Revision: 0.2, December 6, 2003 (under construction)

Author: Joseph S. Wisniewski

Home: http://www.swissarmyfork.com/digital_photography_ifaq.htm

Copyright 2003, all rights reserved

Brief Introduction, what is an iFAQ?

1) Why isn't there a Leica "Digital M"?

2) What are "digital lenses"?

4) Will "full frame" replace "APS sized" digital?

5) What are "Focal Reducers" and "Wide Converters"?

6) What's a Foveon sensor?

7) What's Sony's RGBE sensor?

8) Why aren't there digital backs for film SLRs?

9) How about a DSLR with interchangeable sensors?

10) How many megapixels are enough?

11) How about electronic viewfinders?

12) Which sensors gather the most dust?

13) Does a CCD get hot or use lots of power?

14) What's an AA, anti-aliasing, or blur filter?

15) Why is 50mm "normal"?

16) What's the "three year rule" of lens design?

17) Why do they call some DSLRs "APS sized"?

18) What's a "Frankenstein's camera"?

A) Glossary

Acknowledgements

Disclaimer

Revision History

An iFAQ is a bit like a regular FAQ, a list of "frequently answered questions", and useful answers to those questions. In the case of the iFAQ, it is "infrequently asked questions". There aren't as many questions, and the answers are a bit longer.

Because! OK, that's not good enough. There are no Leica M style digital rangefinders because it's very hard to make one. One big advantage of a rangefinder (for normal and wide-angel lenses) is that it uses symmetrical lens designs. These are simple and compact, yet achieve very high quality. They have the disadvantage of causing the light to strike the film at angles far from perpendicular. Light from the 12mm Voiglander Heliar (a rather extreme example) actually reaches an angle of 60 degrees from perpendicular in the corners of the image. For such lenses, the "exit pupil" (the point in space from which light seems to originate) is very near the "optical center" of the lens.

Film is very tolerant of such wild angles of incidence. Modern digital sensors are not so tolerant. Many show very objectionable vignetting if light deviates from perpendicular by just 10-15 degrees. Few are usable past 20 degrees. That means that no "rangefinder style" normal or wide angle will work well. So, just trying to adapt a DSLR style sensor and electronics package (which can be made remarkably small) isn't enough, unless you plan on doing nothing but mild telephoto work.

Dream on. The digital rangefinder market is small, smaller than the film rangefinder market. And that is so small that film rangefinders are a decade or two out of date with respect to technology. None of the sensor companies are going to fund a wide acceptance angle sensor for this. The most you could hope for is that they develop sensors with wide enough acceptance angles to work better with the more troublesome DSLR lenses. This will solve the problem for normal lenses (50mm exit pupil on a full frame camera) but won't begin to address wide-angle lenses, the forte of the rangefinder.

There are several things a small manufacture could do to an existing sensor to make it "rangefinder friendly". The most obvious is to use some sort of optical device in front of the sensor to make it compatible with the light angles from a rangefinder style lens. Making it accept a wider range of angles (say with a diffusion or scattering plate) is out, obviously. The device would have to alter the acceptance angle range without distorting the image, or removing substantial light. This means a focusing lens, of some sort. That would help, within limits. The lens needs to be very near the sensor, and therefore, flat, like a Fresnel lens or a diffractive optical plate. But it can be done. Either approach could be etched or pressed into an optical coating on the front surface of the chip.

Now, if we say +/- 15 degrees is a reasonable limit for the acceptance angle on a sensor, what can we do with a single plate (Fresnel or diffractive)? Leave the center of the image alone, of course, and gradually direct the acceptance inward as we get to the corners of the image, until we're applying a full 15 degrees of correction. Now, a lens that has a 15 degree angle at the corners is corrected perfectly. One that is 30 degrees is knocked down to 15, and is still acceptable. And a telephoto that is close to 0 degrees becomes -15 degrees in the corners. Still acceptable. This gets us working with any lens from 35mm to telephoto.

