Focusing Canon 100-400L for astrophotography

I have recently tried my new Canon 100-400L for astrophotography. It is a great lens, but it is much less forgiving to misfocus than a wide angle one. For example, with my Canon 17-85 lens at 85 mm I have made many sharp shots. The problem I encountered on the very first night is that 100-400's focus is rather sensitive - a small rotation of the focusing ring results in a large variation of the focus. Besides, the stars of the first magnitude looked too dim in the viewfinder thus preventing any reliable focusing. My 350D does not have the LiveView function, but I found out that my friend's 40D LiveView was not too helpful either because we were always missing the sharpest point as the focusing ring is overly sensitive and the stars were not very well visible. From the Photography-on-the-net forum I learned that somebody encountered similar problems with the 100-400. On the other hand, the 100-400L is known to be used successfully in astrophotography, see for example [1], [2], [3].

Background

I have looked for information on focusing and found a few good links, for example [4]. I decided to test with my lens the "Star Trail Test" (proposed in 1931) in combination with the "Digital Zoom Trial and Error". The star trails can be examined with the digital zoom on the camera LCD with no need for multiple trial-and-error exposures. The referred web page also describes the techniques published by Chuck Vaughn in his article "Astrophotography with Telephoto Lenses" in Sky and Telescope, February 1991 - scale on the focusing ring, minimizing backlash by approaching the focus point always in one direction. In summary, the techniques I have tried with my 100-400 are not new, but they have worked very well.

Setup

The problem of focusing can be considered from the system or black-box viewpoint with the input (argument, position of the focusing ring) and output (performance function, focus quality). Absolute values of the input and the output do not matter. What matter are the relative values at the input and the output and the input-output function, which has to be a one-to-one mapping. The linearity of the function does not matter either. Now when we manipulate the input, the output changes too much, which is visible on the final image, but not with the eye. We have to make sure that (1) we change the input in fine steps and (2) we carefully monitor the output so that we do not miss the extremal point (=best focus).InputI drew a scale with a computer (in Visio), printed it on paper (making sure that the scale is 1:1) and attached it to the lens focusing ring with transparent tape. Then I made a small carton pointer and attached it on the fixed part of the lens.Accuracy (absolute values and linearity) is not important in our case. After all, the focusing procedure has to be performed every night anyway and the absolute numbers do not have any special value. The USM (ultrasonic motor) focusing does not have any limit or reference point. If one turns the ring "beyond" the farthest infinity mark, then the focus point is lost when the ring is turned back. In fact, the procedure has to be repeated several times in cold nights as the focus depends on the temperature of the lens elements. Luckily the whole procedure takes only a few minutes.What is critical here is the ability to distinguish small rotations of the focusing ring. This property has three components. First, manipulation sensitivity - the ability to manually turn the ring by small angle (+/- 0.5 mm is not a problem). Second, the scale granularity (I used 1 mm grid and made readings in 0.5 mm steps). Third, repeatability, which is determined by the mechanical quality of the focusing mechanism (as my lens is pretty new, the backlash is less than 0.5 mm).OutputThe resulting image is analyzed visually on the camera's LCD with digital zoom. It may help to use some special masks and to check the image on the computer screen (probably with some focusing software calculating the thickness of the star trails).The camera and the lens were attached to my telescope equatorial mount in a usual way. No guiding is needed in this method though.Procedure

• 1. Select a star of 1-2m sufficiently high above the horizon (to have good image quality and the ability to see fairly faint stars), but not too close to Polaris (the star trails have to be clearly visible). It is good that in the same field of view of the lens will be stars of various magnitudes

  2. Set the lens to the infinity mark (slightly "before" the rough focus point), while setting the carton pointer to somewhere in the middle of the scale. Take the reading from the scale

  3. Aim at the selected start, set bulb exposure. Release shutter and after 10-15 sec close the lens with a piece carton or cloth (I used my hat). Turn the focusing ring by 1 mm (or 0.5 mm when fine focusing), open the lens and keep it open for another 10-15 sec. Repeat this procedure several times

  4. Analyze the star trails. As one moves closer to the sharpest point, the trails become thinner and the thinnest point is the focus.

Results

In this example I focused my 100-400L at 400mm f5.6. I started from mark "20" and took 1 mm steps until mark 26. As it can be seen in the figure, the sharpest point is mark 23 or 24.

Then I made the second series starting from 22 in 0.5mm steps until mark 25. Obviously, the sharpest point is 23.5. When approaching the focus, fainter stars start to appear thus helping is detecting the sharpest point.

Just a 0.5mm turn of the focusing ring results in a great difference in focus!

Here is the Moon photograph at the best focus (March 8, 2009 21:50 UT; 100-400L at 400mm, f5.6, ISO 100, 1/160s, no guiding, 100% crop, no postprocessing, no stacking), which looks decently sharp. There are two reasons for some softness in the image: (1) the night was quite foggy (hence the glare around the limb) and (2) at the 100% crop the camera pixels are becoming visible (motion blur due to Earth rotation is much smaller than one pixel). I should try it with the 1.4x extender on a better night. By no means this is the best photo, but I just wanted to illustrate the results: the image quality for the +/- 0.5mm focus was noticeably worse (one needs to flicker the images to find the subtle difference).

Future steps

The quality of the focusing procedure can be improved through better input manipulation and output observation:

• fine tuning of the focusing ring, for example, installing a vernier. With the 1mm grid I have not observed any prominent repeatability or backlash, but I expect these problems when working with the 0.1mm grid. I have prepared a 0.1 mm vernier, see the attached TIF file, if you print it in 600dpi it should appear in the correct scale. I will need a small LED and a magnifying glass to make reliable readings. This is the only improvement possible (scale granularity) because the other two components (manipulation sensitivity and repeatability/backlash) cannot be easily affected

  • employing a more sensitive in-focus point detection. Now I use LCD, but the camera can be also connected to a laptop (at home observatory) where it is easier to analyze the star trails visually or with the help of software. I will also test Hartmann mask as it was suggested by Chuck Vaughn