International Comet Quarterly
CCD Photometry of Comets
From the ICQ's 1997 Guide to Observing Comets, pages 94-104. Copyright 1997.
Below is a section of the ICQ's 1997 Guide to Observing Comets on CCD observation of comets, provided here due to the frequent requests for such information (and the fact the the 1997 edition is out of print, pending the forthcoming, greatly revised, second edition). Section 5.1 was written by D. W. E. Green (D.W.E.G.), with input from Petr Pravec (Ondrejov Observatory, Czech Republic), Stephen M. Larson (Lunar and Planetary Laboratory, University of Arizona), Gareth V. Williams (Smithsonian Astrophysical Observatory), Herman Mikuz (Crni Vrh Observatory, Slovenia), Bojan Dintinjana (Department of Physics, University of Ljubljana), Akimasa Nakamura (Kuma, Ehime, Japan), and Alain C. Porter (formerly at the California Institute of Technology and the National Optical Astronomy Observatory, when D.W.E.G. had considerable correspondence with Porter on the subject of extended-object photometry with CCDs). This was written in the mid-1990s, and the revised version of this chapter will have some revisions, but much of this text should still be very useful. Critical comments on this text are welcomed (send to icq@cfa.harvard.edu).
5.1. CCD observations Charge-coupled devices (CCDs) are having a tremendous impact on cometary astronomy, particularly with regard to astrometry and photometry. CCDs have already pretty much replaced the photograph in terms of astrometry and photometry (though a few observers will continue astrometric measurements from photographs for many years into the future). CCDs were developed as a new type of computer memory circuit in the effort to develop a "Picturephone" at Bell Laboratories in 1969-1970, and were soon seen as having immense promise as image detectors [cf. Janesick and Blouke 1987; McLean 1989]. While they were becoming an important factor in professional astronomy during the 1980s, it was not until the 1990s that CCD cameras became mass-produced at a rate inexpensive enough for amateur astronomers to be able to easily take advantage of. This point is key, because --- as a result of widespread CCD cameras in the hands of amateur astronomers --- far more comet observing (in the way of astrometry and photometry) is being done in the mid-1990s by amateurs than professionals! Prior to the 1990s, astrometry of comets was dominated by the professional community, and little photometry was being done outside of visual work by amateurs. [Professional photometry of comets prior to the 1990s was pretty much divided into two areas: (1) limited spectrophotometry of select comets, usually when they were brighter than magnitude 10 or so; and (2) rough, usually-underestimated total and "nuclear" magnitudes of comets contributed in a generally casual manner by cometary astrometrists.]
But by 1994, the market had exploded in CCD cameras --- led by the Santa Barbara Instrument Group (SBIG) in California [cf. di Cicco 1990, 1992, 1996] --- and one of the primary astronomical publishers (Sky Publishing) began issuing a quarterly magazine known as CCD Astronomy. [CCD Astronomy has ceased publication at the end of 1996, and is being incorporated into the pages of Sky and Telescope in 1997; the serious CCD observer should subscribe to Sky and Telescope to keep abreast of a field that is continually advancing.] It is difficult to know where electronic imaging technology will be in ten or twenty years, as microelectronic technology is advancing (and will continue to advance) at a good pace. So while some of this material may be outdated in only a few years' time, it will be of use to some observers to have a review of the current status of CCD observation of comets.
The underlying basis regarding CCD applicability to astronomy comes from Albert Einstein's discovery of the photoelectric effect, in which certain materials will eject an electron (generating an electric current) upon absorption of a photon of light. A semi-conductor is basically a piece of matter in which crystals exist that have both properties of good electrical conduction and good electrical insulation. An 'integrated circuit' (or 'chip') is made in such a way that impurities are injected into the semi-conductor crystals to produce tiny electric circuits and components. CCDs can be made from crystals of semi-conductors such as silicon or germanium, though most optical CCD cameras are silicon-based; in the infrared, silicon ceases to be sensitive, so that material such as InSb or HgCdTe is used for such longer-wavelength photon-detection arrays, often in "hybrids" that contain a layer of 'doped silicon' [McLean 1989; Hanner and Tokunaga 1991]. A CCD is photoelectronic imaging device componsed of many (250,000 or more) individual light-sensitive picture elements (pixels) that can be considered comparable to the silver grains of photographic emulsions. Each pixel can store electronic charges ('photoelectrons') that created by the absorption of light photons. The charges from each pixel are stored in the chip in sites defined by 'gates' (electrodes) on the surface of the CCD. After the exposure, these charges are transferred or 'coupled' from one pixel to the next by the controlled collapse and growth of adjacent storage sites or 'potential wells', by systematically changing the voltage across each gate. Ultimately, the 'charge packets' are counted by an output amplifier and converted to a form that can be used by computers for storage, analysis, and image display. [For a short, clear introduction to a CCD's design, see Janesick and Blouke 1987. For additional technical details on the construction of CCDs, see McLean 1989; Buil 1991; Sterken and Manfroid 1992b.]
CCDs are remarkable for their detective efficiency and their linear response to a wide range of intensities of light radiation. The term "quantum efficiency" --- a quantity that refers to the percentage of incident photons actually recorded --- has been used to compare the sensitivity of CCDs with other astronomical detectors. For example, the human eye has an overall "detected quantum efficiency" (DQE) of only about 15 percent, meaning that for every 100 photons of light striking the eye, only about 15 photons are "registered"; the human eye's rods (used in low-light night vision) have a DQE more like 25 percent, sometimes reaching nearly 50 percent [Hallett 1987], but the eye's rods can only "store" or "integrate" such photons for about a tenth of a second [Green 1985; Sterken and Manfroid 1992b, p. 20]. Integration/storage of direct photons from direct sources became possible in the photographic era, and 20th-century electronic detectors such as photomultipliers increased the DQE to 20-25 percent --- compared to a peak DQE of only about 1-5 percent for photographic emulsions [Sterken and Manfroid 1992b; Reipurth 1996]. Meanwhile, CCDs have a DQE of > 40 percent from x-ray wavelengths around 0.5 nm, through ultraviolet (10-300 nm), up to the near infrared around 900 nm [Janesick and Blouke 1987; McLean 1989]. Thinned, back illuminated and coated CCDs with good blue/ultraviolet sensitivity are still pretty expensive, and they will remain so for some time. Most good CCD suppliers will provide spectral sensitivity DQE curves for their chips. Table 5.1 shows the DQE for the standard BVR filters [Table produced by Mikuz; sources: Texas Instruments CCD Sensor Specifications; Kodak CCD Sensor Specifications] that are used in widely-available CCDs; the TC-211-255 series is used in SBIG ST-4, ST-5, and ST-6 CCD cameras, and they are more efficient at all wavelengths than the Kodak chips.
