain cross-correlation redshifts.


Redshift status

Only 1% of the objects have an "unknown" classification, usually because of low signal-to-noise ratio or completely featureless spectra. The redshifts of all but a few tenths of a percent of the remainder are believed to be correct. To identify the few objects with unreliable redshifts, be sure to consider the confidence we have assigned to each redshift (z_conf in the spSpec*.fits primary header, and the status zStatus and zWarnin of the redshift measurement, which may have failed.). A useful cut on the redshift confidence is z_conf > 0.35 (or zConf > 0.35 in the SQL database).

Galactic extinction correction

The EDR and DR1 data nominally corrected for galactic extinction. The spectrophotometry in DR2 is vastly improved compared to DR1, but the final calibrated DR2 spectra are not corrected for foreground Galactic reddening (a relatively small effect; the median E(B-V) over the survey is 0.034). This may be changed in future data releases. Users of spectra should note, though, that the fractional improvement in spectrophotometry is much greater than the extinction correction itself.

Night sky emission lines

The night sky emission lines at 5577Å, (when there is auroral activity) at 6300Å, 6363Å, and in the OH forest in the red can be very strong, and leave significant residuals in the spectra whose amplitude is occasionally underestimated by the noise model. Be cautious about interpreting the reality of weak features close to these lines.

Plates with not-quite-perfect spectrophotometry

A small number of plates, given in the list of not-quite-perfect plates, suffered from a variety of minor problems. The CCD frames for several plates suffered from a transient electronic problem in the red camera in Spectrograph 2, causing the columns of the CCD to be misaligned on readout. This was fixed in software, and we believe the data to be reliable. Another set of plates labeled "Spectrograph Collimation Problem" suffered from having the spectrograph collimator improperly focused. This problem caused a mismatch between the flatfields and the science exposure instrumental profile shapes on the CCD in both the spatial and wavelength directions, causing the optimal extraction process to reject an excessive number of pixels. This problem was fixed in software, and comparing overlapping objects from adjacent plates confirms that the redshifts from these problematic plates are unbiased. However, the spectra themselves should not be used for precision work or spectrophotometry. Finally, during the exposure of one plate, light from an LED somewhere on the telescope found its way to the spectrographs, resulting in an artificial excess of light centered roughly at 6500Å; the spectrophotometry of this plate is quite poor.

Mismatches between spectra and photometric data

In a few cases, the fiber mapping failed which identifies which fiber has been plugged into which hole. When this happens for two or more objects on the sample plate, there is the possibility of wrong matches between spectra and photometric objects. There are 62 objects for which the mapping between object and spectrum cannot be established ("unmapped fibers"); their ra/dec is listed as -9999 in the spectroscopic data set.

Errors in the deblending algorithm in the target reductions caused spectroscopy to be carried out occasionally on non-existent objects (e.g., diffraction spikes of bright stars or satellite trails). Many of these objects no longer exist in the best imaging reductions with its improvements to the deblender. In other cases, the photometric pipeline timed out during the best imaging reductions in fields for which target imaging proceeded without problem, so that the best photometry is missing for bona-fide objects. This predominantly happens in fields close to a few very bright stars. We expect to recover objects from these ``timeout holes'' in future data releases.

A total of 663 spectroscopic objects therefore do not have a counterpart in the best catalogs, 0.2% of the total. Of these, 80 (including the 62 unmapped fibers) can only be retrieved from the table specObjAll in the database. The remaining 583 objects are contained in the default spectroscopic table specObj, but will not be found by queries requesting both photometric and spectroscopic information.

Galaxy velocity dispersion measurements

The velocity dispersion measurements distributed with SDSS spectra use template spectra convolved to a maximum sigma of 420 km/s. Therefore, velocity dispersion sigma > 420 km/s are not reliable and must not be used. There is a postscript file showing the quality of velocity dispersion error estimates.

We recommend the user to not use SDSS velocity dispersion measurements for:

  • spectra with median per-pixel S/N < 10
  • velocity dispersion estimates smaller than about 70 km s-1 given the typical S/N and the instrumental resolution of the SDSS spectra

Also note that the velocity dispersion measurements are not corrected to a standard relative circular aperture.

See the velocity dispersion algorithm for details.

"Bonus" plates beyond the survey limits

A few plates in DR2 target objects beyond the survey limits for a particular survey stripe, stripe 10 (see survey coverage page). These plate/MJD combinations are:

34351692 (straddles stripe limits)

Their corresponding targetimaging data does not contain any PRIMARY objects. Objects from this region are therefore not available via a normal search of the target object lists. To find the target data, select on status & 0x402 (both GOOD and OK_SCANLINE) instead of selecting on the PRIMARY flag being set in status. The information is also contained in the spObj-*.fit files, which are available for all plates. These files are available via the DAS by requesting the data product tsObjFromMap (or spObj) for the plates you wish. They are also available for bulk rsync or wget download via DAS through http (in spectro/ss_SPRERUN/PLATE/spObj-*).

This problem does not occur for DR2 best data.

Accuracy of stellar radial velocities

The accuracy of stellar radial velocities in DR2 is described on a separate page.

Last modified: Mon Apr 19 11:33:28 CDT 2004