The Point of Sufficiency

I thought for sure I'd already written a post called "The Point of Sufficiency," but a search doesn't turn it up, and I can't seem to find it under another name. So I thought perhaps I should just write it again, at the risk of repeating myself. This was triggered by one of Eolake's comments to the "High ISO High Noon" post, in which he said that when it comes to ISO speed, we always want a little more, and then a little more.

That's true—until it's not.

The idea is one that I've at least mentioned many times, for instance here. It's that in the history of photographic equipment and materials there have been various "competitions" or "races" to see which company can provide the most/fastest/best, but then in most cases a point of sufficiency is reached and customers decline to pay for further improvements. 

This happened with film formats (film basically got down to 35mm size, which was small and convenient enough—further efforts to popularize still smaller formats all failed); lens speed (ƒ/1.4 was deemed sufficient by the market, with further increases in maximum aperture being rare and not especially popular*); telephoto reach (the "Olympia Sonnar" for the 1936 Olympics was a widely admired and popular achievement, but now, 1000mm and 1200mm lenses are mere curiosities); shutter speed (consider that the Nikon N8008's 1/8000th shutter speed was a big selling point, but the Minolta 9xi's 1/12000th was met with a yawn); fine grain in B&W films (Panatomic-X was enough; only a few fanatics followed the Tech Pan craze); reducing the size of SLRs was a brief race in the '70s (the Pentax ME was small enough); even build quality in cameras and lenses (the current market supports—barely—one carriage-trade brand. Beyond that, better build quality has not been rewarded with sales). 

And so on.

Note that this is not a technical matter—it's a marketing matter. It has to do with meeting the need for what people want for as long as enough of us are willing to pay for it. The trend is for the race to continue among the manufacturers for as long as the buying public is rewarding the improvements. But sooner or later the buying public has what it needs, and has had enough, and stops wanting to pay for further improvements, and the manufacturers (save perhaps for one or two niche products in each case) basically give up the chase, and go try to find some other idea that will sell.

My prediction (which I've mentioned many times) is that this will happen with ISO speed in DSLRs, too, and with megapixels in Bayer-array sensors of the current type**. I don't know where the end point of either will turn out to be, and I don't know whether we have arrived at those points yet (is 24 MP enough? Is ISO 12,800 enough?), but history predicts that sooner or later, you will have enough and won't want to pay more for more.

Mike

*People will trot out the example of the Noctilux or the Noct-Nikkor to refute me, but those never sold in significant numbers. One could even argue that with rangefinders, ƒ/2 was the point of sufficiency: the Leica 50mm Summicron always outsold the 50mm Summilux greatly, at least until Leica became primarily a status symbol. Before the introduction of the current Summilux-M, Leica decided that a new 50mm Summicron would have to be priced beyond even their market, but that a new 50mm Summilux could be supported. That's why the 50mm Summilux got a remake but the old Mandler Summicron soldiers on.

**The size of sensors is not just a matter of marketing, because there's a formidable technical stopping-point at 24x36mm that has to do with 35mm-legacy camera and lens design. So far, customers—core customers, at least, meaning enthusiasts—seem to find great appeal in larger sensors, up to at least 24x36mm. The next couple of years should indicate whether sensors larger than that are going to be rewarded with sales sufficient to support their continued development.

Send this post to a friend

Featured Comment by Eolake: "That's a very good point, and I'm sure you're right. One factor though, is that with many of the examples you give, further gain got punished by excessive size or price of the product. With megapixels or speed, this does not have to be the case. But of course, it will still be a case at some point of: how many people will need cameras which can make pictures that can be pin-sharp even blown up to wall size? Personally, with megapixels, I feel that 12 is a sweet spot. I do have a 21MP camera, but it's heavy and I only use it if I go for pictures I think I may want to blow up really big for my wall. With speed...this can go on a while yet, I won't mind. I'd like to never have to worry about shutter speed."

Comments

[Winder]

I think the article fails to mention that cost is a huge factor in when is enough enough. No, the Noctilux has not sold in vast quantities, but at the price it never will. If the Noctilux was a $1,000 lens and not a $10,000 I think you would see a huge difference it demand.

It is a simple matter of diminishing returns. More is always better as long as price does not increase. If two cameras cost the same, but one has twice the ISO range, then people will buy the camera with the better ISO range even if they never need 12,000 or 100,000 ISO.

Camera equipment (like everything else) is still subject to the laws of supply and demand.

[Mike Johnston]

It's true in many cases that "more costs more," but then, it did *before* the point of sufficiency was reached, too--but people still bought. People wanted faster 50mm's and longer telephotos until they'd just had enough, and price might have had a lot to do with that. But in some cases it's not true--a finder-grained film doesn't necessarily cost much more, nor does thicker enlarging paper, or a smaller SLR, or a smaller-than-35mm film format.

Mike

[Michael]

5 years or so ago I wondered aloud why They didn't make a EVF P&S with interchangeable lenses. Sounds like the Micro 4/3 system.

