DSLR sensor sizes and Image Quality 40D vs 5D

pathfinder

I recently was PM'd by LarryL, asking me about the 5D versus the 40D with regard to large prints, as follows - (As mods, we prefer to answer these sorts of questions in the forum where the answers can be shared and commented on by all our readers.............)

"

LarryL wrote:

It's great that I found someone who shoots with both the 5D and 40D. I currently shoot with a 20D and the 17-40L, 24-70L and 70-200L lenses. I was thinking of upgrading to the 5D in hopes that when printing and cropping a photo I would be able to continue to generate large prints (nothing over 20 x 30) with minimal noise. Additionally, I was also hoping for color improvement. If you don't mind, what are your thoughts?

Thanks,

Larry

I think the answers to these question are rather complex to answer simply. I looked through my files to verify the image sizes in pixels of several of Canon 's DSLRs that I have owned.

5D 2912 x 4386 = 12.772 Megapixels


20D 2336 x 3505 = 8.18 Megapixels


40D 2592 x 3888 = 10.1 MegaPixels

1DsMkll 3328 x 4992= 16.61 MegaPixels

A "20 x 30" inch print at 200 pixel per inch ( conservative rendering of image necessary for 20 x 30 in print at a 3 foot viewing distance ) = 4000 x 6000 pixels

By these standards, even the 1DsMkll might be thought to be challenged as it is only 4992 pixels on its long dimension. But I have several prints from SmugMug at "20 x 30" inches shot with a 1DsMkll that are all but grain-less at a distance of 1 foot. They look marvelous.

I believe that either a 5d or a 40D will also give good images if used with care and skill, and technical perfection. I think they may give good 20 x 30 inch prints, but neither will allow significant cropping with its loss of pixels first.

The 17-40 L, the 24-70L, and the 70-200L are all first rate glass and capable of images at this level of magnification, I am certain, if used with technical skill.

I am not certain that I believe the 40D or the 5D will automatically increase the quality of the color over that captured by the 20D. I base this statement on my simple experience, not on any study of the technical differences in the cameras. Good color depends so much on exposure and lighting that I think they are much more important than the technical differences between a 40D and a 5D, or a 20D. Uprezzing one step in Adobe RAW 4.3 converts an 8 MPixel 20D image to the size of a 40Ds.....

The 5D will have lower noise in the shadows than the 20D, that I agree with. Not that much difference between the 5D and the 40D in my opinion - I am not pixel peeping at 300% here, just thinking of prints I have from both cameras. Noise in images from a 40D or a 5D usually responds very nicely to one of the noise reduction programs like NeatImage or NoiseWear.

I do not think of different cameras as better or worse than another, so much as better for a specific task by virtue of file size, camera weight, depth of field ( more for APS, less for full frame ), ability to stretch a telephoto, or widen a short focal length, water resistance, speed of AF, etc. The 40D is lighter and faster to AF than the 5D. The 5D reacts better to a wide angle lens and may be better for landscapes, but is not as good for birds as the 40D. The 5D files will tolerate cropping a little better perhaps.

In Florida recently at MI V, I shot with a 5D, a 20D and a 40D, and a G9. I posted a number of shots from the G9 of birds recently. I have good images from each camera. All will make excellent 10 x 15 images I am sure. 20 x 30, the 5D is probably a better choice and a little more forgiving in framing than the others. But an excellent image from a well exposed shot in a 40D will surpass a poorly exposed OOF image from a 5D or a 1DsMklll also.

Not sure if this has really answered the posters questions completely. I await the folks who will disagree with my thoughts here with aplomb:thumb

My short answer is that both the 40D and the 5D offer significant advantages over the 20D, but all three are capable of quality images if used within their appropriate boundaries. The 40D has a higher frame rate and faster AF. The 5D has larger files sizes and slightly lower noise than the 40D. But in good light, the 20D works nicely also, I still use mine almost interchangeably. I prefer to leave a camera body on a lens and not swop lenses that often unless required.

Tee Why

I'd have to concur,
I've owned the 300D, 350D, 400D, 30D, 1DMIIN, and have used the 5D at times (the 1DMIIN and the 5D have the same processor-digic II, and the same pixel density) and found subtle difference among the bodies. When I was shooting the 1DMIIN with the XT (which has the same IQ as the 20/30D) I did find some small difference in noise with the 1D being less noisy at higher ISO's but nothing that will make or break an image.