Say you still want the rangefinder to have a wider range of compatible lenses. There are a number of different ways. How about some sliding plates (either motorized, or manually operated). For telephotos, don't put in anything. That covers the 0 to 15 degree range. Now add a sliding -30 degree plate, and we're covered from 15 degrees to 45 degrees. If you really want that 12mm Heliar to work, add a second sliding plate, for another 30 degrees, and we can cover 45 to 75 degrees. That's a bit silly, so if we were going to have three plates, it's probably best to make them 20 degree instead of 30 degree plates.

Or better yet, have one glass disc that has all the Fresnel or diffractive optics etched onto it for rectangles of 0, -15, -30, and -45 degrees, has a tough optical coating, and rotates into place for each lens. The tough optical coating would allow it to have a stationary "wiper pad" and become self-cleaning!

Well, there have been hints of this. The Leica Digilux 2 (announced November 2003) is often hailed as a "Digital M". But it lacks several "M" characteristics. First and foremost, it doesn't have a rangefinder (or even an optical viewfinder). It has an EVF (electronic viewfinder) with all the inherent limitations (see section 11.2 of this FAQ). And it sports a largish permanent zoom, instead of smaller, interchangeable lenses.

In it's day; the "M" was a classic reporter's camera, very fast, unobtrusive, and great for indoor, low light work. It is possible to bring these characteristics to a digital camera.

Rangefinder: this requires a mechanical focus coupling between the lens and the camera. The lens has to have a "focusing cam" that can convert barrel rotation to distance. This is easy with small prime lenses, but more difficult with wide range zooms.

Interchangeable lenses: this is not particularly difficult. It's done now with DSLRs, and the rangefinder (not having a mirror) has better possibilities for a permanent "dust shield" to protect the sensor when the lenses are changed.

Better image quality: only possible with a bigger sensor. It's quite possible to have a sensor that's APS sized (like a DSLR) or "four thirds" sized (like an Olympus) instead of the 2/3 sized sensors common to point and shoot cameras today. This would not result in a prohibitively large camera.

Well, there's a really nice "Digital Lens FAQ" floating around. Here's a quick summary. There is no real agreement among lens makers and digital camera users as to what constitutes a "digital lens". A "digital lens" could be described as a lens with one or more of these characteristics: long "throw" (strong retrofocus or telecentric design), reduced coverage circle, reduced chromatic aberration, or increased resolution. It may also have other "digital friendly" features, such as a double lens cap or lens shade that provides coverage for both a full sized film frame and a reduced size digital frame.

This is also covered in more detail in the Digital Lens FAQ. Some consider "four thirds" to be an entirely new, digitally optimized system. Others consider it to be a simple repackaging of existing Olympus OM lenses. We’re still sorting it out. Right now, no one that I know of has access to the "open specification" for the 4/3 system (as announced jointly by Kodak and Olympus) so this section will refer specifically to characteristics of the Olympus "E-System" flavor of "4/3". This appears to be a rather advanced digital lens system, having virtually every desirable "designed for digital" characteristic.

The short answer is no. Film photography has proven that there is plenty of room for different formats to coexist, such as APS, 35mm, and medium format. Each format tends to dominate certain niches of photography, but that isn't a "carved in stone" rule. "Full frame" and "APS sized" digital are settling into a similar pattern.

APS sized DSLRs excel in a number of applications. They deliver image quality that exceeds that of most 35mm film (except the slowest color and B&W films, and digital SLR should be at that quality level within the next 2 years). There is a market for film SLRs ranging roughly from $300 (Canon Rebel, Nikon N65) to $2000 (Nikon F5, Canon EOS 1V). At the high end, people buy them because of all the capabilities of an SLR, speed, AF capability, ruggedness, accurate finder, etc.

The camera manufacturers are finally catching up to this, and making some of the DSLR components (mirror, pentaprism, etc) smaller and lighter to match the APS sized sensors. This frees up more weight and space budget for things like electronics, storage, and batteries. In late 2003, the Pentax *ist is the best example of this. Its APS sized pentaprism shaves about 80g off the weight of a prism built for "full frame" 35mm.

The Olympus E-1 stands out as an example of the ergonomics possible in an APS sized SLR. Being the first camera with a body that isn't a film camera design, its control placement is very well thought out for allowing access to the number of controls typical of a high performance digital camera, in a logical fashion.

