5D or 50D

[Walter Schoenlein]

Just wanted to drop a line on the DLA. As long as you are not doing macro photography, the difference in DLA is irrelevant. The most common reason to stop down is the higher depth of field (if you're stopping down for longer exposure times consider an ND filter). Suppose you need f/11 on the 50D to get enough DOF. Framing and distance being the same, you will need f/(11*1.6) = f/18 on the 5D to get the same DOF. In both cases, diffraction limits you to effectively about 9 megapixels. if you would use a point and shoot with 1/1.8" sensor (crop factor 4.55), you would need f/4, again being limited to 9 megapixels by diffraction. There are only two theoretical solutions to overcome this physical limit:

1. Focus bracketing. Shoot several photos with a bigger aperture and several different focal distances and cobine them with appropriate software. Of course only applicable when using a tripod for a non-moving subject.
2. Make the sensor big enough to get in the domain of macro photography (higher magnification). Then the simple maths above will change a bit in favor of the bigger sensors. Of course, this solution is impractical and doesn't tell much about a difference between 5D and 50D.

So what do we learn from that? The better DLA of the 5D is only an advantage if you are doing macro photography at really high magnifications (say at least above 0.5 to be of pratical relevance). If this isn't be your main interest, don't care about DLA at all.
<p style="padding-left: 30px;"]-- Walter

[Walter Schoenlein]

Just wanted to drop a line on the DLA. As long as you are not doing macro photography, the difference in DLA is irrelevant. The most common reason to stop down is the higher depth of field (if you're stopping down for longer exposure times consider an ND filter). Suppose you need f/11 on the 50D to get enough DOF. Framing and distance being the same, you will need f/(11*1.6) = f/18 on the 5D to get the same DOF. In both cases, diffraction limits you to effectively about 9 megapixels. if you would use a point and shoot with 1/1.8" sensor (crop factor 4.55), you would need f/4, again being limited to 9 megapixels by diffraction. There are only two theoretical solutions to overcome this physical limit:

1. Focus bracketing. Shoot several photos with a bigger aperture and several different focal distances and cobine them with appropriate software. Of course only applicable when using a tripod for a non-moving subject.
2. Make the sensor big enough to get in the domain of macro photography (higher magnification). Then the simple maths above will change a bit in favor of the bigger sensors. Of course, this solution is impractical and doesn't tell much about a difference between 5D and 50D.

So what do we learn from that? The better DLA of the 5D is only an advantage if you are doing macro photography at really high magnifications (say at least above 0.5 to be of pratical relevance). If this isn't be your main interest, don't care about DLA at all.
<p style="padding-left: 30px;"]-- Walter

[gunslinge]

I am also wanting to upgrade my XTi, ( keeping it for a second body, it also was in for repairs by Canon once ), I have been looking at the reviews, the 5d would make by 24 to 105 into a real wide angle lens, I like to take wide angle shots and I love my 24-105 L. But the 50D has a better LCD with Live view, for macro ( another passion of mine, auto focus does not work well really close ), and it also has 1.6 crop factor which is better for racing cars and macro pics. What to do, a cheaper new rumored Rebal would also leave money avaible for another L series lens, a big plus.

[Stephen Probert]

Walter:


Hey, thanks, I didn't really understand this. I needed to go play with a DOF calculator for it to make sense to me. So the DLA is pretty much equal on the two bodies in question. I think in general I enjoy playing with shallow DOF, which would be a plus for the full-size sensor.

[Colin]

I think there's kind of two parameters going on here...


Sensor size, and pixel density.


In terms of pixel density, it seems like the DLA is really more of a matter of when the sensor can start resolving the diffraction itself. Above the DLA, your extra pixels don't get you anything more, but I don't believe that they hurt anything in terms of the resolution of the image itself. The most significant downside of having higher pixel density seems to be noise levels. Otherwise, higher resolution of the sensor doesn't make the image blurrier, it just means that as you surpass the DLA, you're not using the resolution the sensor provides.


In terms of the sensor size, what the article points out quite well is that what you gain, really, is the ability to carry lighter, less expensive glass, because you don't need as much glass to get the same field of view at a given aperturewith a smaller sensor. And, if you're using a lens for full frame sensors/film on a camera with a smaller sensor, the sensor uses the best part of the lense. However....


A given amount of CA, blur, measured in absolute units of distance/size/whatever, nanometers, whatever, will be greater in proportion to the image with a smaller sensor. In other words, a large sensor allows you to maximize the optical resolution of the entire lens, even if that resolution isn't as good towards the edges.


I keep my 30D for the pixel density. However, if my 5D had the same pixel density (and frame rate), I'd never use the 30d, EVER. A full frame sensor can do EVERYTHING a 'crop' body can, if the pixel density is the same (except mount EF-S lenses, but that's for physical reasons). Want a full frame camera to perform like a 'crop' sensor? Just crop the image. Take a picture with both, with the same focal length, same aperture, same distance, and crop the full frame image down, and you'll have the same image. I was reading on how the Nikon D-700 accepts lenses designed for 'crop' bodies.... It just internally crops the image from the sensor, that's all, throwing out all the information from the sensor where the smaller lens couldn't deliver anyways. It only uses the 'crop' area. You could do the exact same thing for macro, landscape, whatever...

[Daniel Browning]

My general dillema is that I am considering the 50D (new) and the 5D Mark I (used).

If you absolutely need thin DOF at wider angles or improved performance in low light, no matter the cost, then I would recommend the 5D Mark I. For everyone else, the 50D is a far better choice.

