Basics of SDSS spectrograph data

Here we describe the general principles around which the SDSS-III data is organized, as well as the major parameters associated with the data set. See the spectroscopic data access documentation for how to access the data.

Keep in mind that the DR8 SDSS spectroscopic data was executed in several different programs. Each observation in each program corresponds to a particular plate, an actual physical aluminum plate that sits in the focal plane, with holes in it into which optical fibers are plugged. In many cases several observations on different dates were performed using the same plate.

The largest program was the SDSS Legacy survey, consisting of around 1800 plates. The next largest were SEGUE-1 (407 plates) and SEGUE-2 (204 plates). There were a number of smaller special programs executed as well. The information below describes the bookkeeping we use to keep track of these observations.

[ID and Version Numbers] [Plate Info] [Object Info]

Identification and version numbers

An SDSS plug plate being plugged.

The spectroscopic data in SDSS-III is organized according to the "plate" associated with each observation. To uniquely identify a spectrum in the survey, one must specify three quantities:

PLATE, a unique integer indicating the plug plate used
MJD, an integer denoting the night of the observation
FIBERID, an integer denoting the fiber number

The same plate may be observed on multiple MJDs. If so, the the same fiber numbers will not in general match the same objects on different MJDs, since the plate will have been unplugged and replugged. So all three of these numbers are necessary to track.

In addition, the pipelines used to reduce the data can change over time. To denote which pipelines have been used, in DR8 we use two extra parameters:

RUN2D, an integer denoting the version of extraction and redshift-finding used
RUNSSPP, an integer denoting the version of the SEGUE Stellar Parameters Pipeline used

For DR8, there are three possible RUN2D values: 26, which corresponds to the pipeline version used for DR7; 103, which is a special version of the pipeline to handle stellar cluster plates; and 104, which is the version of the pipeline run on SEGUE-2 plates. There is only one version of SSPP released in DR8, version 116.

Finally, in CAS, there is a quantity SPECOBJID, which is a 64-bit hash of the above parameters, and is unique within the database. It should be used for joins of spectroscopic data and photometric or targeting data tables.

Plate plugging image for an SEGUE-2 plate.

[Back to top]

Plate information

In both SAS and CAS, we provide information about each plate, including design information, observing conditions, signal-to-noise and overall quality. We discuss in more detail elsewhere the targeting criteria and data quality flags. Included in the latter is how to find which plates are "primary" observations.

Each plate has a radius of 1.49 deg. In addition, fibers cannot be placed more closely together than 55 arcsec due to collision constraints. Finally, no target can be closer than 100 arcsec to the center of the plate, because of the center post on the plate.

The general design information about each plate is given by two parameters:

SURVEY, a string denoting the survey this plate is associated with within SDSS-III. In DR8, SURVEY is one of "sdss", "segue1" or "segue2".
PROGRAMNAME, a string denoting which program within a given survey this plate is associated with. For example, within each survey, some plates were designed for different purposes than others.

For the "sdss" survey plates, the most common PROGRAMNAME is "legacy", which indicates the plates used for the primary spectroscopic survey. The SDSS also had a number of special programs. SEGUE-1 and SEGUE-2 were broken down into a number of different programs with somewhat different targeting parameters.

There are also some parameters giving more detailed information, in particular CHUNK and PLATERUN. Most users can ignore these parameters; they give internal information about how exactly the plate design steps were executed.

Additional information for each plate is available, including the position of the plate on the sky, the date of observation, and other design information. A complete roster of this information can be found in the schema of the plateX table on CAS, and the plates file datamodel page.

[Back to top]

Object information

In both SAS and CAS, we provide information about each object within each plate as well. Each spectrum is labeled by its FIBERID, which for DR8 runs from 1 to 640. Note that if a plate has been observed on two different MJDs, the FIBERID in each does not necessarily correspond to the same location on the sky. The plates are plugged by human beings, and so there is naturally variation each time a plate is plugged.

We discuss in more detail elsewhere the targeting criteria, the spectra themselves and catalog information, including data quality flags. Included in the latter is how to decide which spectra among repeat observations of the same object are "primary."

Finally, we distribute separate catalogs describing galaxy parameter measurements and stellar parameter measurements from the spectra.