The Sunday Morning Photographer
October 24 , 2004
Monthly Column By
Thoughts About "Full Frame"
The last column I published on this site is of a type almost tailor-made to make me look like a fool one day. In "Mo' Bettah" I wrote about "How Much Is Enough," and actually specified some parameters of digital cameras that I thought would amount to "enough."
This, of course, is tantamount to saying "further progress isn't needed," and all such pronouncements run the risk of being simply blind – blind to progress, blind to future innovation – or just wrong.
There are many examples in history, but the story that comes to my mind is one told by the great automotive writer Ralph Stein about a car called the Wills Sainte (((SIC))) Claire. C. Harold Wills was Henry Ford's chief designer, the man who was instrumental in many of Ford's early engineering successes, and who designed the classic, and lovely, Ford script logo.
The Ford blue oval, one of the most recognized logos in the world, was originally designed by C. Harold Wills.
Wills, a multi-millionaire from his work with Ford, wanted to build his own car. So he left Ford and founded his own manufacturing company on the banks of the St. Clair river. His first product had a V-8 engine and was said to be ten years ahead of its time, and cost eight times what a Model T did. Although Wills went on to build somewhere between 5,000 and 12,000 cars, only fifty of which survive today, Stein says that the future progress of his successful company was torpedoed in 1927 when a group of investors got cold feet and hired a consulting firm to analyze the automobile market. The consulting firm – and this is the point – concluded that there were three million cars on the road in the United States already, and the market was saturated. The investors pulled their money, and Wills Sainte Claire went belly-up in 1927.
C. Harold Wills admired geese, and made a gray goose the symbol of his superbly
engineered cars. Photo courtesy Wills Sainte Claire Club.
Apocryphal or not, the Wills story does illustrate the perils of making predictions (as well as the perils of consulting). Especially with a technology as vibrant as digital cameras are today, it's probably wrongheaded to say that anything is already good enough, or that we can guess already what good enough is going to be.
Accordingly, you can also take my thoughts about "full frame" DSLRs with a dose of skepticism if you like.
I first began thinking of this when my friend Oren Grad posed a riddle: What do Canon digital SLRs and view cameras have in common? I gave up on guessing the answer (and none too happily, I might add), but it seemed obvious once Oren clued me in: users of both have to deal with issues of lens coverage and angles of view for different formats.
View camera photographers are perfectly used to this. They commonly use the same lenses to serve different purposes on different cameras. A 210mm lens, for instance, is moderately long on 4x5, normal on 5x7, and would be quite wide-angle on 8x10 if it covered the format, which few of them do. Canon, similarly, has the equivalent of three different formats in its DSLRs today. Its APS-sized sensors on the Digital Rebel and 20D. being a "smaller format," make 35mm lenses effectively longer, by a "multiplication factor" or "field of view crop" of 1.5X. But the 1Ds and 1Ds Mk. II has a "full frame" sensor (meaning, in this case, the same size as 35mm), giving 35mm lenses the same angles of view photographers are used to from shooting film 35's. But there's an in-between sensor size: the one in the 1D and 1D Mk. II, which is 19x28mm and gives a field of view factor of 1.3X. So the same Canon 50mm lens, say, has a "normal" angle of view on a 1Ds, a slightly long 65mm equivalence on the 1D Mk. II, and a short telephoto 75mm equivalence on the 20D.
Nikon, meanwhile, with the introduction of its professional über-kamera the D2x, has shown its firm allegiance to the APS-size sensor that gives a 1.5X effective crop. Its digital "format" is uniform, from the entry-level D70 to the D2's (yes, there's the choice of a further crop with the D2x, but that doesn't count because the whole sensor is still 1.5X). Nikon has also quietly begun building up a collection of its DX lenses, made expressly for its digital format.
Not to beat around the bush: I firmly believe that what Nikon is doing will prove to be smarter in the long run.
The path Canon is taking gives it an advantage right now, because so many photographers want, or think they want, "full frame" (i.e., 35mm-sized) sensors, so they can use their old 35mm lenses on their DSLRs. I've opined before that this is "oldthink," and I still think so. What sensor development up till now has firmly demonstrated is that sensors are getting exponentially better as time passes. Right now, in my opinion, we're beginning to enter the stretch of time when the "enough is enough" influence will begin to kick in. That is, in five or six years, we're going to know approximately what the market demands in terms of ultimate pixel size from a DSLR-type camera. And we're also going to have a pretty good idea of what it doesn't demand.
