Succesfull Photometric Observation of Exoplanet HD 209458b
at CBA Belgium Observatory


This is shortened and slightly reworked version of an article on exoplanet astronomy that will appear in a future edition of "Heelal", a monthly publication of the Belgian Astronomical Association VVS. The article is called "Sterbedekkingen door exoplaneten" (authors : Siegfried Vanaverbeke and Tonny Vanmunster).

1.  Introduction

The American astronomer Otto Struve was the first person suggesting (in 1952) the possibility of observing exoplanets through stellar occultations. In a short article in The Observatory [1] he concluded that it should be possible to indirectly detect massive short-period planets by observing their transits : 'We know that stellar companions can exist at very small distances. It is not unreasonable that a planet might exist at a distance of 1/50 A.U., or about 3,000,000 km. Its period would be about 1 day. There would of course also be eclipses. Assuming that the mean density of the planet is five times that of the star (which may be optimistic for such a large planet) the projected eclipse area is about 1/50th of that of the star, and the loss of light in stellar magnitudes is about 0.02. This, too, should be ascertainable by modern photoelectric methods, though the spectrographic test would probably be more accurate. The advantage of the photometric procedure would be its fainter limiting magnitude compared to that of the high-dispersion spectrographic technique.'

Until 1995, exoplanets remained hypothetical objects. That year, however, astronomers discovered a first exoplanet, orbiting 51 Pegasi, based on Doppler measurements [2].

In November 1999, two independent teams of astronomers succeeded in discovering the first transit of an exoplanet, through photometric  observations of HD 209458, a star of magnitude 8 in Pegasus. One of the teams was headed by Prof. Tim Brown of Harvard University. His PhD student David Charbonneau obtained transit observations using a telescope with an aperture of 10-cm only. Evidently, this opened perspectives for amateurs to enter the domain of exoplanet transit photometry.

Drawing by David A.Hardy of a planetary system 
orbiting Tau Bootis. The exoplanet has
approximately 4x the mass of Jupiter.


2.  Transit observation of HD 209458 at CBA Belgium Observatory

The Finish amateur astronomer Arto Oksanen, member of CBA Finland, was the first amateur to observe a photometric transit of HD 209458b, at his Nyrola Observatory in September 2000. In general, observers with small telescopes, a CCD camera and a high-quality photometric procedure are able to make exoplanet transit observations, provided they can reach a photometric precision of minimal 1% (see further). To foster pro-am collaborations on exoplanet research, has been created a few years ago (see further). The goal of is to coordinate and direct a cooperative observational effort which will allow experienced amateur astronomers and small college observatories to discover transiting extrasolar planets.

The exoplanet of HD 209458 is creating an eclipse every 3.52 d. At first sight, it therefore seems there are plenty of opportunities to observe such an event. However, if we take into account that preferably observations cover a complete transit from ingress till egress, and that the altitude of the star has to be 30 degrees or more (a lower altitude will result in too much noise being introduced in the photometric lightcurve), one immediately realises that chances are rather limited. Add to this the poor weather conditions of Belgium in general, and in practice only a few occasions per year remain. 

Eclipses of HD 209458 were discovered using the STARE telescope (STellar Astrophysics and Research on Extrasolar planets).


Successful observation on 2003, September 3rd

I made a first attempt to observe a transit of HD 209458b on August 27th, 2003, using the 0.35-m f/6.3 telescope at CBA Belgium Observatory. The CCD camera was an SBIG ST-7, without filters. I made differential photometric observations, comparing the magnitude of HD 209458 with one or more constant comparison stars. The attempt was unsuccessful, mainly because I did not have access to a good comparison star on the CCD images (see further). 

The night of September 3rd, 2003 was clear again, and allowed me to successfully measure a complete transit. Armed with my experiences of the failed trial on August 27th, I started the observing session on 22h42m UT (Sep 3.946 UT), i.e. 1 hour prior to the predicted time of ingress. After 4h42m and 1022 CCD images, the session was concluded. At that moment, HD 209458 was less than 30 degrees above the local horizon.


Analysis of the observations

Over the years, I have built up quite some experience in analysing photometric observations of cataclysmic variables, by participating in many of the CBA campaigns. My first approach therefore was to apply these same techniques to the analysis of my HD 209458b observations. The day after my observing session, I made a first attempt, using AIP4WIN to perform aperture photometry (with flat fielding, bias and dark frame reductions properly applied). The result was rather discouraging : the light curve did not show modulations that could point to a successful transit observation. I reached a photometric accuracy of 0.01 - 0.02 mag only, and knew from literature that about 0.005 mag (5 millimagnitudes) would be required.

It would take several more weeks before I arrived at a useful analysis method, that allowed to process all CCD images and to finally deduct the light curve below.