Table 5.1. Quantum efficiency of some CCD sensors in standard photometric passbands Standard Wavelength DQE (%) filter (nm) TC-211-255 KAF-0400 KAF-1600 ------------------------------------------------------------------- B 450 28 12 12 V 550 45 36 39 R 650 62 35 32
CCDs are exceptionally linear in their detection capability (unlike the human eye or photographic emulsions), meaning that the output signal from a CCD is directly proportional to the incoming light signal to 0.1 percent or better. This translates into the possibility of obtaining high-quality photometry, for example, of faint comets from relatively long CCD exposures compared with short-exposure images of much brighter standard stars --- but with the requisite action of timing the exposures accurately. This does not mean that CCDs are linear as a function of wavelength; while having a large spectral range (potentially from x-ray to near-infrared regions), the DQE varies widely throughout that range with numerous 'peaks' and 'troughs'. Unfortunately, different CCD chips can have widely-different spectral responses, leading to problems in unfiltered photometry and even in adapting filters for specific-bandpass photometry. CCDs are still rather small in surface area in comparison with photographs; the larger the CCD, the more expensive it is and the more computer memory it needs to cope with the processing of the data (though both of these problems may be improved upon in the near future).
The CCDs discussed here are "integrating" CCDs [di Cicco 1990], as opposed to the CCDs used in camcorders (video cameras; see section 5.1.1) and in still cameras. One seemingly-obvious project for a CCD-equipped observer is to hunt for new comets. Relatively few such discoveries have been made in the first 15 years of CCD availability to astronomers, and those have all been made by professional astronomers (with the exception of the recent discovery of P/1997 B1 by Takao Kobayashi of Japan) using electronic cameras in scanning mode (see section 4.4). Once difficulties involving small fields-of-view (and subsequent small amounts of monthly coverage), along with the massive software complexities of automatically scanning large areas of sky, are solved, regular all-sky CCD searches of the sky for comets may become routine in the first half of the 21st century [Marsden 1994a; Scotti 1994]. The amateur astronomer who is careful and dedicated may become successful at wide-field CCD "manual" scanning for comets with either telephoto camera lenses or small Schmidt telescopes; in such cases, multiple images (preferably a minimum of three, spaced by at least 30-60 min each) need to be taken on each night, to rule out the many "flaws", "ghosts", etc., that are so readily present on even well-processed CCD images.
Both beginning and advanced observers need to be aware of the many pitfalls that can be encountered regarding CCD photometry and astrometry of comets. Larson notes that observers need to be very critical of every step of their work. Looking out for such things as ghosts, poor guiding, poor focus, bad flat fields, incorrect time, standard-star identification, and noting sky conditions must be exercised at all times, or garbage data will surely result. Observers must know what they are doing well enough to assess if their results make sense (otherwise one can get erronenous results such as the splitting of comet 22P/Kopff as reported by Cremonese and Rembor [1996] and Cremonese [1996]).
5.1.1. Instrumentation, equipment
The beginning (novice) CCD observer needs to be knowledgeable about astronomy and about visual and/or photographic observation of comets to get a proper start in making useful astrometry and/or photometry of comets. A starting CCD observer has to put the camera to some telescope on an equatorial mount with a good clockdrive, which can be directed to a particular target easily (using a finder, divided circles, or a computer). Petr Pravec recommends starting observations (keeping a computer for the camera control in a room rather than in the telescope dome) by trying a lot of easy targets such as galaxies. When he/she is familiar with targeting, the observing procedure, getting flat fields, processing images, etc., he/she can consider starting observations of comets. The observer should decide what kind of work he/she wants to do. For astrometry (and also for photometry of small comets), those telescopes with focal lengths around 2 meters are suitable. For photometry of brighter comets, various focal-length instruments (from 0.2 to 1 meter) can be suitable. He/she has to find a good reduction software for the kind of work he/she wants to do. Then he/she must simply start observing and obtain experience through trial and error. Stephen Larson adds that the real challenge in getting even an experienced observer into serious CCD photometry and astrometry is to have all the tools available and easy to use. If a lot of time has to be spent figuring out poor documentation, surfing the net for software, etc., the enthusiasm is likely to quickly wane. Most observers will have relatively little trouble getting good images, but if it becomes more laborious to reduce the data, the potential may be lost.
CCDs are present in common camcorders that are normally used by the general public to videotape terrestrial places and events. The detectors in such camcorders have very limited response time, however --- similar to that of the human eye in terms of integration time (though not as dynamic in range or sensitivity as the human eye). Such videotaping can record the brighter stars at night, but not much more; as such, they may be show interesting images of bright comets only. One such project is to utilize a "time-lapse" feature on many camcorders, by mounting it on a tripod and taking images of a bright comet over time (say, one frame per minute), to show the comet's rising or setting with respect to the local horizon (or, in cases of close-approaching, fast-moving comets such as C/1996 B2, to show the comet's motion with respect to background stars over the course of an hour or two).
