Quantitative image noise level measurements: Is the 50D really this bad?

[stephenB]

I've recently acquired a 50D as a replacement for my 40D, as part of a complicated insurance claim/saga (another story!)

I have now made some initial quantitative measurements of the noise levels on raw images from this camera, and compared them with the values from the 40D, and my older 350D.

I am dismayed by the results for the 50D - the noise appears to be almost exactly one stop worse than for my 350D. This means that the noise on the 50D at ISO 400 is very similar to that of the 350D at ISO 800, and similarly for the 50D at ISO 800 the noise is the same as on the 350D at ISO 1600! Not good!

Here are some details to substantiate these findings - see especially the attached graph.

To make these measurements, it is important to note that I used images in raw format from all 3 cameras, and for the 50D I disabled all the settings which might affect the basic output from the sensor. So I made sure the following were off:

* high ISO speed noise reduction
* highlight tone priority
* auto lighting optimiser

(i.e. CF II functions 2, 3 & 4). This is effectively the same as for the 40D and 350D, and so the raw images should be basically what is coming off the sensor, with minimal processing. In other words a like for like comparison between the cameras.

The graph shows noise levels measured for all 3 cameras as a function of image grey level (brightness). All show a characteristic decrease in noise for higher image brightness, and all show the expected increases as the ISO setting goes up.

Note how the lines for the 50D agree very closely with the 350D but one ISO setting higher. It is also interesting to see that the 40D out performs the 350D by about 1/2 stop in these terms, which makes the comparison between the 40D and 50D even worse!

To make these measurements I took slightly out of focus pics of a grey scale test chart I had printed out for the purpose (see pic attached), and then used some free software to measure the standard deviation of the image values in small (30 x 30 pixel) regions for each step in the grey scale.

I'm wondering if there could be something wrong with my 50D, but I suspect not.

I note that the DP review site shows a 50D raw image which is visually substantially noisier than a 40D image. It also has the interesting comment that "Considering the 50D's much more tightly packed sensor (4.5 MP/cm² vs 3.1 MP/cm² on the 40D) this comes hardly as a surprise. It would have been unreasonable to expect Canon's engineers to overcome the laws of physics."

Indeed! In fact, the higher noise levels on the 50D compared with the 350D are almost exactly what would be expected given the factor of 2 difference in the number of pixels on the sensor [theoretically, if the noise is dominated by photon counting statistics, then halving the area of the pixel will increase the noise by sqrt(2), which is exactly the same factor that increasing the ISO setting by one stop produces].

But many other web sites appear to suggest that the 50D is no noisier than the 40D, or is only slightly worse. But these are generally based on qualitative comparisons of general images - useless for assessing noise levels properly in my opinion!

I suspect the key to all this may be switching off the "High ISO noise reduction", which is suspiciously set on "Standard" by default. This is likely to low pass filter the image, and reduce noise at the expense of a loss of sharpness. So have Canon set it up like this to try to mask the basically poor noise performance of the 50D sensor, by any chance?

I'd be interested in any comments from others on this. Do other more experienced users of 50D raw images suffer from high noise levels?

[Chrysophylax]

50D Noise level on 800iso. Took a picture of a Bullfinch in the garden yesterday evening. The light was poor so I cranked the iso up to 800 and stopped down by one and a third to get a 125 shutter speed. I was in RAW/Large Jpeg so could compare the 2. The results were terrible and the noise very bad indeed. I was using my Bigma from only 6 feet away with zoom around 300/350 which would have been f/5.6ish. Will post a picture later if I remember. Neil.

[Keith Reeder]

It's certainly noisier than the 40D in my experience.

I should add that my "experience" relates to having downloaded or been sent a number of 50D RAWs, which I've converted using what I know to be an effective workflow for noise management (I have no real concerns about 3200 ISO on my 40D).

And in every file I converted (about a dozen in all) - and doing my best to process them well - they were noisier than I'd expect from a 40D in the same circumstances.

The choice of conversion software is critical here though: I will not use ACR/Lr to convert 40D files unless they're at a very low ISO (and given that my "baseline" 40D ISO is 400, the opportunity rarely occurs) and use Capture One 4 or Raw Therapee instead - Cap One 4 all the way for high ISOs.

Even so, the 50D produces noisier images than I would want. There's not always much in it, but the difference is unquestionably there.

That's precisely why I don't - and won't - own a 50D.

[pduxon]

It's certainly noisier than the 40D in my experience.

I should add that my "experience" relates to having downloaded or been sent a number of 50D RAWs, which I've converted using what I know to be an effective workflow for noise management (I have no real concerns about 3200 ISO on my 40D).

And in every file I converted (about a dozen in all) - and doing my best to process them well - they were noisier than I'd expect from a 40D in the same circumstances.

The choice of conversion software is critical here though: I will not use ACR/Lr to convert 40D files unless they're at a very low ISO (and given that my "baseline" 40D ISO is 400, the opportunity rarely occurs) and use Capture One 4 or Raw Therapee instead - Cap One 4 all the way for high ISOs.

Even so, the 50D produces noisier images than I would want. There's not always much in it, but the difference is unquestionably there.

That's precisely why I don't - and won't - own a 50D.

Keith

I read people on the Nikon forum wanting MORE MP. What you are saying is you want better MP? correct? which makes sense to me.

[tdodd]

Here's a shot with my 50D, at 1600 ISO. First the whole image without edits, then a crop, exposure boosted by 0.7 stops in Lightroom and a little NR and sharpening. Aperture was f/5.6 and shutter speed was 1/250. Taking the exposure boost into consideration, all in all that means the light levels for this were 6.3 stops dimmer than bright sunshine. Considering this was an opportunistic shot, shot at 1600 ISO and underexposed, I am not unhappy with the results.

EDIT : For kicks I've thrown in three more images, this time at 800 ISO, two with no edits and the last with minor tweaks but no sharpening or NR except LR defaults.

I totally agree, when comparing images at the pixel level the 50D is noisier than the 40D. That should surprise nobody. But if you compare equivalent areas of an image, magnified by equal amounts - same physical size in mm/cm - then the differences should be small, with the 50D having the edge for detail - if you use sharp glass, focus accurately and avoid shake/blur.

For anyone unhappy with the output from their 50D I would be interested in seeing their original raw file to see how bad the problem really is. Maybe I can offer some advice. There's no rocket science here, and apologies to those of you for whom this is teaching you to suck eggs, but for starters :-
- Do not use HTP;
- Do not underexpose and then push the exposure in post. In preference, Expose to the Right (http://www.luminous-landscape.com/tutorials/expose-right.shtml) and then pull back the levels/curves in post.
- Do not expect small pixels to have noise characteristics as good as larger pixels. Do not pixel peep at all. Look at pictures, not pixels. People do things today, with digital, that they would not dream of doing with film. Who, seriously, would blow up a 35mm neg from an 800 ISO film to a 6' wide print and then examine it from 12-18" away? That's the sort of magnification you are effectively applying when you view a 50D file at 100% with a typical monitor. On my 17" 1920x1200 monitor, viewing a 50D file at 33% yields an image that is physically exactly 12x8". Viewing at 50% yields an image too large to fit the screen, but it would be ~18x12" if it did fit, which I think is plenty large enough for viewing from 12-18" away. Viewing at larger magnifications than that just seems a little bit unhelpful.
- The higher the ISO the less you should expect to make aggressive crops successfully. If you are using high ISO then you need to fill the frame more. A 100% crop at 100 ISO may very well work nicely. At 1600 ISO it almost certainly will not.
- At higher ISOs expect to use noise reduction. Normally I scrape by well enough with Lightroom, often on default settings, but if I need "proper" NR I use Neat Image.
- If you are sharpening then make sure you are not sharpening noise as well as details. Use a mask to avoid sharpening the noisiest, least detailed areas.
- Sometimes there just isn't enough light to make a good image. Be realistic in your expectations. Sometimes you need to know when not to bother releasing the shutter. The same argument applies for a subject that is too small in the frame. Combining a small subject with poor light = disappointing results.
- For those of you who simply can't help comparing 50D pixels with 40D pixels, may I suggest you turn the tables and shoot the 50D in SRAW1 and then compare pixels with pixels, or pictures with pictures.

[tdodd]

Here is another example - the same subject shot with my 50D at 400 ISO and at 1600 ISO. No edits, no NR other than Lightroom defaults. I've then cropped in to 2400x1600 and resized that to give an 800x533 image as a 33% crop. This simulates a reasonable degree of cropping for a bird that is small in the frame. The background looks a little noisier in the 1600 ISO crop, which NR could fix, but otherwise I think you'd be hard pushed to be unhappy with the 1600 ISO example. EXIF is in the files.

[tdodd]

Last example - this is a 50% crop at 400 ISO, 1600 ISO and 1600 ISO with NR using Neat Image. EXIF is in the first two images. Neat Image has stripped EXIF from the third image file.

Also a sunflower at 1600 ISO, in its own shade but a little bit backlit on the top left. No edits other than Neat Image NR. First the full frame and then a 43% crop. It's not a good photo but I can live with the noise.

[stephenB]

Last example - this is a 50% crop at 400 ISO, 1600 ISO and 1600 ISO with NR using Neat Image. EXIF is in the first two images. Neat Image has stripped EXIF from the third image file.

Also a sunflower at 1600 ISO, in its own shade but a little bit backlit on the top left. No edits other than Neat Image NR. First the full frame and then a 43% crop. It's not a good photo but I can live with the noise.

I am sure that with appropriate post processing using NeatImage noisy images can be made to look quite acceptable. That wasnt quite the point of my original post! That was to try to make a like for like comparison, without any noise reduction processing, of the sensor outputs from the 350D, 40D and 50D. I was expecting the 50D to be somewhat poorer than the others, due its smaller pixels, but was suprised by how poorly it came out in the comparison.

Given there must have been various improvements since the days of the 350D, including the claimed larger micro-lenses in the 50D, I was suprised to find no evidence for them in the comparative figures.

In fact the difference in noise levels between the 350D and 50D is probably even worse than 1 stop, because according to the DXOmark website the 50D is setup so that ISO 400 isnt actually ISO 400, but is notably less - much more so than for the 350D. So, for the same amount of light coming in, the noise would be even higher on the 50D!

I should maybe add I used PSE 6 and ACR to do my raw to tiff conversion, taking care to turn off anything which might affect noise levels (like sharpening). It might be interesting to try DPP and compare results. I'll post the result if I get round to it.

Having said all that, I'm quite surprised how noise free many of Tim's pics are, especially those at ISO 1600. But then is he using "High ISO speed Noise reduction" on the camera, and if so at what level? Also if NeatImage has been applied, then this can easily reduce noise levels by large amounts depending entirely on the "Noise reduction amount".

[tdodd]

Stephen, I have openly declared where I have used NR and it certainly isn't in all the images. The swans have no NR (excpt LR defaults). Ditto the butterfly, the blue ball thingy and two of the three images of the pylon.

I do not use in camera NR settings and, since I shoot raw and process in Lightroom, any such setting would be ignored.

If you want to see comparative examples from a 1D3, 40D and 50D on a like for like basis - identical crop (in terms of subject/composition, not pixels) resized to identical output image size - then please see this online album....

http://picasaweb.google.com/EezyTiger/ResolutionTestLR23CropToSubject# (Processed in LR 2.3)
http://picasaweb.google.com/EezyTiger/ResolutionTestCropToSubject# (Processed in DPP)

The images in this album were shot precisely to help me determine which camera produced the best results, from a tight crop, at various ISOs, specifically for bird photography where the subject is small in the frame.

The original, uncropped images are in an album here....

http://picasaweb.google.com/EezyTiger/ResolutionTestFullFrames#

In my opinion, when comparing images from these cameras on an equal footing the 50D is in no way the inferior performer. Noise is close enough to make no odds and the 50D image has superior detail, at least to my eyes.

If you have a raw image or two that you are disappointed with then please make it available and I will happily take a look and draw any conclusions I can. If you can host it yourself then all well and good. Otherwise you can use the free service from http://www.yousendit.com/ to "email" it to yourself and then post the link to the file in this thread or in a PM.

[stephenB]

Stephen, I have openly declared where I have used NR and it certainly isn't in all the images. The swans have no NR (excpt LR defaults). Ditto the butterfly, the blue ball thingy and two of the three images of the pylon.

I do not use in camera NR settings and, since I shoot raw and process in Lightroom, any such setting would be ignored.



Yes Tim they look pretty good and I'd agree there isnt a big difference between the 40D and 50D in noise terms on your images.

Just to be sure, can you confirm you have set CF II -2 to a value of 3 (disable) on the 50D? I dont know if this is ignored in ACR or not when using raw. The value of zero is the default which is described as "standard" in the manual, i.e. some noise suppression is presumably applied at all ISO speeds.

[tdodd]

Yes. Please see attached screen print.

I've used DPP simply to show the settings. The pictures I posted earlier were processed in Lightroom.

[stephenB]

Yes. Please see attached screen print.

I've used DPP simply to show the settings. The pictures I posted earlier were processed in Lightroom.

Thats pretty conclusive! I wonder if the difference between your images and my graphs can be due to your showing pictures at the same size, not on a constant pixel scale? If there are more pixels on the 50D images than for the 40D, then maybe there could be a certain amount of averaging going on either in the eye when viewing them, or in whatever has been used to scale them to the same size. This averaging will reduce apparent noise levels.

My noise analysis has of course been applied on 1:1 pixel level on all images. Also I dont have Lightroom, I just use PSE 6 and ACR. But surely that wouldnt make much difference?

[tdodd]

I strongly disapprove of comparing output at pixel level. It proves nothing except what we already know - small pixels collect fewer photons and therefore obviously have a worse SNR. What matters to me is which camera produces the better image, not the better pixel. Look at the end game - we want to produce pictures, so let's see which camera produces the best picture. Leave the pixel comparisons to the pixel peeping number crunchers and lab rats. I'll judge results by the images, thanks.

That is exactly why my album makes comparisons of equivalent surface areas of the sensor - i.e. crops to the same composition - not equivalent numbers of pixels. That is, in my opinion, the only meaningful way to compare. The question should be.....

Does a 50D print at 12x8 look better or worse than a 12x8 from a 40D?
Does a 50D print at 18x12 look better or worse than an 18x12 from a 40D?
Does a 50D print at 30x20 look better or worse than a 30x20 from a 40D?

The question should not be....
Does a 50D print look as good as a 40D print when the 50D image is enlarged by a factor 1.22X greater? That is simply a daft comparison to make, completely unfair, or unequal if you prefer. The 50D sensor may have 50% more pixels than the 40D but it doesn't have 50% more surface area and therefore is not likely to be able to capture 50% more light. To stretch its image over an area 50% greater area than that from a 40D and expect equal levels of noise and sharpness is just silly. Sadly, that is exactly what people do when they compare 100% crops, or 50% crops etc..

The same, or similar, comparisons can be made for display on a screen, but they need to be based on equivalent levels of physical linear dimensions, not equivalent numbers of pixels. e.g. a fair comparison would be to make a crop from a 50D of 1,222x815 pixels and a matching crop from the 40D (by area) of 1,000x667 pixels. Then, so both suffer equally from the artefacts of resizing, resize both to, for argument's sake, 800x533 pixels. Then you will have a comparison based on equal terms. In truth those are pretty high viewing magnifications, so maybe it would be better, and more realistic, to double the dimensions of the crops and then resample those crops down to 800x533.

Lightroom and ACR should be the same, but you do need to use the latest ACR release. Adobe does have a history of screwing up on the first release or two of ACR for each new camera release. I've witnessed the problem myself with the 40D and 50D and read about similar issues when the 5D2 was launched. I don't know whether the ACR that supports PSE6 is maintained in sync with the latest releases for PSE7 and CS4. I think perhaps it isn't. Quite what DPReview was doing when they first reviewed the 50D with a beta release of ACR, God only knows. I doubt things were improved much when they retested with the first "live" release either.

Disclaimer : Of course, the above rationalisation is just what I think. Others are free to make whatever comparisons they choose to, but I, for one, will stick to looking at image quality, not pixel quality.

[stephenB]

Tim

You make some interesting points about like for like areas, but with more pixels on a 50D there is a natural temptation to hope that you can push more, i.e. crop more severly to get a better result in the all too frequent case when the bird is further away than you would like. Otherwise, what is the point of having more pixels?

Returning to the raw conversion theme, I've just tried comparing the noise on an image converted with DPP with the previous ones which used PSE6 and the latest version of ACR (5.4).

To my considerable surprise the DPP noise levels are noticeably lower than with ACR - roughly half a stop over most of the grey level range! The difference was even more marked in the dark image areas (<50). See attached revised graphs

Any ideas anyone why?

[stephenB]

Blimey! I seem to have hit a few nerves here!

This isnt quite what I intended when I posted these quantitative results.

I would however observe that I'm not in full agreement with Tim's views on numbers of pixels and constant areas. Logically, if you always examine images on like for for like area basis, doesnt it just depend on the area you decide you want to make the comparison on?

In the past in the days of 2Mp or 3Mp pixel cameras, you would have probably decided that the area was a lot smaller than the one you would now choose, because cameras now have more pixels in them.

With more pixels, the useful print area should go up, but there is a limit beyond which increasing noise makes it pointless.

The question is then at what point is there any justification in creating sensors with ever higher numbers of pixels?

There are quite a few suggestions I have seen elsewhere to suggest that with the 50D Canon have taken a step too far, and a better compromise between noise and resolution is around 10-12 Mp for a x1.6 crop sensor. But maybe they have decided that more pixels = better sales! However, I cant speak from any experience yet as I've hardly used the 50D "in anger" yet.

[njlarsen]

Tim,
I have not followed photography theory closely lately, but when I did, almost all lenses performed better at f8 than at f5.6. Isn't it a problem for a camera not to be able to realize that advantage because of diffraction softness? It sounds to me like there is NO advantage of the added pixels in 50D over 40D (but as you say, there may be other advantages usage-wise).

Niels (I don't even use a dSLR, so going at this from a more theoretical standpoint).

[stephenB]

I had no idea that this post would stir up quite such strong emotions!

All I was trying to point out was that my quantitative and fully objective measurements appeared to show that the noise levels on the 50D were surprisingly high (to me), being somewhat poorer than what you would expect on the basis of photon counting statistics if you took the 350D sensor and doubled the number of pixels.

The finding that DPP processing of raw files seems to reduce the noise on the 50D compared with PSE6 and ACR was also surprising to me, and suggests that maybe this should be used by others concerned with comparatively high noise on the 50D. But the improvement was not that large!

Finally, the 50D does have one key feature that the 40D lacks, and that is microAF adjustment. This was the main point of switching to it, due to marked back focus errors on my 40D + 400mmf4 DO when used with a x1.4TC (but not without it). But that is another story, possibly for another thread in due course!

[tdodd]

Sorry, Stephen, I think I got the wrong end of the stick in this thread. I'm more interested in how my images look than how my pixels perform in lab measurements. Quantitative and objective measurements have their place, but where the 50D is concerned I have moved past that point. I have removed the posts of mine that I am still able to, in order to reduce clutter in the thread.

I wish you luck with your 50D.

Regards,
Tim.

[stephenB]

Tim

It seems a shame to remove all your interesting contributions Tim. Nothing wrong in my view to having several contrasting views expressed in one thread. Isnt a heathly debate what these forums are all about?

[hollis_f]

and stopped down by one and a third to get a 125 shutter speed.

And then raised the exposure in PP? Thats a sure-fire, guaranteed, method of getting noisy results.

[stephenB]

Got almost my first field shots with the 50D today at Farmoor - my local reservoir, despite the dismal conditions. There was a distant Y L Gull, and an underexposed ISO400 raw pic provides an interesting example of both my themes so far in this thread:
(a) 50D images can be noisy!
(b) DPP appears significantly better than ACR in the darker areas

Have a look at the attached. This is a 1:1 blow ups of both pics (sorry Tim but this bird was distant, and I wasnt going to get any closer without a boat or swimming!).

The top one was converted with ACR, and the bottom one with DPP. There was no noise reduction, either in the camera or in both conversions to tiff. The chroma noise on the bird's back is noticeably worse on the ACR image.

Having said this, some better exposed shots of the same bird are much less noisy.

Looks like I will need to get used to the DPP user interface which I find far less friendly than ACR.

[Nikon Kid]

Stephen, keep at the DPP, once you get used to it, I think its a great raw converter made for canon so cant be bad. I like alot about it, all my gallery images have been through DPP and then AE6....

[tdodd]

Well, Stephen, your results do not come as a surprise, but I think they beg a number of questions to be asked....

- Is the 50D quite simply a poor performer when it comes to noise?
- Is ACR less than satisfactory, at times, when it comes to processing 50D raw files?
- Is it reasonable to expect clean files at raised ISOs when underexposing the image?
- Is it reasonable to expect clean files at raised ISOs, when underexposing the image and pixel peeping at 100%?
- Is it reasonable to expect clean files at raised ISOs, when underexposing the image, pixel peeping at 100%, and using poorly optimised conversion software?
- Is it reasonable to expect clean files at raised ISOs, when underexposing the image, pixel peeping at 100%, using poorly optimised conversion software and not throwing in some additional NR?

I'm not in any way having a dig, but I'm sure I've said it before, elsewhere if not in this thread - people need to have realistic expectations. If you push the boundaries on all fronts then something will burst.

From my own experience I have learned not to waste my time shooting things that are too small in the frame, especially with the 50D, at raised ISOs, if I am then going to need to use the file at 100% magnification to get a large enough subject. If that is my intention then I only have myself to blame for poor results, not the camera. Sometimes it is simply not worth taking the shot. When in doubt, I like to remind myself of the boundaries, or expectations we had with 35mm film. Compared to a typical print from a 35mm neg the magnification factors involved when viewing a 50D file at 100% are simply vast. Expect to see flaws unless you take great care with your capture and have favourable shooting conditions.

That said, I strive in my photography not to underexpose, ever, and often aim for that perfect ETTR exposure if I can. I use manual exposure almost always so that I can take charge of what, if anything will be clipped and what will be touching the right hand edge of the histogram. I usually do pretty well. I find that when I play my part right then I am more often than not able to get results from Lightroom (ACR if you like) that satisfy me, occasionally with a little help from Neat Image. When I can't get the results I need I tend to find room to blame myself before the equipment.

Remember, with the 50D you get the options of a very useable SRAW1 and SRAW2, the first of which gives you a resolution almost equal to a 30D (and by implication almost the same pixel density as a 5D2), and the second of which still gives you more pixels than SRAW from a 40D. If you can frame your subject large enough, and noise is a bother, then try shooting in one of the SRAW formats. It should save you some processing time later.

[hollis_f]

- Is it reasonable to expect clean files at raised ISOs when underexposing the image?

No. Indeed you're better off shooting at ISO 1600 than shooting ISO 800 with the same settings and boosting the exposure in post-processing.

Here's (http://photography-on-the.net/forum/showthread.php?t=730218&highlight=iso) a thread from another forum that explains. But it ain't for the faint-hearted.

[stephenB]

I wouldn't call ISO 400 raised! In fact with the 350D and 40D it was my standard setting given the usually less than brilliant light levels in the UK. With the 40D and even the 350D, ISO 800 often gave quite reasonable results.

Given these results, I'm wary about using the 50D at ISO 800, but will have to see how I get on.

Applying NeatImage to the Y L Gull pictures of course cleaned up both images, but it did a good deal better job on the DPP one. See attached. Again the DPP one is at the bottom.

I'd also agree that aiming for higher exposures is a good idea, but in the field often things dont turn out quite as planned. In this case for example, by the time I had upped the exposure, the bird had moved so it was more head on. It then flew off!

[postcardcv]

I find your results quite interesting, I've been reading a fair few reviews of the 50D as I'm thinking about getting one... while most reviews aren't raving about high ISO performance they seem to suggest that noise is on a level with the 40D. Your sample shots suggest it's a good deal worth. Here's one of the reviews I was looking at - http://www.trustedreviews.com/digital-cameras/review/2009/01/31/Canon-EOS-50D/p6

[tdodd]

Tim,
I have not followed photography theory closely lately, but when I did, almost all lenses performed better at f8 than at f5.6. Isn't it a problem for a camera not to be able to realize that advantage because of diffraction softness? It sounds to me like there is NO advantage of the added pixels in 50D over 40D (but as you say, there may be other advantages usage-wise).

Niels (I don't even use a dSLR, so going at this from a more theoretical standpoint).
Niels, Take a look here - http://www.bobatkins.com/photography/digital/canon_eos_50D_resolution_confusion.html

and here - http://www.luminous-landscape.com/forum/lofiversion/index.php/t30752.html

[stephenB]

Yes there does seem to be quite a lot of contradictory stuff on this subject on the web. The respected(?) DP review has a page which supports my contention that the raw images from the 50D are substantially noisier than from the 40D when analysed and viewed 1:1 - see http://www.dpreview.com/reviews/canoneos50d/page18.asp

I suspect some of these differences between reviews are to do with whether or not High ISO noise reduction is on or off, and possibly differences between jpeg and raw images, but I dont know for sure.

Also I'd say that just looking for noise in general images with lots of detail in them can be very subjective. Nothing beats looking at flat areas of grey, and doing some quantitative analysis!

[tdodd]

Here's a shot taken today at 400 ISO and processed in Lightroom. Everything on default settings except I adjusted the sharpening settings to Masking = 50, which reduced the sharpening effects on smoother areas. This is the same shot at different crops and viewing sizes of approx 17%, 25%, 33%, 50% and exactly 100%.

EXIF is 400mm, 1/640, f/7.1, 400 ISO and it looks like I am about 1/3 stop short of the right hand side of the histogram, so not a perfect ETTR exposure, but not bad either. At that exposre setting it appears the lighting is 2 1/3 stops dimmer than a bright, sunny day. I really do not see a problem with noise in this example, even at 100% viewing.

Now, it's not quite sharp, but at 1/640, handheld, and magnified to this level I am not really surprised.

[tdodd]

Here's another example from today - this time different shots of the same subject in the same lighting but 100% crops at 800 ISO, 1600 ISO and 3200 ISO. This was a full ETTR exposure with a few white feathers clipped and I fixed that in Lightroom. Otherwise no other edits apart from Masking = 50, once again.

The fourth attachment is the 3200 ISO example as a full frame image and the fifth is a 100% crop of the 3200 ISO frame but with Neat Image to tidy it up. EXIF is in all the images.

Looking at other 3200 ISO shots I took today, I would say that noise is a non-issue, or at least one that can be easily dealt with, at viewing magnifications of 33%, which is large enough to almost fill my 17" 1920x1200 monitor.

[tdodd]

Now, back in the world where people do actually fiil the frame, making full use of the sensor they paid for, and where they don't depend on 100% crops to satisfy their cravings, here is a non birding image shot at 800 ISO, no edits at all, plus the obligatory 100% crop. EXIF is 400mm, 1/640, f/7.1, 800 ISO.

EDIT : I've added another image from my 50D, this time shot at 1600 ISO and processed through DPP with no substantive edits - only to change picture style and increase sharpening from 0 to 3. First the full image and then a 100% crop. EXIF is 70mm, 1/100, f/2.8, 1600 ISO, flash at +1/3 EC.

As far as noise is concerned, where is the problem? I don't see one.

[njlarsen]

Niels, Take a look here - http://www.bobatkins.com/photography/digital/canon_eos_50D_resolution_confusion.html

and here - http://www.luminous-landscape.com/forum/lofiversion/index.php/t30752.html

thanks Tim :t:

Niels

[Chrysophylax]

And then raised the exposure in PP? Thats a sure-fire, guaranteed, method of getting noisy results.

I am no expert, far from it. Just an average birder with a camera. Any tips or criticisms are most welcome, it can only advance me to a better level of understanding my gear.

[Nikon Kid]

I think all I can say, go out and enjoy your camera which ever one you own, because just round the corner there's a new one coming.

BTW if you don't like the image you have taken, bin it and take another thats the advantage of digital, when I go out I take about 100 on a normal 2/3 hours walk, if I am lucky I will get half dozen worth the website gallery's

[tdodd]

I am no expert, far from it. Just an average birder with a camera. Any tips or criticisms are most welcome, it can only advance me to a better level of understanding my gear.
I had my eyes opened significantly by this thread on POTN - http://photography-on-the.net/forum/showthread.php?t=705691. This is where I came to understand that long shots and heavy cropping were no way to secure beautiful images. Of course, I knew that heavy cropping, especially with the 50D, was a bit futile, but I guess we all have hopes and expectations until reality catches up with us.

I do have to say that birding seems to be one of the more demanding photographic disciplines because of so many factors that challenge us. It's the only photography I do where cropping seems, unfortunately, to be almost a necessity. With that in mind, I don't always find it reasonable to blame the equipment for poor results. In fact, the more experienced I become (a long way to go yet) the more I realise that it is my failings or unrealistic expectations that are to blame. You need to make your own "luck", so....

- Choose your time of day wisely so that you have cool, undisturbed air and light that comes in lower and lights the side/underside of the bird rather than the top only.

- Get much closer to your subject. This is not just a question of longer glass. You need to physically close the distance to your subject. Getting closer will fill the frame more, get more pixels on your subject, gather more light, record more detail, look sharper and less noisey. You will also avoid issues with dirty/hazy air and heat haze. If you are physically closer you can shoot with shorter glass, which means you can use a lower shutter speed, which means you can use a lower ISO and/or stop down a little to sharpen that lens. As a technique it will also cost you far less in glass and support systems and your back will thank you too.

- Position yourself so that the light is in your favour, somewhere behind you, usually, and not so that you are shooting into a backlit scene.

- Expose correctly, or even ETTR. Do not underexpose, especially if you are at higher ISOs already. I've been shooting in manual mode for around a year now. It makes results so much more consistent and controllable. It's got to the point that I almost cannot face shooting in one of the autoexposure modes. That's not willy waving or anything. I simply find manual exposure easier. The light often does not vary much, from moment to moment, and once I have a good exposure dialed in that may well do me perfectly for several minutes, if not hours. If the light is changing then, quite honestly, I am likely to prefer a shot lit by sunshine than cloud, so I can simply choose not to release the shutter if the sun disappears for a moment.

To sum up - shoot in good light; fill the frame (as best you can) and nail your exposure. If you slip up on any or all of those things then your IQ will go downhill.

Here are a few of my recent shots with my 50D and 100-400. All were shot at 400 ISO

[JohnZ]

I read the POTN article with enormous interest Tim. I also noticed that Brad did not have a 100-400 zoom lens on his gear list, although he did say that he took the Duck shot with the zoom. I think he said he did it to prove to the owner that it was a sharp lens ?
May I ask which section of POTN you posted this article in ? I also noticed that you were not happy with some helicopter shots ? Would they be in the Transportation section.

[tdodd]

I'm afraid the helicopter shots were in an online album of mine which have fallen foul of a massive purge/cleanup. The original raw files are long gone. Suffice to say that sharpness was lacking and, if memory serves, the only rational explanation I could come up with was that atmospheric effects were interfering with IQ. I have to say that good IQ over long distance does seem far more elusive than with closer photography. Apart from the cropping aspect, which is a separate issue, I really believe that warm air over long distances can be a real barrier to image clarity. I do not have such problems when shooting the moon in the cool of night, so I don't think it is a lens issue, just atmospheric "wobble". I've lost far too many birding shots to miserable IQ when shooting at 50m-100m+ on "lovely sunny days", expecting a crop to save me. Of course, when shooting big things with limited fine detail, like race cars, the consequences are worlds apart from the same effect applied to fine feather detail on tiny creatures in the distance.

I've found a couple of examples that sort of illustrate the point. These are of aircraft, the first of which is a long way away and you can see that the image lacks punch/clarity. Viewing closer, at 100% everything looks awful. Sure, noise is there, but the more disturbing thing is the distortion. If you look at the tyres they do not appear round. If you look at the tail fin the red stripes are very fuzzy. In the second example, shot 6 weeks later, so I don't know how the heat/haze compare, the plane is a lot closer and these effects seem absent. EXIF is 1/1600, f/8, 800 ISO for the first and 1/1600, f/7.1, 400 ISO for the second, so the first shot has a disadvantage there, but one that affects the shape of things? Edits are minor and identical (picture style and sharpening).

The "article" - it's really just a discussion thread, just like this one - is in the birding sub-forum. See fifth attachment below.

[hollis_f]

Tim,
I have not followed photography theory closely lately, but when I did, almost all lenses performed better at f8 than at f5.6. Isn't it a problem for a camera not to be able to realize that advantage because of diffraction softness.

This is yet another thing that 'everybody knows' about the 50D. The truth is that diffraction effects are nowhere near is bad as the FUD merchants would have you believe.

Here's a PotN posting I made with images shot at various apertures. Far from diffraction effects showing up at f8, it's only when you get to f16 that I can see anything.

[njlarsen]

I'm afraid the helicopter shots were in an online album of mine which have fallen foul of a massive purge/cleanup. The original raw files are long gone. Suffice to say that sharpness was lacking and, if memory serves, the only rational explanation I could come up with was that atmospheric effects were interfering with IQ. I have to say that good IQ over long distance does seem far more elusive than with closer photography. Apart from the cropping aspect, which is a separate issue, I really believe that warm air over long distances can be a real barrier to image clarity. I do not have such problems when shooting the moon in the cool of night, so I don't think it is a lens issue, just atmospheric "wobble". I've lost far too many birding shots to miserable IQ when shooting at 50m-100m+ on "lovely sunny days", expecting a crop to save me. Of course, when shooting big things with limited fine detail, like race cars, the consequences are worlds apart from the same effect applied to fine feather detail on tiny creatures in the distance.

I've found a couple of examples that sort of illustrate the point. These are of aircraft, the first of which is a long way away and you can see that the image lacks punch/clarity. Viewing closer, at 100% everything looks awful. Sure, noise is there, but the more disturbing thing is the distortion. If you look at the tyres they do not appear round. If you look at the tail fin the red stripes are very fuzzy. In the second example, shot 6 weeks later, so I don't know how the heat/haze compare, the plane is a lot closer and these effects seem absent. EXIF is 1/1600, f/8, 800 ISO for the first and 1/1600, f/7.1, 400 ISO for the second, so the first shot has a disadvantage there, but one that affects the shape of things? Edits are minor and identical (picture style and sharpening).

The "article" - it's really just a discussion thread, just like this one - is in the birding sub-forum. See fifth attachment below.

The conditions you describe here is an example where non-mirror cameras have an advantage. With such a camera you can fire series of shots (5-10 in a row) and hope for the one where the air aligns perfectly and gives a sharper image than the rest. I have seen the effects of that using digiscoping with a P&S camera. In a SLR, I have been told that mirror slap creates wobble that to some extent negates the advantage of shooting in series?

cheers
Niels

[tdodd]

The conditions you describe here is an example where non-mirror cameras have an advantage. With such a camera you can fire series of shots (5-10 in a row) and hope for the one where the air aligns perfectly and gives a sharper image than the rest. I have seen the effects of that using digiscoping with a P&S camera. In a SLR, I have been told that mirror slap creates wobble that to some extent negates the advantage of shooting in series?

cheers
Niels
I like your thinking, Niels, and you have a point, but I've even struggled with longer range IQ when using a tripod and Live View (no mirror slap). Despite perfect focus, using Live AF (contrast detect) on a stationary bird, there are times (hot days?!?) when the IQ just will not come when shooting at longer distances. Indeed, even looking through the viewfinder, sometimes things just don't look right.

For anyone who didn't follow the link to the POTN thread, you might like to see my video example of how much the air really does wobble, even in warm conditions (20 centigrade), never mind hot - http://www.youtube.com/user/eezytiger#play/all/uploads-all/1/MsOSBy8P7eY. The wobble is clearly evident, and while it may not matter much on a featureless TV aerial, it will kill fine feather detail dead. I would guess the distance to be ~130m.

[Chrysophylax]

I had my eyes opened significantly by this thread on POTN - http://photography-on-the.net/forum/showthread.php?t=705691. This is where I came to understand that long shots and heavy cropping were no way to secure beautiful images. Of course, I knew that heavy cropping, especially with the 50D, was a bit futile, but I guess we all have hopes and expectations until reality catches up with us.

I do have to say that birding seems to be one of the more demanding photographic disciplines because of so many factors that challenge us. It's the only photography I do where cropping seems, unfortunately, to be almost a necessity. With that in mind, I don't always find it reasonable to blame the equipment for poor results. In fact, the more experienced I become (a long way to go yet) the more I realise that it is my failings or unrealistic expectations that are to blame. You need to make your own "luck", so....

- Choose your time of day wisely so that you have cool, undisturbed air and light that comes in lower and lights the side/underside of the bird rather than the top only.

- Get much closer to your subject. This is not just a question of longer glass. You need to physically close the distance to your subject. Getting closer will fill the frame more, get more pixels on your subject, gather more light, record more detail, look sharper and less noisey. You will also avoid issues with dirty/hazy air and heat haze. If you are physically closer you can shoot with shorter glass, which means you can use a lower shutter speed, which means you can use a lower ISO and/or stop down a little to sharpen that lens. As a technique it will also cost you far less in glass and support systems and your back will thank you too.

- Position yourself so that the light is in your favour, somewhere behind you, usually, and not so that you are shooting into a backlit scene.

- Expose correctly, or even ETTR. Do not underexpose, especially if you are at higher ISOs already. I've been shooting in manual mode for around a year now. It makes results so much more consistent and controllable. It's got to the point that I almost cannot face shooting in one of the autoexposure modes. That's not willy waving or anything. I simply find manual exposure easier. The light often does not vary much, from moment to moment, and once I have a good exposure dialed in that may well do me perfectly for several minutes, if not hours. If the light is changing then, quite honestly, I am likely to prefer a shot lit by sunshine than cloud, so I can simply choose not to release the shutter if the sun disappears for a moment.

To sum up - shoot in good light; fill the frame (as best you can) and nail your exposure. If you slip up on any or all of those things then your IQ will go downhill.

Here are a few of my recent shots with my 50D and 100-400. All were shot at 400 ISO

Thanks Tim. I know my gear is up to it, it's just my lack of understanding that needs to be worked on. I must say though, I have been looking at some pictures taken with my 40D 400mmf/5.6 and I was more happy with those than I am now with my 50D and sigma 50-500. (dont say it, I should have kept the 400mm Canon lens) Better glass is needed so I have put the sigma on ebay. Thanks again. Neil.

[postcardcv]

here's another review page with some nice ISO test shots - http://www.steves-digicams.com/2008_reviews/canon_eos_50d_samples.html to my eyes the images upto ISO400 are all spot on at ISO800 noise creeps in but I doubt it would need any NR for prints at ISO1600 noise is visable but it is not degrading the detail in the image... even ISO3200 looks quite usable so long as you're not printing too big.

[stephenB]

Just to let readers of this thread know that I have written a fuller account of my measurements of noise level on raw images from the 350D, 40D and 50D, together with the methods used. These were the original subject of this thread and I so far have not seen anything in this thread or elsewhere that contradicts them. This account can be found at http://www.stephenburch.com/noise/noise.htm

[postcardcv]

Just to let readers of this thread know that I have written a fuller account of my measurements of noise level on raw images from the 350D, 40D and 50D, together with the methods used. These were the original subject of this thread and I so far have not seen anything in this thread or elsewhere that contradicts them. This account can be found at http://www.stephenburch.com/noise/noise.htm

Well it's an interesting read, though I must confess that I don't really understand the method you've used well enough to make real sense of the results. It's a bit too forensic for me, I want to see the sample images as they'd make more sense to me.

[RJM]

What I find more interesting is that Canon's sensor performance seems frozen in time when comparing these three camera's on DxOMark (http://www.dxomark.com/index.php/eng/Image-Quality-Database/Compare-cameras/(appareil1)/267%7C0/(appareil2)/180%7C0/(appareil3)/183%7C0/(onglet)/0/(brand)/Canon/(brand2)/Canon/(brand3)/Canon). On the otherhand, when the comparable Nikon cameras released nearly at the same time (http://www.dxomark.com/index.php/eng/Image-Quality-Database/Compare-cameras/(appareil1)/294%7C0/(appareil2)/295%7C0/(appareil3)/270%7C0/(onglet)/0/(brand)/Nikon/(brand2)/Nikon/(brand3)/Nikon) are examined, steady and significant improvements in the sensor performance have been made.

cheers,
Rick

[hollis_f]

Well it's an interesting read, though I must confess that I don't really understand the method you've used well enough to make real sense of the results. It's a bit too forensic for me, I want to see the sample images as they'd make more sense to me.

Yup, because all this test does is compare the per-pixel noise. Which is really important if you happen to view all of your images at 100%, or print them out at A0. For you, me and, I imagine, the vast majority of user it's the per-image noise that's really important.

[stephenB]

I'm afraid I disagree! Its the per pixel noise that determines the inherent quality of the image, not some measure that happens to depend on the magnification you want to use on any particular image. The magnification is highly variable from one image to the next and depends on how big the subject is in the field of view, how large you want to view or print the picture etc etc. We arent all fortunate or clever enough to have birds filling the frame all the time...

[stephenB]

By popular request, I have also added to my website page (http://www.stephenburch.com/noise/noise.htm) the Yellow-legged gull example, to illustrate that the effects seen on the graphs can also be quite significant on real photos.

[hollis_f]

By popular request, I have also added to my website page (http://www.stephenburch.com/noise/noise.htm) the Yellow-legged gull example, to illustrate that the effects seen on the graphs can also be quite significant on real photos.

Described as....

a somewhat underexposed pic of a Yellow-legged Gull

If you want to accentuate noise then underexposing the shot and then boosting the exposure in PP is the ideal way to do it.

Which version of ACR was used? Your web site only says - "I initially used Adobe Camera Raw, as downloaded for running in PhotoShop Elements versions 3 to 6".

[stephenB]

If you want to accentuate noise then underexposing the shot and then boosting the exposure in PP is the ideal way to do it.

Which version of ACR was used? Your web site only says - "I initially used Adobe Camera Raw, as downloaded for running in PhotoShop Elements versions 3 to 6".

I didnt intentionally underexpose! The first few shots just came out like that. But surely it is interesting to see if something reasonable can be recovered from them?

For the 50D shots, I used the very latest version of ACR V5.4.