Note: Since this article was written, a great deal of work has been done using negative color filters with CCD imaging. This topic is not discuss here but will be the subject of a future article. Also, new filters are available from several sources that have the infra-red rejection required for CCD imaging built in. See the SBIG web site for the latest information on these filters.
For 50 years I have used Wratten color filters for photography and have done everything from minor color correction to color separation photography. This includes everything from minor color correction for color films to three color separation photography. I believe that I understand color correction and three color photography quite well. I have reviewed the literature regarding color correction and three color work for astronomical imaging and summarize what I have found in the following. The topic of color filters has become of interest as a result of questions about the thickness of filters and the exposure times required in order to obtain correct color balance. A list of the thickness of some filters is appended at the end of this note.
I became interested after I purchased a set of color filters from Optec and found a rather large ratio of exposure times for the red, green and blue filters. Additionally, I have seen rather wide ranging recommendations for the sets of Wratten color filters to use one of which included the choice of an 80A. I was quite astonished by this recommendation since the 80A is a color correction filter and not a color separation filter at all. It must be noted that the choice of filters depends upon the application including elements about the spectral sensitivity of the film in the case of photography and of the CCD chip in the case of electronic imaging. There is probably no best set of color filters to use in general but there may be a best set to use when all factors are known.
In a recent experience, I heard a talk at the Florida Keys Winter Star Party in which astronomical images very much more blue that I have ever seen were presented with the comment that these images were more accurate than most of those seen in the past. They were quite beautiful. We do not know, in terms of what the eye might see exactly what color the many objects in the sky really are. Most dim extended objects appear gray to our eyes. I recently saw M57 through a 40" scope and for the first time I saw slight color. For bright objects, like stars and planets, we do of course see color. The lack of color is simply that the light from most extended objects is too dim for the eye to register color. The objects certainly have color and it may be captured with film and CCD imaging.
Color filters may be of two types. Multi-layer dichroic filters can have very sharp and well defined pass bands. Wratten filters are dye controlled filters and have pass bands which are in general much less sharp. As a starting point, and to limit this discussion, we will assume that a three color system of imaging will be used. This will require a set of three filters that are broadly Red, Green, and Blue. This contrasts to false color systems in which a band of the light spectrum may be assigned any color desired.
There are numerous sets of RGB filters that are traditionally used for three color imaging. The reconstitution of the image might be on a computer screen with RGB colors of their own hue and saturation or by printing processes which all have their own color representation and balance.
Common sets of RGB filters for photographic purposes are Wratten filters:
Some possible sets are: (in order RGB)
Separation filters 29, 61, 47
Tricolor filters 25, 58, 47
Because the filters will be represented by numbers, a brief note is interjected at this point to describe the filter colors as they are described in the Kodak manual:
23A, Light red.
25, Medium red. Red tricolor. For color separation work.
29, Deep red for use with No. 58
56, Light green.
57A, Medium green. (but lighter than the 58)
58, Medium green. Green tricolor. For color separation and tricolor printing.
61, Deep green tricolor. for use with Nos. 29 and 47
80A, Light blue. Color correction to convert from 3200 K to 5500 K light
(tungsten to daylight)
38A, Medium blue. Absorbs red and some green.
47, Deep blue. Blue tricolor. For color separation work with Nos. 29 and 61.
Because of the way these filters, all of which are pass band filters, (Except the 80A which is a correction filter used with color film.) behave in the pass and stop bands and particularly how rapidly they cut off, they will produce distinctly different color balance in the final image. Astronomical, printed color images rarely give all of the information necessary to be able to judge the accuracy of the color displayed. Often though the type of film, if color film, and the exposure times are given as well as the technique used to generate the image. Malin for example, gives great detail about these matters in his book. But even then the variety of photographic techniques used is largely directed to make nice looking images. That is, nice detail, nice color and the like. Photometric color accuracy is not usually the goal.
An example of deviation from the normal (traditional) filter set is suggested by Wallis and Provin, whose color images are quite wonderful. They suggest using the Wratten filter set for photography of 23A, 57A and 47. This set has a slightly lighter red filter and a slightly lighter green filter. Their filter set gives a smoother coverage of the spectrum when used with monochrome film and fills in the spectral gaps where there is appreciable astronomical information. They also recommend an exposure ratio of 1:1.5:2. This seems to me to be a wise choice of filters and exposures.
We must realize that the spectral response of the CCD chip is very different from that of photographic film. CCD chips of the types used in popular amateur imagers are highly sensitive to red and infra red and very insensitive to blue light. This means that images taken with the traditional filter sets used for photography will require very long exposures for the blue filter compared to the red filter in order to obtain good color balance. Additionally, the infra red portion of the spectrum must be suppressed or it will alter the color balance of the composite image. Note that newer blue sensitive chips are being worked on but are currently expensive and not generally available in amateur imagers. This situation has changed dramatically in the past year with the production of the Kodak "E" series of chips which have much better blue sensitivity. Also, it should be noted that many if not most filter sets now have built in infrared rejection. The latest information on these filters can be found on the SBIG web site.