"Full frame" DSLRs will continue to be produced, and will fill the niche normally occupied by medium format. With between 2.2 and 4 times more pixels than an APS sized SLR, these cameras have the resolution needed for product photography, architectural photography, high end portraiture, any field where extremely large final output is needed. And any field where cropping of the pictures (by editors, art directors, etc) is normal. But their size, weight, and cost disadvantages will prevent them from becoming the dominant format.

The focal reducer (or wide converter) is a device that goes between the lens and the camera to increase the field of view of a lens, just as the common teleconverter reduces the field of view. Such devices are well documented. Astronomers refer to them as "focal reducers". (Then again, astronomers refer to "teleconverters" as "Barlow lenses", so they're a bit of a strange lot. Always scurrying about in the middle of the night…).

Here's a nice "focal reducer" link.

http://astro.martianbachelor.com/CB245/ReducerDesign.html

That's the basic principle. In practice, a single lens isn't going to do it. A pair of achromats seems to be about the minimum that will get the job done (just like teleconverters).

Kodak has a patent on the basic concept of the "wide converter".

http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&p=1&u=/netahtml/search-adv.htm&r=1&f=G&l=50&d=CR86&S1=4,591,234.UREF.&OS=ref/4,591,234&RS=REF/4,591,234

Claim 3 is very specific:

The key difficulty with this is that prior art (of which there is much) all deals with astronomical "focal reducers", which all reduce back focus. On a telescope, this is not typically a problem, since their optical systems allow a little extra back focus, for placement of accessories. It is a problem on an SLR "wide converter" because back focus is minimal.

You'll often hear people say that extra elements hurt the picture, adding aberrations, or reducing light. This, of course, is absolute nonsense. While extra elements, applied improperly, will cause image degradation, using them right will improve the performance of the optical system. Yes, each element you add to an optical system adds some aberrations, some attenuation, and some reflection (increasing the risk of flare). But each element you add contributes to the system, correcting some of the aberrations introduced by other elements. If the system is properly designed the beneficial correction from each additional element more than offsets its additional attenuation or distortion. This is why you've probably never taken a picture with any lens having less than four elements.

In the case of the wide converter, reducing magnification brings all sorts of benefits. The prototype 0.66x "wide converter", for example, reduces most of the aberrations of the main lens by 33%. This includes flare (per unit area), chromatic aberration, softness (resulting from coma, astigmatism, or spherical aberration), and any aberrations resulting from centration problems (the lenses elements not all being precisely lined up along the common optical axis). In fact, your simple little 4 or 6 element system would have to increase any of these problems 50% in order to overcome the aberration reducing effect of the reduction in magnification.

As far as light loss, even the single coated prototype has a 2% per surface loss, which is 8% for the four surfaces. Since the focal length reduction provides a 112% increase in light density, an 8% loss is negligible. There's still over a full stop of light gain.

Sure, it's a good question. You've got a nice film SLR, a Nikon F5, F100, or F4, a Canon EOS 1V or EOS 3, a Minolta Maxxum 7 or 9. Why not just slap on a digital back and shoot away? Once you actually start looking at how you'd do it, you see that there are overwhelming reasons why a fully integrated camera is preferable to a digital back. Compared to a full DSLR, a film SLR and digital back has increased cost, decreased strength (or increased weight), poor ergonomics, and a lack of desirable features.

A film SLR + digital back is heavier and bulkier than an integrated DSLR. The back of a film SLR is lightweight. It's only function is keeping light out, and supporting the pressure plate. Neither of these are precision functions, and the attachment mechanism between front and back reflects this. To keep the back and front from flexing independently of each other (destroying alignment) the back needs to be made stronger (heavier, more expensive) than the digital section of an integrated camera. Unlike the integrated camera, a digital back isn't anchored to multiple points of the camera's main chassis. The best it can do is try to grab onto the flimsy back hinge and latch points, and go under the camera, to screw into the tripod socket.

