MPO Connections - Drift Scan Imaging (TDI)
An exciting technique is drift scan imaging. This involves keeping the telescope
stationary and letting the stars drift across the chip. The trick is that the camera is
rotated such that the stars drift from the "bottom" of the chip to the
"top". In other words, say you have the camera setup such that north is at the
top of the image and east is to the left. If you turn off the scopes drive and take
an image, the stars would drift from the left side to the right side of the image. Now,
rotate the camera 90° clockwise (when looking at it from behind) so that west is at the
top of the image and north was to the left. This causes stars to drift from the bottom to
The trick in drift scanning is that it takes into account that as the top line of the
CCD image is read, all the other lines move up one line. If the program reads a line at
the same rate that a star drifts from one line to the next, the star stays on the same
original line of pixels as it moves up the chip. When the given line reaches the top,
its read into a buffer. Every so often, a group of lines is written to a file as a
With this technique, you can let the sky drift into view of your camera for hours on
end. The net result is a series of images that form a strip 15° long for every hour you
keep the exposure going (if shooting at the equator; as you move away from there, the
stars move cos(declination) x 15°/hour. How wide that strip is depends on the width of
the chip and the focal length of your telescope.
As you might have guessed, the further from the equator you go, the more curved the
trails of the stars. There is a limit as to how far above or below the celestial equator
you can go before the trails are curved enough to show the effect. The field of view of
the camera, which is determined by the chip size and focal length of your scope determine
the limit. Only experimentation will tell you the real limit.
Automatic Image Parsing
From the above, you might guess that one could get a very large image, i.e., 765 pixels
wide by thousands of pixels high. Its pretty hard to work with an image under such
circumstances. Connections removes this difficulty by automatically writing the lines that
have been read to the camera to a standard FITS file at given intervals. That interval is
equal to the height of the chip, in pixels, at the binning setting used for the exposure.
For example, say youre using an ST-8 at 2x2 binning. The height of the chip is
510 pixels. All images saved by Connections will be 765 x 510 (765 being the width of the
chip at that binning). This is accomplished by forcing the exposure time to be an integral
multiple of the time it takes for a star to drift the height of the chip (with one
additional frame for "Dead Time"). See below for additional information on how
the exposure time is adjusted.
This process of writing to a file while still reading from the camera is accomplished
by copying the read buffer to a write buffer before the next image is started. The write
buffer is handled by a low priority "thread" (process within a process) that has
a minimum impact on the critical timing required for reading lines from the chip. Because
the next image is not started until the read buffer is copied over to the write buffer,
its possible that one line will be missed in each exposure. This depends on the
speed of your CPU, the amount of RAM (the more RAM, the less likely Windows is using
"virtual memory - a hard drive substituting for RAM), and if you have other programs
running that are demanding CPU time. The latter reason is the most likely to cause
problems so you should not be running other program while drift scanning, especially a
Each time you take a drift scan image, an initial set of lines is discarded. The reason
is that not all lines in the image would have the same amount of total integration time.
For example, say the chip is 100 lines high and it takes 100s for a star to drift across
the chip. If the lines were saved immediately after the exposure started, line 1 at the
moment the shutter was opened, would have 1s of integration, line 2 would have 2s, and so
on. Only the last line, line 100, would have the fully expected integration of 100s.
So, Connections automatically adds the time it takes for a star to drift the height of
the chip to the overall exposure, in this case, 100 seconds, and throws out all the lines
on the chip when the exposure started. This assures that all lines had the full
When the exposure first starts, the status line on the exposure progress form shows
"Dead Time", indicating that the program is reading lines from the camera at the
proper rate but not saving them.
As you change the settings for the focal length during drift scan setup, Connections
automatically calculates the scale of the image (arcseconds/pixel) which, in turn,
determines the time it takes a star to drift the height of the chip. Note that changing
the binning doesnt affect the time because the time is based solely on the physical
size of the chip, i.e. the Field of View.
When you enter a value in the Min. Exposure field, youre telling Connections you
want the exposure to be no less than the indicated time. Going back to the example that
the drift time is 100 seconds for a given setup, the minimum exposure is going to be 100s
(not counting the Dead Time mentioned above well get to that in a second). If
you enter anything less than 100, the program forces the exposure to be 100s. If you enter
110, thats 1.1 exposures. The program always goes up to the next highest integral
number, so the minimum is two exposures or 200s.