Quick Look Summary of MSX Galactic Plane Survey

Links to descriptions for other MSX images:

A more general description is here.

Survey coverage 0 < l < 360, |b| < 5, 300 < l < 30, 5 < b < 6.5
Wavelength coverage: 4 bands (8.3, 12.1, 14.7, 21.3 microns)
Image resolution: 20 arcseconds, 72 arcseconds (low res mosaics)
Positional accuracy: 2 arcseconds rms
Image noise (rms): 1.9-4.8 10-7 W m-2 sr-1 (at 8.3 microns)
5.9-11.8 10-7 W m-2 sr-1 (at 12.1 microns)
3.6-8.9 10-7 W m-2 sr-1 (at 14.7 microns)
13.3-29.8 10-7 W m-2 sr-1 (at 21.3 microns)

Each image is a Cartesian projection about the Galactic center in Galactic latitude and longitude. The image units are of in-band radiance (W m-2 sr-1). The 72 arcsecond images have rms noise levels six times lower than the corresponding 20 arcsecond images. These lower resolution images are intended for examination of large regions of the Galactic plane and comparison with radio surveys and other data of comparable resolution.

General Description of MSX Images — Mini Supplement


The Midcourse Space Experiment (MSX) Galactic Plane Survey has mapped the Galactic Plane for |b| < 5 degrees, the 4% of the sky missed by IRAS (2 IRAS gaps), regions of high source density including 8 star forming regions (Kraemer et al. 2002) and 10 galaxies (Kraemer et al. 2002), and 13 solar system objects (12 comets and asteroids and the Moon) in six mid-infrared spectral bands using the SPIRIT III infrared telescope. An overview of the astronomical experiments conducted with MSX is given by Price (1995). A more complete description of the infrared experiments and data processing is given in Price et al. (2001). The SPIRIT III infrared telescope is a 33 cm clear aperture off-axis telescope with five line-scanned infrared focal plane arrays (Bands A, B, C, D and E) cooled by a solid hydrogen cryostat. The Si:As blocked impurity band focal plane arrays consisted of eight columns of 192 18.3x18.3 arcsecond detectors. Half of the columns for each detector band were offset by half a detector to provide critical sampling in the cross-scan direction. The detector arrays covered a 1 degree cross-scan field of view. The scan and sample rates provided 2-8 data samples per pixel (depending on number of active columns) for an individual scene (final image pixels are the average of 1-7 scene pixels for the Galactic Plane images, 4-36 for the higher sensitivity and out-of-plane images, 2-30 for the IRAS gap images with the coverage being highly location dependent). 

To reduce telemetry rate, not all of the columns are active for each band. Astronomical images were not generated for band B because the short wavelength, centered on the 4.2 micron CO2 feature, and narrow spectral bandpass (0.25 microns) made it too insensitive to provide good quality data for imaging. The characteristics of the other bandpasses are:

Band Active Columns Isophotal Wavelength 50% peak Intensity Isophotal Bandwidth 0 mag Flux Absolute Photometric Accuracy Radiance Accuracy
A 8 8.28 6.8-10.8 3.36 58.55 5% 9%
C 4 12.13 11.1-13.2 1.72 26.51 3% 8%
D 4 14.65 13.5-15.9 2.23 18.29 4% 9%
E 2 21.34 18.2-25.1 6.24 8.75 6% 15%

where the wavelengths are in microns and the zero magnitude flux is in Janskys and is based on the Kurucz model for Sirius (Cohen et al. 1992). The calibration and photometric accuracy are discussed in detail by Egan et al. (1999). It is emphasized that the stated uncertainties are ABSOLUTE. The SPIRIT III instrument on MSX was extensively calibrated both pre-launch and on-orbit. The ground calibration was obtained in a cold chamber against NIST traceable sources. On-orbit calibration used stimulator flashes for the transfer of the ground calibration, and included standard stars, 80% of which were in common with ISO standards, and calibrated reference spheres released periodically during the mission. Independent calibrations of the point source photometry and the extended source radiometry were performed. The larger uncertainty in radiometry arises from conservative estimates of the error in transferring the ground calibration to orbit.

Observation Strategy

MSX images are constructed from data collected as the linear detector arrays are scanned across the sky in a "push broom" fashion. The spacecraft slew rates varied between 0.02 0.125 degrees per second. Multiple scans over a region were used to increase final sensitivity.

The Galactic plane images are produced by co-adding the data from 72 individual scans that mapped the plane. Each scan covers a one degree wide by 182 degree long swath in Galactic longitude at constant latitude. Forty-three observations covered quadrants I and IV of the Galaxy with the remainder covering the outer Galaxy. For |b| > 3 degrees, the scans were offset by 30 degrees in longitude. The number of overlapping observations decreases as |b| increases. As many as seven scans cover the inner Galaxy for |b| < 0.5 degrees, and as few as one at the latitude extremes of the survey. Galactic Plane survey observations were obtained between May 1996 and January 1997.

