Massive number of Hot Pixels on 5D Mark II

<p>I went out to take pictures of the night sky with my new 5D Mark II. And found hundreds of red and blue dots on my shots. The 5D isn't my first SLR and these weren't my first star shots. I found out later they were "Hot Pixels" and should be expected. But I had HUNDREDS of hot pixels.<br>

<a href="http://www.boydhobbs.com/wp-content/uploads/2009/06/IMG_0008.jpg">I went home and took a test shot with a two minute exposure at ISO 50 of the lens cap. The results where astounding. I posted the photo on my site here.</a> Take a look and tell me what I should do.<br>

I know these results are abnormal. I have owned a 10D and a 20D before and have never seen anything like this even on 30 minute exposures. Should I take it in, call the manufacturers?</p>

<p>Actually the results are absolutely normal. You need to turn on the long exposure noise reduction feature. Without it you will get exactly the result you describe - with it enabled the 5DII can do very wonderful night photography.</p>

<p>A while ago, I shot night photography with my 20D. The noise reduction feature was on and caused the camera to process a 30 minute exposure for another 30 minutes before it would even preview. I turned it off just to save time, but even then I never saw one hot pixel.</p>

<p>Boyd, that is just plain astonishing. I've shot a lot of night photography - and once or twice when I neglected to enable the long exposure noise reduction feature the results were (from my perspective) unusable.</p>

<p>You are correct that this feature creates a second "blank" exposure after the actual exposure. This is so that it can capture an image in the blank frame that contains only the hot pixel data, which it then subtracts from the first exposure containing your image.</p>

<p>Do try this. I am confident that the noise/hot pixel problem will disappear.</p>

<p>Dan</p>

<p>I have to say Boyd's results look excessive to me. I just shot a 120 second exposure on my 5D, at ISO 100, and I got this -<br>

<img src="http://img.photobucket.com/albums/v490/alanpix/IMG_4928M.jpg" alt="" /><br>

That's without noise reduction. Did I just get lucky and get a particularly clean sensor? I've only occasionally noticed bad pixels, even at high ISO.</p>

<p>I should have mentioned another common problem with long exposures. Many people who haven't done some night photography try for an exposure that looks as dark as the actual subject. This is almost always a big mistake - and one symptom will be excessive noise. Instead, whatever the image looks like in the LCD, shoot for a well-balanced histogram and then make any needed corrections in post.</p>

<p>I have a quick guide to night photography stuff posted here: http://www.gdanmitchell.com/2009/02/11/hints-for-night-photography</p>

<p>Dan</p>

<p>I suggest that we all start posting pictures of a 2 min. exposure with and without dark frame subtraction enabled, on cameras that support it. Post 600x600 100% center crops--or better yet, 200% crops because then it'll be easier to see the hot pixels. This way, we can avoid debates over pixel density and sensor area. A hot pixel is a hot pixel. Posting the same image size and magnification will allow us to better estimate hot pixel frequency.</p>

<p>Include the following information: (1) Camera model (2) Dark frame subtraction enabled/disabled (3) exposure time, if not 2 min., and (3) whether you posted 100% or 200% crop.</p>

<p>I'll post mine when I get the chance, which should be sometime Monday evening.</p>

<p>Don't know if it matters here...but I noticed lightroom removing hot pixels from nights shots I took with my 20D without long-exposure noise reduction. The picture in Lightroom (and the jpeg-export) looked quite clean. But when I opened the original JPG recorded in-camera (recording RAW+Jpeg using RAW in lightroom) I saw many more hot pixels...</p>

<p>It's not difficult. Shoot 10 images. Including some with the body cap on, at different shutter speeds. If they same pixels are red/blue/not-black in each image then they might be hot. If they move from frame to frame, they're noise.</p>

<p>Peter, mine are pretty clean</p>

<p>5D, ISO 100, no dark frame subtraction, 120 seconds, 100% crop from center -</p>

<p>http://img.photobucket.com/albums/v490/alanpix/IMG_4928M2.jpg</p>

<p>30D, ISO 100, no dark frame subtraction, 120 seconds, 100% crop from center -</p>

<p>http://img.photobucket.com/albums/v490/alanpix/IMG_9945M_30D.jpg</p>

<p>I should also add the photo was taken in RAW and processed in Aperture with RAW Auto Noise Compensation off.<br>

