Data model: frame

General Description
The calibrated, sky-subtracted corrected frame plus associated calibration meta-data. The units of the images are in nanomaggies. It is kept compressed under "bzip2", which we have found is the most efficient compressor of this format. In addition, there is a lossy compression applied to the floating point values (which retains accuracy at the 0.1 percent level). The IDL routine "" in photoop will back out the calibration and sky-subtraction from this file if necessary, in steps explained below. Also explained below is how to calculate the noise in the image.
Naming Convention
frame-[ugriz]-[0-9]{6}-[1-6]-[0-9]{4}\.fits\.bz2, where [ugriz] is the filter, [0-9]{6} is a zero-padded six-digit number containing the run number, [1-6] is the camera column ('camcol') and [0-9]{4} is the zero-padded, four-digit frame sequence number.
Approximate Size
3 Mbytes
File Type
Read by Products
Written by Products

HDU0: the corrected frame, what is normally in the "fpC" files

The "image", a 2048x1489 array of floating point values, the calibrated and sky-subtracted version of the fpC "corrected frame" files produced by photo. Units are in nanomaggies.

The header has additional WCS information. These have been altered from the fpC versions to correct for any offsets from the astrom solutions found in the asTrans files.

The image itself has had the FLOATCOMPRESS algorithm (see the idlutils package) applied to it, keeping 10 significant binary digits. This approach guarantees that the precision is kept strictly within 0.1 percent per pixel, and allows a very efficient compression using bzip2. No special scaling or adjustment need be applied to the values when read in.

Required Header Keywords
SIMPLETboolConforms to FITS standard
BITPIX-32int32 bit floating point
NAXIS2int2d image
EXTENDTboolExtensions may be present
CTYPE1RA---TANstrWCS header type
CTYPE2DEC--TANstrWCS header type
CUNIT1degstrWCS header units
CUNIT2degstrWCS header units
CRPIX1floatX of reference pixel
CRPIX2floatY of reference pixel
CRVAL1floatRA of reference pixel (deg)
CRVAL2floatDec of reference pixel (deg)
CD1_1floatRA deg per column pixel
CD1_2floatRA deg per row pixel
CD2_1floatDec deg per column pixel
CD2_2floatDec deg per row pixel
NMGYfloatcalibration value translating counts to nanomaggies (Already applied!)
NMGYIVARfloatcalibration value inverse variance
VERSIDLstrVersion of IDL
VERSUTILstrVersion of idlutils
VERSPOPstrVersion of photoop
PCALIBstrvalue of PHOTO_CALIB environment variable (that is, which version of the calibrations this is)
PSKYstrvalue of PHOTO_SKY environment variable (that is, which version of the sky estimate this is)
RERUNstrrerun of photo associated with this corrected frame

HDU1: the flat-field and calibration vector

The "calibvec", a 2048-element array of "float32" values, encompassing the flat-field correction to apply, multiplied by the calibration. Translates the counts in the original image into nanomaggies.

The calibrations have ALREADY been applied to the HDU0, so this calibration vector is only to be used to decalibrate an image back into counts.

Required Header Keywords
XTENSIONIMAGEstrImage Extension
BITPIX-32int32 bit floating point
NAXIS1int1d image

HDU2: the sky image

The "sky", an approximately 256x192 array of "float32" values (there are some variations around the y-size of the array), with information about how to interpolate it to the full image size. These sky values are determined from the global sky fits across the run (not photo).

There are cases at the end of runs where the y-size of ALLSKY is less than 186, which requires one to extrapolate off the end of the ALLSKY image in the y-direction when using XINTERP and YINTERP. This extrapolation should be a constant extrapolation (not linear).

The sky values have ALREADY been subtracted from HDU0, so this sky estimate is only to be used to return an image to its (near) original set of values.

Required Header Keywords
TFIELDS3intNumber of columns in table
Required Columns
ALLSKYfloat32[256,ny]the binned sky values, in units of counts
XINTERPfloat32[2048]X pixel-values to interpolate "allsky" into to produce a full image of the sky (bilinear interpolation is fine)
YINTERPfloat32[1489]Y pixel-values to interpolate "allsky" into to produce a full image of the sky (bilinear interpolation is fine)

HDU3: the asTrans structure

Detailed astrometric information for the field. Basically the asTrans structure as found in the asTrans files.