I like the smaller DLA of the 5D and the full frame sensor

Diffraction has the same effect on the 50D and 5D: the DLA does not make one better than the other. For a scene with the same angle of view, depth of field, and reproduction ratio, they will both have the same diffraction. (We can start a different thread about it if desired.)

I know there are a lot of other features that the 50D has because it is newer (liveview?).

For me, the many features of the 50D are highly valuable: live view, large LCD, anti-dust, improved autofocus, etc.

The 50D has a much higher pixel density than the 5D, making it a great camera to shoot sports and wildlife with because of the additional cropability it offers. The 5D, however, beats the 50D in IQ across the board. Now if it's bells and whistles you're after, the 50D has got the 5D clearly beat in all aspects (Liveview, MA, hi res LCD, My Menu, etc etc) BUT if its IQ and ISO performance that you're after the 5D has it in the bag.

Agreed

For an older camera it performs just as well if not better, as in this case, than the latest offerings within the DSLR market when speaking of IQ.

If you look only at low light performance with lenses that have over twice the aperture (i.e. same f/stop in a longer focal length), the 5D1 tromps the APS-C DSLR. However, in ample light, or with lenses of the same aperture, the difference is not as big. In fact, I think the 3 MP advantage of the 50D is far more significant in the case of ample light.

It's not the number of pixels that counts--its the quality of each pixel that does.

That's misleading. If the quality of each pixel was all that mattered, then the Powershot 600, with it's very large pixels, would out-do the 50D, even though it's a junky 12-year old digicam. Here is a 100% crop:






Incidentally, that 100% crop is also uncropped. (It's a half-megapixel camera). So the common saying about "quality, not quantity" is misleading.


A more balanced view is that quality per pixel must be multiplied by the number of pixels to quantify the quality of the total image, just as "horsepower per piston" must be multiplied by the number of pistons to know the total horsepower of an engine.

Generally, noise and dynamic range tend to scale with sensor size, so that a larger sensor with small pixels has more dynamic range (and less noise) than a small sensor with big pixels. Resolution, though, is less correlated.

From Bryan's review of the 50D, it seems he found he didn't like to shoot below f/8-f/11 whereas it seems that the 5D can push f/13- as far as diffraction is concerned. Anyone else have this experience with actual shooting?

It doesn't matter. They're both capable of the same DOF and have the same amount of diffraction. Walter had a great explanation, but I'll add my own to see if that helps clarify.

f/8 on the 50D and f/8 on the 5D each have very different depth of field, so comparing them that way makes no sense. The point in deciding on your f/stop is to get a certain depth of field. f/8 on the 50D will project a DOF onto the sensor that is about the same as f/13 on the 5D.

Think of it this way: All cameras are affected the same exact way by diffraction. One is not capable of "deeper" DOF than the other as long as the larger sensor can stop down to match the same aperture of the lens on the smaller sensor. The only difference is that cameras with larger sensors and larger lenses are capable of thinner DOF (if desired).

So the DLA is pretty much equal on the two bodies in question.

No, the DLA tells you what aperture to shoot at to get the *full* increase in resolution over the older, lower-resolution model. If you use a narrower aperture, you wont get the full increase, you'll only get maybe 90% of the resolution increase. (If you use a *very* narrow aperture, you might only get 50% of the expected increase).

The most significant downside of having higher pixel density seems to be noise levels.

Excellent post Colin; I hope you wont mind too much that I disagree with you on this one point.

According to my point of view, pixel density has nothing to do with noise in the total image.

The common mental model for "small pixels are noisier" goes like this: a single pixel, in isolation, when reduced in size, has less sensitivity, more noise, and lower full well capacity. (So far, so good.) Therefore, a given sensor full of small pixels is worse than the same sensor full of large pixels. (Incorrect.)

The correct mental model is to forget about pixels and think about light. The amount of light falling on a given sensor area does not change, no matter the size of the pixel. Large and small pixels alike record that light falling in certain positions. Both reproduce the same total amount of light when displayed; one just does it with greater accuracy.

The principle reasons that the myth persists are the following mistakes made when comparing data from pixels of different sizes:

• Unequal spatial frequencies.
• Unequal sensor sizes.
• Unequal processing.
• Unequal expectations.
• Unequal technology.

One good example is a chart of noise power and spatial frequency. Many claim that the 50D is noisier than the 40D because it has much smaller pixels. The following chart by Emil Martinec demonstrates that it's false:





I hope that helps guide your decision. Kind regards,


--


Daniel

[atticusdsf]

buy the 40d, and use the savings to buy an awesome lens (or two). i highly recommend the 17-55 ef-s with a 50mm f/1.4 to go with it (considering you already have a 70-200) between those three, you will have a very powerful set of lenses. the 50d looks like a nice camera, but not enough extra nice to justify the extra money, considering you can get a 40d body for about $800 these days.

[Colin]

The most significant downside of having higher pixel density seems to be noise levels.

Excellent post Colin; I hope you wont mind too much that I disagree with you on this one point.

I was thinking, and that makes much more sense. A smaller pixel may have less signal to noise PER PIXEL, but you don't necessarily get more noise per sensor area. What the noise filtering process does is average pixel values with their neighbors. Since noise is random, it may often cancel, or at least always diminishes in peak intensityif you average it. When you apply this filtering, you decrease noise, AND you decrease resolution. If you filter in image, and then scale down to lower resolution, the noise performance improves, similar to if you had a sensor with less pixels


My old 1MP kodak was HORRIBLE with noise. However, if you downsized the image 50% each way, and made it a .25 MP camera, the images looked far less atrocious. The colors were still ridiculous, but it was less obviously awful.


Fine, if somebody wants to give me either a 5D mkII or a 1DSmkIII, I suppose I'd be happy to use it. []