5DIII noise question

<p>Hi all,</p>

<p>I'm doing some postprocessing work for a client, and I've noticed he has a lot of hot pixels clusters (maybe a dozen that really glare at you). I've only seen his shots at 30 sec exposure, either ISO 800 and ISO 1000. These clusters are in the same place on each frame and glare in the usual way from the deep shadows. The larger ones are vivid magenta blocks, 7 pixels square. A number of smaller ones are 3x3 blocks, usually white.</p>

<p>He says his camera was set for dark frame noise reduction, and his camera did sit there for 30 sec after each 30 sec shot, apparently taking a dark frame. I do not know the 5DIII modes, but his exif info indicates:</p>

<ul>

<li>Long exposure noise reduction = auto</li>

<li>High ISO speed noise reduction = standard</li>

</ul>

<p>(I don't know what auto and standard are.)</p>

<p>What I am wondering is this:</p>

<ol>

<li> Do these camera settings indicate he was correctly configured for dark frame noise reduction?</li>

<li> Would a dark frame noise reduction operation not nullify these hot pixel blocks?</li>

<li> Is this sort of noise normal from a 5DII at ISO 800 or 1000 for a 30 sec exposure?</li>

<li> Is there reason to believe his camera needs servicing, e.g. to map out the hot pixels?</li>

</ol>

<p>Thanks, everyone!<br>

Sarah</p>

<p>Auto, and standard are: Auto (ie, LENR is activated on exposures where the camera feels it's appropriate), and standard (vs. OFF, LOW, HIGH) . <br>

These are the default settings for both.</p>

<p>So, no. his settings are not correct. Even though LENR was probably active, he was shooting at ISO 800-1000 - never a good idea for long digital exposures (although the problem is usually not this severe). </p>

<p>LENR only works if the noise is genuine noise. A truly hot pixel (and it's charge bleed to adjacent pixels on a long exposure) isn't really 'noise' (ie. it occurs continuously, not randomly), so LENR is not going to affect it. </p>

<p>To determine whether the camera needs servicing, I'd test it more comprehensively, though I'd guess that yes it needs to be mapped properly. Of course I don't know if the mapping is dynamic (ie. applied to images differently based on shooting settings), or is static. </p>

<p> </p>

<p>I'm sorry, I totally mis-thought! <br>

Assuming LENR was active (likely), LENR <em>should</em> affect it, since, by definition, the 'hot' pixel <em>should</em> still be hot (and affecting the surrounding pixels) even when the shutter is closed. It should have been recorded as such and eliminated from the RAW </p>

<p>Perhaps instead of a 'hot' pixel, you have some extra sensitive 'leaky' ones that are triggering those around them? </p>

<p>Either way I would be running that body through a variety of tests to find more data... with such a limited sample, there's no telling...<br>

</p>

<p>Thanks, Marcus!</p>

<p>I get your "leaky" pixel theory. However, some of these exposures are REALLY dark, and there wouldn't be much to leak. Nevertheless, I'll suggest he examine a true dark frame -- turn out the lights, walk into a closet, put the lens cap on, fire off a 30 sec exposure. ;-) Good thought.</p>

<p>I'll suggest he also study his settings and force a dark frame noise reduction to confirm that this does or doesn't nullify the hot pixels. But yes, it seems very strange they're still there.</p>

<p>Does Canon have any specifications for how many hot pixels are allowed before a sensor is deemed defective?</p>

<p>Hi Sarah,<br>

I'm not sure if this will help, but I have noticed that hot pixels can be seen using Adobe Bridge. If I View a raw file (from 5D2) of a star field, i.e. lots of black background, as the picture is loaded the hot pixels appear momentarily and then disappear. I assumed that the raw data was loaded and then the mapped hot pixels were subtracted on the fly. If you then zoom in to 100%, then the hot pixels reappear until it snaps into focus. If you then grab the zoomed view and move around the hot pixels reappear and stay there until you stop moving the frame. <br>

My thought is you might try this and see if you can observe hot pixels coming and going as described. If some of what you think are hot pixels are not behaving this way, then I guess the sensor needs to be re-mapped or there is some other problem. The odd thing is that you describe the hot pixels as clusters of vivid magenta and white. All the hot pixels I see are distinctly red, blue, or green.</p>

<p>View of hot pixels?<br>

Thought I see if I could illustrate the phenomenon - on the left a shot of the screen (taken with camera) and on right a shot of the same area taken whilst moving the frame, you can see 4 hot pixels that seem to have been subtracted as soon as you stop moving the frame.</p><div></div>

<p>Thanks Rick! I didn't see your response until now. (I've been out of town for a few days)</p>

<p>I'm using DPP for the conversion. The hot pixels might glow a bit when panning around in magnified view, but they're still there when I stop panning. The same hot pixels are also evident in the TIF conversions provided by the client. Here's a collection of them, taken from a very dark part of the frame, with the exception of one that was taken from the dusk sky.</p><div></div>

<p>[[and stay there until you stop moving the frame.]]<br>

Adobe is using their processing algorithm to attempt to remove hot pixels. It the version of the image you see while panning is the embedded JPG not the rendered preview from the algorithm. Once you stop moving, the Adobe software takes over again and renders the image.</p>

<p>All your observation tells you is that Adobe software has some secret sauce. It doesn't tell you anything about the sensor that you didn't already know. </p>