Review date: 15 September 2005.
Last modified 03-Dec-2011.


There are several things about digital photography in general and DSLR photography in particular that're difficult to figure out, and the camera companies aren't helping.

Lens numbers, yet again

Because 35mm cameras were the de-facto standard when digitals started arriving, digicams of all types - whether integrated-lens "point and shoots" or interchangeable-lens models - have inherited 35mm lens specifications. The most immediately puzzling of which is focal length.

I've talked about lens specs in detail elsewhere, but, in brief, a "longer" lens - one with a larger focal length number - gives you a more magnified view, of less of the world.

Binoculars? Long focal length. Peephole in a door? Short focal length.

The focal length of a "normal" lens, which roughly approximates a normal human's perception of the world, is by definition the same as the diagonal size of the image sensor (or film). The diagonal size of 35mm film is 43.25mm (the "35mm" frame is actually 36 by 24mm in size), so that's the focal length a normal lens for 35mm should have.

There's a bit of a fudge factor involved here; people usually call a 50mm lens for 35mm cameras "normal". That lens, on a 35mm camera, gives you a horizontal field of view of almost 40 degrees. That aside, the 35mm-equivalent lens specs for other cameras are actually quite helpful; the real focal length of a lens in a small-sensored consumer digicam may be very small, but when it's set to "50mm-equivalent", it'll show you the same amount of the world as will a 50mm lens on a 35mm camera.

Anyway, a 50mm lens has a 40-degree-wide field of view, and a 100mm lens has a 20-degree field. This may trick you into thinking the mathematical relationship between focal length and field of view is simple. Naturally, it's not.

A 25mm lens, you see, only has a 71.5 degree horizontal field of view (a bit more diagonally, considerably less vertically, thanks to the shape of the frame). As the focal length approaches zero, the horizontal field of view approaches 180 degrees.

The widest "rectilinear" lenses (i.e. not fisheyes) are about 11mm; they've got a horizontal field of view of about 117 degrees. Fisheye focal lengths aren't directly comparable with rectilinear focal lengths; an 8mm fisheye on a full-frame 35mm camera (or its possibly hideous equivalent on some other kind of camera) will give you a 180-degree circular field of view, rendered as a circle touching the top and bottom of the frame.

If you start trying to figure all this stuff out for yourself and get results that don't match your empirical testing of your lenses (measure camera's actual field of view by taking pictures of handy objects and ruler, fail to remember your high school trigonometry, cheat, experience puzzlement at ridiculously small results from equations until you multiply by 180 and divide by pi), that's because lens specs are always fuzzy. One 50mm or 50mm-equivalent lens won't necessarily have quite the same actual field of view of another 50mm, even if they're both from the same manufacturer.

The same goes for aperture settings. One lens set to f/4.5 won't necessarily give you exactly the same image brightness as another (though different examples of the exact same model of lens should be functionally identical - only the very cheesiest manufacturers can't manage that).

Oh, and it goes for ISO ratings, too. Just as one "ISO 200" film formulation isn't necessarily exactly as sensitive to light as another (actually, the big numbers printed on the outside of film boxes can be quite a bit different from the film's actual sensitivity), one model of digital camera's "ISO 200" setting won't necessarily exactly match another's. These things aren't lab instruments; you can't use a digicam picture as a four-significant-digits measurement of illumination, or colour, or indeed anything.

The small picture

Focal lengths are not the only way in which 35mm specifications have sneaked into DSLR-land in confusing forms. There's also viewfinder magnification.

Take a peek through the viewfinder of a real professional SLR - digital or otherwise - and you generally get a quite expansive view of the world. The viewfinder image takes up a decent amount of your field of view.

That view will look like IMAX compared with the view through the finder of even a quite good consumer digicam. And, regrettably, many DSLRs aren't much better.

The viewfinders in non-pro cameras usually show you a bit less of the scene than the camera will actually photograph. That makes framing easier, as you don't have to be perfectly accurate about making objects fit inside the frame. If Aunt Maude's head is touching the top of the viewfinder image, you won't chop the top of it off when you take the picture, even if your hand shakes a bit. Most viewfinders have about 95% coverage, and that's fine.

