The SDSS has, really, two sets of coordinates it uses which are specially designed for the survey geometry. These are described in the Project Book in the section on Survey Strategy.
The first set is the Great Circle (mu, nu) system, where mu and nu are spherical coordinates (corresponding to ra and dec) in a system whose equator is along the center of the stripe in question. The stripes of the survey are great circles which all cross at (ra, dec) = (95, 0). The stripes are defined by their inclination with respect to the equator, and are indexed by integers n such that the inclination of a stripe is -25+2.5n. Thus, stripe n=10 corresponds to the Equator. Astrotools provides utilities to convert to and from these, such as GCToSurvey. The design is that the area covered by the imaging on a given stripe (the so-called OK_SCANLINE area in the north) is a 2.5 degree wide rectangle centered on nu=0 in Great Circle coordinates.
The second set is the Survey (eta, lambda) system. This is simply another spherical coordinate system, where (eta,lambda)=(0,90.) corresponds to (ra,dec)=(275.,0.) and (eta,lambda)=(57.5,0.) corresponds to (ra,dec)=(0.,90.). Note also that at (eta, lambda)=(0.,0.), (ra,dec)=(185.,32.5). For some reason, although "eta" is constant along great circles, it is referred to as "survey latitude," while "lambda" is referred to as "survey longitude." Also, "eta" runs only from -90. to 90.; the back of the sphere is covered by "lambda", which runs from -180. to 180. These conventions defy logic, and please don't blame me for them. The Survey coordinates are defined such that the "primary" area of a stripe in the north is defined by a rectangle in Survey coordinates which is 2.5 degrees wide in eta (coordinate width). You may be interested in SM routines which convert between Equatorial, Survey, and Great Circle coordinates. Again, Astrotools has routines to do this as well.