Required Header Keywords
TFIELDS31intNumber of columns in table
Required Columns
RUNint32Run number
RERUNchar[3]Photometric reduction version
CAMCOLint32Column in the imaging camera.
FILTERchar[1]Filter (u, g, r, i, or z).
NODEfloat64RA of great circle's asending node. (degrees)
INCLfloat64RA of great circle's asending node. (degrees)
NAXISint32[2]The dimensions of the image the astrometry is based on. Usually 2048x1489, but could be different in principle for binned runs.
FIELDint32Field sequence number within the run.
Afloat64A coefficient.
Bfloat64B coefficient.
Cfloat64C coefficient.
Dfloat64D coefficient.
Efloat64E coefficient.
Ffloat64F coefficient.
DROW0float64Zero-order row distortion coefficient.
DROW1float64First-order row distortion coefficient.
DROW2float64Second-order row distortion coefficient.
DROW3float64Third-order row distortion coefficient.
DCOL0float64Zero-order column distortion coefficient.
DCOL1float64First-order column distortion coefficient.
DCOL2float64Second-order column distortion coefficient.
DCOL3float64Third-order column distortion coefficient.
CSROWfloat64Slope in row DCR correction for blue objects.
CSCOLfloat64Slope in column DCR correction for blue objects.
CCROWfloat64Constant row DCR correction for red objects.
CCCOLfloat64Constant column DCR correction for red objects.
RICUTfloat64r'-i' cutoff between blue and red objects.
MJDfloat64MJD(TAI) when row 0 was read.
AIRMASSfloat64Airmass for star at frame center (mid exposure)
MUERRfloat32Error in transformation in mu
NUERRfloat32Error in transformation in nu

Example of use, and calculating errors

As a guide to how to successfully read in and interpret this file, we reproduce here code from "" in the photoop IDL product:

;; 0. find filename of the frame file
framename = (sdss_name('frame', run, camcol, field, $
                       filter=filternum(filter), rerun=rerun))[0]+'.bz2'

;; 1. read in the FITS image from HDU0; the resulting image will be
;;    sky-subtracted as well as calibrated in nanomaggies/pixel
img= mrdfits(framename,0,hdr)
nrowc= (size(img,/dim))[1]

;; 2. read in sky, and interpolate to full image size; this returns a
;;    sky image the same size as the frame image, in units of counts
sky= mrdfits(framename,2)
simg= interpolate(sky.allsky, sky.xinterp, sky.yinterp, /grid)

;; 3. read in calibration, and expand to full image size; this returns
;;    a calibration image the same size as the frame image, in units of
;;    nanomaggies per count
calib= mrdfits(framename,1)
cimg= calib#replicate(1.,nrowc)

Steps (0) and (1) just read in the "image". Step (2) reads in the sky HDU, and bilinearly interpolates "allsky" onto a 2048x1489 sized array at the points on the grid defined by "xinterp" and "yinterp". Step (3) reads in the 2048-element vector defined the calibration-times-flat-field for each row, and expands it to a full-sized image.

If you have performed the above calculations, you can return the image to very close to the state it was in when input into the photometric pipeline, as follows:

dn= img/cimg+simg

These dn values are in the same units as the "data numbers" stored by the raw data files that come off the instrument. They are related to the detected number nelec of photo-electrons by:

nelec= dn*gain

The number of photo-electrons is the quantity that is statistically Poisson distributed. In addition, there are additional sources of noise from the read-noise and the noise in the dark current, which we will lump together here as the "dark variance." Thus, to calculate per-pixel uncertainties, you need the gain and darkVariance for the field in question. The darkVariance comes from the read noise and the noise in the dark current. In fact, these are nearly fixed as a function of camcol and filter (see the table below). You can retrieve the values from the field table in CAS (or the photoField file). With those values in hand the following yields the errors in DN:

dn_err= sqrt(dn/gain+darkVariance)

Statistical errors in the sky values are completely negligible. Finally, to get those errors into nanomaggies, you simply apply back the calibration:

img_err= dn_err*cimg

Finally, there are some areas of the image which are part of bleed trails, bad columns, and the like. If you require to track those in your analysis (e.g. weight them at zero) then you need to use the fpM files. Those files are in a special format, best read using the stand-alone atlas reader software. Use the utility called read_mask.

The gain and dark variance values are given below as a function of camcol and filter. In some cases the configuration was changed during the survey, and those cases are noted below. In detail, some chips had two amplifiers, and in those cases the values below are an intermediate value.

Gain of each CCD
camcol u g r i z
1 1.62 3.32 4.71 5.165 4.745
2 1.595 (for run < 1100)
1.825 (for run > 1100)
3.855 4.6 6.565 5.155
3 1.59 3.845 4.72 4.86 4.885
4 1.6 3.995 4.76 4.885 4.775
5 1.47 4.05 4.725 4.64 3.48
6 2.17 4.035 4.895 4.76 4.69
Dark variance of each CCD
camcol u g r i z
1 9.61 15.6025 1.8225 7.84 0.81
2 12.6025 1.44 1.00 5.76 (for run < 1500)
6.25(for run > 1500)
3 8.7025 1.3225 1.3225 4.6225 1.0
4 12.6025 1.96 1.3225 6.25 (for run < 1500)
7.5625(for run > 1500)
9.61 (for run < 1500)
12.6025(for run > 1500)
5 9.3025 1.1025 0.81 7.84 1.8225 (for run < 1500)
2.1025(for run > 1500)
6 7.0225 1.8225 0.9025 5.0625 1.21