Viewfinder magnification is more important. Magnification tells you how big the image is. High magnification means a big image, like a movie screen viewed from a good seat; low magnification means an image like a distant postage stamp surrounded by a sea of blackness.

The magnification, of course, varies depending on the focal length of your lens. More zoom, more magnification. So the magnification specs are standardised, to whatever you see with a 50mm lens on the camera (or with an integrated lens set to its 50mm-equivalent setting). A 35mm film camera with a 1X-magnification viewfinder and a 50mm lens on it will show you things through the finder the same size as you see them with your unaided eye.

1X is, however, too much to hope for from any 35mm-sized camera. A big bad pro film camera - a Nikon F5, say - will have a 100% field of view finder (so you see exactly what you're going to photograph, no more and no less; pros like it that way), with a magnification of 0.7X. So with a 50mm lens on the camera, things through the finder look 70% as big as they do through the eye that's not looking through the finder.

The EOS-20D has a 95%-view finder, with a 0.9X magnification spec. Cheaper DSLRs, like the 350D, have 0.8X.

Or, to put it another way, they don't.

These magnification specs are filthy lies.

DSLR viewfinder magnification specs, you see, are arrived at with the same 50mm lens as is used to specify 35mm SLRs.

Now, for full-frame DSLRs, that's not a problem. The EOS-1Ds Mark II, for instance, has a 100%-view finder with a perfectly honest 0.7X magnification figure, just like the F5.

But the 20D certainly does not give a bigger viewfinder image than the 1Ds. It just gives more magnification with the same lens, because its smaller sensor (and smaller viewfinder image, to match) gives the 50mm lens the field of view of an 80mm on the 1Ds.

The honest viewfinder magnification spec for the 20D, which actually tells you how big the image is, is about 0.56X. The 350D is 0.5X.

Now, you can buy six 20Ds for the price of a 1Ds-II. Even if twice the pixels and 1.25 times the viewfinder magnification are each worth proportional money to you, the 1Ds-II's still got a 2.4-times price premium. And a 0.56X 95% finder isn't like looking through the wrong end of a telescope, or anything.

But a big, bright viewfinder is an important feature of a true professional camera, and today's (relatively) cheap DSLRs do not make that grade, even though their spec sheets say they do.

The EOS-5D has a viewfinder befitting its "1Ds-lite" nature; 96% coverage, and genuine 0.7X magnification (oh, I'm sorry, 0.71X; that extra 0.01X magnification no doubt makes all the difference). Its full-frame sensor is matched by a bigger mirror shining on a bigger prism; there's room for the bigger prism because, like the more expensive pro cams, the 5D doesn't have a built-in flash.

Applying the same arbitrary value equation I used for the 1Ds-II above, the 5D has 1.56 times the pixels and 1.25 times the viewfinder magnification of the 20D, and costs about 2.9 times as much; if that's all you care about, the price premium's only about a factor of 1.5. That starts looking pretty good even if you don't care about full-frame and the 5D's other perks.

Except you can't, you know, actually buy a 5D yet.

Out, damned spot!

And then, there's dust.

Most cameras are not dust-proof. Zoom lenses have a tendency to suck up dust, because many of them telescope in and out, and few of them are environmentally sealed (that's really expensive, and a half-baked job of it can create a lens from which condensation can't escape, resulting in a perfect environment for the cultivation of award-winning lens fungus.

(Lensular flora can be both avoided and cured, but not every time. If it happens to a $150 lens after five years, not such a big deal. If it happens to a $10,000 lens after one darn trip to India... not so good.)

Integrated-lens digicams seldom have dust problems, because their leaks aren't large, you never take the lens off, and people don't generally use them in very dusty places. Dust inside the lens isn't likely to be a big deal anyway; it's dust on the sensor that'll show up in pictures.