Then there's the problem of making a weatherproof (and dust proof) seal between a heavy, poorly anchored back and the main camera, when the digital back is, literally, hanging off the back of the camera, doing it's best to pull the seal apart.

You cannot build a full frame 24x36mm digital back for a conventional 35mm SLR (or a medium format SLR, for that matter). The SLR is designed to position the film at the focal plane of the lens. To do this, the film (which is wider than 24mm) rides on polished film rails which hold it a fraction of a mm from the back plate of the camera. This back plate has a 24x36mm rectangular cutout leading to the shutter and mirror box.

In a digital SLR, the silicon surface of the sensor itself must be at the focal plane of the lens. The problem is that there are several things in front of the surface of the sensor. First is the cover glass of the sensor. Second is a stack of optical filters: the anti-aliasing filter, the IR "hot mirror filter", etc. And finally, the metal frame that holds it all together. The end result is a stack that it thicker than the film rails.

So, in a digital back such as Leica's, the sensor is mounted to a support structure that presses against the film rails, and the entire sensor protrudes past the film rails into the 24mmx36mm cutout, almost to the shutter curtains. (This is also how the Kodak digital backs for Nikon and Canon bodies worked). The sensor's silicon "chip" is also bigger than the "active area" that actually senses the image. The image area is surrounded by buffers, shift registers, and the metal leads "bonded" to the chip (that connect the chip to the outside world), and a frame for the leads.

When you add the electronics, leads, and cover glass frame all together, you're talking about a 4-5mm perimeter around the sensor. The end result is that 1.36x crop sensor of the Leica back, with a 17.6x26.4mm active area, is about as big as you can cram into the cutout and maintain that perimeter.

Now, if someone were to build a camera that had at least part of the back plate (just past the film rails) removable, this new film camera would be capable of being changed from film to full frame digital. However, consider three "indicators" that no camera company will ever build this new "digital friendly" film SLR. First, Canon say that 2003's DSLR sales have passed film SLR sales. Second, Nikon is now launching all their "flagship" features (fastest 11 zones AF system, fastest mirror and shutter) on a DSLR, the D2H, not on a film body. There may not even be a film "flagship" F6 in the future. Third, Olympus has re-entered the SLR market, with a digital only family. No film body, and lenses that won't even cover film.

And, since the digital back isn't full frame, you need a viewfinder mask. But the readouts in the viewfinder are going to stay where they are now, so you'll have a big gap between the small, masked finder image and the readouts. To make things worse, every integrated DSLR launched since early 2003 (such as Nikon D2H, Olympus E-1, Canon 10D, Pentax *ist) increases the viewfinder magnification, from a film SLR's 0.7x-0.8x, to a more comfortable 0.9x-1.0x. This offsets the "tunnel vision" small viewfinder image of DSLRs.

Since the digital back isn't going to be full frame, you end up with problems in the camera's auto exposure and AF systems. Modern SLRs have metering systems with many zones distributed over the entire 35mm frame, anywhere from the 21 zones of a Canon EOS 1V, to the 1002 zones of a Nikon F5. Sure, you can mask off the finder, but you've also got to "reprogram" the metering system, or it's going to meter things that are outside the smaller "digital frame". No current film SLR has provisions for such reprogramming.

AF is the same way. If you crop the finder, some of the AF zones are either going to end up out of the frame, or uncomfortably close to the edges of the frame. Neither of these are good developments.

And there's the little detail of TTL flash. Most cameras read the flash exposure while the shutter is open, measuring the light that reflects off the film. These sensors don't know how to read the reflection from a CCD (and they may also be built to read an area larger that the CCD, just like the exposure and autofocus systems above). To counteract this, integrated DSLRs use a pre-flash based system, where they fire reduced power pulses from the flash, before exposure, and base the flash exposure on these readings. The Fuji S2 is the only DSLR that attempts to do "normal" TTL by looking at the reflection from the sensor. This introduces other problems that are beyond this discussion.

Aside from the whole "bulkier and heavier" issue, the SLR + digital back can't match the ergonomics of an integrated DSLR. There's no real information flow between SLR and digital back, the SLR thinks it's shooting film. So, any "digital" functions (like white balance, ISO speed, and compression modes) have to be controlled purely by buttons on the digital back and menus on the screen, while the "camera" functions are operated via conventional controls.