Many writers discussing how to get started in CCD observing [e.g., George 1995b, Holmes 1995] discuss the actual cameras, the taking of images and processing, and perhaps the telescope, but they often forget to mention the fact that one cannot do much without a reasonable computer to attach to the camera [I note di Cicco's 1990 article as an exception to this remark]! While some CCD cameras have recently come out on the market that do not need a separate computer, one cannot perform photometry, astrometry, or detailed image analysis without a separate computer. Unfortunately, most CCD software that one can currently purchase for astrometry, photometry, or image processing is only available for IBM-compatible or Apple-compatible 'personal computers' (PCs); fortunately, most amateur astronomers do not seem to mind putting up with the 'user-unfriendly' operating systems inherent in IBM-compatible 'PCs', but observers using other (non-PC) computers generally will be forced to write their own software. The easiest way to start up in CCD observations of comets, then, is to have an IBM-PC-compatible computer with VGA graphics, a 386 (or better) microprocessor, 2 megabytes of random access memory (RAM), MS-DOS 4.0 (or later versions), a Microsoft-compatible mouse, and a CD-ROM reader for the Hubble Space Telescope Guide Star Catalog (although Herbert Raab's "Astrometrica" software can also accomodate reference-star data from normal ASCII files). Furthermore, basic MS-DOS has only 640 kB of memory, and Raab recommends having at least 1 MB of extended XMS memory --- and more memory for larger images [e.g., Rogers 1995, Raab 1996].
In addition, for accurate astrometry, one must have an accurate clock set to Coordinated Universal Time (UTC), and this must be checked regularly against a reliable standard. It is usually necessary to record the time for an astrometric position to 0.00001 day (approximately 1 sec), and the proper recording of time (or lack thereof!) is the single largest source of errors in astrometry today [e.g., Marsden 1984]. One needs to accurately record both the beginning and end times of an exposure, and (if appropriate) report the average of the two times for a single position of an untrailed object.
As with visual and photographic work, if one wants to perform total photometry measurments on a particular comet, one will want a short-focus telescope (to avoid light loss due to aperture and focal-length effects) --- the appropriate aperture being dependent upon (and changing according to) the brightness of the comet. For CCD photometry, one wants a pixel to span 2"-3" or more. Conversely, if one wants to perform accurate astrometry measurements of a comet's changing position, a long-focus, larger-aperture telescope is to be preferred (to obtain the best resolution per pixel). A pixel scale of approximately 2"/pixel is needed; if pixels are much larger than 2", the quality of the positional measures will suffer; if pixels are much smaller than 2", the signal-to-noise ratio may be low --- each image being spread over a large number of pixels [see Williams 1996; Holmes 1995]. The Texas Instruments 'TC' chips that appear in many amateur cameras (including early SBIG cameras) have anti-blooming features built in to help in cases where a whole column of pixels can become saturated by a bright source, but such an anti-blooming feature can cause significant problems for the photometry of extended objects [Holmes 1995; Hawkins 1995].
Dennis di Cicco [1991] showed the astronomy world that SpectraSource Instruments provided capability for wide-field imaging of its Lynxx camera. It has been shown by Herman Mikuz (see images and data tabulation published in the ICQ in the past couple of years) that reasonable total-brightness photometry and imaging of large-scale tail phenomena in comets can be obtained with a CCD camera fronted with ordinary camera lenses. One solution to bright skyglow, which becomes problematical as a result of the typically-fast camera lenses and resulting wide field-of-view --- as employed by Baumgardner and Mendillo [1993], by Dupuy [1994], and by Mikuz [1996] and his colleagues --- is to use filters to narrow the wavelength region that is being recorded.
So we see that one must first consider proper telescopic and computer equipment before seriously considering the purchase of a CCD camera. Various companies produce numerous different models of CCD camera costing from several hundred to several thousand U.S. dollars in 1996, and one should consult the advertisements and the review articles in 1-2 years' worth of issues of the popular astronomy magazines when purchasing a camera. New cameras are being constantly developed and old ones revised, and no recommendations regarding specific brands or models will be given here.
Holmes [1995] notes that for extended celestial objects, one wants a CCD camera containing a chip that has low readout noise and low dark current. Though the DQE for the Kodak chips are lower than for the TI chips, the signal-to-noise ratio may be improved via the use of a binning option.
Note that for extended objects, to reduce noise in the use of CCDs (for photometry, etc.), one should use a short-focal-length telescope system. Fast f/-ratio creates a high-S/N image, as this will determine how fast one records the sky background. In general, the faster the better --- more images to co-add, short observing window, changing conditions: they all beg for fast systems response. Possibly a more important issue is the field size (is it large enough that all of the comet is recorded as well as a representative area of sky?
It is desirable for the derivation of total magnitudes of comets to use short-focal-length telescopes; this can keep the coma size small enough with respect to the field-of-view to have good representation of the background in most of the field. The significant gain can be obtained also from the fact that, for shorter focal length, more light from the coma is included in a single pixel. Since the noise of the background generally increases with decreasing focal length more slowly than the signal on a single pixel increases, the observer gets greater S/N and can resolve faint outer coma easier, and thus trace it to a larger distance from the nucleus. This allows him or her to derive the magnitude for larger diameters (Rlim). And, of course, there is no use in having a good resolution of the inner parts of the coma for the integration of total brightness; the scale of 20-100 pixels per coma diameter is the best range for most CCDs, and for faint comets this could be as small as 10 pixels per diameter. With a 1-meter-focal-length system and 18-micron pixels, this range corresponds to a range of coma diameters from 1'.2 to 6'.2, and means that for bright, large comets, even much shorter focus is suitable. Only for small, faint comets are larger telescopes with 2- to 4-m focal lengths needed.
