Emergency - 20D giving bad long exposure noise!

[Bootscreen]

Having a bit of an emergency on holiday in Florida. I've just got a 20D so I'm not completely familiar with all of it's characteristics, but when I take long exposure night shots they are covered in noise hotspots is this normal?

The ones I've taken so far were with the noise reduction turned off so I guess there should be some noise, but I wasn't expecting this level, it's like the sky has blue, red and white stars all over it (and they definitely aren't stars). Has anyone else tested the 20D with long exposure nights shots (2 sec - 30 sec) and found them very noisy?

I would be grateful for any advice as I wanted to photograph Miami at night over the couple of evenings.

Regards,
David

[Adam Hicks]

Can you post a sample image? That might help us determine if it's anything we've seen or not.

My 20D has done well with long exposures, even though I haven't done any 30 second exposures, I've done some 2-10 second shots which have come out nicely.

Adam

[Bootscreen]

Can you post a sample image? That might help us determine if it's anything we've seen or not.
Adam

This is one of the problem images:

http://uk.pg.photos.yahoo.com/ph/bootscreen@btopenworld.com/detail?.dir=b118&.dnm=49d2.jpg&.src=ph

Please note this is a 5 Mb image so it's probably a bad idea to look at it unless you have a fast connection.

The details for the shot are:

Camera Model
Canon EOS 20D
Shooting Date/Time
11/10/04 00:17:49
Shooting Mode
Aperture-Priority AE
Tv(Shutter Speed)
8
Av(Aperture Value)
22.0
Metering Mode
Evaluative Metering
Exposure Compensation
0
ISO Speed
400
Lens
18.0 - 55.0mm
Focal Length
18.0mm
Image Size
2336x3504
Image Quality
Superfine
Flash
Off
White Balance Mode
Auto
AF Mode
Manual Focus

David

[djtowle]

I haven't noticed those kind of "stars" on mine, but I haven't looked closely either. For now have you tried setting custom funftion 2 for long exposure noise reduction? It will double the length of time an exposure takes 8sec becomes 16. but it may help with your problem. I believe it works by taking the initial exposure, then a second exposure with the shutter closed. This second exposure should be "black" but in reality contains the sensor "noise" for your camera at those settings, temp, etc. Then the two exposures are subtracted, added or exclusive or'd or something which is supposed to remove the noise... but I haven't played with it yet.

Going to have to be sure it is off and shoot some timed exposures, all those colored dot's seem a bit excessive, but I have no idea what canon might consider 'normal'

[Andy_T]

I'm not quite sure what you are referring to ...

I downloaded your photo and don't see any excessive noise in the picture. :!:

What I see is a picture shot at f/22 and with 8 seconds shutter time...

http://us.f2.yahoofs.com/users/4132fc45zd1a6afd8/b118/__sr_/49d2.jpg?phC4ykBBRUgj0Uxd

... 'aperture stars' (those stars originating from the bright lights) that are a direct function of using f/22 at night on any lens/camera
... lens flare? (the lighter spots in the middle of the photo close to the left pillar)
... fuzziness ... very likely diffraction at f/22, maybe also tripod shake during an 8 second exposure.

In short ... nothing that is not to be expected (IMHO) :roll:

some questions for improvement and consideration...
... did you use a cable release or self timer and mirror lockup (to prevent camera shake)?
... have you tried getting the same picture, say, at f/5.6 or f/8?

Hope this helps you....

Best regards,
Andy

[FlipsidE]

Looks like hot pixels to me. Hot pixels are to be expected, no doubt about that. But, I'm not sure about that many of them for that kind of exposure.

FlipsidE

[drisley]

On a 30sec exposure, I usually get around 4 hot pixels which isnt bad at all. The noise is very low.
Enable Long Exposure Noise reduction almost eliminates all the hot pixels.

[Bootscreen]

Thanks all,

I've set custom function 2 and I will test it tonight taking pictures of Miami, hopefully they will be clean.

The good news is that I can't see any problems on daytime photographs and high-iso without slow shutter, on closer examiniation of some shots I took at epcot it looks like some of what I thought of as hot pixles could be stars :) judging by the way that some are hard edged pure colour and others are fuzzy suggesting that they are an object. I took some pictures with the lens-cap on to test the sensor with function 2 off to attempt to profile the conditions when I took the suspect pictures. There was some noise starting at about 1 second as the manual suggests, but not as much as I am seeing in the photo's, I suspect this is due to me setting 400 iso for a long exposure instead of 100.

Anyway thanks for all your comments and advice, I will take the tripod out into the city tonight and see what we get.

Regards,
David

[emilbev]

Hi David, the example you sent uses a f/22 and 8 sec. Try using f/11. You will cut the exposure down to 2/sec and still get the same depth of field if you stay far enough away from the subject.