How's this for awkward? To set the camera for ISO 400, first pull up the ISO menu on the digital back, and set it for 400. Then "dial in" ISO 400 on the camera's knobs so that the metering system and the digital back will both be set to the same ISO, and you can meter. On an integrated DSLR, one knob does it all.

A digital back won't give you any status readouts in the finder. Integrated DSLRs typically show you how many shots are free in the buffer and/or how many more you can get on the CF card. With a digital back, the camera wouldn't even know that the buffer or CF card was full: it would happily let you continue to take more pictures, which would then be sent off into "electronic nowhere".

Does the sensor have an electronic shutter like Nikon D1X, D1H, Canon 1D, so the mechanical shutter can "relax" and just serve as a between shots "blind" for the sensor? It's wasted on a digital back. The film SLR can't "know" that the sensor has these capabilities, and it doesn't have controls or provisions for selecting a 1/500 sec X sync and a 1/16,000 sec top exposure speed.

How does the digital back even know an exposure has happened? In a DSLR, there's a sequence of events. You clear the sensor's buffers, open the shutter, make the exposure, close the shutter, and immediately transfer the captured image out of the sensor. It degrades rapidly if left to sit in the sensor's storage cells. (This is the same degradation that causes noise in long exposures on a DSLR). So the digital back has to monitor the shutter somehow (infrared reflections, sound, etc) to know when to clear and when to transfer an image.

Most SLRs don't have much provision for powering a large accessory like a digital back, so the digital back has to have its own power source. Now we're into a system with multiple types of batteries and chargers.

And, the last issue, which body do you make the digital back fit. Nikon has two pro and serious amateur bodies, F5 and F100. Canon has both EOS 3 and 1v.

It's nearly impossible. You have to interchange most of the digital camera, not just the sensor. Different sensors have different interface voltages, different numbers of read channels, different clocking rates. So whatever carries the sensor must also carry a regulated power supply and the proper number and type of A/D converters. Now, since the different sensor types have radically different signal processing needs (Fuji diagonal matrix, Foveon stacked color interpolation, Nikon D1X 2:1 aspect compression, etc) you either have to put all that on the onboard processor, or make it uploadable.

This is assuming that a processor that can do one of those algorithms in reasonable time can do all the others in reasonable time, too. If not, you need to make the processor bigger (more expensive, more heat, less battery life) and waste that power when a particular sensor doesn't need it.

Then you get into shutter requirements. Does the sensor have an electronic shutter like Nikon D1X, D1H, Canon 1D, so the mechanical shutter can "relax" and just serve as a between shots "blind" for the sensor? Or does the shutter need to be a "real" shutter like in Canon 10D or D1s, Nikon D2H, Kodak 14n, etc.

Then there's details like Fuji CCDs reflect enough light so they can be used for TTL metering, but Nikon and Canon sensors don't, so they require preflash systems (Canon E-TTL, Nikon D-TTL).

And if your "new" sensor doesn't have the same size as the old sensor, you need a new viewfinder mask.

As of August 2003, the common answer is six, for most photographers, and 12 for serious studio photographers. This, of course, is not a valid answer, and requires one to accept three myths: "6 megapixels is adequate for sharp 8x10, 11x14, or 13x19 prints", "current lenses don't have enough resolution for higher pixel counts", and "higher pixel counts increase the cost of the camera dramatically".

This is a very complicated question. Most people agree that about 6 lpm (about 300dpi) is a critically sharp print. This means that 2400x3000 (7.2mp) is a sharp 8x10, 2400x3600 (8.6mp) is a sharp 8x12, and 3900x5700 (22mp) is a sharp 13x19. We're often willing to tolerate a lot less. I make 13x19 prints from a 6mp DSLR that look pretty good (with the right sharpening and scaling) until you compare them to a well made print from a medium format film camera (or even very careful 35mm work on fine grain film). And that current, state of the art 6mp DSLR isn't delivering a sharp 6mp. Because of the AA filter, and because of the way that lens resolution interacts with sensor resolution, there's more like 3 or 4 critically sharp MP. A 36mp sensor would deliver a perfectly acceptable 22mp of "critically sharp" image. The implications of this are staggering. For some, it means 13x19 prints that would look somewhere between "medium format" and "large format contact print" sharp. Prints that you could sell to the most demanding art collector. For others, it would mean "usable" prints up to 36x48 inches (1 x 1.5 m).