Herman Mikuz has made great strides in pioneering the use of normal camera lenses for CCD photometry of comets. This is the best solution (and sometimes the only one) for bright comets. Comparative V photometry, obtained with camera lens + CCD and a 20-cm Baker-Schmidt camera + CCD yields the same results for a given comet. Regardless of comet brightness, the background stars must be obvious. Experiences shows that all background stars must be removed for fainter comets, and at least those of 12 mag and brighter for naked-eye comets. Mikuz and Dintinjana are considering the incorporation of an automatic star-removal option in the next version of the FitsPro software program (see section 5.1.2, below). There is a potential problem when a comet is very bright, in that the images can became saturated even on short exposures; this can be avoided by slightly defocusing the lens (however, the lens should remain defocused to the same amount until the whole photometric sequence is completed!). Larson notes that dynamic range will always be a limitation, so if one exposes to not saturate the central condensation, a bright comet will be less 'bothered' by field stars and background. Pravec adds that good flat-fielding must be done with a camera-lens/CCD setup. A combination of a 50-mm-focal-length lens and an 18-micron-pixel CCD camera is suitable for the range of coma diameters from 0.4 to 2 degrees (corresponding to 20-100 pixels per coma diameter). For a comet like C/1995 O1, this may be quite suitable; if a comet is brighter than mag 1 or so, only the presence of naked-eye stars in the coma are likely to present a problem (considering that the density of naked-eye stars is approximately 0.15 per square degree, it is not probable to have any real trouble with them, unless the bright comet in question is in the Milky Way). However, if there is a need to exclude a significant number of stars from the coma, this may present a problem. Especially for fainter comets, one might have to shoot the field when the comet is not there to adequately subtract field stars. (In this respect, there is not much difference between the photometry of bright and faint comets.)
5.1.2. CCD images and software
Once the CCD camera has collected light from astronomical objects and converted it to an electronic signal, processing must be done to attain a usuable image. Image-processing techniques developed at a fast pace during the 1980s, so that digital-enchancement techniques could be used to gain valuable information from old photographs of comets that were scanned electronically [cf. Larson et al. 1987]. The technique used by Larson and Sekanina to look at the jet activity in the inner coma of Halley's comet from 1910 photographs involves an intensity-derivative algorithm in which intensity gradients and variations are altered via both radial and rotational 'shifts' with respect to the nuclear region [cf. Sekanina 1987]. Some of these techniques have been developed and used to great benefit with more recent observations [Larson and Slaughter 1992].
The software accompanying CCD cameras generally will produce an image in a format that is considered 'standard', such as FITS or TIFF, or developed specifically for the given camera, such as SBIG or Starlight XPress. These formats code the information from each exposure taken with the camera, producing a formatted image that is used for whatever photometric or astrometic measurements that the observer wishes to undertake. For purposes of publication of actual images of comets in the ICQ, observers should send UUENCODEd FITS-, JPEG-, or GIF-format images; do not use Base64 encoding. As noted below under reporting of astrometry, one should never use (or distribute) JPEG or GIF images if any intention of serious photometric/astrometric measuring exists.
With CCDs, one frequently has the advantage of vastly-increased quantum efficiency so that short exposures can be taken, but with fainter comets this will not work, and the observer must either take numerous short exposures and 'stack' them in the processing stages or else track on the motion of the comet. Some cameras (such as the SBIG series) have autoguiding capabilities for moving objects, but Paul Mortfield [Mortfield invites interested observers to contact him for a copy of his program, written for IBM-PC-compatible computers; his address is P.O. Box 2598; Cupertino, CA 95014; U.S.A. (e-mail clockwrk@cup.portal.com)] found (1996) that --- for fast-moving, large comet C/1996 B2 (Hyakutake) --- his best solution was to write his own computer program "that allows [his SBIG ST-4 camera] to step guide when [he inputs] the comet's direction of motion and speed. . . . Step-guiding allows the ST-4 to lock on to a star and make the telescope track the comet's motion." Professional observatories frequently have autoguiding features that allow one to track at a pre-set rate of cometary motion [e.g., cf. Meech et al. 1995].
Determining the background-sky level is very critical for photometry (especially for faint comets). If the procedure is embedded in a software's so-called 'black box', one should be very wary of using such software. It is best to determine sky level manually in terms of reliability, though this can take much time. On the other hand, as most software supplied with the present family of commercial CCDs include basic image-processing routines such as 'bias', 'dark', 'flat', etc. [Newberry 1995a, 1995b, 1996; George 1995a], a large problem may lie not in the software but rather in each observer's knowledge and habits; if such routines are not followed in proper order or manner (or if some routines are omitted, such as flat-fielding), then errors in photometry are likely to result. The best solution is to use image-processing software that is used by professionals or at least widely tested by different observers. Mikuz used the PCVISTA software for 3 years [Mikuz and Dintinjana 1994] before switching to DAOPHOT II for PCs (adopted by B. Dintinjana); both programs were found to give the same results, but DAOPHOT II is more suitable because it works even with large apertures that enable photometry of brighter comets to be performed (DAOPHOT II is found easier to operate). The original DAOPHOT was written at the Dominion Astrophysical Observatory by Peter Stetson [1987]. In the meantime, Dintinjana wrote his FitsPro image-processing software for MS Windows, which includes an aperture-photometry routine. During most of 1996, Mikuz's testing of FitsPro shows that it gives exactly the same results as DAOPHOT II for PCs (providing the aperture was also the same). FitsPro has further advantages such as faster reduction and user-friendly operation through the pop-up menus; it includes also additional image-processing routines that are necessary for processing and reduction of comet images. The program is now available on-line at their World Wide Web site [http://david.fiz.uni-lj.si/astro/comets/ccdphot/ccdphot.html].
Larson notes that it is important for observers to use the same standard reduction algorithms in photometry, as much as possible. The tools in IRAF are the de facto standard among professionals, for example, and for consistency, it would be a good idea if observers used the same software. Now that the average observer has access to affordable and powerful PCs, one can install IRAF on a PC that is running LINUX --- and if an amateur astronomer has a 150-MHz Pentium (or faster) chip, it will rival professional-astronomy workstations. Unfortunately, there is no clear standard among professionals about the parameters to use in sky subtraction, etc. Perhaps it is time to develop a standard for such reductions.