Photometric observations of exoplanet HD 209458b, 
obtained on September 03, 2003 at CBA Belgium Observatory. 
(c) Tonny Vanmunster

A first roadblock in the data analysis was the spread in accuracy of the photometric reductions of individual CCD frames. To achieve a 0.005 mag precision, I had to apply some form of 'statistic sampling'. This was done by adding successive raw CCD images to build a new raw 'master image'.  I then performed aperture photometry on these master images.  I tried several sample-sizes, ranging from 5 to 72 images. I obtained best results by adding 30 CCD images each time to create a master image. Unfortunately, AIP4WIN has no automated way to simply add CCD images. I therefore decided to write some software myself, to automate the whole process.

This first approach almost immediately led to a much improved light curve, showing the 'signature' of an exoplanet transit.

A next optimisation was obtained by aligning the raw CCD images, prior to adding them. This ensures that stars appear at more or less identical (pixel) positions before adding images. This technique compensates for small inaccuracies in the telescope mount. Unfortunately, AIP4WIN again has no automated mechanism for this, so I had to use own (Visual Basic) software and commercial packages to get the work done.

A final step towards further increasing the accuracy of my photometric method was to use 'ensemble photometry'. This techniques aims at grouping a number of constant comparison stars into 1 'virtual' comparison star that then is used in the  photometric reduction. This approach averages noise levels introduced by the comparison stars. AIP4WIN only supports the use of 1 comparison (and 1 check) star. I bought a copy of MIRA A/P, which is a more professional photometry software, that supports ensemble photometry. Using Mira A/P, I arrived at a slightly improved light curve.

The final result of my data analysis efforts is shown in the light curve above. It depicts the ingress, eclipse and egress of the exoplanet. Each dot in the curve shows the (differential) magnitude of HD 209458 for 30 (combined) CCD images, as compared to the comparison star HD 209346. Error bars are included as well. The grey curve shows the theoretical transit lightcurve. Predicted times of ingress and egress are marked with arrows. Note that the photometric accuracy is clearly declining towards the end of the observation session, due to the low altitude of HD 209458.

3. Be prepared ...

Interested in observing an exoplanet transit yourself ? Here are some further tips to share. We already explained some improvement actions to increase the photometric accuracy of your CCD images (e.g., through statistic sampling). However, those are 'a posteriori' techniques. Equally or even more important, are techniques that allow to ensure a good photometric accuracy during image acquisition.

Focus : we know from experience that a slight defocusing of CCD images somewhat improves photometric accuracy, due to the fact that the light of a star is spread over a larger CCD area. The impact of course depends on the focal ratio of your telescope, and on the used CCD binning method. I always use 2x2 binning.

Comparison stars : the selection of appropriate comparison stars is a quite fundamental step (see also our failed first trial on August 27th, 2003). Select a comparison star of a similar magnitude and color as the observed star. This is not always easy to realise : HD 209458 has V magnitude 7.65, and we used HD 209346 (V = 8.32) as a suited comparison star. Color differences are especially important when doing unfiltered photometry of low-altitude objects.

Exposure time : the optimal exposure time for your CCD images needs to be determined by trial and error. Too short exposures will result in insufficient pixel counts, and reduce the photometric accuracy below the required level of 1%. Too long exposures have a risk of saturating pixels, making the photometric reduction useless. In
general, one should aim at an exposure time that results in achieving about 40% to 50% of the so called 'full well capacity' of your CCD camera. For an SBIG ST-7, this corresponds to a pixel count of approx. 30.000.

Scintillation : evidently, atmospheric turbulences have a quite negative influence on photometry. Scintillation is most evident when exposure times are below 5 seconds. With longer exposures (10 sec or more), the effect almost disappears. Scintillation probably accounts for most of the error range introduced in our HD 209458b light curve above. Too avoid saturation, by having to use exposure times of more than 10 seconds, one might consider the use of a V-filter.

Instrumental errors : a final source of errors is due to mechanical or electronic limitations of your setup. This category encompasses for instance small inaccuracies in a telescope mount, causing CCD images not to perfectly overlap. These errors can be easily reduced by aligning CCD images before stacking (see above). Another member of this category of errors are temperature changes over the course of a night. Especially with SCT telescopes, this may result in a gradual defocusing. If not compensated for, CCD imaging might become useless after some time.

4. The Transitsearch Project

The study of exoplanets meanwhile has become one of the fastest growing research areas in contemporary astronomy. Amongst others, this has resulted in the set up of an  international collaboration project to search for, and study exoplanets, called Transitsearch (see for full details). Transitsearch has been
initiated by Gregory Laughlin (University of California, Santa Cruz) and Tim Castellano (NASA Ames Research Center). The Transitsearch website has a list of target stars to monitor and predicted times of transit, including indications on the uncertainties of the predictions, on the probabilities of a transit, etc. In addition, the website contains interesting information on observing techniques (CCD photometry) and a discussion forum.

Good luck with your transit observations !



I wish to thank Siegfried Vanaverbeke (Belgium) for the many inspiring email exchanges in preparing for the exoplanet observations, and during the analysis afterwards. 


[1] Stryve, O., The Observatory, 72, 199-200 (1952).
[2] Mayor, M., Queloz, D., Nature, 378, 355 (1995).





Copyright © 2004 - Tonny Vanmunster.