And Nikon, by getting a jump on standardization in formatting, will quietly build for itself an advantage for the future, while Canon, though giving itself a decided advantage now, is inadvertently acquiring a disadvantage for the future with its multiple "formats."
And what of Michael Reichmann's argument that real estate in sensor size is always going to win, no matter what the technology? Well, I don't think he's wrong: but what shows us the shape of the future in that respect is the just-announced Mamiya ZD, with its gargantuan 36x48mm, 22 million pixel sensor. In other words, for those who really need high quality and large image size, the old 35mm frame isn't going to cut it. It won't be "full" frame in that sense either.
Mamiya's ZD gives us an idea of what the future holds for those who want the
ultimate in image quality.
My bet is that whatever we end up thinking of as ideal in terms of pixel count and image quality in an SLR-style camera for sports, editorial, and news work is going to fit just fine on the 16x24mm, APS-sized sensor area. Gradually, lenses purpose-built for the format will replace 35mm lenses as the standard line, and 16x24mm will be full frame.
Just one opinion.
All best and good light,
— Mike Johnston
A few notes: There is currently no further news about my book, unfortunately. If you've ordered a copy, I sincerely apologize for making you wait so long. This wasn't the way it was supposed to happen, believe me.
The 7th issue of my newsletter "The 37th Frame" has just shipped. If you're a subscriber and still haven't gotten issue #6, please contact me in the next week or two at email@example.com and we'll get you set up to receive the new issues. Issue #7 is more than 40 pages long, illustrated, and features an extended appreciation of the late Henri Cartier-Bresson, new lens reviews, a "rant" about the camera-police in the wake of 9/11, and an extensive "backgrounder" article on the new Leica 50mm Summilux. And of course my peculiar slant on the news. Please go to http://www.37thframe/subscribe.htm if you're interested in ordering your own copy.
My blog, "The Quotidian Meander," is at www.quotidianmeander.blogspot.com. For the time being it's mainly political, what with the election being around the corner, but that's more or less temporary. There are a few amusing "short stories" buried in the archives.Coming on my website: How to test a lens for "bokeh," and subjective bokeh rankings for dozens of lenses.
Please visit my website at www.37thframe.com and subscribe to The 37th Frame photography e-newsletter.
The Quotidian Meander: I now have a non-photographic blog (web log) at: www.quotidianmeander.blogspot.com. Warning: political content! Warning: irreligious content! Warning: smart-alecky content! You have been warned.
SUBSCRIBE to the newsletter and get issue #6 and #7! Click HERE. Do it NOW, while it's fresh in your mind, and before Mike decides to start talking politics again!
See Mike Johnston’s website at www.37thframe.com. Also, check out his monthly column in the British Black & White Photography magazine! (Usually available at Barnes & Noble bookstores.)
Mike Johnston writes and publishes an independent quarterly ink-on-paper magazine called The 37th Frame for people who are really "into" photography. His book, The Empirical Photographer, has just been published.
Some Further Thoughts on the Full Frame Issue
By: Michael Reichmann
I've been publishing my good friend Mike Johnston's columns on this site for some two and half years, and though I disagree with him from time to time, I figure that this is his soapbox, and so I rarely if ever add my two cents. But this time, because I am mentioned directly above, and with Mike's indulgence, I'll add some further thoughts. I don't so much disagree with Mike as I believe that he doesn't paint a complete picture.
The analogy with large format is a good one. The words "large format" in fact don't tell us much about the size being shot. It could be 8X10", it could be an 8X10" camera with a 4X5" reducing back, or a 4X5" camera with a roll-film back. So then this begs the question – why do people shoot these different formats? Why shoot roll-film on a 4X5" camera? Why put a 4X5" ground glass and film holder on a large and unwieldy 8X10" chassis?