Software packages such as "Astrometrica" are widely used by amateur cometary CCD photometrists. Point-spread-function (PSF) rountines are designed for point sources such as stars, and should not be used for cometary photometry. Likewise, CCDs with anti-blooming gates are not advisable for cometary photometry. At the Ondrejov Observatory in the Czech Republic, Petr Pravec uses his own unpublished software for photometry; his software determines the background level in a different way from that of the software package MIDAS, but Pravec finds that there is a good correspondence between them. MIDAS can derive the background as the mean of pixel signals in user-specified area(s), which is good for photometry of comets; though it is also possible to derive the background as the mean level of pixel signals in the annulus around the aperture centered on the object, this is less suitable because it is difficult to keep it far enough from the comet so as to have no contribution from faint outer parts of the coma (and to not include background stars in the anulus). Pravec's software package generally determines the background in an iterative way, such that in each step it rejects pixels that it recognizes as not corresponding to the background (assuming the Gaussian distribution of the background signal around the mean level), until all remaining pixels are considered to represent the background; Pravec finds that this works well for most fields.
5.1.3. CCD astrometry
The most remarkable effect regarding cometary observations resulting from the wide-spread availability of CCDs to amateur astronomers in the 1990s has been a great increase in both the quantity and quality of astrometric measurements of comets [Marsden 1995d]. Observers in light-polluted areas, and those who were daunted by the laborious procedures involved in photographic astrometry, have now found a relatively easy way to make significant contributions to cometary science. Comets are now being followed much more extensively at fainter magnitudes via astrometry than at any time in the past. Details on astrometric measurement and reporting of observations will be found in Chapter 7.
One issue that has come up regarding CCD observations of comets is that of magnitude determination; this is mentioned here because many astrometric observers are reporting magnitudes with their observations, especially as derived from the Raab and Rogers software noted below. Some work needs to be undertaken to assess the reliability of the photometry aspects of these PC-type astrometry programs. In general, those astrometrists who contribute photometry of comets tend to give magnitudes that are fainter than those who are solely doing CCD photometry of comets; this is largely because smaller integration times are usually employed for astrometric uses, so as to not "burn out" the nuclear region (i.e., the astrometrist wants a small nuclear condensation for an easier and more accurate measurement of position). Also, in order to obtain as many observations of as many objects as possible, astrometrists usually do not use filters because of the longer integration times needed to collect a useful quantity of light (good signal-to-noise ratio) for measurement. Prior to the wide availability of CCDs, photometry of faint comets was nearly non-existent; as such, astrometrists are encouraged to contribute magnitudes under the philosophy that "some magnitude is better than nothing", but longer integrations (with or without the use of V filters) are encouraged where possible to gain a better measurement of the total coma magnitude --- meaning that separate exposures need to be taken ideally for astrometry and photometry.
Akimasa Nakamura relates that CCD observations do not give a real centroid, because almost all comets have a 'skewed' central condensation due to tails, jets, or other asymmetric structure in the inner coma; the centroid derived from several pixels tends to be pulled toward any brighter position angle regarding inner-coma material. To avoid this effect, measurers should use as few pixels as possible to calculate the centroid. Pravec warns that the observer should use a good and consistent procedure of centroiding point-source images (stars) and extended sources with condensations (not necessarily radially symmetric), and understand the gnomonic projection fitted to the reference stars, with regard to their distribution in the field.
5.1.4. CCD photometry
CCD photometry of comets is much more complex than is astrometry, for two chief reasons. First, there are problems both with proper integration times and with the proper subtraction of non-comet light (background, contribution from stars near or within the coma, etc.). Second, one ideally wishes to measure a physically-meaningful amount of light from the comet. Unfortunately, no two comets are alike in physical composition, and one gets widely-varying dust-to-gas ratios, which appears in the spectrum as gaseous, fluoresced emission lines superimposed on an underlying 'continuum' (see Figure 5.1).
Photoelectric photometrists need to correct properly for atmospheric extinction using a procedure more rigorous than that used for visual observers [cf. Chapter 4 and Appendix E]. Numerous widely-available books on astronomical photometry describe the procedures [e.g., Henden and Kaitchuck 1982; Buil 1991; Sterken and Manfroid 1992b; Straizys 1992] and should be consulted. [See also the concerns of Carsenty et al. 1987 regarding standard extinction models.]
5.1.4.0. CCD bandwidth response and the use of filters
Michael A'Hearn [1981a, b] proposed a standard filter set for cometary photometry that was designed to isolate primary emissions in the optical region. The fluoresced gaseous emissions of CN at 387 nm, C3 at 406 nm, and C2 at 512 nm were chosen, each with wavelength bandpasses of approximately 5-12.5 nm. Two regions were also chosen to sample the solar-reflected comet dust, called the spectral 'continuum', at 365 and 485 nm [though this can be problematical due to 'contamination' of the continuum from numerous potentially-present gaseous emission lines in this region; cf. A'Hearn 1982]. Sets of these filters were made and distributed as part of the International Halley Watch.
Other filters for amateur comet observing were introduced in the 1980s by a California-based company known as Lumicon [cf. Morris 1984, Marling 1984]. To my knowledge, these have been used visually by most observers. Some interesting photometric results might be obtained from a dedicated program of observing many comets on many nights via CCD with such filters (with detailed records, of course!).
Holmes [1995] notes that some thought may need to be put into the type of CCD camera if one uses narrowband filters, because of the dark current in some chips. But for broadband photometry, this should not be as much of a factor. It is possible to use standard UBVRI filters [cf. Bessell 1995] in CCD photometry of comets, and indeed various observers are already routinely publishing (in the ICQ) V total magnitudes of comets that are obtained with CCDs. Due to widely-varying sensitivities of CCD chips as a function of wavelength, as Bessell further notes, the same filter combination that yields a standard V bandpass for an old 1P21 photomultiplier tube will not yield identical magnitude data when used in front of a CCD chip.
In the visible region of the spectrum, this continuum predominates in comets that are far from the sun (where dust emission overwhelms the gaseous molecules in the visible coma). Closer to the sun, the intensity of a comet's gaseous emissions can be comparable to, or much greater than, the intensity of the continuum --- and this ratio varies from comet to comet. Comet C/1996 B2 (Hyakutake) was a very gaseous comet with relatively little dust emission, and in the visible region of the spectrum, the visible coma was due primarily to the so-called Swan bands diatomic carbon (C2). The problem with choosing filters of any type for comets is that one cannot be certain (without accompanying spectroscopic analyses) of what components are being measured (gas vs. dust). The issue of narrow-band photometry is discussed in section 5.3.
