Need advice please about posssible 40D noise problem

NeilL

From the time I first got the 40D I thought I was seeing way too much noise in images shot in very normal conditions. So I got together recently with a friend who has a 10D to shoot some comparison shots. Here are a couple where the difference in noise is easily seen. The shots of the lights and pole are not totally identical, but perhaps close enough to allow for a comparison of noise to be done. ISO of both was 100. The ones of the building were set up to be identical, ISO 100, 1/6 sec, f2.8. Identical brightness adjustment was made to both to uncover noise. All were converted from RAW to tiff, and brightness adjustment of building shots made, in Aperture (his Mac), and from tiff to jpg in CS3 (my Win). All are at 100%.

Do you think the much greater noise in the 40D images is normal? Would very much appreciate your comments!

Thanks.
Neil

10D






40D





10D



40D

Scott_Quier

I'm only going to look at the last two images as the exposure parameters for the others are not consistant.

My Foxfire exif plug-in is reporting the following:
Last 10D image:
# Exposure Time (1 / Shutter Speed) = 3/10 second = 1/3.33333 second = 0.3 second
# Lens F-Number/F-Stop = 14/5 = F2.8
# Exposure Program = normal program (2)
# ISO Speed Ratings = 100
# Original Date/Time = 2008:03:10 16:28:26
# Digitization Date/Time = 2008:03:10 16:28:26
# Shutter Speed Value (APEX) = 2965/1707
Shutter Speed (Exposure Time) = 1/3.33 second
# Aperture Value (APEX) = 4281/1441
Aperture = F2.8
# Exposure Bias (EV) = 0/1 = 0
# Max Aperture Value (APEX) = 95/32 = 2.97
Max Aperture = F2.8
# Metering Mode = partial (6)
# Flash = Flash did not fire
# Focal Length = 100/1 mm = 100 mm
# Colour Space = 65535
# Image Width = 1024 pixels
# Image Height = 768 pixels


Last 40D Image:
# Exposure Time (1 / Shutter Speed) = 1/6 second = 0.16667 second
# Lens F-Number/F-Stop = 14/5 = F2.8
# Exposure Program = shutter priority (4)
# ISO Speed Ratings = 100
# Exif Version = 0221
# Original Date/Time = 2008:03:10 22:57:16
# Digitization Date/Time = 2008:03:10 22:57:16
# Shutter Speed Value (APEX) = 21/8
Shutter Speed (Exposure Time) = 1/6.17 second
# Aperture Value (APEX) = 3/1
Aperture = F2.83
# Exposure Bias (EV) = 0/1 = 0
# Metering Mode = partial (6)
# Flash = Flash did not fire, compulsory flash mode
# Focal Length = 100/1 mm = 100 mm
# Last Modified Subsecond Time = 04
# Original Subsecond Time = 04
# Digitized Subsecond Time = 04
# FlashPix Version = 0100
# Colour Space = 65535
# Image Width = 1024 pixels
# Image Height = 768 pixels
# Focal Plane X-Resolution = 261863/59 = 4438.36
# Focal Plane Y-Resolution = 288988/65 = 4445.97
# Focal Plane X/Y-Resolution Unit = inch (2)
# Custom Rendered = normal process (0)
# Exposure Mode = auto exposure (0)
# White Balance = auto (0)
# Scene Capture Type = standard (0)

Reading the above, it appears the exposures were not the same. The 40D exposure is 1 stop less than the 10D. That, alone, might account for the differences in the noise level. Plus, under-exposed/plugged shadows will exhibit noise much more easily/readily.

Oh, and you have to get the exposure right in-camera. Attempting to boost exposure in ACR will only increase the noise you see.

NeilL

Many thanks for taking the trouble, Scott! Yes, a shutter speed difference snuck in there! However, considering the 40D shots in isolation, I think this level of noise at ISO100 is terrible. My friend and I did a series of test pics @ ISO100, 200, 400 of this building and another, all with the unlucky exception of the 10D one here with identical exposure, and the 10D images are consistently cleaner. But I had already noticed the differences between his images and mine - that was what led me to suspect the 40D in the first place. I would go look at his images on his computer and be struck by their cleanness, and this was before he did any postwork.

The building images here are RAW converted to tiff and jpeg, no post processing was done except the hiking of brightness in Aperture for the sole purpose of accentuating the noise.

Given that the 40D has Canon's latest sensor and processor, and the 40D is considered by some to have superior noise performance to the D300, I find the noisier images from my 40D compared with my friend's 10D puzzling, to say the least!

It really means that night shots are out of the question, and even any shadows in normal light at ISO100 must be heavily treated for noise.

Scott_Quier

NeilL wrote:

The building images here are RAW converted to tiff and jpeg, no post processing was done except the hiking of brightness in Aperture for the sole purpose of accentuating the noise.

And, that's what I was alluding to in my first post. Boosting exposure in post of an under-exposed shadow will always result in noise in those shadows, regardless of your ISO.

I'm thinking (and we all know that's the fastest way for me to get into trouble ) a better test would be be capture a properly exposed image (similar to the last couple in your original post) and then not boost the exposure/brightness in post. Inspect the result. If you, then, still have perceptible noise, then I think you might have an issue.

NeilL

Scott_Quier wrote:

And, that's what I was alluding to in my first post. Boosting exposure in post of an under-exposed shadow will always result in noise in those shadows, regardless of your ISO.

No argument with that, but still to be explained is the greater amount of noise in the 40D shot compared with the 10D shot given the same treatment.

I'll find a nonbrightened sample of this building. Stand by... In the meanwhile, I draw your attention to the noise in the metal post in the 40D shots - these were not brightened, and as I said need heavy noise reduction at ISO100!

Thanks, Scott.
Neil

swintonphoto

Another important difference is the different megapixel count. The pixel sites need to be smaller on the 40D because of the larger megapixel count (one of the reasons I wish the companies kept megapixels below 10). Thus, when you look at images at 100% they may appear to have worse noise in the 40D than the 10D. It is not good to compare noise on cameras with different megapixel counts by viewing at 100%. I don't know of the exact percentages, but it would be better to view the 10D image at 100% and the 40D image at 50% (Or something) - the key is to view the same percentage of the sensors image on the 10D and the 40D because then you will see noise comparisions using the exact same percentage of each sensor.
Am I making sense?

Another thing to consider - In order to actually view a 40D image at 100% in print you are going to have to blow up a REALLY big print. When you do this the viewing distance will be further away - so you will not notice the noise. Rarely are there situations where people will ever put their nose up to a large print to see each pixel. Viewing pixels at 100% on high megapixel count cameras is really not an accurate way of seeing the quality of the images. A better idea would be to take the same image with 10D and a 40D and then print a 16x20 of each - I guarantee the 40D image will look like it has less noise.

We need to be very careful about pixel peeping on the computer screen at 100% (in my opinion) - it can really send a false message. I think your 40D is doing a great job!

NeilL

swintonphoto wrote:

Another important difference is the different megapixel count. The pixel sites need to be smaller on the 40D because of the larger megapixel count (one of the reasons I wish the companies kept megapixels below 10). Thus, when you look at images at 100% they may appear to have worse noise in the 40D than the 10D. It is not good to compare noise on cameras with different megapixel counts by viewing at 100%. I don't know if the exact percentages, but it would be better to view the 10D image at 100% and the 40D image at 50% (Or something) - the key is to view the same percentage of the sensors image on the 10D and the 40D because then you will see noise comparisions using the exact same percentage of each sensor.
Am I making sense?

Another thing to consider - In order to actually view a 40D image at 100% in print you are going to have to blow up a REALLY big print. When you do this the viewing distance will be further away - so you will not notice the noise. Rarely are there situations where people will ever put their nose up to a large print to see each pixel. Viewing pixels at 100% on high megapixel count cameras is really not an accurate way of seeing the quality of the images. A better idea would be to take the same image with 10D and a 40D and then print a 16x20 of each - I guarantee the 40D image will look like it has less noise.

Pixel peeping on the computer screen at 100% (in my opinion) can really send a false message.

Thanks. Yes what you say makes perfect sense. I guess our test comparisons are flawed, but I wonder if they are completely worthless? We did indeed look at his images at 200% and still they were cleaner!

Whenever I look at my 40D images at normal viewing size on a monitor I easily see noise. Whenever I look at my friend's images at normal viewing size they are very much cleaner. Furthermore, since I got the 40D I have had the opportunity of seeing images from other 40Ds, as well of course as hundreds of other images from other cameras, online when I don't normally have 100% view. I realise that all these are jpgs which have been post processed to some degree or other, but those that are said by the photographers to be minimally post processed look superior to anything from my 40D with the same minimal post processing. To have to remove visible noise in images at ISO100 at normal monitor viewing size from a 40D is not what I bought the 40D expecting. And to have to remove noise from every photo degrades the image. None of the images from my 40D can avoid this, it seems.

Would a better test be to shoot a gray card, perhaps?

claudermilk

To be bluntly honest, I don't think any of the shots shown can tell us anything. There is no pair of 10D/40D shots that are of the same subject exposed the same. I would suggest finding a static subject with consistent lighting, use a tripod and the exact same lens so the framing will be the same, and finally determine proper exposure, then run the cameras set exactly the same in Manual mode. That will eliminate as many variables as possible. You still have all the differences in the sensors mentiond, but no going around that.

NeilL

claudermilk wrote:

To be bluntly honest, I don't think any of the shots shown can tell us anything. There is no pair of 10D/40D shots that are of the same subject exposed the same. I would suggest finding a static subject with consistent lighting, use a tripod and the exact same lens so the framing will be the same, and finally determine proper exposure, then run the cameras set exactly the same in Manual mode. That will eliminate as many variables as possible. You still have all the differences in the sensors mentiond, but no going around that.

We did something like that for those building pics. We used the Canon 100mm lens on both cameras. The focus was a little astray on my camera, and that anomalous shutter speed snuck in once, only!

Putting comparisons aside, I would really like to know if other 40D users see the same noise as my copy is producing at ISO100 at 100%, like in the lamp post above. I have seen images at ISO100 at 100% where such noise is not apparent.

My friend and I saw that the noise from the 40D seemed to be most excessive in the lower quarter tones.

Here are two more examples, the first is ISO100 at something like ~30%, there is noise in the black leather jacket (note too the ugly patchiness of skin tone gradations on the subject's neck - splotches of light and splotches of dark), the second is a crop at ISO125 at 100% of what were black pants, not rusty as here, where noise is readily apparent.

I am reacting to the difference I see in almost all images I see online which don't have such IQ failures, as well as to my friend's shots on his computer. But if other 40D users can show me this is normal, I'll gladly change my mind!

Neil

pathfinder

NeiL,

I received your PM and have reviewed this thread. I am sorry for your difficulties with your 40D, which I generally think is an excellent camera.

I doubt that I can make reliable estimates of the noise levels with 800 pixel jpgs posted here on this thread in dgrin.

I would suggest you obtain some RAW files and some jpgs from another 40D, that is working reliably and compare them to your files.

A 10D file should not have lower noise than a 40D - I am certain of that. The 10D had fewer larger pixels on the sensor, but none of the newer noise reduction technology that Canon builds into their newer cameras since the 10D. I have numerous images from my own 10D from several years ago - I loved the camera when it was introduced, but it is definitely not of the 40Ds quality level seen today.

I will be busy today, but if you cannot find access to original RAW and jpg 40D files on the WEB, I will send some to you if your email will accept that large a file ( some ISPs balk at 12Mbyte files:D ) Otherwise I can load a couple on my mac drive on line and give you access to them.

Noise is a inherent function of the camera and its sensor, but it is affected to a great deal by how it is used also. A very warm climate will increase noise. A hot camera increases noise dramatically. Keep your camera cool and never leave it in a parked car in the sunlight. Long exposures increase noise. High ISOs, of course, do too, but under exposure increases noise dramatically.

Making sure that you keep your histogram "to the right" ( but not so far to the right that you blow all the highlights ) will help to decrease noise when processing images from RAW to a jpg.

To be honest, I can find noise in the lower quarter tones in images from my 5D or 1DsMkll also, so noise is a fact of life in digital images viewed on screen at 100 or 200%. Almost none of that noise will be visible in a properly executed print. I am not bashful about using anti-noise software if needed in areas of an image where noise is objectionable.

Glenn NK

I don't doubt the comments posted to date - comparing the results from two cameras isn't the simplest exercise.

That being said, I recall reading a post on dpreview by one John Sheehy about the results from a "blackframe noise" test he did on some bodies. I specifically recall wherein he stated that he'd tested a 30D and found it to be out of spec, returned it for another which tested quite well. As soon as I read the thread and his post, I tested my 30D and found it gave the same values as his "good" one. The thread is a long one, but for the technically inclined, quite interesting.

http://forums.dpreview.com/forums/readflat.asp?forum=1019&thread=19721647&page=1

He had posted a chart showing "signal to noise floor" data; his first 30D had lower values (meaning noise was more predominant) than the second 30D. I am not able to post this chart, but I will e-mail it to anyone interested; it's only 43 KB in size.

The point I'm making is that even under the strictly controlled manfacturing processes utilized, there are variations from body to body, and it may be this is what NeiL is encountering.

ziggy53

Neil,

I also received your PM.

I haven't responded in this thread before because on my monitor, there is no excessive noise present in these images. While I suspect it might be a monitor setup issue, possibly an LCD display set too bright, it could also be partly a misinterpretation of the scene you are photographing.

In addition to what Pathfinder said, right on as usual, do watch for how the camera is setup to record JPGs. For testing, use the "Neutral" Picture Style, but by all means stay away from increases in sharpness or contrast.

Like Claudermilk said, if you want to test two cameras, lock down a tripod, mount the two cameras one by one on the tripod with the same lens transferred between the cameras so you capture the same scene. Use manual exposure to eliminate "auto" errors.

If you determine that you need to do a more quantified noise testing, start with the ISO test on this page:

http://www.i3a.org/virtual/eye_on_standards_v1_n5.html

Note that this requires absolute attention to detail and will take considerable time to implement and test.

Finally, do some large print testing on glossy paper, viewed under controlled lighting or diffused sunlight. That should give you fairly accurate qualitative and comparative results between the two cameras as well as a fairly definitive evaluation whether the problem is too great for you to accept. (It will either look good or it will look bad.)

Good luck,

NeilL

Glenn NK wrote:

I don't doubt the comments posted to date - comparing the results from two cameras isn't the simplest exercise.

That being said, I recall reading a post on dpreview by one John Sheehy about the results from a "blackframe noise" test he did on some bodies. I specifically recall wherein he stated that he'd tested a 30D and found it to be out of spec, returned it for another which tested quite well. As soon as I read the thread and his post, I tested my 30D and found it gave the same values as his "good" one. The thread is a long one, but for the technically inclined, quite interesting.

http://forums.dpreview.com/forums/readflat.asp?forum=1019&thread=19721647&page=1

He had posted a chart showing "signal to noise floor" data; his first 30D had lower values (meaning noise was more predominant) than the second 30D. I am not able to post this chart, but I will e-mail it to anyone interested; it's only 43 KB in size.

The point I'm making is that even under the strictly controlled manfacturing processes utilized, there are variations from body to body, and it may be this is what NeiL is encountering.

Good point and interesting ideas. Thanks for the links, and I'll PM you my email address for that chart!

The first copy of a 40D I got would almost never accept any lens I put on it and would not trip the shutter when it did. It performed all of its tricks nicely in the store when I returned it and it was swapped then and there for my present copy, the serial number of which was just one digit different from the bad original! And you know the problems with this one. Maybe a faulty batch. But obviously not all Canon stuff leaves the factory perfect!

kdog

Neil,

Only the last picture appears to exhibit excessive noise on my monitor, and that's the one that was under exposed and boosted in conversion. As previously stated, that introduces a LOT of noise.

Regarding setting up futher comparisons, aside from the other great comments you received, like shooting manually, I might also mention that you really need to use the same s/w to process the images from both cameras. Use the same settings as well. Workflow can make a big difference in the amount of noise in the finished product.

Cheers,
-joel

NeilL

kdog wrote:

Neil,

Only the last picture appears to exhibit excessive noise on my monitor, and that's the one that was under exposed and boosted in conversion. As previously stated, that introduces a LOT of noise.

Regarding setting up futher comparisons, aside from the other great comments you received, like shooting manually, I might also mention that you really need to use the same s/w to process the images from both cameras. Use the same settings as well. Workflow can make a big difference in the amount of noise in the finished product.

Cheers,
-joel

Thanks for your input, Joel, much appreciated. What you say is true, and my friend and I got pretty close to what you and the other posters advise in our test shots of the building at night. We got initial exposure settings with our cameras in P mode, and then duplicated them for A, T & M, for each of ISO100, 200 & 400, using the same 100mm lens. We weren't sure if the problems were general or confined to one or other mode. All were processed and converted to tiff in Aperture, further converted to jpg in PS.

We did this night shoot because a few nights previous he had done some night shots of this building with a 10D and I had remarked to him how at 100% the IQ was remarkably good - no noise. As you can see, the boosted brightness version of the 10D shot of the building has comparatively far less noise than the 40D version, even at 200%.

We were fairly sure we were onto something with our testing and wanted to get other pairs of eyes to check what we were seeing. But the absolute noise level of the 40D shots is also interesting in its own right, and I hope some 40D users will show us here some samples of the beautiful noise-free images at ISO100 at 100% their cameras are giving them!

Of course, some level of noise is inherent in all images from any camera, but what we are talking about here are comparisons at ISO100 at 100%.

Neil

NeilL

pathfinder wrote:

NeiL,

I received your PM and have reviewed this thread. I am sorry for your difficulties with your 40D, which I generally think is an excellent camera.

I doubt that I can make reliable estimates of the noise levels with 800 pixel jpgs posted here on this thread in dgrin.

I would suggest you obtain some RAW files and some jpgs from another 40D, that is working reliably and compare them to your files.

A 10D file should not have lower noise than a 40D - I am certain of that. The 10D had fewer larger pixels on the sensor, but none of the newer noise reduction technology that Canon builds into their newer cameras since the 10D. I have numerous images from my own 10D from several years ago - I loved the camera when it was introduced, but it is definitely not of the 40Ds quality level seen today.

I will be busy today, but if you cannot find access to original RAW and jpg 40D files on the WEB, I will send some to you if your email will accept that large a file ( some ISPs balk at 12Mbyte files:D ) Otherwise I can load a couple on my mac drive on line and give you access to them.

Noise is a inherent function of the camera and its sensor, but it is affected to a great deal by how it is used also. A very warm climate will increase noise. A hot camera increases noise dramatically. Keep your camera cool and never leave it in a parked car in the sunlight. Long exposures increase noise. High ISOs, of course, do too, but under exposure increases noise dramatically.

Making sure that you keep your histogram "to the right" ( but not so far to the right that you blow all the highlights ) will help to decrease noise when processing images from RAW to a jpg.

To be honest, I can find noise in the lower quarter tones in images from my 5D or 1DsMkll also, so noise is a fact of life in digital images viewed on screen at 100 or 200%. Almost none of that noise will be visible in a properly executed print. I am not bashful about using anti-noise software if needed in areas of an image where noise is objectionable.

Thanks for this, pathfinder. I have PMd you, and will again to give you an email address to try for that large file.

Once again, thanks! Very helpful.

Neil

NeilL

ziggy53 wrote:

Neil,

I also received your PM.

I haven't responded in this thread before because on my monitor, there is no excessive noise present in these images. While I suspect it might be a monitor setup issue, possibly an LCD display set too bright, it could also be partly a misinterpretation of the scene you are photographing.

In addition to what Pathfinder said, right on as usual, do watch for how the camera is setup to record JPGs. For testing, use the "Neutral" Picture Style, but by all means stay away from increases in sharpness or contrast.

Like Claudermilk said, if you want to test two cameras, lock down a tripod, mount the two cameras one by one on the tripod with the same lens transferred between the cameras so you capture the same scene. Use manual exposure to eliminate "auto" errors.

If you determine that you need to do a more quantified noise testing, start with the ISO test on this page:

http://www.i3a.org/virtual/eye_on_standards_v1_n5.html

Note that this requires absolute attention to detail and will take considerable time to implement and test.

Finally, do some large print testing on glossy paper, viewed under controlled lighting or diffused sunlight. That should give you fairly accurate qualitative and comparative results between the two cameras as well as a fairly definitive evaluation whether the problem is too great for you to accept. (It will either look good or it will look bad.)

Good luck,

Very much appreciate this, ziggy, especially the link to ISO testing which I'll follow up asap!

Regards.
Neil

Mr. Quiet

NeilL wrote:

From the time I first got the 40D I thought I was seeing way too much noise in images shot in very normal conditions. So I got together recently with a friend who has a 10D to shoot some comparison shots. Here are a couple where the difference in noise is easily seen. The shots of the lights and pole are not totally identical, but perhaps close enough to allow for a comparison of noise to be done. ISO of both was 100. The ones of the building were set up to be identical, ISO 100, 1/6 sec, f2.8. Identical brightness adjustment was made to both to uncover noise. All were converted from RAW to tiff, and brightness adjustment of building shots made, in Aperture (his Mac), and from tiff to jpg in CS3 (my Win). All are at 100%.

Do you think the much greater noise in the 40D images is normal? Would very much appreciate your comments!

Thanks.
Neil

Hi,

What mode did you shoot these picturs in

Lw

NeilL

Mr. Quiet wrote:

Hi,

What mode did you shoot these picturs in

Lw

Thanks for looking in, Lw!

On the 40D, the night building pic was shot in Standard, the lamp posts in landscape, the people pics in portrait (I think!)

Neil

claudermilk

If I may gently suggest, forget the P&S modes and learn to use the Av, Tv, and M mode preferably, or at least P. Be in control of your image.

NeilL

claudermilk wrote:

If I may gently suggest, forget the P&S modes and learn to use the Av, Tv, and M mode preferably, or at least P. Be in control of your image.

I chose the P mode on the night test shots because I wanted the camera to expose as it saw fit, and then I used its settings to standardise A & T while ISO was being changed. My friend did the same with his camera. The relationship between exposure, ISO and noise might not be straightforward in digital, so right now I'm not sure if that method was appropriate.

However, I have found that in good photographing conditions (maybe other times, too!), the camera produces perfect histograms - most times better than me! It's not really surprising, is it - why shouldn't a camera know what it's doing? It doesn't expose to the right as much as me, though ; )

Thanks for your comments, claudermilk.

Neil

NeilL

I have been looking through all my shots with the 40D, and I discovered that the raw files are all good, noise is negligible up to 400 ISO. The noise is happening at conversion and PP, and to the shadows in underexposed shots.

I would not have known how to make this assessment I now make without this discussion, and the upgrading of my knowledge here and in other places I was obliged by my need for an explanation to visit.

So, thanks to you all for helping me get to the point where I could know what was going on to produce the noise I was so objecting to. Very generous of you all!

Neil

Cameron

Did you apply any significant exposure adjustment in raw processing in the images you posted?

NeilL wrote:

I have been looking through all my shots with the 40D, and I discovered that the raw files are all good, noise is negligible up to 400 ISO. The noise is happening at conversion and PP, and to the shadows in underexposed shots.

I would not have known how to make this assessment I now make without this discussion, and the upgrading of my knowledge here and in other places I was obliged by my need for an explanation to visit.

So, thanks to you all for helping me get to the point where I could know what was going on to produce the noise I was so objecting to. Very generous of you all!

Neil

NeilL

CSwinton wrote:

Did you apply any significant exposure adjustment in raw processing in the images you posted?

Hi CSwinton,

yes, and to those images which were underexposed, of course, aggravating the noise already there!!

Now I understand the circumstances that produce excess noise, I am going further and looking closely at the noise results of conversion using Digital Photo Professional, Bibble and DxO, and processing in LR, PS and Corel PSP. Do you have a preferences?

Neil

Cameron

I think Digital Photo Professional does a fantastic job, but the noise tends to be a bit more exaggerated at high ISO than with LR or PS (using Adobe Camera Raw) - this is mostly due to weaker noise reduction and stronger default sharpening I think. The newer version of DPP has more control over noise reduction than previous versions. Personally, I like my workflow with lightroom so I tend to stick with that.

You can get great results from all of the programs you listed. Knowing HOW to use the software is more important than which one you choose - pick one you like and learn it well.

NeilL wrote:

Hi CSwinton,

yes, and to those images which were underexposed, of course, aggravating the noise already there!!

Now I understand the circumstances that produce excess noise, I am going further and looking closely at the noise results of conversion using Digital Photo Professional, Bibble and DxO, and processing in LR, PS and Corel PSP. Do you have a preferences?

Neil