In removable-lens film cameras, dust is also usually no big deal. You get a new "sensor" for every shot, and you can open the whole back of the camera, remove the film, and assault the works with a puffer bulb if needs be.

DSLRs, in contrast, just have the one non-removable sensor - which, thankfully, is covered by a sealed glass filter - and the only way into them that doesn't involve tiny screwdrivers is through the lens hole, with the mirror locked up.

Read some of the pages talking about the dust problem and you could be scared away from DSLRs altogether. Most DSLR users don't actually have much of a problem with dust at all. If you're not working somewhere super-dusty, and you take care to change lenses speedily and keep unconnected lenses and camera bodies capped, you won't have to clean the sensor very often, and it probably won't be very difficult when you do have to do it.

It's seldom a matter of urgency, either. Jobbing DSLR shooters who become aware of sensor-crud in the middle of a job just sigh and resign themselves to Photoshopping it out of the results, rather than breaking everything down and puffing and wiping on site.

And because sensor crud is on the filter on top of the sensor, rather than sitting right on the sensor cells themselves (no modern DSLR has a "bare" sensor; there's always glass on top), it's less and less visible as you use bigger and bigger apertures. The classic dust-discovery shot is a picture of a blue sky or nearby white thing (laptop screens work well) at infinity focus and the minimum possible aperture - commonly f/22. That'll turn up a speck or two on practically any DSLR that hasn't been cleaned recently.

But most people hardly ever use that small an aperture for real photos. F/22 or even higher is handy to get maximum depth of field for macro shots, but even at f/16 overall image sharpness will be suffering thanks to diffraction. As you make the aperture smaller and smaller, your camera acts more and more like a pinhole camera, with infinite depth of field but lousy sharpness everywhere.

If you can't see any crud on your images, you should of course leave the sensor completely alone. It is very definitely possible to make things worse - like my friend with his cheap rubber puffer bulb, which unlike my own cheap rubber puffer bulb, happened to be full of little bits of perished rubber, and blew them all into the camera when he used the bulb for the first time. Oh, how I laughed.

EOS-20D mirror

You also shouldn't stress over crud on the mirror of a digital (or film) SLR. If you see dust through the viewfinder but not on the images, then it's somewhere on the mirror or later on in the viewfinder optics, and doesn't matter at all unless there's so much of it that everything looks like a snowstorm.

You certainly can clean the mirror, but it'll be a fragile front-silvered type that won't appreciate enthusiastic scrubbing.

There are lots of DSLR cleaning products out there for those times when a humble puffer bulb won't cut it. Some gunk just won't shift unless you use a "wet" (which actually means "a somewhat damp lint-free swab") cleaner. Photographic Solutions' "Eclipse" fluid is well regarded, but it's just clean methanol, plus however much atmospheric water it's soaked up. Any similarly clean fast-evaporating fluid ought to work just as well.

Personally, I use the lightly-static-charged-clean-brush technique. I tried it with a couple of nicely baggied brushes I bought on eBay (to avoid the whole finding your brushes, washing your brushes, testing your brushes rigmarole) for about $AU25 delivered. They work very well, considering how little DIY effort was required.

(Don't let any of this dissuade you from buying a battery powered blower shaped like a phaser, though.)

The only DSLRs that don't need sensor cleaning are Olympus' E-1 and E-300 "EVOLT". They have a "Supersonic Wave Filter" that vibrates the sensor to shake dust loose. By all accounts, it works.

Dark to light

And then, there's dynamic range.

Dynamic range, in the photographic sense, is the light intensity range of a scene, or recording device. If the dynamic (or "luminosity") range of the scene - say, a brightly sunlit landscape with some deep shadows in it - exceeds the dynamic (or "light sensitivity") range of the recording device, then some of the light areas will come out solid white, or some of the shadows will come out solid black, or both.

In the olden days, even expensive digital cameras had poor dynamic range compared with film. This was a problem, especially for wedding photographers, who're high volume shooters who love digital now, but couldn't use it then, because their basic product is photographs of one person wearing dark black standing next to another person wearing radiant white.