Because the DQE of the Kodak-KAF-series CCDs is only half or less in the standard bands R and B, and about 80 percent in the V band when compared to the ST-6 CCD, the Kodak detectors are less suitable for CCD photometry (though the Kodak ones, with 9-micron pixels, are suitable for astrometry, where high resolution is more important than sensitivity). Furthermore, the TC-241 chip (used in the SBIG ST-6 camera) has an additional advantage due to larger pixels, which are able to collect more photons on a single pixel, resulting in additional improvement of S/N ratio. Herman Mikuz and Bojan Dintinjana conclude that the combination of a short-focus telescope and large-pixel detector is more suitable for the CCD photometry of extended objects.
Signal-to-noise (S/N) is very important --- especially for faint comets. Readout noise has improved since the CCD "explosion" into the astronomical community in the mid-1980s, and this is now usually less significant than shot noise. Aperture and exposure are important, but a good guide is the sky background level. Larson and colleagues find that the best way to improve S/N is to co-add several exposures with sky background at a few times the bias level, to reduce the "random" shot and readout noise by the square root of the number of images used; this greatly improves discrimination of the comet over the sky. It also allows exclusion of poor images due to temporary bad seeing, guiding errors, etc.
5.1.4.1. CCD magnitudes (broad-band photometry)
There is a long history of total visual magnitude estimation of comets [cf. Green 1996c], and there is much value in a continued archiving of the perceived total brightness of a comet's coma in the optical (as well as infrared) bands. A problem still in need of some resolution surrounds the question of how CCD magnitudes --- whether filtered (say, V or R) or unfiltered --- relate to visual magnitudes [e.g., Green 1996b]. The recent return of Halley's comet was the first cometary apparition to yield a large amount of CCD photometry [Green 1986]. Near the 'transition area' for comets, which is when they are near mag 12-15 (toward the limit of visual observations), it has been noted that CCD total magnitudes are typically 1-3 magnitudes fainter than visual estimates. This may be partly due to a combination of short astrometric exposures and generally longer focal lengths and higher apertures employed by CCD observers --- resulting in larger coma sizes for visual observers than their CCD counterparts --- and also to problems with true background vs. outer coma contamination (see discussion below), but it may be heavily related to the R-band vs. V-band responses of the CCD and human eye. The scotopic and photopic responses are much closer to the solar peak (and includes the C2 Swan bands if they are there). Larson notes that a useful project for visual observers would be to use solar analog stars in deriving magnitudes, and then perform systematic photopic and R-band wide-angle CCD photometry to quantify the difference.
Note also that broad-band magnitudes have limited use in the study of cometary brightness and activity, except in cases where a large historical archive makes such data useful (as with V vs. visual); narrow-band magnitudes are discussed more fully in section 5.3. While it is recommended that observers use V filters for their CCD photometry of comets, exposure times a few times longer are generally required with a V filter to get the same coma diameter as obtained with an unfiltered CCD. It has been found by some observers that V and unfiltered magnitudes of comets do not differ by more than several tenths of a magnitude, and it is far better to obtain unfiltered magnitudes than none at all. Some professional astronomers routinely perform cometary photometry with R filters, but there are not good stellar catalogues of R magnitudes spread around the sky in the way that Vmagnitudes are. Furthermore, many professional astronomers who concentrate on cometary photometry tend to use small 'diaphram' sizes to obtain the brightness of only the inner coma [e.g., Jewitt 1991] --- and indeed with large coma diameters, it may be impossible to integrate the entire brightness with a telescopic CCD camera. But such studies are limited to a very few comets on a very few nights, and more useful and more meaningful physical results are likely to be obtained by integrated the entire visible coma.
It is recommended that, whenever possible, observers doing CCD photometry of comets perform a variety of exposures with V filters. Note that standard V photometry of stars must be used for calibration; if R filters are employed, then standard R star photometry must be used for calibration (and so on for other broad-band photometry). Larson notes that unfiltered thick, uncoated CCDs most closely approximate an R-band for a reasonably flat spectral distribution like the solar continuum. For comets, derived magnitudes will be reasonably accurate for only intrinsically dusty comets, or distant comets with no gas emissions. For dust-poor, bright comets --- such as C/1983 H1 (IRAS-Araki-Alcock) --- C2 emissions may really skew the data. Larson routinely uses a Cousins R-band filter to ensure that the continuum dominates (even though there will be C2, NH2, O I, and H2O+ in the bright comets), and to maintain a more standard band that others observers' data can be correlated with.
Exposure may have significant effect on CCD magnitudes, and the ICQ has instituted a way to allow one to record the exposure information (in seconds) in the tabulated data (in the same columns as used for magnification for visual observations); all CCD observers are encouraged to report exposure times (whether filtered or unfiltered). Those CCD observers without filters are strongly encouraged to only use comparison stars of solar-type (G) or earlier (bluer). Faint comets should have longer integrations, perhaps up to 15 min or longer, but one again needs to experiment to see how much more coma is detected with various exposure times. There is a problem with dense star backgrounds, because a reliable procedure is needed to subtract stars that may be present inside the integrated coma of the comet. Akimasa Nakamura uses software named 'IP-Lab', in which the contrast of the image can be changed easily, but he is very careful in deciding the coma size (i.e., the aperture size); when Nakamura changes the contrast to be as high as possible, very faint outer coma can be shown. With his f/6 system, Nakamura uses exposures that are usually 2 min long (4 min for faint object), though he is interested in performing astrometry (and thus not overexposing the central condensation) and in taking as many exposures of objects per night as possible.
