Supernova Search

Before PEST was built, from 2007 to 2009 I designed and ran a supernova search program that discovered 3 supernovae.  The equipment was pretty basic, consisting of a C9.25 telescope, Vixen GPD mount, and kit-built Artemis ART 285 CCD camera (see picture below right, which also shows the home-made power supply unit).  These had to be carried outside and set up each night, and stowed away again in the morning.  Because of this, polar alignment and therefore go-to pointing accuracy was not very good.

Artemis camera
Kit built Artemis camera and power supply. Image credit: TG Tan

Automation of the the search was with scripts written in the AutoIt scripting language.  The components were;

1.  GalaxyGen – generates a list of galaxy targets for the night.  Also generates a list used to download DSS images as references.  The search algorithm has to compensate for poor polar alignment.  This makes for poor pointing accuracy if the scope is significantly away from the initial alignment point azimuth and altitude.  GalaxyGen overcomes this by picking targets within a narrow band of declination, but with increasing RA through the observing run, such that the telescope always points to approximately the same alt-az point.

2.  TargetPoint – controls the mount and camera and successively points the telescope to each galaxy on the list, acquiring a set number of images of each target.

3.  Examiner – enables quick comparison of these images to reference images downloaded from the Digitized Sky Survey (DSS).

Typically between 150 to 200 galaxies were observed in an automated overnight run.  Using individual 20s exposures, supernovae down to mag 17 are detectable.

Proof of concept was achieved when a supernovae was detected in PGC51820 (ESO385-32) on 14th March 2007.  Subsequent checking revealed that this SN had been discovered about a month previously and designated SN2007X.  More information on the search program is in the article below.

Supernova Discoveries


Designation and Discovery InformationOther Info
SN2007rv, discovered 7th Nov 2007 in NGC 689.
RA: 01h 49m 52.86s  DEC: -27° 28' 4.1", 14"E and 4"S of host nucleus.
mV = 15.9 on 2007/11/11.51 UT
Type 1a
~200mln light years away.
SN2008ff discovered 29th Aug 2008 in ESO284-32 (PGC 64319).
Location:  RA: 20h 13m 59.96s  DEC: -44° 21' 7.8", 39"E and 1"S of host nucleus.
mV = 15.5 on 2008/09/1.52 UT
Type 1a
~260mln light years away.
SN2009gg discovered 16th June 2009 in ESO235-35 (PGC 65919).
Location:  RA: 21h 01m 29.91s  DEC: -52°01'00.2", 1.4"W and 3.4"S of host nucleus.
mV = 15.8 on 2009/06/15.591 UT
Type 1a

S&T Article

A condensed version of the following appeared as the article “Searching for Supernovae on a Shoestring” in the July 2009 issue of Sky and Telescope magazine.

Supernovae have always fascinated me.  These violent events seed the universe with the elements that go on to make planets and life, and they continue to be important to cosmologists as standard candles that may help to explain how the universe has evolved since the Big Bang.

My supernova search setup – pre PEST. Image: TG Tan

And so it was that when I started to think about a science project that would use the skills and equipment I had acquired from several years of astrophotography the possibility of discovering supernovae was immediately appealing.  The idea that I could be the first to see a star that exploded many millions of years ago was compelling.  But could it be done, and how much time would I need to spend to get results?

This article describes my supernova search program which, running from February 2007 to August 2009, has discovered three supernovae.


Supernovae are rare.  A typical galaxy, such as ours, hosts a supernova only once or twice a century.  To stand a reasonable chance of discovering one, I would have to observe hundreds of galaxies a night.  This implies that the search has to be automated, and that set-up time, particularly after dark, must be minimised.

I do astrophotography from the backyard of my suburban home and do not have an observatory or permanent mount for the telescope.  The gear I have is pretty modest – a Celestron 9.25” SCT, a tripod mounted Vixen GPD mount with SkySensor 2000, and a kit built Artemis 285 CCD camera.

Given this, the main challenge to be overcome in the design of the search program was to find a way to automatically and accurately point the telescope at several hundred galaxies over many hours each night.  The GPD and SkySensor combination is very capable, but I know from experience that accurate pointing requires good polar alignment, and that accuracy deteriorates away from the point that the mount has been synched to. I definitely could not afford to go through the time consuming process of polar alignment each night.

The solution to this turned out to be elegant and conceptually simple.  I ensure that each galaxy on the target list will be at about the same altitude and azimuth at the time that its image is acquired. With the target list constructed in this way, exact polar alignment is not needed for good pointing accuracy.  Imagine a mount that is not motorised. The telescope on this mount points at the same place in the sky as the stars wheel around.  Someone observing through this telescope would see the stars drift by, but he would always be looking at exactly the same declination albeit with increasing right ascension (RA).  This is true even if the mount is not polar aligned.  Of course in reality, some flexibility in choosing targets is desirable so that approximate polar alignment is still needed.

