TN 0026: General Considerations for Drift Scanning

Author: Arne Henden
Date: 970307
Revision: #1 970307
Key Words: CCD, astrometry, observation, techniques

As opposed to stare-mode observations, drift scanning uses the rotation of the earth to scan strips in the sky. Since what the CCD sees is a projection of the curved sky into the flat plane of the CCD, a mapping takes place. Like a Mercator projection of the earth, there are distortions in this mapping that affect both the image shapes and the astrometric quality of drift scan images. This note will discuss some of these affects.

The scale of a CCD camera is given by

        scale (arcsec/pixel) = [206.265*p]/f
where p = pixel size in microns, and f = focal length in mm. For the TASS cameras, where p = 9 microns and f = 135mm, this gives a scale of 13.75 arcsec/mm.

Strictly speaking, drift scanning has no distortion if the scan follows a Great Circle (a line in the sky that forms a plane that passes through the center of the earth). There are two obvious great circles at any given location: the celestial equator and the meridian. If you used a one-dimensional CCD (that is, a 1xn format) and lined the long axis up so that it faced and was parallel to the celestial equator, it would image a 1-pixel high strip along the equator with no distortions and a scan rate time equal to

     rate (sec/pix) = 86164.09 (sec/sideral day)
                      ---------
                      [[360*3600 (arcsec)]/13.75(arcsec/pixel)]
or 0.9142 seconds/pixel at the equator. For 512 pixels as in a TASS camera, the integration time would be 0.9142*512, or 468 seconds. If you used a one-dimensional CCD and lined the long axis up so that it faced and was parallel to the meridian, then it would image an n-pixel high strip with no distortions but with only a one-pixel integration time.

If you are scanning away from the equator, the star takes longer to move the same angular distance on the chip. At the North Celestial Pole (NCP), a star takes an infinite amount of time to transit. The basic relation is

      rate{dec} (sec/pix) = rate{equator}
                             ------------
                             cos[dec]

Obviously, we have two-dimensional CCDs, and therein lies the problem. If you project the sky onto a two-dimensional flat surface, then near the equator you get nearly a square grid, but that becomes more and more distorted as you approach the NCP, where the grid lines become complete circles. For an arbitrary location away from the equator but nowhere near the pole, stars follow curved tracks across the CCD. For example, a star entering the CCD field of view at the northeast corner will track south until it crosses the meridian, and then track back north until it exits the CCD field of view at the northwest corner (assuming the chip is properly aligned). What you see in the final output image, in the declination direction, is a star that is spread out in the declination direction, where the length of the spread is proportional to the declination.

Likewise, in Right Ascension, stars near the top of the chip are at higher declination and therefore move slower across the chip than stars near the bottom of the chip. The scan rate is only correct for one declination in the field of view; all others will be incorrect. This means that there will be a smearing of images in the RA direction, with the smearing proportional to the declination distance away from the correctly scanned declination, and this smearing is further proportional to the center declination of the chip.

These two affects have the result of an output image that is smeared out, with a larger fwhm in both RA and Dec from equivalent images taken with the same camera in stare mode.

However, that is not the full story. Note that the Dec smearing is always towards the NCP, and is worse the closer you get to the NCP. This means that stars are less smeared near the bottom of the chip than the top, and therefore there is a scale magnification in Declination that is dependent on the central declination of the chip. Likewise, stars near the top of the chip are transiting slower than stars at the center, but are being 'pulled' forward by the selected scan rate. This results in the RA smearing to be towards the east at the top of the chip, and towards the west at the bottom of the chip. In other words, the image looks like it has been sheared.

The equations for the magnitude of the RA and Dec smearing are given in Stone, et. al., 1996 AJ 111, 1721. I've run a typical TASS camera through the equations with the results given in Table 1. At Dec=0 (the equator), the RA distortion (drne,drse) is of the same sign since the north edge faces the NCP and the south edge faces the SCP. As soon as you get above 1.5 degrees Declination, then both edges are facing the NCP and the smear starts having different signs. A similar affect occurs for the Declination smear (ddne,ddse).

The smear becomes the same size as a pixel by the time the central declination reaches 4 degrees. This is one reason why scanning +-3 degrees from the equator is the preferred strip with these cameras. Note, however, that the astrometric performance is compromised right from the beginning. The FASTT system described by Stone limits astrometric observations to the region where smear <= 1.7FWHM.

There are several methods to circumvent this astrometric problem:

  1. The RA and DEc smear shears can be analytically removed from the frames, so that, while the images are smeared, they at least reside at the proper central position. Note that this requires knowledge of which column has the correct scan rate.
  2. The smear is less for the central declinations of the chip, so you can 'shorten' the height of the chip in software.
  3. If you had a shutter, you can shorten the exposure time by only leaving the shutter open the same time duration as an equatorial scan. This means the stars follow a shorter arc on the CCD.

However, only options 1 and 2 are available to TASS cameras.

COMMENTS

Tom's February data showed a fwhm of around 3 pixels. Since his scans were at -1 degree on average, the smear due to drift scanning would be around 6 arcsec max, which is less than one pixel and probably does not contribute to the large observed image size.

The astrometric accuracy measured on Tom's nights was around 3arcsec. Note that the RA shear is about +-3arcsec, and so may be a major contributor to the error budget. I noticed in my reduction that there was a correlation between Declination and RA error, which is probably related to this shear.