Although the spectral response curves of CCDs peak generally in red, they are very wide (some 400 nm FWHM). Thus, considering them as R-band detectors is a very rough approximation; this is true even for observations of a more-or-less continuous spectrum (such as a stellar spectrum), and much more so for emission spectra. For example, when a very cool star is imaged with CCD through a clear filter, its infrared energy excess makes it brighter with respect to other field stars than its corresponding R magnitude might suggest. Emissions in various spectral regions can make the "unfiltered CCD magnitude" even more different (greater or smaller) from R magnitude. On the other hand, if someone can assume that a spectrum of the particular object is continuous with no significant excess in any part of that spectrum between 400 and 1000 nm (as with a cometary dusty coma), then considering CCDs as R-band detectors is possible. Moreover, as long as the spectral reflectance of the coma is independent of wavelength (i.e., it has a neutral color and the spectrum of the reflected light is the same as the solar spectrum), then it is possible to use a nearly-solar-color star (a solar analog is the best choice) to calibrate the cometary images and estimate comet's V magnitude even from the unfiltered CCD images. Observations indicate that spectral reflectances of comae are really almost neutral, so using the solar-color V standards for calibrations of unfiltered CCD images of dusty comets (which often means those not too close to the sun) is probably a good way to get their V magnitude estimates. Petr Pravec's experiments indicate that the differences between V and "C" (unfiltered) magnitudes are not greater than 0.1 mag in most cases of observations at r > 1.5 AU (he lacks data for comets at smaller heliocentric distances); thus, for faint comets at greater heliocentric distances, the use of properly-calibrated unfiltered CCD images for estimations of comet V magnitudes may be possible. Though most (but not all) CCDs are dominantly sensitive in the R band (and generally, one may consider them as a R-band detectors), this is of little or no use when we are dealing with the standard UBVRI photometric system, where each passband is strictly defined.
Regarding the importance of S/N ratio for broadband photometry of comets, Petr Pravec provides the following thoughts. In the following example, let us assume that the signal of individual background pixels has a Gaussian distribution with a mean value of b and standard deviation of Nb. Let us further assume that the surface brightness in the observed coma decreases as R-1, where R is the distance from the nucleus. The signal from the coma in the given pixel at distance R is S(R) = S0/R. The integrated signal over all pixels from 0 to distance R from the nucleus is then I(R) = 2(pi)(S0)R. We can trace the cometary coma in the CCD image, when it is properly displayed on the screen, out to the distance (Rlim) at which the signal S from the coma on individual pixels decreases down to S approximately Slim, corresponding to Slim/Nb approximately 1. (The observer can trace the coma sometimes even below the level of S/N = 1, but let us take Slim = Nb as a reasonable limit in the following.) The distance Rlim is the observed radius of the coma, and we take it here in units of pixel size. The quantity S0 (the signal in the unit distance from the nucleus) is then S0 = (Nb)Rlim. The total integrated signal from the coma is then Itot = I(Rlim) = 2(pi)(Nb)Rlim2. The signal-to-noise ratio at which the integrated coma is detected is then Itot/[(Nb)[(pi)(Rlim2)](1/2)] = Itot/[(Nb)(Rlim)[(pi)(1/2)]] = 2(pi)[(0.5)Rlim] = 3.545(Rlim), which means that if Rlim is greater than 10 pixels, then the error of the integrated signal caused by background noise is smaller than 3 percent --- i.e., negligible from the point-of-view of determination of the total magnitude.
Although the above example shows that, while the error of the total integrated brightness caused directly by the background noise is unimportant, there are more critical parameters, that are also related to the backround. A very significant error can be caused if the mean value b of the background level (which is subtracted from each pixel signal value before integrating the net signal of the coma) is determined with even a small error. Let us assume that the value for b derived by the observer (b') is greater by Nb/2 than b (so that b' = b + Nb/2). Then, considering the above example, the observer derives the total brightness of Itot' = Itot -- (pi)(Nb/2)Rlim2 = 0.75Itot. Thus, the total integrated magnitude derived by the observer is fainter by 0.3 mag than the true total magnitude. Moreover, if the observer integrates the coma behind Rlim --- in the hope of including the faintest parts of coma behind the radius where he can trace the coma on the screen --- the error caused by incorrect determination of the background level b increases quite rapidly. So the accurate determination of the background level is of principal importance for cometary photometry.
If the observer cuts down the noise (e.g., by using a lower-noise camera or by taking a longer exposure), he can trace the coma to a greater limiting radius. For the R-1 coma profile in the above example, one has to increase Rlim by a factor of 2.512 to get the total integrated magnitude brighter by 1 magnitude. This is also the factor by which the observer has to decrease the noise in the individual pixel (Nb) to get Rlim so great. Assuming that Nb increases with the square root of the exposure time (while the signal increases proportionally to the exposure time), one has to increase the exposure time 6.3 times to get Rlim greater by that factor. However, it is not so straightforward in a real situation, because (1) there are also other noise sources present, (2) the size of the integrated coma is limited by the size of the field-of-view, and (3) sometimes also the saturation level does not allow the observer to increase the integration time.
In CCD photometry, the information on the radius to which the coma brightness was integrated is of the same importance as the derived value of the magnitude itself. Attempts to derive the magnitude for radii of Rlim, Rlim/2, and 2Rlim (although the last one can be difficult sometimes) should always be welcomed, as they allow one to assess the slope of the integrated brightness increase with R, and also to check if the reduction procedure seems proper.
Another important problem is that the distribution of the background signal may be non-Gaussian (in fact, it really is, because there are stars projected in the coma). This non-Gaussian background distribution is usually corrected for by exclusion of the pixels in which the background stars are located from the integration of the coma. One must remember, however, that a residual effect can still remain there in the case of dense star fields. Another problem appears when the image is not correctly flatfielded; this can change both the distribution of the background signal and the mean background level b in the area of the comet. So correct flatfielding is quite necessary for cometary photometry.