The answer is, practicality, cost and convenience. You can work more quickly and inexpensively with roll-film, yet the large format camera gives you movements that a fixed medium format camera doesn't. Someone with an 8X10" camera schleps the beast around because they are after ultimate image quality. But there are times when the lower cost and convenience of 4X5" makes its use worthwhile. In other words, compromises in image quality are made, not because of image quality considerations, but because of practicality, cost and convenience.
Konica Minolta Maxxum 7D with Apo Tele Zoom 100-400 f/4.5-6.7. ISO 200.
Do We Really Want APS?
Now let's look at reduced frame (APS-C) sized digital cameras vs. full frame. Why would one choose a smaller format over a larger one, when the size of the equipment used, and the cost of shooting a frame, are essentially no different?
The answer is, simply because at this point in time large sensors cost more to make than small sensors. There are no other compelling factors.
Image sensors are made on silicon wafers, using semiconductor technology. Producing a chip has a fixed costs regardless of the number of sensors it contains. There are inevitable manufacturing defects on a wafer. If you have dozens of sensors on an 8" wafer, that costs $1,500 to make, and there are a few rejects, then the average cost of each chip is only slightly increased. If you have only 4 chips on that same wafer (as is the case with almost full-frame 645 sensors), and defects cause two of these to be rejected, then costs have doubled. (By the way, the reason why 645 chips are about 10% smaller than nominal 645 size, is because by this slight size reduction four chips can be made to fit on an 8" wafer).
Smaller chips mean lower per unit cost as well as reduced rejection rates. Those are the economic realities of why your Nikon D70 or Canon 10D has an APS-sized chip rather than a full-frame sized chip. If sensors could be produced as inexpensively as different sized potato chips, we never would have had reduced frame sensors in the first place.
Bigger is Better
No one wants a smaller sensor without a commensurate advantage. If you like the magnifying effect of a 1.5X sensor when shooting with long lenses, great. But you get the identical effect from cropping a larger image. But these smaller sensors mean that existing wide-angle lenses won't give the intended and often required coverage, and therefore new lenses need to be purchased. These lenses carry with them the curse of not being usable on full-frame cameras as well, if and when prices of these come down, (as I believe they will).
We saw with the Olympus E1 that smaller sensors don't necessarily produce smaller cameras. The E1 (a 2X factor camera over full-frame 35mm) is roughly the same size as a 1.5X APS sized camera, and a full frame camera needn't be any larger either. Right now Canon's and Kodak's full-frame cameras are large, because in the case of Canon they have chosen to put this larger sensor in their 1 Series pro bodies, and in the case of Kodak because they aren't very good at camera design. (Why do I think I'm going to hear no end of feedback on this?)
So, let's look at the complete situation? Smaller chips produce no practical advantage for the photographer, and smaller chips don't necessarily make cameras smaller (at least within the 35mm format).
Now let's look at the flip side.
Smaller chips have no quality advantage. Indeed they are disadvantaged. Bigger chips can have more photo sites (pixels) and therefore will produce images that have higher resolution and which can be printed larger. Bigger chips also allow for photo sites that are themselves larger (for any given Megapixel count), and therefore image quality will be higher, because all other things being equal, bigger pixels are less noisy than smaller ones.
I know what Moore's Law says, and I know that it has to do with chip density, not chip size. But, nevertheless, I'll bet dollars to doughnuts that as time passes large imaging chips will continue to cost less and less to produce, and as this happens the cost differential will be reduced to the point where full-frame cameras become more affordable for the mainstream photographer.
Look at medium format digital. Full-frame 645 backs cost $30,000 a few months ago. Some still do, but now because of the Mamiya ZD announcement, prices are dropping. Phase One has effectively reduced their price for the 22 Megapixel P25 for a third, by offering up to $10,000 on a back trade-in. Why? Possibly because the ZD will likely come in at well under $15,000 for a full 22MP camera.
The Mamiya ZD uses the Dalsa chip. For Mamiya to be able to drop the price this much they clearly have to be buying these chips for considerably less than they would have a year ago. Kodak will likely follow suite with their 22MP chips.
Canon is now starting to ship the 16 Megapixel 1Ds Mark II for the same price as the 11 Megapixel 1Ds of two years ago. Final 1Ds bodies are being discounted by at least a couple of thousand dollars. Prices are dropping.