TABLE 1.  SCAN DISTORTIONS FOR A TASS CAMERA
TASS camera Dec=768 RA=512 Scale=13.75
  Dec     drne      ddne      drse      ddse     rate
    0    -2.305     0.769    -2.305    -0.769   0.91416
    1    -5.448     1.293     0.839    -0.244   0.91430
    2    -8.593     1.819     3.986     0.280   0.91472
    3   -11.741     2.346     7.137     0.804   0.91542
    4   -14.894     2.874    10.294     1.329   0.91640
    5   -18.054     3.403    13.460     1.854   0.91766
    6   -21.223     3.935    16.636     2.381   0.91920
    7   -24.403     4.470    19.824     2.909   0.92103
    8   -27.596     5.007    23.027     3.439   0.92315
    9   -30.804     5.547    26.246     3.971   0.92556
   10   -34.029     6.091    29.483     4.506   0.92827
   11   -37.274     6.639    32.742     5.043   0.93127
   12   -40.540     7.191    36.023     5.584   0.93459
   13   -43.829     7.747    39.329     6.128   0.93821
   14   -47.144     8.309    42.662     6.676   0.94215
   15   -50.488     8.876    46.026     7.228   0.94641
   16   -53.863     9.449    49.422     7.785   0.95100
   17   -57.270    10.028    52.852     8.347   0.95593
   18   -60.714    10.614    56.321     8.914   0.96121
   19   -64.195    11.207    59.829     9.487   0.96684
   20   -67.719    11.808    63.381    10.067   0.97283
   21   -71.286    12.417    66.979    10.653   0.97920
   22   -74.901    13.036    70.626    11.247   0.98596
   23   -78.566    13.663    74.326    11.849   0.99311
   24   -82.285    14.301    78.082    12.459   1.00068
   25   -86.061    14.949    81.897    13.078   1.00867
   26   -89.898    15.609    85.777    13.706   1.01710
   27   -93.800    16.281    89.723    14.344   1.02599
   28   -97.772    16.966    93.742    14.994   1.03535
   29  -101.816    17.664    97.837    15.654   1.04521
   30  -105.938    18.377   102.013    16.327   1.05559
   31  -110.144    19.105   106.275    17.012   1.06649
   32  -114.437    19.849   110.628    17.712   1.07796
   33  -118.823    20.611   115.079    18.425   1.09002
   34  -123.309    21.392   119.632    19.155   1.10268
   35  -127.900    22.192   124.296    19.900   1.11599
   36  -132.603    23.013   129.077    20.663   1.12997
   37  -137.426    23.856   133.981    21.445   1.14466
   38  -142.375    24.723   139.018    22.246   1.16009
   39  -147.460    25.615   144.196    23.069   1.17631
   40  -152.689    26.534   149.523    23.914   1.19336
   41  -158.071    27.482   155.011    24.782   1.21128
   42  -163.616    28.461   160.670    25.676   1.23013
   43  -169.337    29.472   166.512    26.597   1.24996
   44  -175.245    30.519   172.548    27.547   1.27084
   45  -181.352    31.604   178.794    28.528   1.29282
   46  -187.674    32.729   185.265    29.542   1.31599
   47  -194.226    33.898   191.976    30.591   1.34042
   48  -201.024    35.114   198.946    31.678   1.36620
   49  -208.088    36.380   206.194    32.806   1.39342
   50  -215.437    37.701   213.742    33.978   1.42219
   51  -223.095    39.082   221.615    35.197   1.45262
   52  -231.086    40.526   229.838    36.468   1.48485
   53  -239.436    42.040   238.441    37.793   1.51901
   54  -248.178    43.630   247.457    39.177   1.55527
   55  -257.344    45.303   256.921    40.626   1.59380
   56  -266.972    47.066   266.875    42.145   1.63479
   57  -277.104    48.928   277.363    43.740   1.67848
   58  -287.787    50.899   288.437    45.419   1.72510
   59  -299.076    52.990   300.154    47.188   1.77494
   60  -311.029    55.215   312.581    49.058   1.82833
   61  -323.716    57.586   325.792    51.037   1.88561
   62  -337.214    60.122   339.872    53.136   1.94722
   63  -351.614    62.841   354.920    55.370   2.01362
   64  -367.017    65.766   371.048    57.752   2.08536
   65  -383.543    68.923   388.388    60.300   2.16310
   66  -401.330    72.344   407.095    63.032   2.24756
   67  -420.541    76.065   427.348    65.972   2.33962
   68  -441.365    80.129   449.361    69.146   2.44033
   69  -464.030    84.590   473.390    72.586   2.55091
   70  -488.806    89.512   499.742    76.328   2.67284
   71  -516.018    94.974   528.787    80.418   2.80790
   72  -546.062   101.072   560.981    84.908   2.95830
   73  -579.426   107.930   596.888    89.864   3.12672
   74  -616.711   115.705   637.215    95.365   3.31655
   75  -658.679   124.597   682.859   101.511   3.53206
   76  -706.294   134.874   734.977   108.426   3.77876
   77  -760.808   146.894   795.088   116.269   4.06383
   78  -823.864   161.149   865.224   125.246   4.39688
   79  -897.670   178.336   948.167   135.629   4.79099
   80  -985.253   199.475  1047.834   147.782   5.26446
   81 -1090.867   226.117  1169.906   162.211   5.84375
   82 -1220.669   260.742  1322.963   179.629   6.56854
   83 -1383.847   307.574  1520.575   201.084   7.50118
   84 -1594.551   374.406  1785.543   228.174   8.74560
   85 -1875.014   477.311  2159.283   263.467  10.48885
   86 -2258.602   655.075  2725.519   311.350  13.10507
   87 -2771.249  1026.156  3682.059   379.978  17.46722
   88 -3119.750  2088.530  5630.935   486.316  26.19418
   89   583.626  5578.479 11630.207   671.843  52.38039


Back to Index of Technical Notes
Back to TASS home page