You would be surprised. A high quality lens built for full frame 35mm film has a resolution on the order of 200 lpm on the film plane (2.5 um pixel pitch). There's an interesting relationship between sensor resolution (including any AA filter) and lens resolution, and how they combine to deliver the total (system) resolution.

They cost virtually nothing, just a little bit of time. Moore's law (more properly "Moore's observation") states that every 18 months, the number of transistors you can cram onto a given chunk of silicon (and their speed) doubles. This has held true for a couple of decades now, and shows no sign of changing. And, as far as a digital camera sensor goes, it's priced by the overall size of the die (rectangle of silicon) it's on, not the number of transistors you "draw" on that die. It wouldn't be that big a problem to take existing APS sized sensors (currently in the 6mp range with 8um pixels) and quadruple their pixel counts, to the 24mp range. The only problem is that the rest of the silicon in the camera hasn't caught up with that pixel count yet. Without quadrupling the speed of the camera's processor, you'd be looking at one picture every 2 seconds, instead of the 2-3 pictures/second rates of today's entry level DSLRs. And you'd be looking at 25 shots on a 1 gig flash card, instead of the 100-150 we get with a 6mp camera.

So, the cost is a little time. 36 months will definitely see enough reduction in processor and flash memory prices to support those 22mp APS sized cameras, and expect frightening 50mp full frame machines.

Ah, that's the interesting part. There are so many things.

Enough megapixels can completely replace teleconverters. If you get the sensor resolution so high that it can extract essentially every bit of resolution your lens has to offer, the sensor makes a better teleconverter than any "add on" optical device. It's at the focal plane of the lens, the point where aberrations are corrected as well as they can possibly be. There are no compromises, like a conventional teleconverter. No question of the converter matching some lenses better than others.

Enough megapixels will make the AA filter unnecessary. If the sensor is sampling at such a high resolution that it's beyond the Nyquist limit for a particular lens, you no longer need an AA filter for that lens. The lens performs all the high frequency limiting that you need.

A strange issue, it takes on almost religious aspects. Electronic viewfinders, either a 1.5-3 inch LCD display on the back of the camera, or a smaller display with an optical system that lets you bring it to your eye, are common. Some cameras share both viewing systems, optical viewfinder + back panel LCD.

There are many. In some ways, it's an improvement over the "what you see is what you get" design of SLR viewfinders.

Exact framing: An EVF can easily display exactly the image that you will take. For a conventional SLR camera, 100% accuracy requires painstaking (and therefore expensive) alignment procedures. This is why all but the most expensive SLRs have viewfinders that cover between 85% and 95% of the actual image.

If you want to shoot black and white, an EVF lets you see and compose in black and white.

If you shoot with an infrared filter, you get a bright, clear, visible preview of your "invisible" picture.

You can stop down to do a depth of field preview without having the display go dim. (I don't think anyone is actually implementing this yet, but with an EVF, they could.

You don't have the parallax and barrel distortion of a modern SLR focusing screen (those ultra bright screens they have these days only scatter light over a fairly narrow cone, it's a cross between old-fashioned ground glass and a pure ariel image). When you move your eye behind such a finder, the framing actually changes.

Instant zoom in for critical focusing. You don't have to buy an extra focus magnifier.

They provide a bright viewfinder in moderately dark situations.

On demand grid lines, crop lines, centered crosshairs, reticule, etc. An EVF can display as much information, guidelines, etc. as desired, or give you a "clean view". Want a superimposed histogram? It's available.

They are movable, and remote operation is possible. Some "SLR style" EVFs tilt up and down, so you can look down into the camera for low angle shots. "Back panel" LCD displays frequently tilt and swivel.