Thus, the derivation of the correct background level is of principal importance for cometary photometry. This is a problem concerning both the observational technique and the reduction procedure. The coma should always be small enough in comparison with the field-of-view (FOV) to have a lot of pixels far from the comet available for proper derivation of the background level. Pravec's experience suggests that generally Rlim should not exceed 0.1 of the FOV in size, so that (for example) a small FOV of 15' by 15' realistically limits the CCD photometry to comets that are not larger than several arcminutes in diameter. For comets with coma diameters approximately 5' or greater, the observer with a small FOV cannot decide reliably where the apparent coma ends and cannot be sure that the pixels around the coma do really represent the background level. As both effects make the magnitude estimates of larger comae generally too faint, this is likely to be one of the reasons why some CCD observers are consistently producing fainter magnitudes than are visual observers. If a CCD diameter is much smaller than the visual diameters for a comet, it could mean that the CCD observer is stopping the integrations at diameters that too small. Thus, is important to know that in the case of large coma diameters, the observer might be easily fooled into thinking that the surface brightness of the outer coma is the "sky background" --- in which case, the derived magnitude will be fainter than the real one. Nakamura notes that if the coma size is larger than half of the field-of-view, he does not measure the magnitude for fear of mismeasurement.
At Crni Vrh Observatory in Slovenia, Mikuz and Dintinjana [1994] found that a 15th-magnitude comet requires a 300-sec exposure with their 20-cm f/2 Baker-Schmidt camera and V filter to achieve a S/N ratio of 18 (corresponding to an error of 0.06 mag), which they consider still acceptable for a good photometry. Similarly, other CCD observers should determine the necessary exposure times to achieve a reasonable S/N ratio. Mikuz's CCD V magnitudes are generally approximately 1 mag fainter than visual m1 data, varying somewhat from comet to comet, even though his CCD V coma diameters tend generally to be larger than the visual estimates (though the visual estimates may have large uncertainties). Thus, the different response to Swan-band emissions, for example, may be an issue; the dark-adapted human eye sensitivity curve is almost centered on the Swan band, while the V filter transmits only part of this emission (see Figure 5.1).
Diagram showing the differences between the spectral sensitivity of the human eye at night vs. the standard V photoelectric filter (as used with a CCD), with the main C2 Swan-band emissions noted at the bottom; from Mikuz and Dintinjana [1994].
Large variations in CCD vs. visual magnitudes are due to many factors, as noted at the beginning of this section; the use of unfiltered CCDs, unsuitable comparison stars, saturation, errors in image processing, and data reduction are all issues that need to be considered. Unfiltered CCD photometry should not be used if at all possible (though such photometry is certainly better than nothing for faint comets!). Standard filters are now available from several manufacturers; prices range from US$100 upward, depending on the size and quality of optical surface. Loss of limiting magnitude via the use of filters is obvious, but this will contribute to better quality of data; CCDs with higher DQE would have less problem with the loss of limiting magnitude. Having acquired filters, the observer needs to calibrate them (e.g., determine the color transformation coefficients) using some standard sequence. In this way, we get information on how well the filters conform to the standard passbands [see Bessell 1995; Henden and Kaitchuk 1982, 1990; Sterken and Manfroid 1992b].
Nakamura, who prefers unfiltered CCD astrometry/photometry, estimates that he needs to take exposures that are up to 4 times longer with a V filter, posing problems both for his clockdrive and for his program of obtaining as much astrometry/photometry of as many comets as possible on a given night. As CCD chips have different spectral responses from human eyes, observers should be careful in their choice of comparison stars. Nakamura advises the use of V magnitudes from catalogued F or G-type stars for unfiltered CCD magnitudes of comets, to more closely approximate the V or visual band. As noted elsewhere in this Guide, the usage of GSC stars should be avoided for any photometry, especially those from southern-hemisphere (SERC-J plates).
However, considering the spectral response of most CCDs, the R magnitudes of comparison stars for unfiltered magnitude work should be used (the resulting comet magnitudes being therefore similar to R magnitudes), though there are not widely available good catalogues for R maginitudes fo stars (the Landolt sequences are possible the best, although the numbers of comparison stars are still too small). Larson recommends Cousins R-band filters (which do not have to be expensive), noting that for typical CCDs, not that much light is lost --- whereas V-band filters will waste CCD sensitivity, though this may be justified because they better approximate visual estimates and the huge amount of archival data so accumulated over the years. Given the ambiguity in V magnitudes (emissions plus continuum), an R-band may lend itself more readily to physical interpretation.
Pravec questions whether it is better (1) to have an accurately-defined spectral band, while integrating a coma to smaller apparent radii (as in the V-band, a CCD is significantly less sensitive than in the case of using a clear filter), or (2) to have a vague definition of the spectral response curve (which may cause no significant problem for a dust coma via the use of a solar-color standard-star V magnitude for calibration), but better S/N in the faint outer parts of coma, and thus being able to integrate the coma to larger radii. The latter (unfiltered) may be better in the sense that an integration of too small a radius can cause much greater error than that caused by a difference between the spectrum of a nearly-solar-color star and the actual spectrum of the coma. Pravec advsises that only for bright comets at smaller heliocentric distances (perhaps 1.0-1.5 AU), which can exhibit significant emission features, should the use of V filters be preferred and the clear-filter observations avoided.
5.1.4.2. CCD surface photometry
During the past decade, coinciding with the advent of large-format linear CCD detectors, surface photometry of comets has been widely utilized. As noted in section 2.5.0, the surface brightness profile of a comet can be calculated by measuring the brightness (in mag/arcsec2) in concentric annuli centered on the nucleus. When the brightness is plotted on a log-log scale, the slope is d(ln B)/d(ln p), where B is the brightness and p the distance in arcsec from the center; this slope should equal -1 for radial and isotropic outflows of dust grains. Karen Meech notes that these type of measurements are extremely challenging, because it is essential to correctly subtract the contribution from the sky from under the comet before computing the profile. This requires that the sky brightness be estimated well away from the comet image and interpolated underneath the comet. Correct subtraction requires the linearity of the CCD detector, and the ability to flatten (i.e., remove detector sensitivity variations) the data extremely well. With care, it is possible to detect low-surface-brightness comae that are a factor of 102 to 103 times fainter than the night-sky background.
(reference list to be added later)
ICQ
