Sensor Sizes Part II

As I wrote in Part I of this post, below, it seems to have become a "forum truism" that the 4/3rds sensor size (13 x 17.3 mm) is woefully inadequate and can't compete with the allegedly more desirable APS-C sensor size (14.8 x 22.2 in Canon cameras). I made the argument that assertions for the superiority of the larger size were dependent on direct, controlled comparisons, and that any differences are actually probably too small in magnitude to reliably detect if people were simply looking at pictures.

Readers commented that not all visible differences can be attributed to sensor size alone—implementation and sensor design have a lot to do with it too.

That's certainly true, and to demonstrate it you only have to look at sensor design and implementation with a time dimension. In the year 2000, Canon introduced the landmark D30 (n.b.: not the 30D, a more recent descendant), a camera that was remarkable for any number of reasons. It was the first all-Canon DSLR, and the first to use a CMOS sensor. At 1.6 pounds it was considered almost amazingly compact and light for a DSLR, and, by the time it was reviewed there, Steve's Digicams said, brightly, "With an estimated street price now less than $2600, D30 sales have been and continue to be brisk!" Michael Reichmann made his bones by famously proclaiming that the D30's image quality matched that of film.

The D30 caused tremendous, almost unprecedented excitement at the time. Photographers had been accustomed to a sleepy, saturated market in which progress came mostly as refinements, and most refinements had to do with convenience for the user (rather than image quality) and more efficient manufacturability for the cameramakers. Slow-evolving companies with little R&D could usefully compete alongside companies that made haste somewhat less lazily. In that context, the pell-mell development of DSLRs for several years surrounding the turn of the Millennium seemed like a continuing succession of revolutionary shocks. (It was, among other things, a lot of fun to watch.)

The D30 had 3.25 megapixels. (The "camera to beat" at the time was the 1999 Nikon D1 at 2.74 megapixels, which Digital Photography Review called "Nikons [sic] answer to Kodak's domination of the professional SLR's [sic] market.") You could buy an IBM microdrive card with as much as 1GB of storage, although they were not cheap. Diagonal edges showed visible jaggies at 100%. Regarding noise, dpreview concluded that "Shooting for the web, smallish prints or a family album and [sic] you could certainly get away with ISO 800." It was considered "very fast" because you could take a shot every 1.5 seconds.

Eight years later, we take for granted factorials of two to four, all in our favor, in terms of megapixels, ISOs, shooting speeds, write times, and prices. 1GB flash cards are so cheap they're sometimes given away. The improvements and capabilities of 4/3rds sensors shadow those of APS-C sensors. Digicam sensors routinely exceed the D30's specs—or most of them—and a 4/3rds camera with the D30's specifications would be laughed off the planet.

You don't know
So, anyway, here are a few common assumptions about sensor sizes, closely paraphrased from actual statements I've encountered: "The days of 4/3rds are numbered. It won't be around in a few years." "All else being equal, bigger is always better." "The whole market is now moving to full-frame." "Every advance in quality and capability that is available in 4/3rds is also available to larger sized sensors, so the larger sensors will always be better."

This amounts to "the conventional wisdom" (CW), and it might be correct. (The CW sometimes is.)  But ever since I took a debating class in 8th grade it's been part of the repertoire of my thinking processes, so let me just bring up a few "debating points" that perhaps oppose the CW as set out above. Together these don't constitute a coherent argument; you might rather think of them as "nodules" of evidence, little wet balls of fact and assertion zapped at your pretty blackboard diagrams when you're not looking like spitballs through a straw.

Nodule: You don't know. The future isn't always a linear continuation of the assumptions of the past and the present. Progress is not always overtly logical, even if it follows someone's idea of logic. Ten years from now, 4/3rds might be dead. Also, ten years from now, all cameras might have 4/3rds sensors, from pro DSLRs to pocket digicams—at which point such an evolution will be back-constructed to seem logical and inevitable 'twixt here and then. Not saying I know. But you don't either.

Nodule: Despite the public pining of pontificating pundits such as moi demanding small cameras with large sensors, only one is even on the horizon. Is this because tiny sensors the size of fingernails are simply getting so good that Joe Sixpack and Jill Boxed White Wine don't want or need anything more? Is the manufacturers' strategy simply going to be to keep improving the fingernail-sized sensors until people like me have to shut up?

Richard Man (Imagecraft.com), My Wife, Karisu. Olympus E-3 and 35–100mm ƒ/2 lens, shot wide open.

Nodule: Bigger was better when it came to film, too, but that didn't mean that bigger always won out. In fact, film sizes got smaller and smaller until they fetched up against 35mm (24 x 36 mm), at which point there were several concerted attempts, most led by Kodak, to make the sizes smaller still (110, the disc camera, APS), none of which remained viable more than temporarily. So-called medium- and large-format survived, supported mainly by professionals and a loose confederation of aficionados and artists who together comprised an active but numerically minuscule segment of the overall market.

Nodule: "Post-Bayer," as Thom Hogan likes to call it, is going to change the game completely. Don't think for a second that there aren't engineering teams at Canon, Nikon, and elsewhere working feverishly to figure out how post-Bayer is going to play when it hits the mainstream.

Nodule: The market is moving to full-frame, check...based on three extant cameras (Canon 1Ds, Canon 5D, and Nikon D3) and three more (Canon 5D replacement, Sony "flagship," Nikon's competitor to the 5D) supposedly in the pipeline. These will doubtless exploit a market demand that currently does indeed exist. The Sony Mavicas, which wrote directly to CDs, and the digital "bridge" camera, like the Olympus E-10 and Sony F-717, also exploited markets that existed in their time. Where are they now? There's nothing that says any trend has to continue.

