Gordon Laing wrote:
...Quick question concerning the star images. I noticed in your (and my own) long exposures that star images can often look quite electronic, and sometimes almost square-like. I realise the actual pixel locations on the sensor are smaller than the spread of the star using our optics and atmosphere, but wondered what your take was on this.
Interesting question - long answer! As you know CMOS sensors, such as are used in my 40D, are supposed to be immune to the "pixel bleeding" that some CCD sensors suffer from along their "rows".
This phenomenon of "square stars" affects the brighter stars in my images and can be seen in both the processed RAW and original JPEG files. This can't be put down to atmospheric "seeing" as that would tend to create round stars. One of the nice features of the IRIS software is that you can read the camera's RAW files using 1x1 binning. Here is a much magnified portion from the centre of one of the original images.
The biggest difference between this and the processed images is the way the processing has converted the square stars into squares with longer diagonal "extensions". As for why the bright stars are square this seems to be referred to in the trade as "crosstalk". Here is an example of the way sensors are constructed, courtesy of Fuji Corporation.
If some light is scattered by each microlens then it can find its way across to other microlenses and hence to the adjacent pixels. My take on this is that the proportion of scattering from microlens to microlens adjacent in the horizontal, vertical and
diagonal directions is such that the resultant "shape" is likely to be square. If the incident light isn't white then not all of the sensors in the Bayer matrix will respond to this scattering and that is nicely shown in the largish bright star in the upper right corner of my image above with its lattice like structure. Of course, once the image is de-mosaiced this lattice structure disappears.
As with so much information on the Net, figuring out the correct search string is key. Once I realised that doing a Google search for "microlens crosstalk
" was the way to go things became much clearer. In particular, I like the article Crosstalk challenges CMOS sensor design
. This also addresses the issue of electrical crosstalk.
), here's a question for you. Does this crosstalk and the resultant loss of spatial resolution around bright objects affect the way you test lenses?