Why is the depth-of-field preview in the optical viewfinder of my Canon 500D inaccurate?

Question

I have noticed that in my Canon 500D, the depth-of-field preview in the optical viewfinder is inaccurate with large aperture settings.

If I press the DoF preview button, there is very little difference between, say, f/1.8 and f/3.5. In particular, pressing the DoF preview button with f/1.8 vs. f/2.8 seems to make no difference at all.

Obviously, there is a huge difference in the photo, and certainly I can also see the same difference if I use live view (LCD screen) and the DoF preview button. And even with the optical viewfinder, the DoF preview button seems to work as expected with smaller apertures (say, the difference between f/4.0 and f/8.0 is clear and what I see in the viewfinder matches what I see in the photos).

What is going on? Exactly what limits the performance of the DoF preview button with the optical viewfinder, and what is the largest aperture with which it still produces "correct" results? Are there differences between different camera models regarding this aspect?

---

After a lot of googling, I was able to find this page which suggests that the focusing screen in the optical viewfinder might be the limiting factor:

"Oddly, these modern screens get no brighter when you're using a lens faster than f/2.8. Try it: put on an f/1.8 or other fast fixed lens and flick the depth of field button. You'll see no change in anything until you stop down to about f/2.5!"

Sounds familiar – but the above quotation is about Canon 5D, which is obviously a very different thing from my 500D.

I also found this page which is specifically about 500D, but the discussion thread seems to give few conclusive answers.

Answer 1

Many confused answers here... Eruditass got it right, it's all about the viewfinder. Actually it's mostly the "ground" glass, which is not a ground glass anymore: it's a microstructured glass, optimized for light transmission with slow lenses, not for ease of manual focusing. Something a bit like a Fresnel lens. The eyesight, has nothing to do with this problem, nor the viewfinder coverage, nor the pentamirror or whatever.

Ken Rockwell suggests a simple experiment: "Look through the front of your fast lens at the focus screen. It's black outside the area of the lens that corresponds to f/2.5!". Try it! You will clearly see that no light comes through the outer part of the lens. If light cannot travel one way, it cannot travel the other way: only the light rays that hit close to the center of the lens can get through the eyepiece.

If you want a focusing screen optimized for actually focusing... you may try one of the KatzEye focusing screens. Never tried myself.

Edit: As a followup to Matt Grum's post, here is a picture of a 85/1.4 seen from the front side:

On the left: the lens alone (with my girlfriend holding the aperture open). You can appreciate the extra large entrance pupil (~ 61 mm). On the right, the lens on the camera. Here the camera is holding the aperture wide open, but you only see light coming out from the center of the aperture. It's roughly f/2.8, although the borders of the effective aperture are not very well defined.

Answer 2

It's to do with the focussing screen, however I don't profess to completely understand all of the effects you mentioned. The focussing screen in modern DSLRs is made of laser etched glass in order to facilitate manual focussing and transmit as much light as possible for slow lenses. With old fashioned ground glass screens, the micro-structure of the glass contains lots of tiny globules, each of which acts like a miniature split prism (what you used to get in the centre of the focussing screen on old manual focus SLRs). This makes the in focus parts look even sharper to help with manual focussing.

This is backed up by the fact that most manufacturers offer darker focussing screens for easier manual focusing, which do get brighter as you open the aperture past f/2.8, or brighter less accurate screens for better visibility in low light which don't.

edit:

I carried out the experiment Edgar suggested using a 50 f/1.4 lens and whilst I originally saw only the middle of the focussing screen, as I held the lens closer to my eye I was able to see more and more until I could see the entire screen. I'm not doubting that the lack of extra brightness at large apertures is due to the screen, and that the way the glass is cut somehow obscures light from the perhiphery, just that I wasn't able to observe the manual vignetting that Ken suggests.

I couldn't get a good photo to prove this as I couldn't get the lens of the other camera quite close enough, but I did get this:

You can see the bottom two corners and if I moved the camera a fraction the top two corners too.

I tried 4 different cameras and always got the same result, that it's possible to see the whole focussing screen through the lens. I also got this shot with a macro lens that demonstrates the Fresnel structure of the standard focusing screen:

The shot also demonstrates some falloff which seems to be responsible for the lack of brightness at f/1.4, but why the edges of the focussing screen aren't darker when viewing it head on I don't know.

Answer 3

Today's viewfinders are designed to have better light transmission at the expense of diffusion. This is because autofocus cameras use a semi-transparent main reflex mirror so part of the light passes through the mirror and to a secondary mirror that is reflected down to the AF sensors at the bottom of the camera. In addition, many cheaper cameras use a pentamirror which offers a less bright image.

I use LiveView and DoF preview at the same time to get an accurate view of the bokeh.

Answer 4

I have the 550D (T2i) which is close to the 500D in many ways. I do not find it credible that the viewfinder could substantially alter the depth of field unless somehow it is managing to refocus the out-of-focus areas and I doubt it's capable of that.

As a check, I took a look at items in my office through an f/2.8 17-55 mm lens and was readily able to detect changes in depth of field even between f/2.8 and f/3.2. The changes are more pronounced at 17 mm than zoomed out at 55 mm. I then mounted an f/1.8 85 mm lens and looked at the same subject. This time it was almost impossible to be sure of a change in depth of field until the aperture reached f/5.

An explanation can be had by calculating the depth of field. With a 17 mm lens focused at 8 feet, for example, the depth of field at f/2.8 extends to 2.48 feet in front of the subject and at f/3.2 it's 2.70 feet in front. That change of 0.22 feet (almost 3 inches) was large enough for me to notice. With an 85 mm lens focused at 8 feet the DoF at f/1.8 extends a mere 0.09 feet in front of the subject. At f/2.2 it increases to 0.11, a whopping 0.02 feet (1/4 inch). I just couldn't see that, because the room is a little dim to begin with, the subject was not extremely contrasty, and as one stops down the aperture the viewfinder not only gets darker but it gets noticeably vignetted (even further darkening the areas that tend to be marginally in focus). By f/5, though, the DoF has expanded to 0.24 feet in front of the subject, a change of 0.24 - 0.09 = 0.15 feet (almost two inches): this is within the range my eye can detect even in a darkened subject.

Therefore I would like to suggest that the combination of your viewfinder (which is small and fairly dark), your eyesight (whatever it may be), and your scene give you a certain threshold of distance over which you can detect a change in focus. (For me, with my scene and my middle-aged eyes, that threshold seems to be around two inches). This threshold translates into a minimum change in f-stop that depends on the current f/stop and, importantly, on the focal length of your lens. Especially with medium to long telephoto lenses you might have great difficulty actually seeing the changes in DoF associated with small differences in f-stop through your viewfinder.

As others have already mentioned, the LED screen gives a much better way to preview DoF, especially because you can zoom in to investigate extreme details. Even with better 35 mm format cameras and better viewfinders (I've had a bunch over the years) I never found DoF previewing in a viewfinder to be very reliable: the most you can hope is to get a rough sense of whether your entire subject might sort of be in focus.

Answer 5

I have a different theory that might explain some aspects of having different depth of field on your viewfinder and on your live view screen / final picture.

The theory is that your eye has its own focusing mechanism. With live view or electronic viewfinder, all of the light rays are seen to come from the same apparent distance. However, with an optical viewfinder, your eye can somewhat correct for the shallow depth of field in the optics.

Based on this theory, the apparent depth of field in the optical viewfinder should be deeper than in the live view screen or final picture. This is exactly what I have been observing, even with the maximum aperture supported by the lens.