Hot Pixels

Introduction:

Last updated October 1, 2003. Revision #21

New! Read about my own Hot Pixel experience with my own camera! A very interesting resolution. Click here.

Although this discussion uses the Nikon CP990 as an example, the basic information applies to almost all other digital cameras as well. I have now added additional information about other cameras throughout the article where needed.

There is a lot of talk about noise and hot pixels on longer exposure shots with digital cameras. Is this really a problem? Do you want to test your camera for hot pixels? I'll give my opinion here. Note that this page deals with hot pixels on the CCD (the sensor in the camera) that actually appear on your pictures, not hot pixels that you might see on the LCD screen, which is a different topic. Hot pixels on the LCD screen seem to be common on many brands of cameras, and most manufacturers will tell you that number can be tolerated and are not grounds for replacement.

What are hot pixels:

Hot pixels are individual sensors on the CCD with higher than normal rates of charge leakage. They can appear as small pixel sized bright points of light on longer exposures. Every pixel on the CCD has some charge leakage, and if you expose long enough, any pixel would light up. On a long exposure, you will see pixels ranging from just barely visible to possibly bright hot starlike points. There might be a few bright hot pixels, more intermediate one, and lots of very faint ones, an entire spectrum of brightness. At the low end, the faint hot pixels contribute to the noise in a picture.  All cameras on the market today have "hot pixels". Because the rate of charge leakage is the same for a given pixel over time, the longer the exposure, the brighter it becomes. This charge leakage is worse at higher temperatures, even a 10 degree difference can be seen on an exposure.  Hot pixels are often just one site, and are one of the colors of the color filter array over that site. In the case of the CP990, hot pixels are usually Cyan Magenta Yellow or Green, although other colors are possible if you have two adjacent hot pixels. Green is the most popular color for hot pixels because there are more Green sensors on the CDD. Most cameras made after 2002 now have a noise reduction technique used on longer exposures that involves taking a duplicate exposure with the lens covered that is then subtracted from the main exposure. Also note that some cameras use software noise reduction that makes hot pixles appear White instead of their normal color (Minolta and Sony do this).

A real size crop of a hot (Pink!) pixel on my Nikon 990 at 1/11 second against a normally lit background (a door). See it in the upper left corner?

 

So hot pixels cause noise in low light pictures?

Most of the noise in a low light exposure is actually not from hot pixels, but is due to other factors. One factor is that not all sensors on the CCD are equally as sensitive, for a number of reasons. One site might be just a little less sensitive or a little more leaky than the ones next to it, resulting in some fine grain noise on the final picture. The main factor that causes noise is a statistical effect, due to the small number of electrons that are actually involved at each sensor site which in turn is due to the very small size of the sites. This is because the entire CCD is very small, (about 8mm per side in the 990), so that each individual CCD element or pixel is only about 3.5 microns across. Not every photon that hits results in a free electron that can be detected later, and not every electron will end up being detected. With the losses through the lens system, the small sensors, the low efficiency of the CCD (about 10-20% in consumer CCDs), and the loss in the small colored lens that is placed over each site for color interpolation which excludes at least 1/3 of the light, there do not end up being that many electrons produced by photons hitting each individual site, it might only be a few hundred that are finally detected. There is statistical variation in any small randomly distributed sample that says that the number of electrons will vary by the square root of the total number, so if there were 100 electrons you would expect to see +/- 10 or a 10% variation, 1000 electrons +/- 31 electrons or a 3 percent variation, etc. You can see that when the numbers are small, the variation gets big. This contributes to some of the random noise that you see on a low light shot too. The less the light, the worse the effect is.

This is why some cameras such as the Canon D30/D60/10D can produce such low noise pictures, they have larger CCDs. The sensor sites on the D30 CCD have about 10 times the area of the sites on a typical 3.4 or 5.2 Mpixel CCD. In addition the Canon CMOS sensors have a special noise readout that is tested for each exposure.

That being said, it is true that if you have some very hot pixels, they can be particularly noticeable because they stand out.

Hot pixels vs. Stuck Pixels

The terms Hot pixels and Stuck pixels are sometimes used interchangeably by people on discussion groups. A Stuck Pixel is just a pixel that for practical purposes appears on all of your shots, where as a Hot Pixel is one that only appears on longer exposures. If you expose long enough, you will find Hot Pixels on any camera. Obviously there is some overlap and the Stuck Pixel can just be an extreme case of a Hot Pixel, one that has so much charge leakage that it appears on all exposures. Most people would agree that a pixel that appears on many of your normal exposures could be called "stuck", there is no formal agreement as to what the terms mean.

