Creative Commons License Copyright © Michael Richmond. This work is licensed under a Creative Commons License.

Analysis of possible occultation by (20000) Varuna on Dec 31, 2005

Michael Richmond
Jan 8, 2006

Table of contents:

Reduction of the video file

On Dec 31, 2005, Michael Vasseur watched to see if the Kuiper-belt object (20000) Varuna would occult the star 2UCAC 40675770. Maps of the event are at Steve Preston's occultation predictions site. This was a very challenging event, because the target star was very faint (around mag 15), the occulting body was completely invisible (near mag 20), and the uncertainty in path location was pretty big.

Michael's location for observing this event was at the Galaxy Blues Observatory, IAU code H89:

You can see his messages about the event at

His equipment included

Michael set up his camera so that it co-added 128 fields into a single summed image; since each field is 1/60 of a second, each summed image covers about 2.1 seconds. He sent the video through a KIWI OSD unit, which placed a time signal on each frame of the video. That means that the resulting movie shows

I stupidly didn't read his message describing this automatic co-adding carefully, and so wasted a day or two chopping the movie into individual 1/30-second frames ... which turned out to have exactly the same stellar information as their neighbors. Silly me.

Michael made a copy of his movie available on his web site. The movie is in DVR-MS format, which is a version of MPEG-12, I believe. It's a bit harder than usual to break a movie of this format into individual images, but I managed to figure out how to modify my software to do so.

First, get the latest version of MPlayer. Next, grab a copy of codecs.conf file and edit it to add one line to the stanza for "videocodec mpeg12", as shown below.
        videocodec mpeg12
        info "MPEG-1 or 2 (libmpeg2)"
        comment "with postprocessing"
        status working
        format 0x10000001  ; MPEG-1
        format 0x10000002  ; MPEG-2
      ; added this line to play dvr-ms Digital Video files
      ; as instructed on MPlayer E-mail archives
        format 0x20525644  ; dvr-ms
        driver libmpeg2
      ;  dll "libmpeg2"
        out YV12,I420,IYUV
        out 422P
Then, invoke MPlayer with the following command-line options to play the movie "".
     mplayer -demuxer 35 -codecs-file /etc/mplayer/codecs.conf file.dvs-ms

Each frame of the movie is 720x480 pixels. Although the movie file is 17948 frames and 220 MBytes in size, there are really only 280 distinct images of the sky and stars (1 out of every 64 frames). The entire dataset is much more manageable once one has extracted these 280 frames. Below is an example, frame number 75 out of the 17948 frames in the big movie.

Each pixel in the image can hold one of 256 greyscale values (called "counts" from here on). The background sky was high, at about 200 counts; that left only about 56 counts for all the stellar data before a star would saturate. Setting the gain parameters in this fashion was probably necessary to detect the very faint target star, but it leaves very little dynamic range for measurement.

I processed this movie file in the following manner:

  1. split into 280 individual frames (each represents 2.1 seconds of detection) and converted to 8-bit JPEG format by mplayer

  2. converted from JPEG to FITS format via ImageMagick tools

  3. analyzed with programs from the XVista astronomical image processing package

In addition to the target star, there are several other stars visible in the field. Here's a stretched version of frame 75.

Here's the same image with a couple of labels. Note how faint the target is, and how close it is to a much brighter star.

And here's a chart from Aladin with the stars of interest labelled.

     star          UCAC2  ID        UCAC2 mag
   A               40675796           9.9            

   D               40675774          13.8   

   target          40675770          15.5           

Most of the stars were saturated in the 8-bit FITS images which I extracted from the movie file. The FWHM was hard to measure, but was around 3 pixels for stars like C and D.

For each of the 280 frames, I

  1. placed a small box around the position of star "A", subtracted a local value of the sky from the box, convolved the box with a 2-D Gaussian of FWHM 6 pixels, and searched for the brightest peak
  2. noted its position;
  3. computed a position for all other stars, B, C, D, E, F, G and the target, using fixed offsets from the position of "A" (as measured on one frame)
  4. placed a circular aperture of radius 2 pixels (see below) around each position
  5. added up all the light within this aperture
  6. calculated a local sky value for each star based on pixels in an annulus of inner radius 15 pixels, outer radius 25 pixels
  7. subtracted the contribution of the background sky to the light within the aperture

After looking at the resulting light curves, I decided that the smaller the synthetic aperture, the better. I chose a radius of 2 pixels for the data shown below.

