Is the 20D crap or am I doing somthing wrong?

[Bootscreen]

Yes it's an inflammatory question especially since most of the postings I read on this forum about the 20D are glowing. Having said that I appear to have had two 20D bodies that exhibit an enormous amount of noise for long exposures.

It is very possible that the method of testing the camera is flawed so I will describe what I have done and perhaps someone could show my why my tests are wrong or unreasonable.

The tests have consisted of 2 20Ds and 1 300D. Using the same 1GB CF card (reformatted between uses) I took two images with the lens caps on at 100 ISO and 30 seconds, the 20D had noise reduction off as I believe the 300D doesn't have any, both images were in RAW format.

These are the results I after opening them in Elements 3 and moving the slider all the way to the left in levels.

http://uk.pg.photos.yahoo.com/ph/bootscreen@btopenworld.com/detail?.dir=f721&.dnm=70eb.jpg&.src=ph

http://uk.pg.photos.yahoo.com/ph/bootscreen@btopenworld.com/detail?.dir=f721&.dnm=dc07.jpg&.src=ph

As you can see the 300D has no noise at all, not a single pixel is non-black, whereas the 20D is covered in blue dots.

So have I been enormously unlucky and got a second body that is faulty, or is the test unreasonable, if it's an unreasonable test why is the 300D so good?

David

[Jon]

Try shooting in JPEG. There was a thread here earlier on a similar issue; I think that the original (at least) 20D raw converter software had a problem; whether DPP 1.5 and EV 1.2 have fixed this, I don't know. Anyhow, the poster was seeing speckles in his long exposure RAWs when he converted them using Canon's software, but not when he used PS CS for the conversion. But since the 20D does have noise reduction, I'd use it in the test. The goal of taking test photos is to get the best possible out of the camera in question so you can get a good assessment of what its capabilities are.

[davidwegs]

Yes it's an inflammatory question especially since most of the postings I read on this forum about the 20D are glowing. Having said that I appear to have had two 20D bodies that exhibit an enormous amount of noise for long exposures.

It is very possible that the method of testing the camera is flawed so I will describe what I have done and perhaps someone could show my why my tests are wrong or unreasonable.

The tests have consisted of 2 20Ds and 1 300D. Using the same 1GB CF card (reformatted between uses) I took two images with the lens caps on at 100 ISO and 30 seconds, the 20D had noise reduction off as I believe the 300D doesn't have any, both images were in RAW format.

These are the results I after opening them in Elements 3 and moving the slider all the way to the left in levels.

http://uk.pg.photos.yahoo.com/ph/bootscreen@btopenworld.com/detail?.dir=f721&.dnm=70eb.jpg&.src=ph

http://uk.pg.photos.yahoo.com/ph/bootscreen@btopenworld.com/detail?.dir=f721&.dnm=dc07.jpg&.src=ph

As you can see the 300D has no noise at all, not a single pixel is non-black, whereas the 20D is covered in blue dots.

So have I been enormously unlucky and got a second body that is faulty, or is the test unreasonable, if it's an unreasonable test why is the 300D so good?

David


So its a black sky just after the last bit of light left the scene? Or perhaps a charcoal sketch of the above in a black matte?

Ok, so I see your point but its not always the camera, Its normally the processing used. If you can, try PS CS, I don't get the results you do but I do get a little color in there. Its slight, but there.

[CyberDyneSystems]

Your post did not indicate if you are using the long exposure noise reduction function?

I am sure the above replies regarding post processing RAW files are more pertinant.. but the 20D's CF setting still may be of some help.

[Bootscreen]

The noise reduction was off since the 300D doesn't have noise reduction. I have repeated the tests with NR on and they appear worse for some reason

[Bootscreen]

The images are actually pictures of the inside of the lens cap. What I'm attempting to achieve is a zero signal images so that the only detail will be that created by the CCD itself. As you can see from the posted images the 300D doesn't produce any noise due to sensor leakage, whereas the 20D looks like a star-field. Since the 300D is effectively a 10D in a plastic body and the 20D is supposed to have much better noise reduction the test images should be at least equal not worse by a huge margin.

[Bootscreen]

Could I ask everyone that reads this thread and owns a 20D to try the test I have just done, that will at least show if this is normal or not.

Please do the test with the camera from cold, that is turn it on after it has been left for a while and immediately take the picture so that any effects caused by heat can be reduced.

Simply take a picture with the lens cap on at 100 ISO with a 30 second exposure using the RAW format. Then just copy the image to your computer and open it with an editor that can read the Canon RAW format, please don't apply any adjustments to the image as we are interested in as close as possible an image that is a map of the CCD noise.