Some terms.

 Stuck pixel = a pixel that always reads high (maximum) on all exposures.
 Hot pixel   = a pixel that reads high on longer exposures.
 Dead pixel  = a pixel that reads zero (black) on all exposures.

What is so bad about hot pixels?

If a hot pixel is bad enough, we call it "stuck", and it might appear in even bright exposures as a small starlike point. If a hot pixel is bad enough, it might even be annoying on regular shots. While everyone can agree that a pixel that appeared on all of your shots would be bad, most of the concern over hot pixels involves low light shots, shots over 1/4 second exposure.

Sure, you could "clone out" a stuck pixel on all of your shots. Sure, it would not bother everyone. Most people though can agree that a stuck pixel would bother them.

Factors that influence hot pixels

Temperature:

The warmer the CCD, the brighter the hot pixels will be, and there more noise there will be. I have even heard that the noise might double every 10 degrees C or so.  This is due to thermal effects, the hotter the CCD is, the more electrons wander off into the substrate all by themselves and are detected as false positives later.

Gain:

The higher the gain, the more pronounced the hot pixels are. Gain in a camera is just the "ISO". Because of this, a camera will show more hot pixels at ISO 400 than ISO 100.  Like amplifying any signal, at some point there is no more information that can be gained by amplification, and you begin amplifying noise. This may be the case in the current crop of cameras with ISO 400 and 800.

Testing your camera for hot/stuck pixels:

First of all, ask yourself... do you really care?  If you have had your camera for a bit, and have not noticed a problem yet, is it something that you need to worry about? Why go looking for trouble. Interpreting the hot pixels that your camera certainly has (they all do) and deciding if it is "bad enough" to do something about, is an agonizing can of worms and is totally subjective.

How to test your camera:

You test by taking an exposure in pitch dark, with the lens covered for 1 or more seconds. This is called a "dark frame". THE EXPOSURE TIME IS CRITICAL! To do this right you have to...

* Set the camera to ISO 100. You can take some shots at ISO 200 and 400 i if you want also, I'll guarantee that they look bad right now. DO NOT USE AUTO ISO, the default setting the CP990 and other cameras. This is because when the camera "sees" that you are taking a picture that is pitch dark, it will try to boost the ISO up to 400 (200 on the CP995), and you will have horrible noise on even short exposures.

* On the CP990/995/5000 and other cameras that offer this feature, turn off auto contrast. Set this to Normal. This is again, because when the camera "sees" your shot, which is mostly dark with a few bright points, the auto contrast feature will go nuts and make things look worse than they really are.

* Turn off auto sharpening and set sharpening to "Soft" if your camera has that. Again, the camera will "go nuts" when it sees your super "unsharp" all black picture.

* The camera should be cool, not having been on for hours. The air temp should be less than say 75 degrees. This is because the effect is so much worse at high temps. Don't take your camera from a hot car and try and test it.

* Set the camera for Normal or even Basic resolution. For just a first eyeball look it is not necessary to use Fine or TIFF or Raw if your camera has that, unless you want to, it will make it harder to show others your dark frame if it is real big, and side by side tests show no advantage as far as just seeing if hot pixels are present. It is true that you should shoot higher quality if you want to process hot pixels out of your shots, but for this eyeball test there is no advantage to shooting larger shots. There is no harm in using TIFF if you want to, go ahead by all means if you want to be "sure", but it will make for a very large shot. The effect of using Normal compression is some very slight blending of hot pixels into the adjacent pixels that will not affect your test. If you find some pixels that bother you you can take a high quality shot later.

* Set the camera for shutter priority or manual mode. It does not matter what aperture the camera uses because it is totally dark! No light will be coming in the lens anyway. All that matters is the shutter speed.

* Head into a dark room, cover the lens tightly, and take some shots. Be sure to cover the lens well, as light from the monitor or even faint stray light can mess up your shot. I'd set for ISO 100, and take 1/4, 1/2, 1, 2,4, and 8 seconds to start with..