Photometry: plots and a table

Here are quick views of the results.

The entire light curve for the target and several other stars.

Here are the curves for the target and nearby star D only.

The target star does not appear to show any clear signal of an occultation. It bounces up and down a lot on short timescales, but so do the other faint stars "F" and "G".

I then performed ensemble photometry on the group of stars D, E, F, G, and target, using my ensemble photometry package. The idea is to remove any systematic shifts in the overall brightness of all the stars from frame to frame; if, for example, a cloud covered the field for three frames, it might make all the stars appear 0.20 mag fainter. Or if a strong wind shook the telescope for several seconds, it might blur the images, again making all the stars fainter by a similar fraction.

The bright stars D and E dominate the ensemble solution. Here are the amounts by which each frame had to be shifted up or down, in magnitude units, to make all the stars best match their average brightness. The vertical scale has an arbitrary zeropoint.

You can see that the average brightness of all the stars varied by up to about 1 magnitude at most; that corresponds to a factor of about 2.5 in brightness. I think that the brief excursions to fainter brightness were caused by gusts of wind shaking the telescope, as Michael Vasseur mentioned in his messages.

After removing those systematic shifts in brightness, we obtain the remaining light curves of each object. In the graph below, I have used magnitude units and shifted several of the stars vertically for clarity.

We are looking for a time during which the target star becomes fainter or disappears, but none of the other stars becomes fainter. There isn't much evidence for such an event. Maybe the best candidate time is around frames 80 to 100, which correspond to times UT 06:17:51 to 06:18:33.

You can grab the data in a multi-column ASCII text file below. The columns are

 col              quantity          
  1              frame index

  2,3            flux of target star in 2-pixel-radius aperture, 
                     and estimate of uncertainty in that flux
  4,5            ditto 2.5-pixel aperture
  6,7            ditto 3-pixel aperture
  8,9            ditto 4-pixel aperture
  10,11          ditto 5-pixel aperture

  12,13          flux of star D in 2-pixel aperture,
                     and estimate of uncertainty in that flux
  14,15          ditto 2.5-pixel aperture
  16,17          ditto 3-pixel aperture
  18,19          ditto 4-pixel aperture
  20,21          ditto 5-pixel aperture

  22,23          flux of star E in 2-pixel aperture,
                     and estimate of uncertainty in that flux
  24,25          ditto 2.5-pixel aperture
  26,27          ditto 3-pixel aperture
  28,29          ditto 4-pixel aperture
  30,31          ditto 5-pixel aperture

  32,33          flux of star F in 2-pixel aperture,
                     and estimate of uncertainty in that flux
  34,35          ditto 2.5-pixel aperture
  36,37          ditto 3-pixel aperture
  38,39          ditto 4-pixel aperture
  40,41          ditto 5-pixel aperture

  42,43          flux of star G in 2-pixel aperture,
                     and estimate of uncertainty in that flux
  44,45          ditto 2.5-pixel aperture
  46,47          ditto 3-pixel aperture
  48,49          ditto 4-pixel aperture
  50,51          ditto 5-pixel aperture

Photometry of co-added frames

Because the photometry of the original 2.1-second exposures was so low in signal-to-noise ratio, I decided to try co-adding several consecutive frames. Michael Vasseur also suggested doing this in one of his messages. I simply added together each set of N=5 consecutive frames, reducing the 280 original frames down to 56 coadded images. I did not attempt to register the stars in each set of 5 images before adding them together (well, actually I did try that, but it turned out to be very difficult, so I gave up).

I then performed exactly the same sort of aperture photometry on this set of 56 images as I did for the original dataset. The results are shown below.

Here are the measurements, in linear units, for stars in the coadded images.

Here are just the target and the nearby star D:

Again, it appears that the target star may become a bit fainter during the same period as before (original frames 80 to 100, coadded frames 16 to 20), but it doesn't look like a real occultation.

I again performed ensemble analysis on photometry from the coadded images. The systematic shifts are a bit smaller than they were before, reaching only about 0.6 mag peak-to-peak instead of roughly 1.0 mag peak-to-peak.

And here are the ensemble light curves from coadded images:

The case for an occultation looks even weaker to my eyes in this coadded dataset. But feel free to analyze the numbers if you wish.

Creative Commons License Copyright © Michael Richmond. This work is licensed under a Creative Commons License.