On the old "free lunch" argument, here's some common disadvantages of the EVF.

They're slow, 15-30 frames/second update, there's a noticeable finder lag when shooting action. This is improving.

The resolution currently is pretty poor, about 200,000 pixels. Images look grainy. This should improve, but slowly. Most EVF finders use displays that were created for the much larger camcorder market, and 200,000 pixels is adequate for such devices.

They have limited low light performance. Right now, a dark-adapted eye with an SLR or optical viewfinder outperforms an EVF by a significant margin, basically due to the way an image sensor works when sampling a reduced resolution image. You have an annoying transition, for relatively low light the EVF is brighter than conventional finders. But in very low light situations, the EVF breaks up into noise.

They use power. A comparable optical viewfinder can operate all day without draining a camera's batteries.

LCD displays (especially the low cost ones used for EVFs) have limited color gamuts, and a tendency to have colors shift dramatically with temperature, viewing angle, age, etc. while an optical viewfinder accurately portrays the scene colors.

CMOS and CCD sensors all accumulate pretty much the same amount of dust. Yes, CCD stands for "Charge Coupled Device", but the charges are small charges (volts) behind a layer of optical glass, which is behind a metal coated AA filter. The Sony ICX413AQ in the Nikon D100, for example, operates on 15 Volts.

The big difference is that, in a Nikon camera, the filter is closer to the sensor than in a Canon camera. So, a Nikon shows the effects of dust as visible spots when you use small apertures. The Canon filter is a little farther from the sensor, so it takes a pretty big clump of dust to show up as a spot. Dust does affect the picture by reducing contrast. By the time you notice this, there's an awful lot of dust on the sensor. Because Nikon uses a CCD and Canon uses a CMOS sensor in their high volume cameras, people associate dust accumulation with CCDs.

CCDs do not get hot. People constantly talk about how hot CCDs get, or how much power they use, citing those as major advantages of CMOS. This is a misconception. The CCD uses the least power of any major component of the camera. The display lighting, CPU, shutter mechanism, auto focus mechanism, and memory all use much more power. Look at the Sony ICX413AQ CCD used in the Nikon D100. It operates on 15V, and draws 7mA at 15V. That's 105mW, 1/10 watt. For a device that size, you cannot feel 1/10 watt of heat. I you touched the CCD, you wouldn't be able to tell if it were on or off.

In just about every format, weather it's 35mm, medium format 645, large format 4x5 or 8x10, etc. the "normal" lens is about the diagonal of the image frame. This gives a very pleasant 60-degree view, and looks "normal" on a print viewed from a "normal" distance. For full frame 35mm film, the image diagonal is actually 43.3mm, not 50mm. When rangefinder and viewfinder cameras walked the earth, before SLRs, the "normal" lens was a bit shorter, 35-40mm. A 50mm lens actually has a little bit of magnification, relative to a "normal". This "flattens" perspective a little: you have to get back just a little farther from your subject than you would if you were actually standing and looking at it. This was not done to make a more pleasing image. It was done to save costs in SLR designs. Low cost lenses (normal lenses on either rangefinder or SLR cameras) can be made symmetrical (front and back lens groups are mirror images of each other). This means a relatively inexpensive lens design corrects a lot of aberrations. For a rangefinder, a symmetrical 35mm lens might approach within 25 or 30mm of the film. For an SLR, this wasn't possible. So the camera makers had a choice, make the lens retrofocus (non symmetrical) which usually makes it more expensive and worse performing, or make it longer, to clear the mirror yet keep costs down and performance up. So the SLR normal is the "slightly telephoto" 50 or 55mm lens.

So, 50mm on a 35mm full frame camera has become the "dominant" measure of "normal". When you talk about viewfinder magnification, it's relative to a 50mm lens mounted on the camera (even for smaller digital formats such as APS sized Nikon and Canon cameras, or the 1/2 sized "4/3 system"). When you talk about a lens's "magnification" or "power", it's relative to 50mm, so you call a 400mm lens the equivalent of an 8x telescope or binoculars.