* Note that some cameras such  as the Nikon 995/5000/4500 have auto dark frame subtraction and noise reduction that kicks in at a certain exposure (for example at 1.3 seconds on the Canon G1 ) that will subtract the hot pixels from your longer exposures. Only the shorter exposures are meaningful in these cameras unless you turn off the noise reduction feature. On some cameras such as the 995 you can disable the dark frame noise reduction for your test. If you can disable dark frame subtraction, you should. Some cameras, such as the Minolta D7 series, use the deark frame subtraction technique on all "bulb" exposures, and you can't disable it.

* Note that there are some freeware software packages that will test your camera for hot pixels. I say, FORGET THEM. Just use your eyes.  I don't personally think that these programs are better for a first look than the test of just examining your dark frame with your own eyes; after all you are going to be using your eyes to examine your pictures in the end, however there is nothing wrong with using these packages if you interpert the results correctly.

     These packages have the ability to show you hot pixels that you didn't know existed. If you are a new user, you will see hot pixels with these packages, and they provide no guidelines as to what matters and what does not. If you use one, it is VERY VERY important that you consider the exposure time and ISO of the dark frame that you use; the software will show you lots of hot pixels on every camera if you get the exposure long enough. The one thing that the software can't do for you is tell you if the hot pixels that it finds are "bad enough" to bother you. It may have some preset threshold values, but only you can make a decision about what is bad with your own eyes, so be sure to take a look for yourself. There is posting after posting after posting from people who have looked at the results from these program on their new camera, see that it shows hot pixels, and think that they have a problem when they don't. For example, a Pink hot Pixel right in the middle might be more objectionable than a Green one off to the side, even though the software might mark them as being equally intense in your test. Some cameras such as the Sony series and the Minolta Dimage series seem to have a software noise reduction routine that removes the color from hot pixels, making them less visible on actual photos.

 

Interpreting your results. Should you be worried?

This is the part where you have to form your own opinion. All 3.3 Mpixel and up cameras will have some noise.  I will give you my opinion as a guideline, but you have to temper this with what you will actually be doing with your camera.

First, look at the 1 second ISO 100 shot. It should be pretty much black, with maybe a few faint pixels showing. If there is a very hot bright pixel, then you might have a problem and you will want to do further testing. As for the other shots, they are just for fun. The 8 second shot will almost certainly show some noise. If you are lucky, just a few will show at 8 seconds, everyone will have some hot pixels. Really, if you are clear on 1 second, you have a good camera in my opinion. People can compare longer exposures, but I don't really think that it is meaningful. For fun look at some long ISO 400 exposures, it looks like the center of the Galaxy.  What you should really be worried about are pixels that intrude on normal shots.

 

I see one at 1 second! What should I do?

Well, it would help to do more testing. Take some shorter dark frame exposures, 1/2, 1/4, 1/30 seconds at ISO 100. The goal is to find out if the hot pixel (s) are visible on shorter exposures.

If you have one that seems to be there on even very short exposures, you still have to decide if it is a problem for you. Do you really take that many 1/4 second shots of a pitch dark room? When was the last time you got out the tripod to take a 1 second exposure? Would it kill you to have to edit the pixel out on those shots? Take some normally exposed shots in dim light that you are likely to take in real life. The further the hot pixel extends into the "real life" shooting range (1/4, 1/15/ 1/60, etc.) the more you should worry, especially if it appears on normally exposed shots. The existance of hot pixels on longer exposures is of less concern if your camera has auto dark frame subtraction built in (usually called noise reduction mode), in that case you are just worried about hot pixels that are in the normal exposure range where the noise reduction can't help you.

Obviously, if you have a hot pixel that appears on every shot you take, it is going to be annoying, you should not have to edit every shot that you take. I don't think that you should put up with that.

I've got one or more that aren't that bad, but they are marginal, showing up at (say) 1/4 second. I think that my $1000 camera should be perfect

It is your right to exchange your camera if you want to. On the other hand, consider that every 3.3 or 4 or 5 Mpixel camera and up has some hot pixels, just varying in the degree. Judging from the frequency of the reports, you might even have one on the "good" end of the spectrum, and get a worse one when you exchange. You have to weigh how often you are going to actually use exposures long enough and of faint enough objects to see your hot pixel(s), and how easy it would be to edit it out to fix it in a second against the "risk" of exchanging your camera. I don't want to tell you what to do, but I will say that I like low light photography a lot, and I have some hot pixels at 2 seconds ISO 100, but I still take 99 percent of my shots in light where the hot pixels do not show up. I've had a hot pixel at 1/2 second on my CP950 for a year, and it has never bothered me.