A 3 Megapixel APS-sized sensor camera, the Canon D30, cost over $3,000 a few years ago. A couple of years before that a 6 Megapixel Kodak DCS camera cost $30,000. Today an 8 Megapixel Canon 20D, using an APS-sized sensor, costs under $1,500.
Can anyone deny that sensor chip prices are dropping? And knowing the semiconductor industry, would anyone deny that prices will continue to drop?
I'm no prophet. My crystal ball is fuzzy, and I rarely pick the right stocks. But it doesn't take a clairvoyant to see that chip sizes and chip prices are headed in opposite directions – the first one up, and the second one down. At some point the marginal cost differential between putting a full-frame chip in a 35mm format camera vs. putting in an APS-sized sensor will have diminished to the point that it won't represent a significant manufacturing cost factor.
So, who is right? Mike or me? Canon or Nikon?
Time, obviously, will tell. But won't the ride be fun?
I certainly respect Michael's argument, and it's well worthwhile for the reader to remember that Michael is an expert on digital imaging, whereas I'm not. And it's quite possible we're not arguing at cross purposes, merely seeing the same facts somewhat differently.
Still, I'm not convinced. MR says:
"Why would one choose a smaller format over a larger one, when the size of the equipment used, and the cost of shooting a frame, are essentially no different? The answer is, simply because at this point in time large sensors cost more to make than small sensors. There are no other compelling factors."
But there is a compelling factor: optics. Currently, 35mm's 24x36mm frame size is considered "full frame" because it's the smallest size that has proven robust for film (and because of compatibility issues that are far more important for film than they will be for digital). There's simply nothing magic about it other than that.
The great limitations in the design of lenses for photography are the flange distance, objective size, and exit pupil size. A few basic definitions:
Flange Distance: The distance between the plane of the lens mounting ring and the "pickup plane," i.e. the film or sensor plane where the image is focussed.
Objective Size: In the technical sense, "objective" is the outermost element, or the lens element closest to the image object (what you're taking a picture of).
Exit Pupil: The diameter of the lens element closest to the film, or the theoretically largest possible diameter of that element given the physical and mechanical limitations of the lensmount.
In 35mm cameras, the flange distance is dictated by the physical space needed for the reflex mirror. It greatly complicates the design and efficiency of wide-angle lenses for SLRs, many of which have traditionally had to be retrofocus type. Exit pupil size puts an ultimate constraint on the speed potential of the lens design. And while objective size is theoretically unlimited with 35mm lenses, in practice it is limited by the size, weight, and cost that photographers can bear.
Right now, we're assuming that we have to be stuck with the old 35mm "full frame" size simply because it fits our existing lenses. Our use of existing lenses is also dictating (for Canon and Nikon, among others) the flange distance. Even retaining the flipping mirror of conventional film-camera design, as DSLRs have done so far, the flange distance doesn't have to be as great as it is for APS-C as it is for 35mm, because the physical size of the mirror is smaller. But the fact that we're all still using 35mm lenses makes it mandatory, for optical reasons, that the manufacturers retain their existing flange distances. You could argue, in this sense, that "DX" and other APS-C specific lenses aren't totally designed for digital after all. They're designed only to cover the smaller format of the APS-C sensor, true, but they're also designed with the old 35mm flange distances as a given.
What does it matter?
So why are smaller flange distances, exit pupils, and objective sizes pertinent? Here, we have only to look to digicam lenses for the lesson. With a radically smaller sensor format, much smaller lenses that are also faster, higher resolution, and less costly are possible. The best digicam lens at the moment is probably the lens of the Olympus C-8080. It is a 5x zoom, 28-140mm equivalent, f/2.4-3.5 lens. While it's a fixed lens and I can't give exact size and weight, when extended it's smaller than a beer can, and the whole camera weighs 25 ounces, so we could estimate the weight of the lens at perhaps 10-15 ounces. And Olympus can sell it and the camera for $1k list and make a profit; so call the cost of the lens $450-600 and we probably aren't too far wrong.
A lens of the same specification and roughly similar quality made to cover 24x36mm format would be roughly twice as large, twice as heavy, and three times as expensive.