[Persian-Rice]

I would go ahead and do a hot pixel test.

-Turn Camera Off and let it cool for 1/2 hour.
-Place the lens cap on and go into a dark room, ie. washroom with lights off.
-In Tv Mode, ISO 100, take six shots. 0"5, 2", 8", 15", 30" & bulb 90".
-In Tv Mode, ISO 200, take six shots. 0"5, 2", 8", 15", 30" & bulb 90".
-In Tv Mode, ISO 400, take six shots. 0"5, 2", 8", 15", 30" & bulb 90".
-In Tv Mode, ISO 800, take six shots. 0"5, 2", 8", 15", 30" & bulb 90".

Take the images in Med/fine.jpg

The ISO 100/200 ones should all be fairly clean, they will have hot pixels though. ISO 400/800 more. Especially in the Lower ISO shots, look for pixels that return consistently in the same spot, known as stuck pixels.
If you have stuck pixels or excessive hot pixels, Canon will fix this for you. Don't be surprised to find many though, if there are any images that concerns you, postem up and we can take a look.

BTW. You will see all hot pixels in either red, pink, purple, blue, yellow, orange or green.

[drisley]

The good news is that I can't see any problems on daytime photographs and high-iso without slow shutter, on closer examiniation of some shots I took at epcot it looks like some of what I thought of as hot pixles could be stars :) judging by the way that some are hard edged pure colour and others are fuzzy suggesting that they are an object. I took some pictures with the lens-cap on to test the sensor with function 2 off to attempt to profile the conditions when I took the suspect pictures. There was some noise starting at about 1 second as the manual suggests, but not as much as I am seeing in the photo's, I suspect this is due to me setting 400 iso for a long exposure instead of 100.

Yes, it is possible that you are seeing stars.
However, you just noted that you used ISO400 instead of ISO100. That will DEFINATELY increase the number of hot pixels.

[evilenglishman]

at iso 800 30 seconds I see 9 spots full size image here (http://www.electricvenus.com/dump/ISO800_30sec.jpg)

is that good or bad?

[Persian-Rice]

at iso 800 30 seconds I see 9 spots full size image here (http://www.electricvenus.com/dump/ISO800_30sec.jpg)

is that good or bad?

Thats fine. The only situation I would call something bad would be if there are tons all over the place. Even at 3200, it will be bad, but even then it won't be awful.

Sometimes hot pixels show up because of the weather you are shooting. Shooting in a hot, dry summer night will cause you problems compared to if you were shooting on a cool spring night.

[evilenglishman]

thanks for the reply. :D

[Bootscreen]

Thats fine. The only situation I would call something bad would be if there are tons all over the place. Even at 3200, it will be bad, but even then it won't be awful.

Sometimes hot pixels show up because of the weather you are shooting. Shooting in a hot, dry summer night will cause you problems compared to if you were shooting on a cool spring night.

So would I be correct in thinking that hot pixels are an inherant problem with the technology when taking long / high-iso pictures and that as long as the pixels reset so that they aren't stuck it is not an indication of a camera fault, just another challenge to digital photography?

I did black photo test at various speeds to see how many hot pixels were generated. Interestingly there were less than when I took the original problem photograph. I'm guessing that this is caused by me turning the camera off for a couple of hours after I'd used it so the electronics were cooler.

[Bootscreen]

I've discovered that the problems I was worrying about appear to be caused by the Canon software. I recently bought Elements 3 and discovered that the problem photos when converted using Elements didn't exhibit the same colour noise. Breezebrowser also uses the Canon SDK and shows the same artefacts so I'm guessing the problem is in the Canon software.

[phili1]

I do not know about hot pixels but you get it with film as well. I happens with slow exposures and small F stops. I used to know the name for it but have not used it in so long I forgot.

There is nothing wrong with your 20D.

Here is an explanation of lens flare which will help explain that it is the light and lens causing this not the sensor. No matter how well a lens is coated in certain situations you can get it. Your type of flare is exagerated by the small aperature.

Here is an article that give a somewhat expalnation.

http://jonathanclark.com/diary/flare/

[Jon]

I do not know about hot pixels but you get it with film as well. I happens with slow exposures and small F stops. I used to know the name for it but have not used it in so long I forgot.

Reciprocity failure. The problem occurs in the same conditions, but itdoesn't look the same. It's a whole-frame problem rather than an individual pixel problem. Symptoms are - under or overexposure at metered settings (ISO shift) or colour balance shift (different layers of the film respond differently at lower speeds).

[phili1]

Reciprocity failure. The problem occurs in the same conditions, but it doesn't look the same. It's a whole-frame problem rather than an individual pixel problem. Symptoms are - under or overexposure at metered settings (ISO shift) or colour balance shift (different layers of the film respond differently at lower speeds).

Jon you are correct about Reciprosity being a color & exposure shift.

His picture has star like light patterns which is lens flare but there is a specific name for it that I can.t remember.

This type of flare can be controlled by adjusting the F stop, in fact Adobe Photoshop allows you to put it in.

[Jon]

Oh, that.That's just starburst diffraction patterns, and is going to vary depending on the number of blades in the diaphragm, but happens only at small apertures independent of shutter speed. You said "long exposures and . . . ".

[Bootscreen]

I'll post a screen-shot of the problem later tonight when I get home. The problem looks like hot-pixels in that there are a number if single bright pixels that are usually red or blue. I don't think they really are hot-pixels though as they don't show-up when I use Elements to convert the RAW file, though I can get a similar effect if I use very high sharpening on the image.

[Jon]

The various different RAW conversion programs use slightly different algorithms to convert to JPEG, so it's not really surprising that one shows different characteristics than another. Try shooting directly in JPEG and they'll show up just as well.

[phili1]

Now I see what he is talking about. Is next to the left tower. That is flare weither caused but the lens or the program. If you follow it down it goes to the star light. that is caused by the angle of the lens to the light as it hits the film or sensor. It can also be caused by dust on the lens.

Sorry I di not sse that I thought you were talking about the star pattern.

[DamienB]

This thread is a perfect demonstration of how useless discussion forums can be sometimes, with every other poster talking about something different and utterly ignoring the evidence of the posted image!

Now given the original poster's complaint and the image, I thought it was perfectly clear he was talking about the individual red, white, blue and green dots that speckle that image - hot pixels. On top of those there are several very clearly defined X's made up from one bright white pixel surrounded by four slightly darker ones. Look to the left of the middle crane thing - a red dot, a blue dot below it and a white X in between and to the left of the red/blue pair. Look at the foliage in between the cranes, where there is even a white pixel surrounded by a black circle. These are all pretty nasty additions to the image and I find them pretty surprising for the newest Canon DSLR on a mere 8 second exposure.

My D60s need a good 15-20 seconds to beat that level of malfunctioning pixels (though unlike the 20D they begin to show up a purple bloom along one side from about 5 seconds on). I guess this is why Canon reintroduced the dark frame substraction!

[Chazs]

Enough of all the technical stuff. It's got to be a bad sensor. I performed an extreme leveling on it in PS (take the white slider all the way to the left), and on a Jupiter photo of mine with a 10 second exposure on a DRebel. Not scientific. But if my sensor had that many ufo's I'd get a new camera.

His photo
http://members.isp01.net/snevetsc/photos/hisimage.jpg

my photo
http://members.isp01.net/snevetsc/photos/myimage.jpg

[Jon]

This was RAW, converted, I believe he said, but when he ran PS Elements as the converter the hotspots weren't there? What happens on a straight JPEG, not RAW converted? If they're hotspots, they'll show up in the JPEG. If it's an artifact of the Canon CR2 converter, they won't be there.

[Bootscreen]

OK, as promised I've uploaded a few images:

The first is a screendump of the image as seen through DPP 1.1

http://uk.pg.photos.yahoo.com/ph/bootscreen@btopenworld.com/detail?.dir=b118&.dnm=af1a.jpg&.src=ph

The second is the same image as a screendump viewed through Elements 3

http://uk.pg.photos.yahoo.com/ph/bootscreen@btopenworld.com/detail?.dir=b118&.dnm=c304.jpg&.src=ph

And the third is the RAW file converted in Elements with auto-level, auto-contrast, and auto-colour correction applied. Note there is no sharpening.

http://uk.pg.photos.yahoo.com/ph/bootscreen@btopenworld.com/detail?.dir=b118&.dnm=c026.jpg&.src=ph

Obviously the images look different but does it indicate there is a sensor problem or a software problem? Don't forget that I didn't have the noise reduction option set for this image so I would expect a certain amount of hot-spots.

The settings for the image are:

ISO: 100
F8.0
30 Seconds

[Jon]

What happens on a straight JPEG, not RAW converted? If they're hotspots, they'll show up in the JPEG. If it's an artifact of the Canon CR2 converter, they won't be there.
Have you tried shooting in JPEG, and do you see the problem there? Have you used DPP 1.5?

[Bootscreen]

Dear all,

I'd like to thank you all for your suggestions and help. I finally compared my 20D with a 300D taking 30 sec images at 100 iso with the lens cap on, the 300D won hands down so I am now convinced there was a problem with the camera.

I've since taken the camera back to the store and they have replaced the body with a new one. Early tests are encouraging, images are now coming out of the camera looking reasonably sharp (previously the noise made them look very soft) and the huge number of hot-pixels seems to be reduced. I can't be sure that everything is OK yet as I haven't had the opportunity to take many pictures but I think I would be enormously unlucky to get two bad 20Ds in a row when there are so many other owners who haven't had any problems.

David