It's sort of like "Moore's Law" for optics. Several years ago, a lens designer (still trying to remember the name) pointed out that, because of increased computing power and optics design software availability:

"In the last three years, more lens design computations had been performed than in the entire, centuries long history of optic design before them"

And that this will also be true of the next three years, and the three years after that. This is why we're seeing zoom lenses that rival the performance of prime lenses from a decade before, and an explosion of versatile, high performance optics.

Most DSLRs have sensors that range from 13.1x17.4mm (Olympus) 13.8x20.7mm (Sigma), 15.1x22.7mm (Canon), 15.6x23.7mm (Nikon, Fuji, Pentax), up to 17.8x27.0 (Canon again). There are several different terms that can be used to describe this. Nikon calls it "DX format", where DX means "digital". It's the same label they put on their flashes meant for digital cameras. Olympus calls theirs the "four thirds system", an obscure reference to the "Vidcon tubes" used in TV cameras decades ago. Those two definitions are specific to their respective companies, and not useful for general conversation.

Other terms for an image in those size ranges includes "half frame", which is an 18x24mm format used by cameras such as the old Olympus Pen-F and Yashika Samurai, and "cropped". These terms define the digital formats in relation to film formats: "half of what?" or "cropped from what?"

APS film has, by coincidence, a 16.6 X 27.4 film size (APS-C, the most common of the three APS formats). This is close enough to the larger formats (Nikon and Canon) so that it's a useful name.

It's a term applied to cameras that lack a "finished product" feel. This is normally because the companies that designed these cameras chose to use limited engineering resources for other aspects of the camera design. For example, my two favorite "Frankenstein cameras", the Fuji S2 and Sigma SD9/SD10 have exceptional image quality due to the Fuji "Super CCD" in the S2 and the Foveon X3 in the SD9/SD10.

The main characteristics of "Frankenstein's cameras" are a lack of mechanical engineering effort, and awkward ergonomics. For example, compare the Nikon D100 to the Fuji S2. Both are comparable cameras, based on the Nikon N80.

In the case of D100, the film drive mechanism has been removed. The space where the take up spool used to be accommodates the CF card and allows an increase in the N80 battery compartment size to accommodate the rechargeable lithium "camcorder style" battery. Electronics fills the film cassette compartment. In S2, the N80 film camera mechanism is virtually unchanged. All the digital gear is outside the film camera "envelope". So the S2 ends up 18mm (3/4 in) bigger than D100, 90g (3 oz) heavier, and less reliable (the interconnection between digital component and film component is a well known failure point in S2. S2 retains the original "film camera" disposable lithium battery to power the "film camera" components (shutter, flash, exposure, and AF systems) and adds an AA battery tray to power the digital "add on".

The N80 firmware is totally rewritten to accommodate a mirror "prefire" anti-shock mode for macro photography, increased x-sync speed from 1/125 to 1/180, cleaning mode with "mirror lockup", and integration of the camera's existing controls with the new controls added for digital. The S2 designers didn't have access to N80 internals to make such changes. So you have none of those modes. D100 lets you change things like ISO, WB, compression mode, etc. just like you do with "film camera" parameters like shutter speed, aperture, bracketing, exposure compensation. Or, you can set those parameters from the back LCD buttons and menus, if you choose. Fuji S2 has no such integration. You change digital settings on the back LCD, film settings on the top knobs and buttons, and never the twain shall meet.

D100 has a new finder LCD, with markings more appropriate to the digital camera. Like any "crop factor" SLR, the viewfinder is masked to reflect the smaller sensor size. In D100, the LCD has been relocated 4mm, so that there is no gap between the viewfinder image and LCD. S2 uses the original N80 LCD in its original position, so there is a gap. Aside from this being simply annoying, you have to have your eye closer to an S2 finder to be able to see both the LCD and the image than you do with the D100. So D100 is a better camera for people wearing glasses or moving actively.

In the case of SD9, it's very similar to S2, another "Frankenstein camera". There's very little mechanical engineering; just enough to attach a digital component to a film camera. SD10 is an improvement, they've removed a few grams and relocated the battery. But it still doesn't have the "finished product" feel of a D100 or 10D.