Nodule: Film speed. Lens speed. Film grain. Motor drive speed. Shutter speed. What  do these and many other technical specifications of photographic devices have in common? I'll tell you: they were once hotly contested fields of competition between manufacturers that were followed avidly by the market...until they reached "points of sufficiency" when further development just seemed to no longer make sense. (Okay, so maybe film speed never really quite got there). Lenses got to ƒ/1.4 and the market pretty much decided that was plenty; faster lenses exist, but they're mainly curiosities that never sold in anything but very low numbers. The Nikon 8008 had a huge market advantage when it came out, with its spectacular 1/8000th top shutter speed and 1/250th flash sync speed; but when faster shutters than that came along, the market pretty much yawned—when you can already freeze moving helicopter blades, you pretty much don't need shutter speeds faster still.

Okay. Fast forward to now, when high ISOs, capture rates and write speeds, shutter lag, resolution, and dynamic range are  hotly contested fields of competition between manufacturers that are followed avidly by the market. Will they be forever? Not likely. Much more likely: all parameters will reach "points of sufficiency" where people just won't need or want more. Where those points will be, I can't say. But there's one thing about it that's probably true....

It's not up to you
Idle speculative follow-up question: Let's say (for the sake of argument) that sensors reach this mythical "sufficiency" in every conceivable parameter the market demands by, oh, say, 2015. Then the manufacturers figure out they can develop sensors with the same performance, but smaller, cheaper, and with higher profit margins. Care to guess what your chances of buying a sensor of any particular size are going to be, if there is no demand for such a sensor in the market as a whole and if the manufacturers don't want to make them?

Upshot: Don't write off 4/3rds, and don't assume anybody knows what cameras or digital capture is going to look like fifteen years from now. If the lessons of the last fifteen years would seem to indicate any one thing, it's that this ain't over yet.

_______________________

Mike

Comments

[Tom V]

Not a trick, just a nice bokeh demonstration from a largish sensor and an f/2 lens. This is not a picture a current compact could take.

For me, low light capability is supreme. I am trying to get good at shooting dance under theatrical lighting. I am ALWAYS stuffed in the low-light corner of the envelope: too high an ISO, too wide an aperture, too slow a shutter speed. Autofocus performance in low-light is crucial as well.

For the average snap-shooter, low-light performance is key too! That picture of your mom and your sister laughing in the kitchen as they clean up the dishes after Christmas dinner? Do you want your camera to record the scene as you remember it, or with a little crappy flash that records it "deer in the headlights"?

[Cateto]

Very interesting point of view, and I mostly agree with you, Mike.

Yet I see a flaw in your reasoning; not big, but a flaw:

In most electronics gear in the XX century, the "sufficiency point" concept made sense.

Enter computers/processors. And that idea lost all its meaning: there is, by definition, not such a thing as "sufficiency" in computers: there is, every three months, something bigger, faster, and cheaper.

For good or bad, photography (and many other kinds of products related with electronics) entered into a dynamics which resembles a lot that of computers. And that means that they also are into a huge competition to do bigger, faster, and cheaper. So I don't expect, at all, to see "sufficiency points" reached in dSLR cameras, at any time.

(Or else the market, such as we conceive it today, will be lots.

But mind you: I WOULD ADORE to see the comeback of reason in the form of sufficiency points being something logic again. ;-) )

[dwig]

Excellent points. You could also note the small difference in image quality between cameras like the current Canon G9, the older Nikon CP8400, and other similar cameras which sport a rather small sensor and those from 4/3's or APS-C cameras. The size difference between 4/3's and APS-C is all but trivial.

There is one trivial slip in your list of Kodak's "smaller" formats. You should have listed 110 and omitted 828/Bantam. The standard Bantam image size is actually somewhat larger than 35mmDF (aka Full Frame) due to it being a 35mm wide film with perferations (one per frame) on one side only. If memory holds, the standard was a 28x42mm image. There was at least one split format camera (the Foth Derby, of which I own a sample) that shot two images per standard frame.

[Mike Johnston]

"For me, low light capability is supreme."

Tom,
For me too. I would say that the real difference to me with digital over film--my ability to shoot in MUCH lower light is much greater.

"Autofocus performance in low-light is crucial as well."

Right. This to me is a more legitimate criticism of the E-3 than its sensor size. Accord to Pop Photo's review, its AF is very fast with the right lenses but slows markedly in low light--the D300 and A700 do better when the light gets low.

Mike J.

[Mike Johnston]

"There is one trivial slip in your list of Kodak's "smaller" formats. You should have listed 110 and omitted 828/Bantam."

dwig,
Thanks for catching that. I'll go change it now.

Mike J.

[Jabog6]

Great post Mike (as I've come to expect).

My beef with the small sensors is the failed promise of smaller and cheaper lenses. This advantage, IMHO, does not really exist. Maybe the OEMs know that the photo equipment market is in much quicker flux than it has ever been, and thus will not assume that they can sell their new lens model for the next 30+ years to recoup the development costs.

So, other than the cost, why not get a larger sensor (ie "full-frame")?

Right now, cost dictates a large difference between APS-C and FF cameras, but this difference won't always be this big.

[Mike Johnston]

"there is, by definition, not such a thing as 'sufficiency' in computers"

Cateto,
Good point, and I'm sure you're right in some respects, possibly many. But in some ways the camera features we are talking about DO admit of points of sufficiency. For instance, dynamic range. We know that we need more than 7 stops. Do we need more than 14? We certainly don't need 30, or 50. So there is a point in there where it will be "good enough" and more will not be needed. I think the same thing is likely of ISO speed. I think we're already approaching saturation on frame rate--my survey a few weeks ago showed only about 11% of our readers consider high frame rate to be important. We didn't ask those 11% how high a rate they need, but presumably it is not infinitely high.

I don't think the megapixel wars will go on forever either, but maybe I'm wrong. There is much more information that could be included in pictures and maybe, once we have it, we will want more. Now, I want more highlight separation and information, for instance. Then there are the cameras that you can focus after you take the picture, which will be very data hungry and processor-intensive when they become viable. So maybe you're right about the size/storage aspect.

