chi and h site logo

Home
About χ & h

Webcam amateur astronomy

Object
Detector
Optics
Camera
Mount
Computer and operating system
Driver software
Acquisition software
Data reduction
Presentation
Theory and tests

Object

Bright extended object: Moon, Sun, planet

A bright extended object is defined here as one that delivers enough light to overexpose frames and that is so large as to occupy an area on the detector at least several tens of pixels in size in both directions. There are three possibilities:

  1. a moderately sized disc, often roughly circular, but could be disc with elliptical ring (Saturn),
  2. a disc much larger than the detector but with its limb or terminator (bright/dark boundary of the surface of the celestial body) running through the field of view,
  3. a disc much larger than the detector, so that only part is caught by the detector.

For me the first kind is one of the apparently large planets (Mercury to Saturn) at long focal length (3500 mm), or it could be the Moon or Sun at "normal" focal length (normal photo lens, 50 mm). The second and third kind are Sun and Moon at tele or long focal length (400 mm to 3500 mm).

Moon
Moon on 2002/11/24. Philips ToUcam Pro VGA, f = 3500 mm. Unsharp mask. Image scaled down twofold.

The brightness is important because we have to reduce the exposure time and the gain (signal amplification) far enough so as not to saturate the digitisation (cf. Theory / Calibration). The highest digitisation result is 255 - separately for red, green and blue if we take colour. If the interesting features of the object are too bright they are digitised as constant number 255 and we lose all contrast.

However, at the same time we want the noise to be strong enough to be digitised (cf. Theory / Noise and digitisation). The smallest digitisation step is 1, i.e. after digitisation the brightness values are integer numbers 0, 1, 2, ... 255. If the noise level is much less than 1 then we do not detect the noise. In that case we also cannot reduce noise by averaging several or many frames.

So how can we reconcile the requirement of not overexposing with that of digitising the noise? There tend to be two controls, exposure time and gain. We set the gain high enough to digitise the noise, and then we set the exposure time short enough not to overexpose. The noise does usually not depend on the exposure time (i.e. the number of photons collected), but is in fact generated by the action of reading the image data out of the detector and of amplifying this electric current (cf. Theory / Detectors). The gain widget in the user interface is our handle on controllin the amplification.

Pointlike object: star field

If the object is not extended we may have a problem with compression. The webcam is designed for every day images, not for images that have vast brightness differences from one pixel to the next. Depending on the frame rate, the number of pixels and the bandwidth of the connection between webcam and computer, more or less compression may be required. Hence stars may not look right, they will come out as bright points with an overly dark ring around them.

The USB bandwidth is such that even at a frame rate of 5 Hz a VGA sized image (640x480 pixels) cannot be transferred without compression. The webcam may have to be used in 320x240 pixel mode.

Faint object: cluster or nebula

A faint object is defined as one that even in its brightest parts does not deliver enough light to overexpose the detector. We can therefore use the longest exposure time and an optimum gain setting that is determined once and for all by checking the noise. In addition we may also set the brightness and contrast settings to maximum.

Orion nebula
Orion nebula on 2001/12/25. Logitech QuickCam VC, f = 90 mm. Combined linear stretch (grey) and false colour.

The digitisation of noise (cf. Theory / Noise and digitisation) is vital for imaging faint objects: A faint object may only give 0.5 or 0.1 digitisation unit in each pixel (cf. Theory / Calibration). This can only be detected by

  1. taking many frames for later stacking (cf. Theory / Noise reduction / Average many frames), and
  2. having random noise in each pixel of at least 1 or 1.5 digitisation units so that the 0.1 ADU of signal will be registered in 10 per cent of frames as 1 ADU higher than in the other 90 per cent of frames.

Copyright © 2003 Horst Meyerdierks
$Id: object.shtml,v 3.3 2004/02/21 18:13:39 hme Exp $