Very Long Exposure Problems

<p>I have a Nikon D610 and using it with a Nikon 17-35mm f2.8 and a Breakthrough 10 stop ND filter and shooting in RAW. ISO is 100. I have been experimenting with this new filter. I don't know where this starts, but I know it is there at 20 minute exposures. If I zoom into 100% and definitely at 200% it looks like there are lots of pin holes in the image. These manifest themselves into bright specks. I am shooting with long exposure setting off, because the camera will work for more than 10 minutes cleaning up the image. That's not acceptable. Anyone know what's going on?</p>

<p>Thanks</p>

<p>I think it is probably due to sensors showing significant noise at long exposures.</p>

<p>or hot/dead pixels </p>

 

<p>Hot dead pixels !? That sounds serious and kind of what it looks like. Is there a solution?</p>

<p>The purpose of the "Long Exposure" setting is to follow the exposure with a dark exposure of equal length to subtract the effect of hot pixels and offset the random thermal noise. You might be able to do the same thing using software designed for long astronomical exposures, which use separate dark exposures for the same purpose.</p>

<p>I don't have the camera but my guess is 20 minutes is kind of long and 10 minutes processing is reasonable. Also, maybe you used a very small aperture like F16, F22 which shows many bad things on the filter</p>

<p>You can't subtract random thermal noise, as it is random.</p>

<p>But anything that isn't random can be subtracted.</p>

<p>As well as I understand it, they way camera sensors work is that they come along with a look-up table giving the sensitivity of each pixel. That works well for ordinary shots, but maybe not enough for very long exposures. </p>

<blockquote>

<p>You can't subtract random thermal noise, as it is random.</p>

 

</blockquote>

<p>True, but it averages according the law of squares. The net noise is inversely proportional to the square root of the number of samples.</p>

<p>So, in practical terms, how to do a 20 or 30 minute exposure and optimise the result? Is there a "B" or "T" setting on most digital camera shutters and can they be invoked manually? Yes, I know, I should read my 100 page camera manual....</p>

<p>The nice thing about film cameras is that you can go for long exposures and you need only worry about reciprocity failure unless the color rendition changes substantially with time.</p>

<p>Astronomers learned a long time ago not to shoot very long exposures with digital cameras. They have developed stacking software so that they can take a series of 30 second exposures and combine them. You don't mention the subject matter. Would stacking software work for you?</p>

<p>http://imaging.kodakalaris.com/sites/prod/files/files/products/e4050_portra_400.pdf</p>

<p>says no reciprocity corrections from 1/10000s to 1s, and suggests tests under your conditions for longer exposures. In the case of color film, there could be color shifts. </p>

<p>But since I am mostly guessing in the first place for long exposures, I don't usually find much need to look at tables.</p>

<p>(regarding subtracting random noise)</p>

 

<blockquote>

<p>True, but it averages according the law of squares. The net noise is inversely proportional to the square root of the number of samples.<br>

</p>

</blockquote>

<p>If you average a series of measurement, then the signal/noise improves by sqrt(N).<br>

<br>

That is, the signal increases by a factor of N, the noise by sqrt(N), and so the ratio improves by sqrt(N).<br>

<br>

For digital photography, you could average N images to improve S/N. (As noted later in the thread.)<br>

<br>

But you can't take one image, and average with dark (shutter closed) noise. This will increase the noise, and decrease the S/N. </p>