Mike J.

[Ken Sky]

Mike
You're right about our impotence to predict the future. So we can only deal with the present and that includes the present state of digital technology. So where are we now? Well we've probably reached the point of sufficiency in the number of pixels. Whether it's 10, 12, 16, or 25 megapixels, we now have a choice based on what we are going to do with the file. The present debate is about the quality of the pixel. Size and spacing of pixels are limited by the physics of light. However improvements in post processing can overcome some shortfalls here. We still haven't reach the point of sufficiency here but a pitch rate of 6 mu seems to be the limit for quality pixels presently. That's it for the sensor. The rest of the future innovations will have to be implementation and feature set. That said, how many cameras can you hang around your neck at one time? There will be a leveling off in sales very soon as the point of sufficiency is fast approaching for 90% of photographers.

[CP]

You make a very good argument and have raised very good points. I would go buy an Olympus E- 3 tomorrow rather than a used 4X5 Sinar F1 (with Fujinon 90mm and 210mm lenses) I have been wishing for a long time.

[Gerry Morgan]

'there is, by definition, not such a thing as "sufficiency" in computers'

There's an important difference between general purpose computers and cameras. Computers are used to run a multitude of applications, while a camera typically only runs the software/firmware that was written for it by the manufacturer. Software developers are constantly writing new software that makes old computers obsolete. But a Canon D30 (for example) remains as capable as it was in 2000.

[michael]

I posted similar postulations on dpreview, and was yelled at. Someone blasted me with "there will never be a 5D replacement." Really? I love the poster who declared "FF will never drop to $1K." Really.

Of course you've read the postings on how 4/3 is dead. Really? I own two 4/3 bodies, and a complement of lenses. Guess I should throw them away. Then the person on dp review who told me I don't know what I'm talking about, when I wrote Moore's Law applies to sensors. Really?

Here's my take.

When, not if, FF drops below $1K for non-pro bodies, cropped sensors will drop below $500, and pricing will be like 35mm film cameras. My Nikon FM10 (really Cosina), cost $265, with the kit lens. That's what is coming for non-pro cropped sensors. The top pro bodies, FF, will be priced like the F6. Sony will heat things up at Photokina. I'm sure you saw their FF mock-up - it looks like the old Nikon F (which I used). And don't forget, Dalsa makes FF sensors. There is going to be a lot of price pressure, once Sony is in the market. And Momma-Miya is selling their bundle for $10K. I can't wait for Photokina. It's going to be fun to watch.

And then there's Olympus. They're going to have to rethink sensor architecture, because they can only pack a limited number of diodes on that small real estate. Do I see Foveon in the wings? I don't know if you saw it, but a Korean photographer compared the DP1 with the Canon 5D, and it was favorable.

What about Canikon, and their refusal to have in-body IS? That's going to cost them in the long run. Sony is going to have it with FF.

Lastly, Polaroid. Alas poor Yorick, I knew him well. Sure, you hate to see it. But how many of those crying in their beer actually used Polaroid products on a regular basis? Not many, if at all.

The relentless march of technology. The future is fun.

[Amin]

One of the knocks on Four Thirds is that the DOF "goes on for miles." Of course this is not true. The DOF differences between APS-C and Four Thirds are rather slight. However, the dearth of primes for the Four Thirds format contributes to the misconception that a shallow DOF is unattainable in that format. Not too many people want to pay for and carry the huge f/2 zooms or the Leicasonic 25/1.4, or the rather bulky third party offerings from Sigma. A couple nice f/2 or faster primes at 25mm and below would go a long way towards clarifying the relatively slight divide between Four Thirds and APS-C.

[Dwight]

Point of sufficiency is already here for some of us. I'm totally happy with 4/3 in every respect except dynamic range.

[John Camp]

The 4/3 may linger, but, I think, only as a fringe system.

Some reasons:

Larger generally *is* better. If you're jammed right up against the noise/pixel barrier (wherever that is, and even if it moves) the larger sensors can play it both ways -- offer slightly more resolution at better ISOs, or the same resolution with a lot better ISO performance, and so on. I'm saying that anything you can do with a small sensor, you can probably do better with a larger one.

Momentum/history. For a top-end 4/3 system to grow, it would have to take away sales from the top-end 35s. Which means people would have to sell off their treasured *systems,* and not just upgrade the camera. For them, buying a 4/3 will be extremely costly; to get people to do that, in large numbers, the 4/3 advantage would have to be huge. I don't see any way it will become huge.

Marketing: Why should the Canon/Nikon Duopoly step back from 35, when they've got big lens systems and accessory systems built around that size? I think they're quite content with the present arrangement, and as long as they can continue to lead in R&D, and to pay the pros to endorse them...why would they risk a change?

Size. One of the sales points for 4/3 was smaller size, and the other day, you wrote an article suggesting that the Olympus was just about the perfect size. Well, guess what? The Nikon D300 is less than an inch longer, less than half an inch taller, and 2/10ths of an inch thicker. When I handled it earlier today, it occurred to me that I wouldn't want it much smaller, at all. I suspect a D300 with the ZF Zeiss primes would be an awesome street camera -- as fast, good and light as any Olympus. And...it's got a bigger sensor.

In your list of sub-35 sizes, you didn't mention the Olympus Pen-F, a half-frame camera, which is a case study in the sufficiency theory. I remember newspaper photographers buying them with the idea that they could shoot longer and cheaper (for the paper) and still have more than "sufficient" quality for newspaper reproduction, because newspapers, after all, are printed on toilet paper. And they were right about all of that. Still, Pen-Fs are collector items, aren't they? People didn't buy them because people didn't buy them.

JC

[Joe Henry]

F2 or 1.4 and full frame. I like the limited depth of field in portraits and it would be a big draw to me. I like a 1:1.5 aspect ratio (6x9 in medium format film was nice) but realize that the shape adds to lens size requirements. Rangefinder is fine...what I really mean is, quiet is nice.
Can't we get a "manual" digital camera? Let me adjust ISO, shutter speed and aperture and put any extra money into sensor refinement.