Now, granted, the advantages are not so marked when we're talking about the difference between 35mm full-frame and APS-C. But the potential advantages are still significant. So isn't this an advantage? This demonstration isn't bulletproof, of course, for the obvious reasons, but let's just look at a couple of lens prices. Nikon's new D2x, available soon, will have a 2X "cropped sensor" mode that doubles the effective lens focal length. Nikon's AF-S Nikkor 300mm f/4 lens costs $1,160 (I'm using B&H prices for comparison purposes). With Canon's full-frame 1Ds Mk. II, you can get exactly the same lens functionality in Canon's EF 600mm f/4 IS. For $7,200. That's a difference of $6,040. Congratulations – you've just paid for your D2x, with some change left over.
And if you want a lens that's one stop faster? Nikon's AF-S 300mm f/2.8 will run you $4,400, a savings of almost three thousand dollars over the one-stop slower Canon 600mm. And you can't get a 600mm f/2.8 for any price.
I know, some of the pixel peepers are out there screaming "No fair! It's not a direct comparison!" True. The 1Ds Mk. II gives you a lot more pixels. But remember that the D2x gives you a few alternative advantages, such as higher frame rate, faster buffer clearing, and smaller files that don't fill up your cards as fast. But if they're thinking they can always crop the larger sensor size and still get the same angle of view and as many or more pixels, I beg to differ. That sounds good on paper, and might work okay for landscapes on a tripod. But long experience tells me that this is not a very viable alternative for actual photographers doing actual work in the field. Visual people compose visually, not mentally, and trying to imagine a cropped central part of a larger viewfinder is unnatural in practice, especially with fast-moving subjects or visually exploratory subjects like street photography. Besides, you'll still be recording 16 MP when all you intend to use is 8, and who wants to do that? The fact is, I've known dozens of photographers who have happily set themselves up intending to crop a larger format size for one reason or another, only to discover that it's ultimately not a very satisfying way of working. Few people stick with such a solution in the long run.
As soon as camera makers develop a direct electronic viewfinder (EVF) that makes it possible to eliminate the reflex mirror (and they will, although it may take a while); as soon as they learn to put all the image quality we want into an APS-C sized sensor (and, they will, given current progress in sensor design), then the advantages to us in terms of lens size, weight, and cost will really begin to kick in. This is the crux of my argument: I'm claiming that, in the future, you're going to be able to get all the pixels (and image quality) you want from a smaller sensor, and get the lens advantages too. And at that time, the smaller sensor is going to be the winner.
— Mike Johnston
One Last Thought
By: Michael Reichmann
Most of Mike's arguments are sound, and I have little argument with them. But I feel that this issue needs to be seen beyond the theoretical, and in the light of market realities.
There are two of these which I feel need to be kept in mind. The first is the hundreds of millions of 35mm format lenses that are in existence around the world. These can't and won't be obsoleted in any short period of time. Eventually, of course, but not all that soon.
And, while amateurs with two or three lenses may change their systems every now and then, because some newer technology is appealing, pros are much less likely to do so because of their huge investments in glass, which, barring accidents, will usually last for a lifetime. This, combined with frugal amateurs, will ensure that full-frame lenses will remain in use for decades to come, and therefore there will be an ongoing demand for cameras with sensors that can take full advantage of them.
The second reason is that the cost to the world's camera makers of replicating their entire FF lens lines to some reduced frame is a process that will be both costly and lengthy, if it ever does happen.
Also, one only has to look at the history of photography to see that those seeking the highest image quality always turn to larger formats, and full frame 35mm will simply be seen as a larger format. Medium format digital is and will be of still higher quality, and so it goes. Just because something is good enough does not make it better. Better is better. As I reported on in my October 2004 Photo Plus Expo report, there is now a huge growth of interest in medium format digital by working pros. Look at the cover story on this month's Digital Imaging magazine, about how George Tiedemann, a widely published sports photographer, shot the Summer Olympics using a Hasselblad and a Phase One 22 Megapixel P25 back. Why? Because Zuma, the news agency that he was shooting for, wanted the highest possible quality images.
We are clearly headed toward a fragmented marketplace of multiple formats, but the paradigms of the past will continue to hold. As the say in the auto trade – there's no substitute for cubic inches.