I think careful exposure is much more important than some small differences in noise level under sometimes very artificial testing conditions. I will defer to your evaluation of prints as I don't print that large, but unfortunately, I think too much is being made of minute differences of images viewed at 100% crop.

To me, most if not all current and last generation dslr's if used with good glass and most importantly good skills will produce outstanding prints.

Composition, timing, lighting, and exposure is what makes or breaks a photo in my view, not some small difference in resolution or noise level.

I think with the advent of the internet tests and forums and ability to pixel peep things at 100% crop, things may have gotten out of hand in terms of testing and looking at the specs.

ziggy53

I, of course, agree with both Pathfinder and Tee Why in that subject matter, composition, lighting, lens and camera, as well as processing and post work, all factor into the qualities of the final image and how well any image can be enlarged.

I've seen 35mm negatives that blew into beautiful large prints, and I've seen 4"x5" negatives that could not.

I used to have an image of a dandelion "clock", or seed head, that I shot at 640 x 480 pixels with a 3-chip video camera that made a very nice 8"x10" print. In that case there was something in the subject that translated easily from the small to the large, but that was very unique.

Pay attention to the craft and the art of photography and you will be rewarded, at least until the "next big thing" comes along to give us pause.

LiquidAir

The resolution difference between the 5D and the 40D is nearly irrelevant. 11MP to 13MP is not a big enough difference to make a real difference. That said, there are other differences in image quality which are created by the difference in sensor size. The difference I notice most between the cameras is due to the magnification of the lens image. Here's the issue:

Take the same shot on a 40D and a 5D and print it 12x18 (inches) which, from both cameras, is about 240 dpi (240 dpi is about where pixel level details in the original file become noticable in print). The 5D lens image was 24mm x 36mm and is being magnified about 13x to make that print. The 40D lens image was 15mm x 22.5mm and is being magnified by about 20x to create the same print. If the lens image is sharp enough to outresolve the 40D sensor, then this difference in magnification is irrelevant (except as a difference in DoF). However, when the lens image is not sharp enough to outresolve the 40D sensor, then the lower magnification of the 5D will result in a sharper print.

My general advice is this:

If you are currently using a crop body and you often find that lens sharpness (rather than sensor resoution) is often limiting your print size, a full frame sensor (e.g. 5D or D3) will be a noticable upgrade.

If you like to use shallow depth of field as a compositional tool, a full frame sensor is a considerably more powerful tool in this regard.

If neither of these apply to you, then you are unlikely to see the benefits of a full frame sensor. In fact, the shallower DoF of a full frame sensor can make some styles of photography considerably more difficult, so unless you are prepared to take than on, I generally recommend against full frame.

pathfinder

Excellent points, Ken.

The original poster said he owned a 17-40L, a 24-70L and a 70-200L, so I felt glass, if used with good technique, was not likely to be the limiting factor.

Poor technique, too slow a shutter speed, no tripod more likely to contribute to poorer images than these lenses, I'll bet.

I frequently choose a smaller sensored camera for the increased DOF it gives me for a given aperture, including P&Ss for macro for precisely this reason.
Full frame cameras, and their inherent shallower DOF can be a bane or a blessing depending on what your shot needs.

It all comes down to understanding your tools and using them to their advantage for the task at hand.

LiquidAir

pathfinder wrote:

A "20 x 30" inch print at 200 pixel per inch ( conservative rendering of image necessary for 20 x 30 in print at a 3 foot viewing distance ) = 4000 x 6000 pixels

I believe that either a 5d or a 40D will also give good images if used with care and skill, and technical perfection. I think they may give good 20 x 30 inch prints, but neither will allow significant cropping with its loss of pixels first.

I have one 20x30 print from a 5D hanging on my wall which is sharp enough to look good at 3 feet (if I get closer, I can see some softness in it). It was shot on a tripod with the 200/2.8L and MLU. Even then, I took 4 shots in the field and picked the sharpest one for print. The had a very strong green channel so the demosaicing (Bayer interpolation) gave a better than average result. When shooting with a 5D I never promise larger than 16x24 (180 dpi), but the exceptional image can go larger.

claudermilk

All good info. I agree with all of pathfinder's post.