And this is a 100% crop of the image. The hot pixels are clearer here. These hot pixels do not change from picture to picture.<br>

<img src="http://www.boydhobbs.com/wp-content/uploads/2009/06/IMG_00081.jpg" alt="" /><br>

2 Min, f/22, ISO50, RAW, Noise Reduction OFF, 100% Crop, Of Lens Cap</p>

<p>I'm not going to take the time to engage in this test. I do a lot of night photography and I have made the mistake of doing long exposures w/o the long-exposure (or "dark frame") noise reduction feature enabled. I and the other night photographers I know and work with use this feature by default, and we have seen the effect of not using it.</p>

<p>Dan</p>

<p><img src="http://gdanmitchell.com/gallery/d/2625-3/MINSYWallBrickGreenMoon20090207+copy.jpg" alt="" /></p>

<p>I found one of my friends 5D MK II's today and officially tested the difference myself. The results are undeniable. I took 100% crop comparisons of the two with the exact same camera settings.<br>

On his there is only 1 Hot Pixel visible. On mine, well over fifteen. And these results are multiplied across the frame. Mine has hundreds, his has maybe a dozen total.<br>

<img src="http://www.boydhobbs.com/wp-content/uploads/2009/06/Hot-Pixel-Comparison.jpg" alt="" /><br>

At this point, I think the conclusion is undeniable. My sensor is damaged on some level.</p>

<p>The purpose of the test is not to compare the dark frame subtraction vs. no dark frame subtraction images taken on a single camera. <strong>The purpose of the test is to compare the performance across different cameras, both within a particular model and across different models</strong>.</p>

<p>Doing this test is a (mostly) effective way of determining what is considered "normal" noise performance for long exposures within a camera model. For instance, if one user reports much higher noise than others with the same camera, then we should see that in the comparison shots. The only reason to include both noise reduction on/off options is to establish a baseline "off" setting, and then to see the extent to which the camera successfully "hides" these pixels. Because different camera models may do this noise reduction with varying degrees of effectiveness, it is important to see what the unmodified dark frames look like in order to make an apples-to-apples comparison.</p>

<blockquote>

<p>At this point, I think the conclusion is undeniable. My sensor is damaged on some level.</p>

</blockquote>

<p>I absolutely concur. This is not acceptable performance out of the box.</p>

<p>Alan Bryant, your exif indicates that the file was processed in CS2. Lightroom/ACR will remove hot pixels automatically (notice that they disappear once LR fully loads the full size file?)</p>

<p>That's Photoshop CS2, which does not alter images automatically, unless I am very much mistaken. I didn't use Lightroom; I don't have Lightroom. I used JPEG, not RAW, to sidestep issues with the RAW converter, and also preserve the EXIF. I realize that keeps Digic in the loop, and that will have some small effect - but seriously, the difference between Boyd's and mine is not subtle.</p>

<p>If you within the stores return policy, I suggest returning it for a new one.</p>

<p>Otherwise contact Canon for an RMA.</p>

<p>Yeah that looks pretty bad. My 300D hot heaps of hot pixels before I sold it. When it was new the problem was not as bad.</p>

<p>I found that PS "Filters - Noise - Dust and Scratches" does a pretty good job of removing hot pixels. Small radius, high threshold, and run it on a layer and then use the eraser to bring back areas where the filter has caused excessive smoothing.</p>

<p>All these excuses for a defective sensor! How can totally unpatterned, random, hot pixels be attempted noise reduction. You would see banding or at least a structure to the NR. Its a dud sensor. Black should be black, or are we being conned? Get a new one.</p>

<p>I agree it's not right. It doesn't matter what LR will do in post - what's happening before that is unacceptable. If it were me, my question about the sensor would start to expand beyond night images. Get a replacement ASAP.</p>

<p>Boyd,<br>

I have a MKII as well, it doesn't normally have so many hot pixels at such low ISO, I would return the camera. </p>

<p>I have completed the test shots on my 5DmkII. ISO 100, 2 minute exposure. 100% crops, 600x600 pixels. No post processing applied except conversion from RAW to JPEG in DPP.</p>

<p>Sorry, I screwed up the attachments. This photo attachment method is rather clumsy. You can't change the file after it's been uploaded--you can only delete it. :P</p>

<p>Here it is, dark frame subtraction OFF:</p>

<div></div>

<p>And here it is, ON:</p><div></div>