Camera following the stars - stacking
Sections in this page
On the previous page we saw how it is possible to take decent astrophotographs with a fixed camera and yet avoid stars trailing due to the Earth's rotation. Clearly it would be better to compensate for the rotation by moving the camera.
One easy way to make the camera follow the stars is to mount it piggy-back on a motorised telescope. (This is not the cheapest method - search the web for "barn door tracker" to see a more basic mechanism.) Some quite inexpensive motorised telescopes are now available and here is one with SLR camera mounted on a bracket, looking parallel to the telescope:
The telescope is a Meade ETX 125mm Maksutov-Cassegrain reflector with GO-TO controller (select the coordinates of a star or other object and the instrument drives to point at it). The camera is connected by a USB cable to a lap-top PC. Canon software (free with the camera) runs on the PC to fully control the camera and capture images directly to the PC's hard disc at regular intervals.
It must be admitted that the drives of lower-cost telescopes like this are not designed to cope with the weight of a digital SLR camera. Hence the home-made counter-balance weight which can be seen at the top left of the photo. Users of compact digital cameras will be better off in this respect.
In this configuration the telescope is used purely for guiding. It rotates about an axis which points to the celestial pole, and so is parallel to the Earth's axis. The wooden base in the photo is home-made and angled for the observer's latitude so that polar alignment can be achieved. This tilted-axis arrangement is called an equatorial mount (sometimes German Equatorial Mount, GEM). The drive goes at sidereal (ie, star) rate, which means that one complete rotation through 360° would take 4 minutes less than 24 hours. The 4 minutes take into account the fact that the Earth is also moving along its orbit around the Sun.
Without the home-made base converting to an equatorial configuration, the telescope shown above has a vertical and a horizontal axis. Placed on a flat solid base, that constitutes an altazimuth mount (short for altitude-azimuth, for the two axes). The telescope's electronics can be used in either equatorial or altazimuth mode. In the equatorial case the drive is simpler because it is about a single axis at a fixed rate. In the altazimuth configuration the electronics works out a changing ratio for driving each axis so that the telescope still follows the sky. That is fine for the visual observer but not so good for photography because although the aiming direction is correct, the field of view slowly rotates.
In the equatorial configuration, provided the drive is accurate enough, a series of photographs in piggy-back mode can then simply be averaged to get a result with less noise. In the altazimuth case that will not work because of the field rotation.
A principal reason for developing GRIP was to be able to align and average images automatically even in the presence of the field rotation due to an altazimuth mount. This has been achieved - the astro-process option on the batch menu does exactly that. In fact it goes much further and enables a series of photos from a fixed camera to be aligned and averaged automatically. GRIP finds the brightest stars in each photo and recognises the shapes formed by their connecting lines, from one frame of the series to the next. It is then able to map the photos accurately on top of each other and combine them.
A better way to make the camera follow the stars is to mount it by itself on a robust equatorial mount. I chose the HEQ5 type which is pretty solid but not too heavy to carry around regularly.
I strongly believe that this is the best next step for would-be astrophotographers, after the fixed camera method on the previous page: buy a solid equatorial mount rather than a telescope. There are plenty of things to photograph in this configuration. Several interesting things in the sky, such as the Andromeda Galaxy (M31), are too big to fit into a normal telescope's field of view. I did choose the HEQ5 with a view to adding a sizeable telescope later, but even after obtaining the telescope I still often take photos with the camera alone on the mount.
People have asked how the camera can be attached to the HEQ5 mount, given that it does not have a screw pointing up like an ordinary photographic tripod. My method is to use the removeable attachment plate from a photographic tripod. It is likely to have bevelled edges that can be clamped quite firmly in the bevelled channel on the HEQ5, as the following pictures show. This example is a plate from a light-weight Velbon tripod. I was able to buy spare plates so I have a few of them.
All of these were taken with Canon digital SLR on a basic fixed-rate HEQ5 mount as shown above. No telescope and no fancy guiding equipment were used.
You will almost certainly be trying to photograph objects that are fainter than can be seen with the naked-eye. So the aperture and ISO sensitivity settings will be as high as possible. A long total exposure time will be needed to reduce noise (improve signal-to-noise ratio for faint objects). Even with the camera on an equatorial mount driven at sidereal rate it is still necessary to align and stack multiple images, for several reasons.
- The polar axis of the mount will need to be very accurately aligned parallel to the Earth's axis. Otherwise the field of view will drift. The axis does not point exactly at Polaris, so aligning is not trivial to do and is unlikely to be perfect.
- The drive rate will not be perfect. It is likely to have small periodic variations due to the gear chain and perhaps also to the power supply. You get what you pay for of course but there is no need to buy the most expensive mount on the market because the errors can be managed by taking multiple exposures and using some software.
- Another source of similar errors in long exposures is wobble due to wind. Using multiple exposures we can simply reject bad ones caused by gusts.
- Especially from a suburban site, the image will saturate with background light within minutes. Therefore multiple frames will have to be added. That will have to be done in an accumulator that keeps more than 16 bits per channel (more about this later).
I use my own GRIP software for aligning and adding images. At the end of the process it gives a report on "drive errors". On the right is an example, from the camera-only photo of the Rosette nebula among the examples above.
This shows the average offset of the star pattern in x and y when going from each frame to the next in the sequence. This particular graph shows a slow drift in y due to polar alignment inaccuracy but a more irregular variation in x due to the drive about the polar axis.
The biggest step is about 4 pixels in x but less than 1 pixel in y. That is an offset from the start of taking one frame to the start of taking the next. The timed trigger was allowing 35 seconds for that, so most of the 4-pixel movement took place during the 30s exposure. The stars will therefore be blurred by about 4 pixels (in a frame that is over 5600 x 3700). With the 400m lens I was using that is equivalent to 13 arc seconds.
If I had tried to get the photo from a single very long exposure, other factors permitting, then the blurring would have been about 20 pixels in the x direction and slightly less in the y direction. That is much more noticeable.
You can also see that there is scope in this example for rejecting all those frames that have large steps in them. That would still leave about 38 frames to be processed, out of the original 49.
How multiple exposures can avoid background saturation and reduce noise is explained on my "Exposure settings" page.