The higher sensitivity images of the Galactic plane are made up of 9-25 overlapping scans rastering over fields 1 by 3 to 5 by 5 degrees in size. Most of the images are constructed from scans at a constant Galactic longitude with a small offset applied between subsequent scans.

The non-Galactic plane images of star forming regions and galaxies are also coadditions of multiple (25 on average) scans over the imaged fields from 10 by 10 degrees for the LMC to 1.1 by 1.4 degrees for most of the galaxies. Most of these observations are scans along constant Right Ascension (the observation of M31 is an exception). For some of the larger maps, the scan coverage in the final images is variable with fewer observations at the edges of the fields.

The IRAS gap images were constructed from 160 degree long scans at constant ecliptic longitude (epoch 1984). Two sets of scans were performed covering the IRAS gaps at Ecliptic longitude ~162 (Gap 1) and ~342 (Gap 2) degrees using 37 and 27 scans, respectively. The scans for each gap were offset by 0.45 degrees at the Ecliptic plane; therefore, scan coverage is a function of Ecliptic latitude with the highest sensitvities towards the Ecliptic poles.

Data Processing

The images are generated from the time ordered data by first converting the data from raw counts to calibrated band-integrated radiances in units of W cm-2 sr-1. This process, labeled CONVERT (SPIRIT III DPC 1997), flags anomalous data samples, subtracts the dark current and applies gain and uniformity corrections for each pixel as a function of time and focal plane temperature, and then scales the result by the responsivity. Extensive post CONVERT processing produced significant improvement in dark offset subtraction trends, artifact removal and boresight pointing.

Image Construction

The data in radiance units are position tagged by the attitude history of spacecraft using the POINTING CONVERT software (SPIRIT III DPC 1998). The spacecraft attitude data were taken in the FK5 system and recorded in equinox 2000.0. The spacecraft attitude was improved by registering the observed mid-infrared point sources to the MSX infrared astrometric catalog (Egan & Price 1996). The rms uncertainty of the positional information is about one arcsecond.

The position tagged data for each scan are then placed onto a rectilinear image grid called scenes. For the Galactic plane images, Galactic coordinates and a Cartesian projection with l = 0, b = 0 as the projection center are used. The non-Galactic plane images use a tangent projection about the field center in Equatorial FK5 coordinates. The grid is populated by convolving the data with a Gaussian with a FWHM of 7.06 arcseconds. The pixel spacing is 6 arcseconds, which is the smallest pixel spacing the data sampling can support. The value of each image pixel is the average of the individual data samples weighted by the convolution with the Gaussian. Two arrays are generated: one containing the resulting radiance and one containing the convolved weights. These scenes were visually inspected to identify transients such as optical contamination from dust, Earth orbiting satellites and known asteroids.

Final Image Co-addition

The zodiacal emission was removed from each scene of the individual scans using an updated version of the CBZODY model described by Kennealy et al. (1993). The scenes were then overlaid and a weighted average taken for all the image pixels in common. The final images are on 6 arcsecond centers.

For the Galactic plane, the intent was to have the center pixel of each image on an even grid beginning at Galactic longitude and latitude of (0 deg, 0 deg). A Y2K bug led to the data being processed with pre-flight (default) mirror-fixed positions rather than the in-flight update. The fix to correct the astrometry shifted the CRPIX1 and CRPIX2 keywords from integer values. The image headers contain all relevant astrometric information, the isophotal wavelength of the bandpass and the mean modified Julian data of the observations used for that particular image.

The 72 arcsecond resolution images were constructed by convolving the 6 arcsecond spaced samples with a Gaussian smoothing kernel to produce images with 36 arcsecond sampling and 72 arcsecond resolution. All relevant data processing steps (including zodiacal light subtraction) have been incorporated into these images.

Image Sensitivity and Resolution

The resolution of the images is 20 arcseconds which is the inherent resolution of the instrument smoothed by the image construction kernel. The rms positional accuracy of the final images is 1-5 arcseconds (2 arcseconds rms for the Galactic plane survey images). The rms sensitivity of the survey images depends on the number of scans that were used to create the image, the scan rate and the dark noise for each of the scans. The focal plane temperature and, consequently, the dark noise increased with time during the mission. The effects of these changes were well calibrated; however, the scans taken late in the mission were 2-3 times noisier than early scans. The inner Galaxy scans were executed first and the inner Galaxy was covered by more scans than the outer Galaxy. Therefore, the outer Galaxy images are noisier than the inner. The rms noise (noise equivalent radiance) of the final Galactic plane survey images in units of MJy/sr) was found to be:

Inner Galaxy Outer Galaxy
Band |b0| = 0.0 1.5 3.0 4.5 0.0 1.5 3.0 4.5
A 1.9 1.9 2.2 3.3 2.6 2.8 3.3 4.8
C 5.9 5.9 6.6 9.8 6.6 7.9 10.9 11.8
D 3.6 3.9 4.2 6.5 5.4 5.4 7.1 8.9
E 13.3 13.3 14.4 21.0 16.6 16.6 22.1 29.8

In general, the 1 by 3 degree rasters of the Galactic plane are four times more sensitive than the corresponding Galactic plane survey images. The 3 by 3 degree image of the Rosette Nebula has comparable sensitivity and the 6 by 5 degree image of W3 is less sensitive due to the late date in the mission that those data were taken.