Joe

[David]

Nice writing.

[Marc Rochkind]

The inconsistency that I see is that, on the one hand you make the point very well that size per se doesn't matter, but rather the qualities that it provides, and if "sufficient" quality can be provided by a sensor of a different size, then we won't care.

But, on the other hand, you make such a big deal about a specific size, 4/3.

What I wish designers (or the marketing dept.) would pay more attention to is raw file size. It keeps going up, by 10 - 20% a year, but the IQ doesn't. (I'm talking about the Canon G9, which I love anyway.)

--Marc

[Ctein]

Dear Mike,

A lovely exposition.

A few observations of my own. Both sub- and super-35 mm formats are admissible to your argument. They were all attempts to tinker with 35 mm, over a 50 year span, and they all failed. Steady and remarkable improvements in film made super-35 mm increasingly irrelevant. 25 years after I started using a Pentax 6x7, 35 mm film was producing higher-quality than 6 X 7 cm format had when I started out. Similarly, remarkable advancements in optics, mechanics, and electronics design made sub-35 mm irrelevant. The Olympus Infinity Stylus put the definitive nail in the subminiature coffin.

In the long term (15 + years) this is all even less predictable. Somewhere between 15 and 50 years from now, photography goes seriously nonlinear. I wouldn't try to predict which of several flavors of nonlinearity it will be, but the technical process will not look like anything you've seen before. At some point lenses will stop being recognizable as such. Oh yes, there will be things that will massage the light from the subject before the sensor collects it, but you will not identify what it is doing as forming an image. The science is in the laboratories today; knowing how it will play out in consumer technology would require a crystal ball.

It is possible to build a camera in which overall exposure, shutter speed, and focus are irrelevant. I don't mean they're automatic; I mean that a single universal exposure works in nearly all situations, from the camera's perspective. You, the human photographer, decide which instances of those qualities you want in the photograph, post-capture. Think of it as RAW format on illegal steroids, 50-feet tall and about to trample Tokyo.

The business with Adobe's computational photography allowing user-definable focus that you featured some months ago is a piece of all of this. Or it may not be. Theirs isn't the only approach. But it's a good example of the uncertainty in the future; in their scheme, a collection of smaller sensors in one camera would work better than one larger sensor.

But everything I'm talking about has something in common-- huge amounts of data (and processing) are involved. So while Cateto is wrong for photography-as-we-know it, where there are pretty clearly some limits it makes no sense to go beyond, photography-as-it's-yet-to-be will dictate data collection capabilities way beyond awesome.

pax / Ctein
[[ Please excuse any word-salad. ViaVoice in training! ]]
=========================================
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photo-repair.com
=========================================

[Patrick Perez]

When Michael Reichmann reviewed the Olympus E-1, he added a section about 4/3s and stated then he didn't think the future was bright for it, largely on the basis of it's being so small. I replied then that semiconductor manufacturing constraints will make the economics of reduced-size the dominant format.

On a silicon wafer of a given area you can only get a finite number of good chips (regardless of type). If there are 5 unrecoverable faults on a wafer (many defects can be 'routed around' in a finished chip) then if the wafer was making 'full frame' chips, thats a substantially higher loss of revenue than if they were lower cost smaller chips. As well, every sensor sold is a potential entree for the manufacturer into a new revenue stream (extra lenses, flashes etc). The economics just demand moving to smaller sensors.

Where does this leave photographers interested in image quality? At present, there is a big enough market force to maintain large sensor production. But as time inevitably improves product quality (which, really, is what Mike's post is about), the absolute difference/advantage, I predict, will erode for all but a miniscule segment of the market. I don't know if 'medium format' or full frame will go away first, but I'm fairly certain we'll never see another 1.3x from Canon again.

As for me, I bought an e-1 despite the 'disadvantage' of it only having 5 instead of 6 MP like it's closest market competitors. But a large part of that putative advantage disappears once one prints onto common size papers. The 4/3 sensor is 13 vs. 14.8 mm on the tall side (or a reduction of 13%) not that big a deal. I liked the handling of the e-1 better than the available Canon and Nikon options. I also liked that the lens line, though far from comprehensive, was entirely of new design, specifically for the new sensor format. They indeed have not proven to be 'small' compared to 35mm, but the size doesn't impact my usage. I'm sure I could be happy with a Pentax or Sony (I too really want IS in camera), but a full frame camera isn't something that appeals to this happy amateur.

Patrick

[Eamon Hickey]

"Nodule: Bigger was better when it came to film, too, but that didn't mean that bigger always won out."

To my ongoing puzzlement, this point seems to have completely escaped the Photoforumosphere (Forumverse? Online CameraChatterarium? Webographic Commentariat? Ok, I give up.)

Medium and large format cameras indisputably produced far higher quality images than 35mm film cameras did but sold in comparatively tiny numbers. Why? Because size, price, and convenience matter, too. That doesn't necessarily mean that 4/3rds sensors will survive over the long term, but if they fail it won't be because bigger always wins.

[Craig Norris]

Mike,

I agree with your sentiments and logic, but would like to suggest a supplementary point of view. The first is about marketing. In a nutshell, I've learned that to be Number One in any market for consumer equipment, a manufacturer must include a minimum amount of mediocrity into the product. If it isn't mediocre enough, it won't be Number One in the market - it's as simple as that. Look through the records to find the name and model number of the Number One selling SLR in history. Then ask yourself "Was that the best camera of it's kind?" The answer (from people like us in this forum) is surely "No." The same is true of motor cars, and most other consumer products.

Secondly, the "logic" behind camera design and consumer choice. The logic is confused because of the unspoken agreement that a single camera should be perfect at any task thrown at it. That unspoken agreement is the cause of the basic problem we all have with cameras. The problem goes away as soon as we consider a particular purpose for the camera. Given a single task for which a camera should be great, it's easy to choose.

We should accept the fact of owning a suite of cameras and lenses, and happily use each of them for the specific purpose for which it was designed.

Our angst comes from the combination of the above two factors. We expect perfection, but the marketing people know that mediocrity is important to sales figures when it comes to a market dominated by consumers who mostly don't understand what they are buying. We camera connoisseurs are in the minority. The camera design isn't addressed towards us. We have to live with that fact. That's life.

[Bert]

Bravo, great post.

As for the future, it is true that we cannot know anything for sure. But I would say that the argument on the points of sufficiency is quite valid. When improvements on any given aspect will no longer reflect in increased sales, the attention of marketeers will turn to something else. They may be dumb, but they have the bottom line to guide them.

This does not mean in any way that the market will stand still. New advances in more-or-less related technologies will act as enabling factors for new evolutionary trends.

For example, there is a tremendous amount of research done on batteries nowadays. Of course, photography is not the reason for this, but it will benefit from the results. This means faster AF, on-board fast wireless, better flashes, etc.

But it also means that we will reach a point where it will become possible to pack real computing power on board of cameras, and not just a select bag of tricks hardcoded in silicon to preserve power.

When this happens, you will see cameras appear with support for user-defined filters, special effects, and what not. You may not want any of this now, but it still will happen. Hopefully, this will open new avenues for creative work, and some such features may very well become the must-have of the future.

And then there's the research in optics... but you get my point. We won't see the day where no further evolution is possible.

[michael]

One more comment:

"The Sony Mavicas, which wrote directly to CDs"

I believe the original Mavica, which cost $1,000, wrote to a floppy disc. They did make some later models which had a mini CD drive, but I forgot the model names. Also, in the additional format department, there was the 126 Instamatic cartridge, which flopped.

Kodak used to make the Retina Reflex, which was on a level with Zeiss. They killed it off, and replaced it with an Instamatic SLR, which also flopped. Kodak keeps shooting itself in the foot.

[Adam Isler]

Mike,
I think your basic point is right here. In the future we'll look back at whatever developed and the view backwards will be different from the view forwards (we see ourselves moving forward into the future; the ancient Greeks, more accurately, saw themselves moving backwards into the future). And it's likely the future won't be something we'd have predicted.

To my mind, however, and, of course, you're right - I don't know - that future is unlikely to be 4/3 (or APS-C) - in fact, as your discussion of post-Bayer suggests, sensor dimensions will probably be irrelevant. But if new technologies that drive superior IQ mean that a point of sufficiency for sensor size has been reached with 4/3, that still wouldn't be a good reason for Nikon, Canon, Sony and Pentax to all switch to 4/3 - it just means we'll stop arguing about sensor size and start arguing about "photonic entanglement DSPs" or whatever the new Black Swan technology is that makes CCD/CMOS sensor sizes irrelevant. It seems unlikely that some new higher IQ technology will actually require 4/3 shaped sensors.

And I think the point from your original post in this thread and your discussion of how far we've come since the D30 are the most important: image quality is so good now and so hard to distinguish just looking at pictures that it hardly matters (though I must admit when I got my Dimage A1 I never imagined I'd need anything better; when I got my Maxxum 7D I was sure even if technology improved I'd never need more; and now I can't imagine anything more I could possibly need than my shiny new Nikon D300 has to offer! :-))

[hugh crawford]

I think that in a few years people are going to wonder why photography spent the first decade of the digital era messing around with cameras with one big slow single lens and one big chip. The current generation of digital cameras is treating sensor chips like film. A few years from now people are going to scratch their heads in wonder at how silly we are.

It would be so much cheaper and simpler to have an array of cellphone quality cameras that would be sharper than is theoretically possible now with large format film. We will be able to have what would effectively be a f/0.1 lens that could have the plane of focus changed after the exposure, or multiple planes of focus or just extreme deep focus. While they are at it by using buffers, deciding on the moment and length of exposure can be done after the fact,

Look up synthetic aperture photography or confocal imaging on google.

People have been playing around with these concepts for years, it's just now that imaging chips and computational power has gotten cheap enough to make it practical.

Someday some non camera* company is going to show up with what looks like flat piece of plastic that makes anything with a lens and a shutter obsolete.Then instead of the pixel races it will be the density of holographic zone plates races.

(* unless it turns out to be Kodak, which is getting lots of expertise in the required technologies and has done such a fine job of wrecking it's camera bussiness that the company hasn't got too much to lose )

[Fazal Majid]

Medium and large formats did not win over 35mm, but on the other hand newer smaller formats than 35mm like half-frame (Olympus Pen), 110, Disk or APS itself all died an ignominious death. So the larger mass-market format did in fact win.

You are assuming that improvements in sensor technology will continue forever (or until 2015, which is effectively forever in the DSLR marketplace). That said, DSLR sensors have reached detective quantum efficiency (DQE) levels that are reaching physical limits. The Leica M8's KAF-1050, for instance, has a DQE of 40% in the green channel, and it isn't even particularly competitive as a low-light camera compared to, say, the Nikon D3. There are also limits as to how much noise digital signal processing can remove.

The fact is, Four-Thirds cameras are already a failure in the marketplace. Canon and Nikon are leading by far (about 40% each, 84% combined), followed by Sony and Pentax/Samsung in hot pursuit (Pentax has about 5%), all firmly in the APS-C or larger camp. Sony may have a chance of reaching Canon or Nikon's market share, but Olympus and Panasonic are rounding errors in the statistics. It's not as if their bodies are particularly compact either compared to a D40 or Rebel XSi, so the size advantages of Four-Thirds are mainly on paper.

The DSLR market is starting to reach its own saturation point, just like point-and-shoots, and there isn't room for either to introduce a product so revolutionary it trumps their minimal channel presence and allows them to take over significant market share.

We'll see if the Sigma DP-1 validates the small-body, large sensor, premium concept just like the Contax T or Nikon 35Ti did in their time, but the market has settled on cheap-and-cheerful pocket digicams for the masses, DSLRs across a wide spectrum of price and quality for enthusiasts and pros.

[Stephen Scharf]

One of the principal issues I have with the 4/3 sensor is that I quite simply do not like the aspect ratio. It is the same aspect ratio on our old montors and TVs. The 35 mm aspect ratio is more amenable to cropping to the 16:9 of HDTVs, which I really prefer. The 4/3 aspect ratio is too "square" for my sensibilities. I find the wider aspect ratios lend themselves to photographs with a compositional "weight" that I much prefer.

[Stephen Gillette]

I am tempted to comment on many TOP posts, but opt out because I simply have little new to contribute beyond the observations already offered.

But I feel compelled to join others in voicing appreciation for Mike's ability to turn a phrase. This post of Mike's covers a lot of ground, but never rambles. The ideas are cogently expressed, and with commanding flair.

Regarding Hugh Crawford's comments, they ring true to me: new technologies have only just begun to impact photography. Todays cameras (and the attitudes of we who use them) are closer to those extant 60 years ago than to where they will be 15 years from now. (I almost wrote "20 years from now" but the paradigm shifts underway will converge sooner, one can predict.)

The great Al Jolson was tagged, midway through the shooting of "The Jazz Singer" in 1927, to introduce the new technology of "talking" pictures. This added dimension to his famous line: "You ain't seen nothin' yet!" So it is today with the state of our art, photography...

[John]

Wow... quite a response to your thoughtful and though-provoking piece Mike. I too think that dynamic range is the bigger issue that size per se and I believe this is one area where digital sensors overall have some distance to go before achieving a level of performance that will pass for sufficiency. This and digital noise associated with smaller pixels are the hurdles that smaller sensors need to overcome.

[Martin Doonan]

It is interesting to read the observations of the photographically literate commenting here. Trouble is, us lot represent (generously) 1% of the camera market.
99% of camera buyers don't even print these days - right now 2MP is sufficient (monitor resolution). Sufficiency in photography may well be driven by TV/monitor formats.
I've not yet met, in the flesh, a camera user who shoots RAW. Most SLR buyers are doing so for the pose value, I reckon.
As for all the other qualities - there are so few people who know what any of this is. When I talking to ordinary people (rather than photo-heads) they are quite happy that what comes out looks like what they saw.
Anyway, who says the future of imaging is 2D. If/when imaging goes holographic, all this talk of flat photos will be seen as laughably quaint.

[John Ono]

Nikon and Sony maybe going full frame. But there are sign that canon is moving from full frame to 1.3 crop factor at lease in 1D mark III. who know which way is the right way?

CCd replced Film and many believe CMOS will replce CCd. Now we have P-MOS and N-MOS which is said to be better in some way over CCD and CMOS. These will somewhat change the landscape on sensor size in some way. Many full-frame users are complainting the buck and weight as they get older and wiser.

What about Single-Pixel Camera that is currently under development in Rice University in Texas? http://www.dsp.ece.rice.edu/cs/cscamera/
The technology now is parallel to what CCD was when it was first use as imaging device. The first Kodak CCD camera took 100 x 100 pixels photo and 30mins to process a picture.

Who can tell what this single-pixel camera will become 10 to 15 years from now? When the time come for this new technology to take over from the current imaging sensor technology. Will we still ask what is the size of your sensor or how many megalpixels is your camera? The answer will be "as you like it sir, I will set my camera it to match whatever you want".

[JeffS]

To reinforce the aspect ratio issue with 4/3: As it is now, we have a choice between the sensor dimensions of 4/3 (1/1.3) and APS-C (1/1.5). If I truly like the 4/3 format size, then why not buy a camera with a slightly wider sensor and have the freedom to crop either way, especially when there is an advantage in dynamic range and no price penalty?

[Rib Riley]

your four thirds sensor size is erroneous
the size you quote is the published image area. You should quote like with like sizes as clearly the image area is less than the sensor size.
The most commonly accepted size for four thirds sensor is 18x13.5mm, however the E1 and E3 have larger sensors. The E1 sensor size is published in Kodak's pdf file KAF-8300LongSpec.pdf, and KAFproductSummary1.1.pdf
the E1 sensor is 19.7x15.04mm

[E]

FF sensors will not become cheap in the near future ($1K FF cameras will not happen, atleast not in the near future, maybe in 5-10 years, definitely not before), because sensor is a fixed-size chip (unlike computer-processors, that rapidly become smaller, hence use less silicon, hence become cheaper yet more powerful), because the sensor is a fixed-size chip it will use the same amount of silicon and thus no advancement in sensortechnology is going to make it cheaper....
The only thing that can make it cheaper is a radical advancement in silicon processing technology, and those are slow and take a looong time...
Also like patrick pointed out above, flaws in the silicon are way more costly for larger chips, meaning that from fixed-size silicon wafer you will get higher % of working smaller chips, that again will mean that the overall price of larger sensors is even higher than simply due to the cost of the silicon...

Now some personal thoughts, for me, it would seem logical that one day there will be only FF and 4/3 dSLR's left. FF (still expencive) for shooting "black cats in a coalceller" so to speak (and perhaps for higher than usual resolution), and cheaper 4/3 for lesser overall weight (cameras are not that different, but the overall weight of cameras and lenses combined in ones backpack is different enough) and adequate quality for most photographers needs...
APS-c is a strange in-between, it has to use heavy FF glass most of time (there are not that many pro-quality APS-c lenses around), yet it is not that much better than 4/3...so you'll end up lugging a lot of unnecessary weight around for nothing (and if you do use APS-c glass, should you want to go FF, you will need a different set of lenses, hence it will be as expencive as switching systems)...so for me, APS-c cameras has all the disadvantages of both the FF and 4/3 cameras without none of the advantages...
But that's just me...

[Bernard]

I for one hope that 4/3 does not go away. I would never go back to shooting longer lenses with a full frame camera.

A Sigma 150-400 lens now gives me incredible range and quality that's good enough for a 16x20, all in the form factor of an old-style 4.0 80-200.
Similarly, a 2.0/150 gives me the same results as an old-style 2.8/300. I know which one I would rather carry all day.

The latest Leica/Panasonic zoom gives me a high quality 28-300 that is the same size as a FF 28-70.

I'm sure that a FF camera could give me prints that go to 20x24 instead of 16x20, or a slightly higher useable ISO (at the cost of slower, non-stabilized lenses), but why would I or the average photographer bother? The best camera (or lens) is the one that you have with you, not the one that you left at home because it's too heavy.

[jeff hartge]

Sorry to be the jerk in all of this...

But hasn't it occurred to anyone that the current state of affairs in digital photography is a short-term aberration?

There has been a significant amount of money pouring into the DSLR market since the beginning of the decade.

However, haven't you all noticed that print media is fast disappearing. Further, notice how much of CNN, ESPN, etc. have become videos? (Who needs a bunch of still photos when you can get the original video of Clemen's testimony?) Even the New York Times has gotten into the act!

That's not to say that news photography is going to disappear. I just think that most of the "news" corporations are going to wake up one day and ask themselves if they really need hundreds of the latest camera for all of their photographers. Maybe some of their photographers can make do with older equipment.

How about wedding photographers? Maybe there will come a point when they ask if the new $5000 camera is really going to lead to an improvement in the bottom line.

Finally, what about all of the "amateur/hobbiest"? DSLRs are new and novel in this decade. What new "fad/hobby" will command their attention in the next decade? Is it possible that they will simply say, what I have is good enough photographically. Maybe I should put the extra money into my retirement accounts.

Anyway...

My point is that I don't believe that DSLRs are going to sell as well in the next decade. Without all of this extra money pouring in, I think the innovations will slow significantly.

This isn't necessarily a bad thing.

I really don't believe most of the people here have utilized the equipment they currently have to the limits of its capabilities. A slow down in camera releases could be a good thing. People can spend more time taking photographs with what they have instead of wasting time fantasizing and wishing for the next big thing!

[Michael Perham]

Everything Mike says about distinguishing between 4/3 and APS-C can be said about the difference between APS-C and FF. So my money is on the APS-C size and I do see an advantage in smaller size lenses. My 50-135 f2.8 is a heck of a lot smaller than the 70-200 f2.8 on my 35mm analogue camera.

Mike P.

[Tom V]

It seems to me that the 4/3 sensor is ideal for DMD compacts.

One of my dream cameras is a camera using 5M pixel 4/3 sensor in a body and lens modeled on the Olympus Pen-EE half frame camera. That light meter surrounding the lens? Its now a multi-colored LED array "ring-flash"!

Oly are you listening?

[Hendrik Bosman]

What a lively subject ;-))

Mike you've done very well. Congratulations.

I`ve been using Olympus for 34 years now. Starting wit an OM-1. I currently own an E-1, E300 & E-3. Personally I cannot see the difference when I take a specific photo with either one of Olympus, Nikon or Canon. So a lot of the arguments are just that..."Arguments" for the sake of being argumentative. I'm obviously a Olympus fan just as somebody else is either a Nikon or Canon fan and each of us will always defend our choice. Happy snappy or should I say "Enjoy your photography" regardless.

[Enzo]

What gets muddy in these arguments is the "general" consumer vs. the pro and advanced hobbist.

The general consumer does not want to lug around a 5lb camera and many do not want the limiations of a point-and-shoot. The smaller lighter cameras win out in this respect. I meet parents every week who want to shoot their kids sporting events and drudge the fact that the guy at the camera store sold him a boat anchor and find themselves either not bringing the camera or setteling for less quality images but easier to take around P&S.

When I show them my E-510, they are amazed. I have turned many converts to an Olympus system. A recent parent even baked me a cake to thank me -- LOL!

There are more general consumers than there are pro's. I cannot believe that the industry will go to single format anytime soon.

Silicon to make chips is NOT cheap and thanks to a commodity boom, it is getting more expensive to make. Bigger sensors means lower yields per wafer which means higher cost. While they can demand 5k from a pro, not so from anyone who is not a pro. Smaller sensors have a higher yield so they can be cheaper. Note that Intel and AMD are going to smaller and smaller chips partly for power consumption reasons but more due to cost.

The argument to size is more unclear when it comes to lenses. The FF crowd rightfully argues that some of the best 4/3 lenses weigh as much as the FF lenses and are about the same cost. Here, I believe we have a physics problem that cannot be solved. There is only so much that can be done to attract and bend light. I do believe we are going to see a move to in camera IS but lenses are going to stay the same.

Is bigger better? No, but in the world of engineering bigger is much easier to optimize for a specific quality attribute. In the case of ISO/Low light attributes, it may be that it is currently easier with larger photosites. However, as we have seen Kodak recently announced a process to do the same in a smaller package first stated to be in cell phones and then move to cameras. So, there could be a day when 4/3 sensor could have the same low light performance as FF. Would manufactures then still use larger format sensors that provide lower silicon wafer yields, or would they go to higher yielding wafer if the attributes are the same?

As you said, nobody knows -- the punduits at this point on either side of the argument are both wrong until proven otherwise -- one thing is constant, things change.

[Patrick Perez]

I appreciate the agreement E gave me re: sensor manufacture economics (I was tired of the voices in my head being the only nes who did).

I'd like to clarify that in my original comment to Mike's posting, that although I was talking about 4/3, the points I make apply equally to APS-C (and I think that is what Mike was getting at originally; 4/3 and APS-C are effectively equivalent). I wasn't making claims about 4/3 being a better development strategy for manufacturers vs. APS-C, just that the two seen as representing 'reduced size' of a similar format together represent where sensors will go long term.

Patrick

[Mikefellh]

"It's not as if their bodies are particularly compact either compared to a D40 or Rebel XSi, so the size advantages of Four-Thirds are mainly on paper." - Fazal Majid

First off look at your hands, camera bodies of a certain size feel comfortable in your hands, and if it's made too big or too small the camera just doesn't feel comfortable and you can't hold it steady...cameras will continue to be made a certain size.

Second, dSLRs can be made smaller for those with smaller hands, the E-400/410 are proof of this.

Last, the benefits of the FourThird format is not in smaller bodies, but smaller lenses...look at the size of the Zuiko 300mm f/2.8 and compare it with the equivalent Canon 135 format 600mm f/2.8 lens, if there is one (I couldn't find one).

[conrad]

Obviously sensor development will be an on going thing for many years to come, but why do we end up with a whole new camera for every new sensor. Can't the electronics be made so that it can easily be changed for the latest version, keeping the same body and lenses?

An easy target for this idea would be Leica. Those guys make a nice camera with superb lenses that just cries out for a removable electronics package to be slipped into it. Instead of coughing up a ridiculous amount of money for a complete camera you would instead buy a body and lenses and chose the sensor/electronics that suit your goals.

I am as keen as anyone for new and better stuff, but alot of my favourite lenses are sitting around collecting dust while camera manufacturerers spend alot of time and money dressing up their latest sensor in fancy new bodies with bad viewfinders and a lot of extremely useless features.

Am I the only one who feels this way?

[Cateto]

Mike:

"But in some ways the camera features we are talking about DO admit of points of sufficiency. "

I agree. But the key point IMO is not that, actually, most of the current dSLR are close (or into) the sufficiency point. And yet, customers play the game as if they were not at that point.

In other words: I (as many others) am convinced that current dSLR outperform the abilities of 99.99 % of their users (including me: I have a K10D and I am sure I can't use its countless possibilites). But still, the customers look forward to further improvements, EVEN if they are not enabled to use the full possibilities of their current dSLR: they always want more.

Call it marketing, call it measurebation, call it whatever: fact is that people start dreaming in the next model, with more megapixels, more fps, better low light performance, etc etc. It does not matter what they ACTUALLY can use: what matters is the theoretical possibilities, although in most cases, they will remain untouched (with the small exception of a few test in the first weeks after the camera is purchased).

In short: I have the feeling that it is less about the reality, and more about the always-better-although-I-won't-ever-care-to-use-it. Which is EXACTLY which happens in computers: they get speedier every month, yet people does the old, boring, and simple text processing, and a few users (1% maybe) are actually power users.

We all are into a course game of electronics. The market is power-driven (or money-driven, if you will), not reality-driven.

[Ken]

Just to clarify the Canon 1.3x sensor cameras: the original Canon EOS 1D had a 1.3x sensor back in 2001. The 1D Mark II, the 1D Mark IIN, and now the 1D Mark III all have the same size chip, called APS-H.

The 1Ds series of cameras have the "full frame" sensors.

Several years ago a Canon official suggested that these two lines would converge at some point, but it hasn't happened yet. The 1D series generally has half the number of megapixels and about twice the shooting speed as the 1Ds series cameras.

So Canon isn't "moving into" 1.3x sensors. That's where the pro Canons started.

[Ctein]

Dear E,

I don't agree with your analysis of cost trends. First,computer chips have not gotten smaller-- in fact,the long trend has been towards bigger chips. The gates on the chips have gotten smaller.

Second, the cost of sensor chips is dominated by yield, speed of fabrication and cost of fabrication (material costs are insignificant at the present time). All of these undergo steady improvement and have not topped out. It didn't require radical change for there to have been massive price drops over 10 years for chips of constant area. We've not hit the wall.

One may not expect sensors prices to drop by a factor of two every 18 months as they do for computer components. But one may expect an ongoing considerable drop in prices.

That said, your underlying thesis is correct -- sensor costs will be a significant factor in total camera price for some time.

pax / Ctein

[Erez]

Very well written.
I would like to add that the price (or possibly the price difference) is a major contribution to the perceived value of the product.
You will rarely see a bad review for a more expensive product just because if "it's more expensive" it must be better. This is true for almost all consumer goods.
In reality the "normal user" cannot tell the difference between different lenses/cameras, once it's printed nicely. Already we magnify our pictures to the "pixel level" to find the faults and compare products.
No one seems to agree to the basic truth that it's the photographer not the camera that take good photos.
My two cents.

[John Camp]

The whole concept of "sufficiency" is intriguing. In computers, for example, I've often thought that there might be a speed/purchase barrier at the point where computers could achieve "real time" generation of fully realistic action figures for gamers. We're very close to that now. Once we get there, why would mass-consumers (as opposed to scientific users, or model-makers) want more processing speed?

With photography, I think there's a point where additional resolution becomes non-economic, and I'd suggest that point is where typical home high-resolution art photographs become big enough to be displayed in-home as paintings are now. That would mean high-resolution prints, after some cropping of the RAW image, at 24x36, or even 30x45. These would be prints that could withstand close inspection without pixel noise becoming obtrusive. We're close, but not there yet. Maybe two or three more generations of high-end DSLRs?

JC

[Donald Ingram]

All I can say is that I currently find the small 4/3 sensor in my E3 'sufficiently' good to get on with taking the pictures I want.
It's far better at low light indoors at ISO 800, with the 25mm f/1.4 Summilux than my Leica M6 with 50 f/1.4 and Fuji 800 neg film.

Overexpose latitude may not in the same class as NPH/NPZ film but seems to keeping up with similar pixel count APS sensors.

[ET]

Four-Thirds is everything you say it is. But the problem I see with it, is that it's an option we don't need, and won't miss.