I think that Nikon/Olympus/Sony should fix my new camera, I can't/won't exhange it

If it is bad enough, they will fix it by mapping the hot pixel out or by replacing the CCD if it is real bad (a group of bad pixels). If you get a new CCD, the new CCD might be just as bad or worse. Nikon has some criteria for deciding if your hot pixel is bad enough to fix, but that information is not public.  I don't know what criteria other manufacturers use. For Nikon, my impression is that if your pixels shows up at 1/4 second on a well exposed shot (not a pitch black lens cap on shot) that they might fix it. You can improve your chances by enclosing a disk with a sample picture on it, and enclosing a printout showing the pixel with the pixel circled. people have reported sending their cameras in and having them returned cleaned but with the same hot pixel.

I really want you to tell me if I have a good camera.

OK, if you were to force me to give my opinion, based on the reports I've seen so far, I would say that 1/2 to 1 second at ISO 100 is probably the break even point, half of the cameras have a stuck pixel at 1 second, half do not. I'm guessing! If your camera does not, I'd say that you have a good camera, I'd keep it. Even if yours does show a pixel at 1/2 to 1 second, I'd say that you have a 50/50 chance of getting a new camera that has one just as bad. If you have a stuck pixel visible on a totally black shot at say 1/8 second or less, then I think that it would be worth exchanging, and if it is visible at 1/60 second I'd definitely exchange it or send it in for repair.

OK I'm twisting your arm, I really really want to know if I have a bad hot pixel

I think that what you have to look at it this way, you don't want something that you have to deal with on your normal shots of course. If you see the hot pixel on pictures that you took of your cat or kids in normal light, then I would exchange/repair the camera.  You also have to consider that EVERYONE, even the very best examples of the 990/995/other, will have to clone or subtract out some hot pixels and deal with noise on very long time exposures, say over a couple of seconds, so just because you have a "hot" pixel or two on shorter exposures your camera would not be any different in the range of longer exposure shots; the same amount of work wil be required. There will be that little range around 1/4 to 2 second where a "perfect" camera might not need any touch up and yours will, but you have to consider how often you would actually take shots in that range, and how big the "my camera needs work and others would not" range is for you vs. how often you would really take those shots, balanced with the trouble of fixing/exchanging your camera.

As an example, I like low light photography, but low light is still only a few percent of all the shots I take. Most of my low light shots are several seconds, it is rare for me to use a 1/4 second on any shot. If you are taking 1/4 second or longer shots of casual subjects around the house, they are guaranteed to be blurry anyway if you aren't using a tripod. In addition, keep in mind that any real life low low light shots will not be completely black ;)  you will have some brighter exposed areas to mask any noisy pixels.

Another thing to consider if you are on the fence about sending your camera in is that you might want to wait in case you have to send it in for some other reason later in the warranty period.

I don't think that the camera should be this noisy at ISO 400/800.

Well, you are right, it can look bad. The problem is that all of the current crop of cameras have considerable noise at high ISO. Some solve the problem by only allowing ISO 100, or limiting the longest exposure to 1/4 or 1 second. The only solution to the noise problem will be bigger CCDs, (like in the Nikon D1) and that is very expensive at the moment. Even when the prices come down, the larger CCD will mean larger cameras, no more pocket cameras with those. Personally I appreciate being able to use 400 if I want, and being able to use the longer exposures rather than have the manufacturer prevent me. Most new 3.3 Mpixel and up cameras now haves an ISO 800 mode which is REALLY noisy and really just for situations where the alternative would be no shot at all.

Don't be fooled by other effects that aren't hot pixels!

Many users have been fooled by the various reflections of the flash that can look like hot pixels. In particular I've had two people send me pictures of swimmers with water drops on them with small spots on the drops caused by reflection of the flash... and several people have sent in pictures with flash reflections from clothing or the areas around the mouth and eyes or dust floating in the air... these aren't hot pixels. See the Spots! page for more details about these. True hot pixels are seen on longer exposures with the lens covered, and not usually on flash exposures (unless they are REALLY bad), will always be in exactly the same place and are just one pixel wide if you zoom in on the image. They are usually the primary colors used by the color filter over the CCD, Red Green or Blue for some cameras, and Cyan Yellow Magenta (Pink!) Green for the Nikon cameras.