Exposure ratios for the RGB of 1:2:4 or even 1:3:6 are not uncommon. It is generally required, for accurate color rendition, to use an infra-red reject filter since leakage of infra-red through the color filters will spoil the color balance of the image. Infra-red rejection filters are of two types. The "hot mirror" or multi-layer type has very sharp rejection characteristics and is generally considered the best. Another type is simply heat absorbing glass which works fairly well but has a very slow cut-off and not complete absorption of the infra-red wavelengths.
There is still considerable difference among professionals about the exact filter sets to use. Clearly it depends on the exact spectral sensitivity of the CCD chip and the spectral accuracy required by the particular application. All manufacturers of filters sets have shown images which are very beautiful. The color balance is controlled by exposure times and the color balancing attained in the reproduction of the images as much as in the original choice of filter set.
I have a set of filters from Optec, a company well known for its excellence in photometry, which has a light red filter, a normal green filter and a very dark blue filter when compared visually to the normal tricolor or separation filter sets. It appears that this filter set requires a very long blue exposure both because of the density of the blue filter and the lack of sensitivity of the typical CCD imager. I have seen a filter set consisting of 25, 58 and 38A recommended. This seems to me to be a very good filter set for CCD imaging since it has a slightly lighter blue filter than the 47 recommended for photographic film. Because the CCD chip is so insensitive to blue, the exposure for the blue image will not have to be unreasonably out of proportion to the red and green exposures with the lighter blue filter.
On the other hand Hoot has recommended a filter set consisting of a 23A,
56 and 80A. The 23A is a light red filter, the 56 is a light green filter and
the 80A is a very light blue filter. In fact the 80A is not really a band pass
filter at all but a color correction filter used to change 3200 K tungsten light
to daylight balance of 5500 K. It passes 80% blue and 20% red and is used to
correct daylight color film for use with 3200 K (tungsten) illumination. This
does not mean that this filter set is necessarily bad. This set has an exposure
ratio of 1:1:1. Such a ratio is useful since it is conserves some time to take
a blue exposure which is shorter than that usually recommended for blue filters.
The total time required for the blue image so often
nearly half of the total time. It was also stated that this filter set gives nice looking images rather than color accurate images. Since the red and green filters are also lighter than usually recommended, the whole set gives overall shorter exposure times. Since the blue filter is very light it gives an image which is heavy in blue but with considerable green and red. The very lightness of the blue filter makes up for the lack of sensitivity of the CCD chip in the blue region of the spectrum.
If this set of filters gives nice looking color images, that is fine. But, we should have no illusions that it is accurate color. For astronomical color imaging a great latitude is allowable since no side by side comparison can be made as it can with photographic imaging of earthly objects. We are happy to see images that are not just gray. A photometrically accurate set of color filters combined with the lack of blue sensitivity of the CCD chip requires an almost excessively long blue exposure. We may have to wait for the general availability of blue sensitive CCD chips to solve this problem.
Addendum (August 1998): Additional discussions about three color imaging has come up with still another set of filters that might be considered. The idea was to retain reasonable passband integrity but lighten the blue filter so that the total exposure time would be shortened. This set is 23A light red, 57A light green and 38A light blue. These filters have fairly broad passbands but still have stopbands which should give reasonable color separation.
Then again what is accurate color? Is it the image in one book or another, the color slides I saw a year ago, the color I see on the computer screen, the color I saw at the Winter Star Party or some other color balance? It's hard to tell. One might now say, "What about the color that color film shows?" Is that accurate? Not so simple to say. Color balance is a strong function of exposure time and reciprocity failure often dominates. Color film can give quite unbalanced color images as well. Professional films are made for short exposures of under 1/10 second and for exposures of several seconds to insure controlled color balance. It is probably best to judge a color image on the basis of its detail its range of colors and its general beauty. These are subjective judgments indeed.
Filter thickness is an important issue when taking three color images. The filters must be exactly the same thickness or there will be an image focus shift at the imager. Filters vary in thickness considerably. I have always made a practice of keeping a record of filter thickness for my photographic work.
For a reference, here are some values for a few 2" filters I use.
Lumicon Deep Sky Filter 2.55 mm
Lumicon UHC Filter 3.30 mm
Lumicon H alpha pass 2.18 mm
Lumicon Minus Violet 1.90 mm
Clear Filter 2.60
Clear filter 3.30
I keep these to match the Lumicon filters above when used in the Optec 2" filter slider.
Mirror (minus IR) 4.25 mm
This filter is used with all color filters to get rid of IR
filters Red, Green and Blue 2.00 mm
This set of color filters are each exactly the same thickness which is essential for color imaging.
Filter 2.5 mm
Hoya Orange Filter 1.94 mm
Hoya Red Filter 1.98 mm
Hoya Blue Filter 2.18 mm
The color filter thicknesses in the SBIG color wheel are the same and those in the Meade color wheel are not according to the information I have seen.
I hope that this information will be of some value to those considering color imaging. (11 September 1997 )
Return to Beginning
Return to Filters Index
Return to CCD Index
Go to Home Index for Doc G's Info Site