Search Design and Execution

The search program consists of three parts – target list generation, telescope pointing and image acquisition, and a means of rapidly comparing the acquired images with references.  The core functions in all three are carried out by scripts that I wrote in the freeware AutoIt language.  GalaxyGen generates the night’s target galaxy list, TargetPoint reads the target list and controls telescope pointing and image acquisition, and Examiner displays a DSS reference alongside the corresponding acquired image allowing visual comparison, and records any result.

A typical night’s observation starts with generating a target list.  As explained above, the target list (see image below left) consists of a series of galaxies all at about the same declination, ordered by increasing RA.  I set a suitable RA for the start of the observation such that targets are at a fairly high elevation (above 60 degrees), and select a declination angle for the night.  GalaxyGen produces a target list that minimises deviation from the desired fixed altitude and azimuth over the duration of the run.  Since pointing accuracy depends on this I spend some time tweaking GalaxyGen’s input parameters to achieve the best balance between number of targets and deviation from ideal target location.   The GalaxyGen script interface is shown below right.

Target list generated by GalaxyGen
GalaxyGen script interface

I carry out the tripod and mount and place it on the pavement just behind my house. Placing the tripod feet in slight indents I have made in the pavement ensures that there is approximate polar alignment. The telescope is then placed on the mount, and the camera installed. I use a desktop PC on a trolley to run the camera and mount. This is wheeled out from its home in my garage and all cables connected. It usually takes me half an hour to set all this up.

Using Cartes du Ciel, a freeware planetarium program, I slew the telescope to a star close to the first target on the list, centre the star and synch the mount to this point. I then focus the telescope.

When I am happy that everything is working fine I start the TargetPoint script. TargetPoint works with the Artemis camera software to acquire images, and with IRIS for telescope pointing (IRIS is primarily an astronomical image processing program but also has the ability to talk to Meade LX200 compatible mounts). TargetPoint sets image acquisition parameters , slews to the first target, does the required number of exposures, waits for images to be downloaded to the PC, then slews to the next target on the list. Images are visible on the screen as they are taken so I usually confirm that at least the first target has been acquired successfully. Thereafter the observation session needs no further intervention. I cover the PC with a plastic sheet to keep dew off, and then am free to enjoy the rest of my evening.

The next morning, I copy all images acquired onto a USB drive, dismantle the equipment and put everything away. This usually takes only 20 minutes, so is easily done before I leave for work.

The next bit is the most challenging and time consuming. I use the next script, Examiner, to compare images acquired against DSS references. When GalaxyGen generated the target list, it had also prepared a separate list suitable for submission to the Canadian Astronomical Data Centre (CADC) DSS interface website. These reference images are downloaded in fits format. Examiner has a very simple interface that enables me to quickly go through the images taken whist recording results.


So what happens when I do see a star that is not in the reference image?  The first step is to confirm that it is indeed a star and not, say, a cosmic ray strike, or a CCD hot pixel.  For each target, I usually take 3 consecutive images.  Because of tracking imperfections there will usually be a slight image shift between images.  A hot pixel will stay in the same place and not shift with the image.  A cosmic ray hit can also be identified in that it will also only affect one of the subject images.

The first supernova that I discovered was SN2007rv on 7th Nov 2007.  It had been a windy night and since, really, my mount is too light for the telescope, imaging is very susceptible to wind, and I had not held much hope of many usable images.  So it was only four days later that I got around to looking at the images.  With much excitement, I saw a star just east of the nucleus of NGC 689 that was not there in the DSS image.  A quick check showed that no known minor planet was in that location at the time.

I then set up the telescope for a confirmatory observation.  The Central Bureau for Astronomical Telegrams (CBAT), which acts as a clearinghouse for reports on transient astronomical events, requires that any suspect supernova discovery be confirmed by a separate observation on a second night.  Yes, the star was still there.  I quickly did some astrometry of the suspect supernova in IRIS, and submitted a discovery report to CBAT.  There followed an email exchange with Dr Dan Green, Director of CBAT which showed how little I knew about this – everything from the number of decimal places for the position of the object, to how photometry was done.  With his help my discovery report was eventually licked into acceptable shape and I got confirmation a few days later that the supernova had been confirmed as a Type 1a, and given a designation.  I was elated!

Since then the observation program has made two other discoveries, SN2008ff in August 2008 (another windy night!), and SN2009gg on 13th June 2009.   The screenshot below shows the discovery of SN2008ff.  Note the wind affected stellar images.

Screenshot showing discovery of SN2008ff