 

How come I just noticed this effect on my shots, I've had my camera for ...

First, you have to understand that all current consumer cameras will show hot pixels under the right conditions (some cameras now have the ability to subtract them out automatically, like the Canon D-30). The conditions that make hot pixels more visible, in order of importance,  are..

1) Long exposures

2) High ISO

3) Dark background over the hot pixel area

4) Hot temperatures

Although you could have had a new hot pixel develop (I did!), the most likely explanation is that you just never had the right exposure, ISO, background, and temperature to notice them on your shots before. Long exposures bring out hot pixels. The longer the exposure, the worse they are. Maybe you have never taken as long an exposure as you did when you noticed them. You might see hot pixels on exposures of 1/4 second if the ISO was high. High ISO also brings out hot pixels. Maybe you had your camera set to Auto ISO, so that the camera is using ISO 400 in low light. Go back and check the exposure and ISO on the offending shots, and compare to other shots. Dark backgrounds make hot pixels more visible. Maybe before you never had the area containing the hot pixel in the shot so dark. Hot temperatures can make hot pixels worse, maybe you had the camera on for a long time before taking the bad shots. Keep in mind that ALL cameras shot hot pixels with the right conditions.

It is possible that a new hot pixel could have developed. There have been sporadic reports of this happening, and it happened to me, three times now. Like all electronic devices, the failures seem to decrease the longer you have had the camera. Apparently all CCDs are going to develop hot pixels as they age.

 

How to test in the store before you buy

There are several advantages to buying a camera in a local store (but one advantage that you might miss out on the price). You might expect that you could test a camera in the store for "hot pixels" before you take it home, but the short answer is that unless it is a horrible defect covering a large area, you can't test in the store by using just the camera alone; you will need some way to view enlarged images.

The long answer is that the reason that you can't test most cameras in the store is that the individual hot pixels are too small to be seen on the full image zoom in (4X on the CP990, 5X on the Canon G1, 6X on the CP995, 4X on the Dimage 7, etc.) on the LCD unless you have a REALLY bad cluster of them. That is because the actual resolution of the LCD is not that good, 240x180 pixels is typical. Even at full playback zoom, you cannot see the individual pixels on your shot to examine them. The only way to examine your new camera in a store would be to take a laptop computer with you, take a test shot, and then download it to the laptop using a PCMCIA card slot and adapter for the memory card and then examine the shots on the laptop (not a bad idea if you can do it!). Since this might be impractical, you might want to sound out the dealer about returning the camera if you have a problem before you buy.

Note that any bright points that you see in the same place on the screen of the camera are usually actually "hot" pixels on the LCD screen, not on the CCD. I'd consider ignoring the LCD pixels if you can get a camera with a good CCD. Nikon says that two or three stuck pixles on the LCD screen are normal and they will not repair them (again, a cost saving measure I'm sure).

How do Nikon and other manufacturers fix hot pixels?

I used to think that Nikon would always replace the CCD if you sent your camera in, however I have now had several reports that they simply mapped out the hot pixels (950 and 990) with software in the camera. One person reported that they charged him over $200! to do this out of warranty on a 950, but that the dark frames from before and after proved that it was actually the same CCD with the single hot pixel mapped out. In addition, he says that the hot pixel is still visible on the LCD screen when the shot is taken, but not on the final shot proving that it is being mapped out in software. Another person from Finland where they have walk in service reports that their hot pixels were mapped out in 15 or 20 minutes while they waited with no CCD replacement.

 Since so many cameras are returned for hot pixels, it sure would benefit Nikon and others if they released the software to let users do this. I've heard of people returning three or four cameas!  I wonder if Nikon has any idea why the return rate is so high on the 990... I guess we can dream. One way that the software might work would be to cause the camera to expose a frame with the shutter closed, then pick the "10 hottest" pixels out from the frame and map them out (depends on how big the internal hot pixel table is).  This would probably require the serial cable, and might require the user to cover the lens manually.

One reason that it might not be possible for Nikon to release their software is that the hot pixel table might be in write-only memory along with other parameters for the camera (such as white balance settings and focus calibration settings). It might not be possible to write just the hot pixel information in current cameras without calibrating the entire camera. I hope that Nikon takes hot pixels into consideration when designing future cameras, and makes the hot pixel table independently and user writable.