One thing I'd liketo touch on is the "better color" thing. I've seen that pop up a few times. To be honest, the color you get is more a function of the computer 6" behind the viewfinder than the one inside the camera body. You need to make sure the camera is set correctly. I shoot a 20D and push it to it's limits most of the time; one thing I don't have too much trouble with is color, now that I've learned how to set the camera properly.

I have shot a 1Ds Mk II side-by-side with my 20D and a buddy shot his 1D Mk III at the same event. Color was not the big difference between the bodies, that was the most constant variable as we both set all the bodies the same (theater lighting, all bodies set to 3200K)--what was a bigger difference was the noise levels, and overall sharpness; obviously my little 20D was hard-pressed there but put up a spirited fight. I very seriously doubt the 40D or 5D would show much difference in color at the same venue with the same setting. Oh to be fair the lenses in use were a 70-200/2.8L on the Mk II, a 135/2L & 85/1.2L on the Mk III and a 24-70/2.8L on the 20D, so glass was not an issue.

pathfinder

LiquidAir wrote:

I have one 20x30 print from a 5D hanging on my wall which is sharp enough to look good at 3 feet (if I get closer, I can see some softness in it). It was shot on a tripod with the 200/2.8L and MLU. Even then, I took 4 shots in the field and picked the sharpest one for print. The had a very strong green channel so the demosaicing (Bayer interpolation) gave a better than average result. When shooting with a 5D I never promise larger than 16x24 (180 dpi), but the exceptional image can go larger.

Ken,

Do you feel that Image Uprezzing software will help with bigger prints? Genuine Fractals, or Big or some of the others?

ziggy53

pathfinder wrote:

Ken,

Do you feel that Image Uprezzing software will help with bigger prints? Genuine Fractals, or Big or some of the others?

I'm not Ken, but I do use both ACR and RAW Therapee to interpolate from RAW to larger sizes. It seems this is a very good time to interpolate since demozaicing "is" a type of interpolation in and of itself. The larger pixel count and 16 bits does seem to make a difference in retaining extra detail through vigorous subsequent manipulation.

gtc

20D 32X20 inch
My 20D images,depending on ISO,subject etc etc will print up very well to 32X20 inches @400dpi, just using bicubic interpolation from PS5.

If a client wanted a larger size,which hasn't happened yet,then the lab could do it for me.

I am still interested in a second hand 5D though....waiting for people to upgrade and the price to drop further....

LiquidAir

pathfinder wrote:

Ken,

Do you feel that Image Uprezzing software will help with bigger prints? Genuine Fractals, or Big or some of the others?

I have not tried anything other than the bicubic interpolation in Photoshop, but here is a link to a discussion of them I saw recently:

http://theonlinephotographer.com/the_online_photographer/blog_index.html

My standard approach to uprezing is to first scale the Photoshop bicubic and then sharpen with USM (or something better; I have just started using Pixel Genius PhotoKit Sharpener). When I sharpen, I keep the radius to 1 pixel or smaller and watch to see whether the sharpening algorithm is still picking up the edges. My conclusion is that there little benefit to scaling an image larger than it can be reasonably sharpened.

When printing at home on my Epson, I target one of three resolutions: 360 ppi, 240 ppi or 180 ppi. The theory behind that is that the printer's native pixel resolution is 1440 dpi so targeting an even factor of that will give me better performance from the print driver. While that appears to be the common wisdom on the net, I have not personally run the tests to see if that is true. Largely I find 360ppi to be overkill when sending files to the printer so I never uprez larger than 240ppi. So when making large prints I try 240 ppi first; if it sharpens well, I use it; if it doesn't I try 180 ppi instead. If 180 ppi does not sharpen well my usual answer is to make a smaller print.

As for ziggy's plan of uprezing in RAW conversion, it makes sense, but I haven't tried it. I usually don't know how big my print is going to be when I am converting the RAW file so converting for a target print size is a bit awkward in my workflow. Currently I am getting the best RAW conversions out of DxO Optics 5. Compared to ACR, DxO has significantly better highlight recovery and appears to have a touch more real resolving power.

bham

I have a 24" x 36" print from an 10D file that is awesome. I didn't upsize the dpi at all, maybe the lab did, not sure. But I wouldn't make a decision based on pixel count, since its only a small difference 10.1 vs 12.7.