Cleaning up our Act

Science Nugget: October 1, 1999

This week we go "behind the scenes" at SXT data reduction central. Follow along as we reveal some of the secrets of the SXT team.

Introduction

The Yohkoh project has tried very hard to reach out to the public with X-ray images of the solar corona. The most recent SXT image is always available at the SXT First Light web page, a similar mirror (identical) site in Japan, and at various other locations on the web. Literally tens of thousands of posters of the X-ray sun have been distributed world-wide. The level of public awareness of the sun is measureably higher than it was in the pre-Yohkoh days.

The images that people are familiar with, though, have been cleaned up quite dramatically from the raw data. Lots of instrumental artifacts with strange names -- CCD bleed, dark current, stray light, to name a few -- are removed before the interested viewer gets to see to the final image. This nugget, hence, attempts to describe, in layman's terms, the long twisty maze that SXT data travels through before it reaches your screen.

The Raw Materials

A 'basic' SXT full-frame, movie-quality image passes through a number of steps before it shows up at your door. Why? Well, let's have a look at the raw data:

Short Exposure Long Exposure

Whoa! Those are some lousy images! The one on the left (short exposure) has no detail to amount to much. It is only useful in those parts of the image where it is really bright. The one on the right (long exposure) has some sort of fog growing in the bottom, and a big spiky hairdo on the top left. This image is useful only in those parts of the image where it is relatively faint. Where the image is really bright the CCD is overexposed, which causes "bleeding" -- spilling of the signal into adjacent picture elements, or pixels -- into vertical streaks (spikes). How is it possible for this sort of nonsense to be turned into a beautiful SXT image?

Well...it ain't easy.

The Method to our Madness

There are numerous things that have to happen to these 'diamond in the rough' images before they're suitable for public (and scientific!) consumption. An approximate version of the process follows:

  1. The raw data are decompressed.
    The digital data aboard the spacecraft are subjected to a simple compression algorithm before being downloaded to the ground. Such compression reduces the number of computer bits used to represent the image, and hence allows more images to be transmitted within the limited daily downlink capacity. (The compression algorithm is simple, but beyond the scope of the current topic.)
  2. The CCD dark current is removed from the images.
    SXT images are collected and digitized via a Charge-Coupled Device, or CCD. (Visit this nugget for a short discussion on CCDs.) Even without any light on them, all CCDs produce some signal, which is called the dark current. During an exposure, the CCD accumulates a background level of electrical charge that shows up on the image as a wedge of intensity, bright at the top and dimmer near the bottom. Witness:
    SXT Dark Current Image
    All manner of evils are apparent in this image. A dark current image such as this is created by taking a picture without actually opening the camera shutter. This allows the dark current to accumulate as it normally would, but since there's no light on the CCD we're able to use this measurement to remove the dark current from the actual data. There are a number of features present in this image: Note the increase in intensity from bottom to top, and also notice the streaks due to hot pixels -- defects in the CCD that just won't stop leaking charge.
  3. The images are corrected for stray light.
    Some unwanted light, called stray light, gets into the telescope. SXT images are subject to a certain amount of stray light. It gets in because, over the years Yohkoh has been on orbit, micrometeorites have poked tiny holes (pinholes) in the thin filters (like aluminum foil) used to let X-rays into the telescope. The level and nature of the stray light is dependent on the filters used for a particular image. The Al.1 images (the ones used for this demonstration) are particularly sensitive to stray light due to a number of pinholes in the thin aluminum entrance filters. How bad is this stray light? Well, the raw images we noted above provide a good clue to how bad things are. Here's the stray light image:
    SXT Stray Light Image
    The fuzzy bright blobs near the bottom and to the right of this image are clearly evident. What's not evident is at least one similar blob near the top left. The relative brightness of these blovs varies depending on the pointing of the telescope. (E.g., if we move the sun up and left in this image, these bright spots will largely disappear but others will pop up.)

    The stray light images are commonly referred to as terminator images. This is because of the manner in which we collect them: as Yohkoh orbits the earth, it passes in an out of spacecraft night, which is when the earth is blocking the sun. When Yohkoh just emerges from behind the earth, the sun's visible rays can reach the detector but the sun's X-rays are blocked by the Earth's atmosphere. This happens to be a perfect time to measure the level of visible stray light in the telescope.

  4. The images are co-registered, and scaled in intensity.
    In this step, we correct the signal from the CCD for exposure time, putting the long and short exposures shown above on the same intensity scale as well as possible. This is done to prepare for making a combined image, which takes advantage of the good parts of each image, and throws away the bad (overexposed or underexposed) parts.
  5. The overexposed portions of the long-exposure image are replaced with the short-exposure image to increase the overall dynamic range.
    Remember how parts of the bright image were saturated and causing CCD bleed? Well, if we compare those parts of that image to the low-exposure image, we find that there is detail clearly evident in the low-exposure image. By virtue of software and ingenuity, if we replace the saturated parts with the other image we get the final product:
    The Final Product
  6. The final product is posted to the web.

Clearly, many steps take place between the raw data and the final product. We've glossed over many details of this operation here, and the opportunity for error looms over the entire process. Chief Observers are continually vigilant, trying to keep the systems of correction in check.

In memoriam, Kiki Hudson (1941-1999).

Brian Handy <handy@physics.montana.edu>
Phil Shirts <shirts@lmsal.com>
October 1, 1999