Reports are that other manufacturers also map out the hot pixels in software. (see below for some software that might help you do this yourself)

 

Some common questions about sending your camera in to Nikon have the hot/stuck pixels mapped out... (this part is for Nikon owners only)

1) If it is under warranty, there is no charge. If it isn't there are both reports of it being done for free, and of being charged up to $200.

2) When the hot pixel(s) are mapped out, you will not be able to detect the repaired spot in the final picture. There won't be a "black hole" there, for example. The hot pixel is ignored and interpolated from the adjacent pixels.

3) If you are in the USA, it usually takes about two weeks for Nikon. (Mine took three weeks)

4) When you send your camera in to Nikon, you don't have to pre-arrange anything with them. I've always just sent my cameras off with a letter explaining the problem. You should print out the repair form from the Nikontech web site and fill it in. Be clear, provide examples either printouts or files on a disk. Just as with any repair, the clearer you are, the more likely that Nikon will identify and fix your problem.

5) Why cameras come from the factory with horrible visible hot pixels, I don't know.

A really good idea!

Well, I think so anyway... An easy solution; Why doesn't Nikon (or other manufacturers) add a "user hot pixel fix" to Nikonview or other camera software. It could allow the user to fill in a table with coordinates of hot pixels, and fix them automatically by interpolation when the image was moved over from the camera. The EXIF data would be preserved. This would be a painless fix, and would take care of the problem for most customers and would not require any changes to the cmaera. If the program were to use exactly the same JPEG compression settings that the camera used, there would be very minimal loss due to re-compression.

An alternative to sending your camera in for repair

Since many older cameras are now out of warranty, I'll mention that a freeware program is now available that will work with many Nikon and some Olympus cameras to remap your hot pixels automatically. Follow this link. Note that this MIGHT damage your camera somehow, so use this at your own risk. The latest update to this program does support the Nikon 990 using the serial cable.

 

Myths about Hot pixels

I have seen some misinformation about hot pixels being propagated on the newsgroups.

1) Myth #1. Taking long exposures with your camera will permanently damage your CCD and "cause" hot pixels by "heating" the CCD.

This is simply not true. CCD heating is mostly not due to the exposure, but due to the readout process. The most heat generated in consumer CCDs is from the continuous readout needed to show you the live preview on the LCD. Taking a long exposure actually generates less heat because the CCD is not being read out during that time. Taking long exposures will not damage your CCD.  Pointing it at the Sun for a long time might ;)

2) Myth #2. Hot pixels are worse if you use the camera in cold temperatures.

Actually, the opposite is true, I'm not sure how this rumor got started.

3) Myth #3. If you see any hot pixels at all you have to return your camera.

I've seen several "I saw a hot pixel so I returned my camera" postings. All cameras can produce hot pixels if you get it hot enough and take a long enough exposure. People who return cameras without qualifying the conditions that hot pixels show up under run a significant risk of getting a camera that is worse. Just putting the lens cap on and taking a shot is NOT a good test, because most cameras will automatically use high ISO and long exposures that will reveal hot pixels on any camera under those conditions. You have to control the conditions of your dark frame shot as shown above, especially the exposure length! This is especially true of software that counts your hot pixels for you, pay attention to the exposure.

4) Myth #4. Dead pixels produce a dark hole in your shot.

A "Dead" pixel is a pixel that always has no response to light. If digital cameras produced their images with a one pixel on the CCD produces one pixel on the final image method, then a dead pixel would produce a dark "hole" in your final shot. However, all modern cameras use a color filter array over the CCD, and each pixel is sensitive to one color (often Red, Green or Blue). The final image is produced by interpolating the color information from adjacent pixels. Because of this, a pixel that does not respond produces an off color pixel or small just slightly darker area in the final picture that will be very hard to notice, even if you take a picture of a uniform surface.

5) Myth #5. You can subtract out "noise" on your everyday digital shots.

Although you CAN subtract out individual hot pixels on very long exposures, the noise that you see on everyday shots is not due to hot pixels. This noise has several causes, but the main one is random variation in the number of electrons captured by the very small and not very sensitive sensors involved. This noise is not reproducable from shot to shot, and you can't subtract it out. You CAN average it out though, if you take several shots of the same still scene and add them together. I wrote an article about doing this with the CP950 back in 1999 that might be interesting to look at, the results are pretty amazing but require a still scene. Click here to see the article.