Cheap and nasty digicams still have this problem.

The 20D's dynamic range, however, is a bit better than that of the 10D, which already managed about 8.5 stops. Each "stop" is a doubling of the amount of light, so 6 stops is twice the range of 5 stops, and the ten stop range of Fuji's current flagship is a significant improvement.

Ordinary colour print film only delivers about seven stops (not all of which makes it to the print...), slide film six stops at best (though all of that can make it to the projection screen), and the best current pro digitals are over 10 stops - only the widest-ranged black and white film can match them.

Like analogue recording media, though, film deals with overloads elegantly. The light response of various film emulsions plateaus off at the top end, so you're more likely to be able to recover detail from mildly overexposed images.

Then again, digital sensors deal better with underexposure, making it easier to recover detail that's all puddled down around the bottom of the intensity range, especially if you shoot in raw format (which you should, if all of this technical mumbo-jumbo actually matters to you at all). The Shadow/Highlight tool that debuted in Photoshop CS makes it particularly easy to rescue even seemingly severely underexposed images, but you can do it easily enough with free software too.

In digital photography, as in digital audio recording, you have to remember to err on the side of caution; a bit of underexposure is better than a bit of overexposure. It's not as easy as just setting the volume knob right when digitising your LPs, though, because a digital image will clip if any of the three colour channels (red, green, blue) is too bright, not only with white light. Clipping in one or two channels creates colour shifts.

It should be noted, however, that the brightness range of carefully made dye transfer prints, which are the finest prints in existence and certainly not your normal minilab photos...) is about 500:1, which is almost nine stops. The maximum possible brightness range for normal photo prints is more like 200:1, less than eight stops. In the real world, prints seldom beat seven stops.

So current consumer DSLRs (and many older ones, back to the D30) are up there with the best pro colour film for dynamic range, and both of them exceed the range of normal print technology. But if you want to do digital processing of your images, film becomes a big problem. You need a scanner that can capture the dynamic range of the negative, and that ain't cheap.

If you need more dynamic range when shooting challenging subjects, and your subject has the decency to sit still, you can get it in the same way as astrophotographers do; take multiple exposures and combine them.

This sort of thing was heinously fiddly in the analogue days, but nowadays desktop computers can do it with ease. Photoshop CS2 has a Merge to HDR function that automates the process very successfully; take a series of shots at different exposures, feed them to Photoshop, have a cup of tea, and there'll be a monstrous 32 bit image ready to clog your hard drive.

(Incidentally, you could use the same technique to make a wide-dynamic-range image from film scanned on a small-range scanner. It'd take a while, but if you're only doing a few images, it'd be worth it.)

Even if you're into arty black and white photography, though, dynamic range is not necessarily going to be a big issue. Many scenes don't come close to testing the limited dynamic range of slide film (which is good news for all of the people who still love it), and you can take great black and white pictures with a DSLR. The 20D has a "black and white mode" (which doesn't actually change anything about the files it captures - it's just a flag that causes RAW files to be decoded as monochrome), but image editor features like the Channel Mixer in Photoshop let you control the contribution each colour channel makes to the final black and white image.

Current home photo printers like the Epson Stylus Photo R800 (or the wide-carriage version, the R1800) produce excellent black and white prints without any futzing around with third party inks and papers. Bingo, the best of the old and the new, all on one desk.

It doesn't let you look as ominous as you do in a real darkroom, and Photoshop doesn't have the same potential for serendipitous screw-ups as film, but you can't have everything.

The rest of this review:
Introduction Resolution Lenses Strengths Alternatives

Review EOS-20D kindly provided by Dirt Cheap Cameras.

Australian digital camera shoppers should check out

Shoppers from the USA might prefer DealTime.

See also

The first page of this review is here.

If the photo-jargon in this review baffles you, my photo tutorial may help.

Or perhaps you'd prefer a book.

Give Dan some money!
(and no-one gets hurt)