22 November 2011

Spurious noise

8 minute exposure
In the night photos from Runde I noticed something odd. There were tiny streaks in my images, looking almost like luminous dust, in some of the shots. The number of streaks increased with exposure time. Nonexistent at thirty seconds, and all over the place at fifteen minutes exposure.  On close inspection, I also found dark streaks of the same size and frequency as the luminous ones. 

In a weird moment, it reminded me of a photo I once saw of a fog chamber used to detect cosmic radiation, set up by a Danish physicist who claimed our local neighbourhood in space had a lot more impact on Earth's climate than had been previously thought.

But then worry crept in, and I wondered whether my camera was faulty (oh, angst). Then I noticed the streaks were never in the same spot from one shot to the next. So if it was the camera, it couldn't be the sensor, at least. Besides, groups of faulty pixels would be detected by either the pixel-mapping feature or the long-exposure noise reduction, wouldn't it?
Detail from the above showing the
bright specks

I was at total loss on ideas for what it could be. First I considered dust. Maybe the sensor heated up so much during long exposures that the dust appeared luminous from reflecting heat back onto the photosites? That didn't fit with the occurrence of both black and light streaks, and it was unlikely that so many dust grains would shift position between every exposure.

A friendly soul on Pentaxforums confirmed to have observed the same, at least, and suggested software to deal with the problem. He said the best of the bunch are programs used by astrophotographers, who are dealing with faint light on a regular basis.

So I checked it out. There are lots of programs available, but their interfaces are somewhat incomprehensible to a regular photographer. Scientific terms and names of mathematical algorithms for noise reduction are tossed about, presumably expecting the user to know them apart. I didn't, but decided not to let that stop me. Eventually I homed in on the terminology and realised that my phenomenon had a name, it was categorised as "spurious" noise. To an astrophotographer, this kind of noise could look like a real signal. I can imagine a meteroite could produce a similar streak.

Armed with this keyword, it took me just one more Google search to hit the explanation. Well at least i hope it is the right one; it still sounds a bit like UFO/grain circle/tinfoil-hat theories to me.

Apparently what happens is that the sensor gets hit by cosmic rays. Sort of. For technical details see this article by Don Groom at Lawrence Berkeley National Laboratory, USA (PDF). Image examples are also available in many astrophotographers' blogs, as are suggestions on how to avoid it and how to recognise it from real stars. Or meteorites.

After a few days of thinking about it, I'm more comfortable with the thought, even if I'm still incredulous about highly energetic particles from space ("relativistic muons" and "Compton recoils" apparently) slipping through the atmosphere and hitting the sensor. But the explanation fits like a glove. Some particles hit at a shallow angle, leaving a streak. Others hit head on and make dots. Since the particles are charged and moving very fast, they can trigger a signal in the imaging sensor like a whole bunch of photons.

The dark streaks can be explained by particles hitting the sensor during the long-exposure noise reduction, which essentially is an exposure of the same length as the picture but without adding any light from the outside. The cosmic rays gets through anyway, however, and creates streaks of false noise that the camera then subtracts from the image.

The bottom line, however, is that cosmic rays will remain an obstacle to really long exposures, and that the only real way to avoid it is to do as the astrophotographers recommend, to take multiple exposures of 30 seconds each and combine them.
There's even a plugin for Lightroom to do that.

So perhaps the fact that my weird association to fog chambers turned out to be pretty close is more incredulous than the explanation itself. Spooky, even.

Go figure... :-)

No comments: