PLease help Canon 50D

[Tell Dickinson]

Just a quick tale about looking at images at 100%....

I was talking to a well known (in the UK) professional photographer and he said something like

"of course you need a faster shutter speed to get a sharp image with digital than you ever did with film"

I said "hey you will have to explain that one to me :) "

he then went on to say

"everyone looks for sharpness at 100% with digital, but you just couldn't look at 100% with negatives/slides, you could look with a loupe but that didn't equate to 100%"

so what he was saying is that perhaps we are a bit obsessed with sharpness with digital and there must be an awful lot of great published film/slide images out there that are not as sharp as a lot of our digital images because we now have the tools to be much more critical than we had before.

Just thought his comments were interesting :)

Tell

[Tim Dodd]

An interesting read - http://www.northlight-images.co.uk/a...al_limits.html

[Jay Gould]

An interesting read - http://www.northlight-images.co.uk/a...al_limits.html

First, thanks Tim!

I am no scientist; I had to read the article twice as I think the author raises some interesting points, and appears to support what Tim was saying in earlier posts.

In order to learn I have never been afraid to stick my foot in the goo; therefore, what I "got" from the article [I am really looking forward to the more senior photographers reactions if they take the time to read the article] is as follows:

1. 5D2 - full use of the sensor is f/10 and anything smaller reduces the use of the full sensor; 50D - limit f/stop to f/8 or larger.

2. DOF has been substantially reduced from the way you used to think of it's limits.

3. The old rule of 50mm = 1/60 SS; new rule for tack sharp is 50mm = 1/125 or better 1/250.

Perhaps, just perhaps, one of the reasons that there are so many comments about the image looking soft is that we are simply not shooting fast enough!

Teach me!! :D

Cheers, Jay

[Tim Dodd]

The following is my opinion. Some of it may be fact, some conjecture, some plain wrong, but it is how I see things. Most (all?) of it is a rehash of stuff I wrote earlier, but it's an attempt to address some of the points Jay has highlighted from the article I linked to, and then to go on a bit. Also, sorry if any statements I make are too literal. This is a brain dump, not a thesis....

All the old rules of photography were developed based on the amount of enlargement from the plate/film/sensor to the final displayed image. In the days of film, when there was no such thing as a pixel, this was all pretty straightforward. If you had a 1.5"x1" frame of film (35mm if you like) then to get a 12x8 print out of that you were magnifying by a factor of 8X. You would probably be viewing the image at a "crooked" arm's length of 12"-18". Rules (guidelines) for shutter speed and DOF were based on those reasonable and relatively standard physical constraints/assumptions.

It was recognised that with the advent of digital, and cropped sensors, the rules needed changing. Certainly the 1/focal length rule for minimum shutter speed needed updating to take into account the crop factor. That is entirely reasonable since to get from a cropped image to a 12x8 print you need a greater degree of enlargement - 13X for a Canon cropped sensor instead of the 8X for a "full frame" sensor/negative. The greater degree of enlargement also magnifies the effects of shake/blur more. Even though this was now within the digital age the matter of pixels did not come into it at all. It was all about enlargement factors from the physical size of the captured image on film/sensor to the end product.

Much the same adjustments were made with respect to DOF calculations, all based on film/sensor size and absolutely bugger all to do with pixels. The adjustment calculation is just as straightforward. For any given subject distance, your focal length must be adjusted by the crop factor in order to maintain the same field of view. Assuming Canon cameras with a 1.6X crop factor, if you used a 50mm lens on the cropper you'd need to use an 80mm lens on the full frame body to match the composition. If you plug numbers into a DOF calculator - I favour the one here - http://www.dofmaster.com/dofjs.html - you can compare the difference in DOF. e.g.....

50D + 50mm lens at f/5 and 10m subject distance give DOF from 7.24m to 16.2m
5D2 + 80mm lens at f/5 and 10m subject distance gives DOF from 8.1m to 13.1m

In order to get the same DOF from the 5D2 and its longer lens we need to stop down more. How much more? 1.6X more. i.e. you'd need f/8 on the 5D2 to match the DOF from the cropper. Let's check it out....

5D2 + 80mm lens at f/8 and 10m subject distance gives DOF from 7.29m to 15.9m

It's not a perfect match but it's pretty close. Certainly, to the nearest 1/3 stop, it is the aperture that matches the DOF figures most closely.

So, we seem to have rules about shutter speed that are directly related to crop factors. We have rules about composition that are directly related to crop factors. We have rules about DOF and aperture that are directly related to crop factors.

Nowhere yet has anybody brought a discussion about pixels to the table. As far as the standard rules of photography are concerned, pixels are neither here nor there. Everything is about magnification factors, whether the magnifying strength of the lens to the enlargement factor to get you from a negative/sensor to a print/VDU image.

If we consider diffraction, it causes light to disperse, thus losing focus, as it passes through the aperture of the lens. The smaller the aperture the greater the diffraction. With a low pixel density you can stop down quite a way before the resolving power of the sensor is sufficent to pick up the diffraction softening. As pixel densities go up, so does resolving power. The higher density sensor can detect/display the effects of diffraction sooner than the low density sensor. There comes a point where, if you stop down too far you destroy the benefit of the higher resolving power of the sensor. The overall image will look no worse at all, from the point of view of diffraction, but individual pixels will start to show the softening. This is why a high density sensor like that of the 50D has a lower diffraction limited threshold than the 5D2, which only has a pixel density equal to the 30D.

The problem that digital has brought about is the wretched pixels, or more specifically, people's obsession with them. The more of them there are the more absurd the enlargement factors are when viewed at 100%. We aren't dealing with 12x8 printed images from 35mm film frames any more. We're dealing with "virtual" images that are potentially enlarged to 3X, 4X or even 5X that size when displayed on a monitor at 100%. All the old rules have been thrown out the window. Whether you consider camera shake, subject blur, misfocus, poor lens IQ, diffraction or anything else, we are now putting the images under a microscope, relatively speaking. This is no 5X or 10X loupe placed over a negative. This is a 50X microscope displaying on a hulking great screen. We're talking about sensor pixels, each only a few microns wide, being displayed at a size visible to the naked eye.

If you want to view things at 5X the "old school" enlargement factor then you will need to improve everything about your photography by 5X, or make 5X the allowance for IQ issues. So, yes, diffraction becomes a real world constraint when you look at the pixels; yes, DOF does not extend anywhere near as far when you look at the pixels; yes, the image will look soft, when you look at the pixels, if there is any flaw in lens, focus or technique.

If you want to record fine image detail you need lots and lots and lots of pixels. Is 10MP enough? 15MP? 21MP? Who can say? It depends on what you want to do with them. It depends how detailed your subject is. It depends how large you want your end product to be. It depends on the acuity of human vision. It depends on how far back from the image you view it. All things being equal, more pixels is not necessarily a bad thing, but it may be way more than you need. Ah, yes, but more pixels is a bad thing if they are too small, I hear you say. They are noisey. Yes, true, so stop staring at the individual pixels. The whole is greater than the sum of the parts. Look at the overall image, not the individual pixels. Look at the fine detail within the image. See how it is not blocky and pixelated. See how it is smooth and detailed. Enjoy. If you have too many pixels then merge them. The noise will diminish and the image detail will still match or exceed what you need for your level of enlargement, viewing distance and acuity. If you simply crop, and crop, and crop some more then you are throwing away the IQ that the whole sensor is able to deliver.

If you really want to benefit from the detail that many pixels bring, and you want each and every pixel to be clean of noise, how do you do that? Well, (a) you do have lots of pixels; (b) you make each pixel large. What are the consequences? You need a larger sensor. Kerching! Ding! Ding! Ding! Ding! Jackpot! Big sensors give higher image quality, so long as you have the glass to fill them with your subject. That's what I've been trying to say all along. Of course, bigger glass costs more money. Nobody said great IQ came cheap.

If you chose to start with the image from a cropped sensor camera, like the 50D, and then crop from that image to, say, an 800x533 image to be viewed/used at 100% then please appreciate that you will only be making use of 1/36 of the total sensor area - just 2.8%. What was I saying earlier about larger sensors yielding greater IQ? And here you are starting with a small sensor and that cropping from it to virtually nothing. Is it any wonder it may not thrill you with the IQ?

Another interesting read - http://www.ophrysphotography.co.uk/p...ameandcrop.htm

Let's view all this from a slighty different angle. Hopefully most people accept that to achieve a high quality (pro grade) print you need to print at around 300ppi. I'm not sure anyone has considered what sort of quality (noise mainly) those pixels should have, but let's take the recommendation at face value. A 50D produces images of 4752x3168 pixels. Dividing those dimensions by 300 should give us the potential print size of 16"x10.5", give or take, to achieve "pro" quality. Now, this may be rough and ready, and not directly comparable, but if you view a full frame, uncropped 50D image on your monitor at a virtual/physical size of more than 16"x11" or so then what are you proving? What you are proving is that if you stretch the pixels too thinly the image won't look "pro" quality.

My monitor is a 17" 1920x1200 display. The visible area is 14.5"x9". If I view an uncropped 50D image on my screen at a size much larger than "fit to screen" then I can expect to be disappointed. Taking the vertical pixel density I have 1200 pixels in 9" of display, giving me 1200/9 = 133 pixels per inch. Even if I view the image at 50% magnification that is effectively the equivalent of viewing at 266 pixels per inch - close to pro print quality but a little under. The point is, if I view one of my 50D files at enlargements above 50% then what am I proving to myself? More than anything it proves that I have expectations that far exceed what is reasonable, based on everything that has gone before in photographic history, or that I really don't understand practical realities very well. If I can get a pin sharp image from my 50D (and zoom lens), viewed at 100%, then I should be ecstatic. However, I would certainly be asking a lot to expect it, especially from every shot I take.

Maybe people should stop thinking so much about pixels, and 100% viewing, or 50%, or any other %, and we should actually re-evaluate the benefit of defining limits based on physical image magnification/enlargement factors once more. How much simpler and more honest and accurate it would be if we agreed that images from a modern digital sensor could be enlarged by a factor of "X" for viewing at a distance of "Y". How much closer to reality would that be? Significantly, I think. Can/should a 50D image be enlarged more than a 40D image? Is it right and reasonable to suppose it can be? No, I'm not sure that is a fair extrapolation at all. Did increasing the pixel count by 50% suddenly capture 50% more photons too? Have we really got more image data to spread around, or have we just carved it up into smaller chunks? Make it bigger, make it worse. That's the likely truth, once you go over certain thresholds, such as 300ppi.

I hope I've covered the points Jay raised, directly or indirectly. Just one more comment related to the article in the link. I think what they are getting at is that in the digital generation, where people do crop, heavily, and do stare at pixels all day long, the rules of the game have changed. The old rules no longer apply. We need new/adjusted rules to satisfy the croppers and pixel peepers.

I'll get my coat.

[Jay Gould]

Thanks Tim, I have found this discussion fascinating and enlightening.

[Flavio Rose]

Very interesting post by Tim, thanks.

If you plug numbers into a DOF calculator - I favour the one here - http://www.dofmaster.com/dofjs.html - you can compare the difference in DOF.

DOF depends on the assumed circle of confusion. The calculator you point to assumes 19 um (um = micrometers) by default. Our dSLR pixel pitches lie between about 5 um and 8 um. Seems to me that if something is blurred to three pixel pitches and you look at it 100%, it's not going to look sharp, i.e. within the DOF.

To compute DOF in digital I think you need to figure out the scale at which you intend to view your image, be it 100%, 50%, or whatever. If you are looking at 100%, so that circle of confusion = pixel pitch, then DOFs are depressingly shallow (and hyperfocal distances are on the order of a kilometer away).

if I view one of my 50D files at enlargements above 50% then what am I proving to myself? More than anything it proves that I have expectations that far exceed what is reasonable

Problem is I think most of us obtain images that are sharp at 100% routinely in static subject photography, at least with 10 and 12 megapixel cameras. The technology has spoiled us.

[Tim Dodd]

Cheers, Flavio. The DOFMASTER website is referenced in a Canon article about DOF - http://cpn.canon-europe.com/content/...d.do#container. The article also rationalises the choice of a CoC of 0.019mm for APS-C cameras, based on the enlargement to a 7x5 (gasp) print, although I did not notice a specific viewing distance stated for that degree of enlargement, just a generalisation that larger prints are normally viewed from larger distances.

[Kaustubh Deshpande]

What a discussion. fun. I havent gone through every line....so probably what I am saying has already been covered.

Kobus....raw file from 50d is 4752 X 3168. Take that file and resize it to 3888 X 2592...thats the resolution of the 40d raw file. Shoot the same photo with 40d....all parameters same. Now compare the two. Look at 100%. Does the 50D one look noisier? that should settle the dispute, right?

Tim....may be you can post 100% crops with the above exercise.

[Tim Dodd]

Kaustubh, I've taken your idea on board but I have modified it a little. I agree with the principle of comparing like with like but it does not seem correct that the 40D is cropped to its native resolution (100% crop) while the 50D file is subjected to interpolation to downsize it.

Using DPP, I cropped my 40D raw files to 800x533 and my 50D files to 978x671. That is the equivalent ratio to 3888 vs 4752. I was about to output the 50D files at 800x533 to equal the crop size from the 40D but that's when I had my change of heart. Thus both crops have been downsized a little, to make the playing field more equitable. The 800x533 crops from the 40D and the 978x651 crops from the 50D have all been downsized to 700x466. That is equal to an 88% crop for the 40D and a 72% crop for the 50D, giving no unfair advantage to either camera.

The full album of comparable results from ISOs 100-3200 is here - http://picasaweb.google.co.uk/EezyTi...velComparison#.

Here are the two examples at 800 ISO



The conditions were not sufficiently controlled to announce a convincing winner but from the results here the 50D appears to have a clear edge over the 40D for sharpness and detail. The differences in noise (if any) are too close for me to call, but if forced to pick a winner I'd give my vote to the 50D on that too.

[Kaustubh Deshpande]

Thx Tim. But I wud still like to see comparison done with equal sized images. I guess what I want to see is that if noise is more "apparent" in 50d, it is just because the files are larger (i.e. have more pixels). Thats just a hunch I have...never used 50d and dont have the money to upgrade from 40D.

I know you dont agree, but can u do it for the heck of it :-) been a little less busy at work today :-)

[Tim Dodd]

Thx Tim. But I wud still like to see comparison done with equal sized images.

The images are equal sized - same total surface area of sensor used within the crop, same composition within the crop. Do they not look the same size to you? Apart from my tweak to ensure that both cameras "suffer" equally from downsizing, rather than just one of them, my comparison is mathematically exactly as you proposed and provides a level playing field for the comparison.

I do not understand why you are unhappy with my approach?

If you prefer I can make a couple of raw files available - one each from the 40D and 50D and then you can compare them however you wish. Just pick an ISO and let me know which pair you would like.

[Jay Gould]

Cheers, Flavio. The DOFMASTER website is referenced in a Canon article about DOF - http://cpn.canon-europe.com/content/...d.do#container. The article also rationalises the choice of a CoC of 0.019mm for APS-C cameras, based on the enlargement to a 7x5 (gasp) print, although I did not notice a specific viewing distance stated for that degree of enlargement, just a generalisation that larger prints are normally viewed from larger distances.

Tim, you have such a wonderful nonscientific way of explaining things that I have reposted this quote in a new thread in the General Photography Discussion Forum entitled Focusing-CoC-Hyperfocal Distance. Looking forward to, hopefully, your comments and contribution. :D

[Kaustubh Deshpande]

Tim, ur methd is fine. sorry. I didn't realise that u finally resized to make them same size. I had checked ur post in my email and had not seen the actual images.

[Tim Dodd]

50D a sleeper camera - according to Peter Gregg

[Keith Reeder]

Kobus....raw file from 50d is 4752 X 3168. Take that file and resize it to 3888 X 2592...thats the resolution of the 40d raw file. Shoot the same photo with 40d....all parameters same. Now compare the two. Look at 100%. Does the 50D one look noisier? that should settle the dispute, right?

No.

If you do this, what's the point of the extra pixels in the 50D?

This is exactly the point Arash is making when he says that noise must be compared at a pixel-for-pixel basis: anything else is obfuscation.

FWIW I agree with those that argue the 50D is noisier than the 40D in the Real World, away from contrived test scenarios: I've seen it with my own eyes and (having very recently been able to borrow a 50D for a while to confirm my earlier conclusions) there's no question that - other things being equal - you get a noisier image from the 50D than from the 40D.

[Flavio Rose]

If you do this [compare pictures at the same size], what's the point of the extra pixels in the 50D?

Seems to me the purpose of the comparison should guide how one carries it out -- pixel comparison or same-size comparison. From what I've read (I'm still thinking of buying the 50D), the extra pixels in the 50D give you more croppability at low ISO but you pay for it in noise at higher ISO.

[Kaustubh Deshpande]

No.

If you do this, what's the point of the extra pixels in the 50D?

This is exactly the point Arash is making when he says that noise must be compared at a pixel-for-pixel basis: anything else is obfuscation.

FWIW I agree with those that argue the 50D is noisier than the 40D in the Real World, away from contrived test scenarios: I've seen it with my own eyes and (having very recently been able to borrow a 50D for a while to confirm my earlier conclusions) there's no question that - other things being equal - you get a noisier image from the 50D than from the 40D.

I am not saying that there is no point of extra pixels. By reading the thread, I just thought more noise is apparent because the picture dimensions are larger and one is viewing from the same distance. BTW I use 40d and have never tried 50d. And I agree that pixel size is more important than the sensor size when it comes to noise...as Arash says. Smaller pixels = more noise is what I have always heard. But the pixels also are improving all the time.