Once you have the image in your editor open the levels control and move the far right arrow (in photoshop) all the way to the left. You should initially see a single black bar on the far right since there shouldn't be any light to generate a signal, as you move the white level to the left you are making anything that is more than black visible i.e. the noise generated by the CCD. Once you have made these adjustments you should hopefully still have a pure black image which would mean that the CCD isn't generating any noise at these settings, anything other than pure black I think suggests a problem since the 300D doesn't produce any noise at these levels and the 20D is supposed to be better.

Just in case you are thinking why is it important? Well, these setting are hardly an extreme case as this ISO is the lowest setting the camera can achieve so everything will be much worse as you increase the ISO, and for anyone who enjoys taking night-time images or star trails you can basically forget it if the camera is producing this kind of noise.

David

[pcasciola]

At 30 seconds I get almost no noise at ISO 100.

Here's my 11 minute test at ISO 100 (Note: full size JPEG ~2.5MB):
http://www.casciola.com/pics/ISO100-11min.JPG

[RichardtheSane]

That is pretty good!

[paulhillion]

Just done a test exactly as you described and ended up with a perfectly solid black image! Hope you can get whatever's wrong sorted.

Paul.

[Whaler]

Yeah . . . Take the frigin lenscap off and go out and have some FUN with you 20.

[Bootscreen]

Yeah . . . Take the frigin lenscap off and go out and have some FUN with you 20.

Thanks for the helpful comment and I'm sure you feel it's pretty anal to get worried about this, but the fact is that long exposure photographs of night scenes are a favourite of mine and the first 20D I had has gone back because all of the night scenes I took on holiday in Florida were ruined by noise, so I need to know if this is another bad body that I have to take back before the shop will refuse to exchange it.

[gramps]

I just did your test. 30 sec. exp. f2.8, cap on and the lens was the 24 - 70L. I shot a large raw image, opened it with Canon's Digital Photo Pro ( the software that came with my 20 D. Saw nothing but black and the dust on my monitor. I opened the imbeded jpg file with PSE (huge file) and saw pure black again. Here is a linke to the pic, it's a bit too big to post here on the forum. Let me know what you think.

http://www.pbase.com/sjh/image/37519507

[phili1]

Well I guess I can't under stand your test. If you do not have any light hitting the sensor or film you get nothing. How can you evaluate someting that has had no exposure at all.

Second long night exposures have been a problem with film as well producing color shifts and heavy grain.

Third I looked at the pics you posted and unless I am going blind I do not see what you are talking about.

I think what you are trying to acheive is going to be a problem with any digital camera you buy.

And if the 20D is crap then it is the best crap I have ever owned and I owned a Hassleblad.

I know in night time photography they changed the normal exposure times to offset the reciprosity failure (color shift), you can try that.

[Radtech1]

MY GOD...IT'S FULL OF STARS!


Sorry, couldn't help it. Actually, I am suprised that no one else said it first.

On the topic, though, I do have a couple of questions:

•I am wondering if you got the same results shooting .jpg rather than raw.
•Also, does the noise insrease exponentially with an increase in ASA?
•Decrease linerally with decreasing exposure time?

Despite appearences, I am just not sure if a sample of only one technique can pinpoint the problem to noise. Could be an issue in the raw conversion, firmware, etc.
Rad

[gramps]

Could your problem be "hot pixels" and not noise??????

[Radtech1]

Well I guess I can't under stand your test. If you do not have any light hitting the sensor or film you get nothing. How can you evaluate someting that has had no exposure at all.
Long exposure random noise has been the bane of digital since it's inception. The problem is the heat generated by the flow of electricity through the image-receptor. This heat can trigger random excitations in the photosites, so, in a sense, the camera "sees" the heat. The longer the exposure, the more the heat, the more the noise.

could your problem be "hot pixels" and not noise??????
Might be, but unlikely. That would be a heck of a lot of hot pixels, AND they would have shown up at more normal exposure levels. No, I think that this is true noise - random fluxuations in the signal.

Rad

[tumb]

When taking night pictures with noise reduction (C.Fn-02=1) I get excellent images, but with it off there's lots of noise. I did read an article recently that spoke of problems with DPP's conversion causing excess noise, and my experience has shown this to be true. EVU is better, and PSCS is better yet. I didn't see if you mentioned which conversion software you are using, but you have to turn noise reduction on or your pictures will look like poo.