The IRAS Gap images have comparable sensitivity to the Galactic plane survey images at the Ecliptic plane. The sensitivity increases by a factor of 2.5 at the Ecliptic pole.

Tabular versions of the relative spectral response (RSR) curves for the 4 MSX bands of interest to astronomy are provided in IPAC table format in the links below. The curves are from The Midcourse Space Experiment Point Source Catalog V1.2 (Egan et al. 1999). Users are directed to Appendix A of that document for more information on the RSRs.

Conversion of Radiance to Flux Density

The images are in units of radiance (W m-2 sr-1); to convert to intensity (Jy sr-1) at the isophotal wavelength, the source brightness function S(nu) must be known. In the mid-infrared, emission lines and bands contribute significantly and S(nu) varies considerably with position. Care should be taken in transforming between radiance and intensity.

The conversion factors for a source brightness function that is inversely proportional to frequency (S(nu) ~ nu-1, isophotal assumption) are

Band  Conversion from (W m-2 sr-1) to (Jy sr-1)
A 7.133x1012
C 2.863x1013
D 3.216x1013
E 2.476x1013

Lookup tables of color correction factors to modify the above conversions for various types of sources are given in Egan et al. (1999).

Residual Artifacts

There are five significant artifacts that have been removed or compensated for in the data:

  1. detector saturation
  2. detector shadowing
  3. hashing
  4. artifacts around bright point sources
  5. Band C ripples

The dark offset from a pixel is reduced after saturation. The effect lasts about 100 seconds and the recovery is well represented by a linear trend whose parameters were deduced from data taken by overlapping scans in the opposite direction. Four Galactic Plane sources severely saturated (Eta Carinae, NML Cyg, RAFGL 2465 and IRAS 15576-5400) as did the center of M42 in Orion. Small residual effects may exist in the images for a degree or two on either side of the saturation sources and would appear as linear depressions constant in Galactic latitude.

Detector shadowing was more prevalent, especially for the Band E images. Shadowing is the depression in dark offset for a column of detectors when an empirically determined threshold of in-column radiance is exceeded. Shadows manifest themselves as depressions of the dark offset for all columns across a degree in Galactic latitude. Shadows for the higher sensitivity images are constant in Galactic longitude; the non-GP images have shadows perpendicular to the scan direction of the telescope. Shadowing has been corrected by applying a shadow correction function to the affected column of data. The shadow correction has been determined by comparing overlapping scans that do not contain the bright object (overlapping scans are nominally offset by 0.25 deg). The representative analytic function derived depends on the filter band and focal plane temperature. Residual shadows do remain in Band E images at the level of a telemetry count or less (< 1.1x10-6 W m-2 sr-1) and manifest themselves as a column of pixel values at a constant longitude with slightly lower values. The effect is well within the quoted radiance uncertainties.

Hashing are short stripes in the in-scan (lines of constant Galactic latitude for the GP survey) direction in the regions around bright sources. The hashing is due to "bleeding" in the detector chip. Glints from the filter bezel produce low level artifacts in both directions. Observations against CW Leo and NML Cyg to calibrate the near field side lobe response of the system show that the amplitude of the effect is about 0.1% of the peak response, at most. Thus, these artifacts are small and were left in the image as no secure model could be derived. The added uncertainty to diffuse background measurements around the very brightest sources is well within the stated radiance uncertainty.

Quasi-periodic low amplitude ripples, indicative of electronic interference, occasionally plague the Band C pixels. The ripples affect from 1/3 to all of the detectors in the array columns with one edge of the detector array usually being much more seriously affected. The frequency of the ripples varies smoothly and they appear as a region of bright and dark bands in the cross-scan direction, which broaden and then narrow again. The cause of the ripples is unknown and the variability of the phenomena precluded correcting for them. In general, we have omitted the worst of the ripples from the data used except where that data is the only data available. Users of the Band C images are cautioned to look for ripples in the images and to be cautious about doing extended photometry in regions affected by the ripples. The amplitude of the artifact is about one telemetry count or 2x10-6 W cm2 sr-1.

Note on Galactic Plane images downloaded prior to 1 September 2000

The MSX Galactic plane initial release images that were available from the IRSA site prior to 1 September 2000 have incorrect CRPIX1 and CRPIX2 values caused by a Y2K bug in the MSX processing pipeline. These images can be identified by the DATE keyword in the FITS header which is set to 1999. These images are offset by +3.77 arcseconds in Galactic longitude and +3.64 arcseconds in Galactic latitude from the true positions. The problem has been fixed and the correct images, with the DATE keyword set to the year 2000, are available. Please retrieve the corrected image if you have downloaded MSX Galactic Plane Survey images prior to 1 September 2000 or the DATE keyword in the FITS header contains the year 1999.


This page last updated on Wednesday, December 04, 2002 13:56