Other cameras and hot pixel solutions:

Almost all consumer grade CCDs have hot or stuck pixels initially. It would be too costly to only use those few perfect CCDs that are produced. Every camera on the market has a hot pixel table, that is filled in at the factory with the worst offenders. You will never notice these interpolated out pixels in your shots.

Some cameras such as the Canon G1, G2 and D-30 and Sony S75, F707 and Nikon 995 and 5000 have special long exposure modes that kick in on longer exposures. They automatically take a "dark frame" with the shutter closed after the real exposure, and then subtract this frame from the exposure. Unfortunately, this is often done in a primitive way, just blacking out the offending pixel, which can leave a dark "hole" visible on lighter backgrounds. For best results, take your own dark frame and subtract it with software that can do real interpolation of the missing pixel from the surrounding pixels. Qimage Pro can do this. Some of the newer cameras such as the Nikon 5000 have very good long exposure noise reduction so that exposures over a minute are possible.

Flash! The Olympus E-10 and E-20 have a new user hot pixel management scheme with the 130 and 133 firmware. Apparently the camera can automatically take a dark frame and map out the hot pixels found! This is a large step forward, and user hot pixel management will someday be standard on all cameras. Unfortunately, you have to send your E-10 back to Olympus to have them do the firmware upgrade and pay if the camera is out of warranty, they don't allow user firmware upgrades. In addition, the Sony 707 and 717 can apparently map out hot pixels if you press the reset button. That is the way that it should be.

Can you get new hot pixels developing after some time? My hot pixel story

Yes, you sure can. Not only have there been many reports, but my own CP990 has now developed a new bright pink hot pixel after almost a year. If you know where to look for it, you can see it on all shots even up to 1/1000 second. This is very worrisome since Nikon will not fix cameras that develop hot pixels for free once the warranty expires.

Update: The camera came back after 17 days. They sent it back UPS ground which is why it took so long.  A very surprising thing, they replaced the CCD! The repair ticket says

REPL CCD

ADJ AUTO FOCUS OPERATION
ADJ A-WHITE BALANCE

GENERAL CHECK &  CLEAN

New dark frames show that the CCD is indeed new; none of the hot pixels on very long exposures match up with the old CCD. It is also VERY clean, no visible noise at 8 seconds ISO 100, and just a light scattering at ISO 400 8 seconds. I can't wait to try this out on some low light shots now!

I reviewed my shots to see when my hot pixel developed.  It was not visible on any of about 20 shots that I took on the morning of April 24th. The last shot, at 11:52am, DSCN6274, was of some flowers through a telescope, and the background in the area where the hot pixel would develop was almost black, so it would have been visible if it had been present.  I unloaded the shots with the USB cable, and the camera then sat on my desk for 5 hours. I next took it in the car in it's bag on a short trip to take some pictures while I dropped my kids off at a lesson. The next shot that I took, DSCN6275 at 4:02pm was of a farmhouse in bright sunlight at 1/1000 second. The hot pixel is visible if you know where to look for it even on this bright shot, although I didn't notice it until a few shots later with a darker background. I checked some older dark frames from the camera, and there is no "proto" hot pixel at that location. So the hot pixel apparently developed suddenly, with no obvious cause.

06/23/2001 Some additional comments on my repaired camera. The camera is MUCH less noisy now at long exposures. Previously, I felt that about 12-15 seconds at ISO 100 was about as long as could be used before noise overwhelmed the signal. Now, I can actually take 60 second long shots at ISO 100!  Although there is plenty of noise at 60 seconds, there is also quite a bit of signal. There does seem to be some random Red channel noise that is not subtracted out by the dark frame. One other thing that I have noticed is that the camera is now much more sensitive to heating than it used to be. Now, if I take several long exposure shots in a row there will be a noticable increase in noise. I had not seen this with the previous CCD.

02/09/2002 Update on the repaired camera. 7 months later the previously perfect replaced CCD is developing it's own hot pixels. I now have two pretty bright ones growing, visible at 1/2 second at ISO 100, and these are now blazing hot on long exposures. I think that this shows how CCDs deteriorate over time, and how it is critical that manufacturers start to address this. A user pixel map out would do this.