Next big change in cameras will be not be greater generalization but mass customization. Retro thinking. Make the bodies and sensors separate. Make a nice general sensor (100-800) but also optimized ones (25-400 or 1600-32000), infrared, UV (with special lenses), cooled, and B&W (pan and ortho). Think of the sharpness with no Bayer filter and 24Mp!

Canon made a special 20D for astro-photographers why not special cameras for everybody?

[Timprov]

Megapixels at least are going to hit a big boundary soon - the current top-end 35mm-equivalent dSLR sensors are starting to outresolve their lens lineups. I think it's unlikely very many people want significantly more megapixels at the cost of having to engineer an entire new lens lineup to use them.

I'm not sure there's a natural inflection point where the cost of higher ISOs will rise faster than their usefulness, other than physical light-level boundaries.

[Player]

Ahhh, maybe that's what Pentax's multi-colored candy store cameras are all about. The company may be trying to tell us that they believe that the "Point of Sufficiency" has been reached with DSLRs. I mean if the cameras were really lacking in important functional areas do you think they would even think about giving them wacky colors?

[robert e]

Is this bell curve week on TOP?

The phenomenon is actually accelerating. As photographic technology advances on many fronts, the benefits of pushing any one front diminishes even further. That is, an f/1.0 lens might have had significant utility when ISO 100 was considered fast. But in an age of high quality ISO 800 with image stabilization, its potential market is considerably smaller.

It works from the bottom, too. I suspect that cell phone cameras achieving mediocrity is what finally pushed major manufacturers to develop high quality compacts in earnest.* Miniature cameras with mediocre picture quality is clearly a doomed market. Probably no coincidence that the producers most dependent on that market are leading the way on high-q compacts.

*I don't mean to suggest that Mike's DMD crusade did not importantly lay the groundwork and establish the parameters, just that it alone wasn't sufficient.

[John Camp]

I think it's driven by use as well as price. The early Leicas provided prints that were good enough for almost any publication purpose, at a time when magazines and newspapers were in their golden age (30s, 40s and 50s.) 4x5 was ultimately better, for art purposes, but not nearly as usable or portable, and provided no advantage in publication until high-speed presses got better. At that point, medium-format was good enough for the best magazines, while small format was still good enough for most purposes. But half-frame wasn't...just not quite. It was fine for snapshots and so on, but you wouldn't want to shoot the front of the Daily News with a half-frame, especially when they were often severely cropped and *then* blown up.

I'm coming to the conclusion that m4/3 and APS-C, in very small cameras like the G1/EP1/GF1 could be playing the role of Leica against the heavy, over-sized beasts like the top-end Canons and Nikons. The big do-everything cameras are heavier than a Speed Graphic, when you bolt on an f2.8 zoom, and are almost as large in frontal area (conspicuousness.)

An m4/3 will work for most magazines that don't need MF, and all newspapers, and all video formats (TV, monitors, etc.) It's these sensors that are critical, as they can't yet touch, say, a D3s in terms of ISO. When they can turn in a good clean ISO 3200, with a price of less than $1000, and especially if an independent lens-maker like Zeiss (or Leica) begins making glass for them, then we may see a widespread shift -- because the price is right, the flexibility is right, and they are "good enough" for almost everything.

[john robison]

Your point on f2 being sufficient with 35mm RF was also true in SLR's. In the mid 70's we sold at least 3/4 of the then popular mid range SLR's with the slower normal lens, ie f2~1.7, the rest went across the counter with 1.4's. As an aside, I really think it's a shame to only offer DSLR's with the stone slow kit zooms and cause the customer to pay an excessive premium for a fast standard angle prime. But then I'm sure that's a minority opinion among potential customers.

[Kent]

This is a fascinating subject.

Sometimes I can't help wondering what it is that people want out of their photography (I'll probably never know, because it's sort of like asking why people like different types of music). But it is clear that technically things can only go so far. The end product of photography -- the photograph -- will never be an exact reproduction of the original scene, and not only because of the huge spatial distortion that occurs in the transition from three dimensions to two. No matter how much detail is reproduced, or now noise-free that detail is, there's still a huge amount of information missing. So there must be, as Mike points out, a point of sufficiency ... or perhaps "balance." How much technical quality is sufficient for a photograph to be the most well-balanced interpretation of reality it can be?

An issue of major importance for the future of photography, methinks.

[Alex Vesey]

This point holds for some kinds of software: the site for PC magazine has an article by a reviewer who will not not upgrade. MSWord was a fine word processor with version 5--since then it has bloated to be something for everyone. The willingness gets a rejuvenation with a paradigm shift, a new device/product that reconfigures perceptions of value and quality enough to start a whole new purchasing trend. Kind of like escalators in a high building: each one can only go so far; continued financial success comes with building stacks of them that connect with little walking time (or a stroll through a really interesting display of other products.)

[ObiJohn]

It's all about price versus performance.

Who would have thought that we'd be able to buy a 35mm-sized camera today that can make photographs that are superior in quality to a medium-format film camera made just a decade ago? Today's digicams make MUCH better photographs than our grandfather's 120 roll film cameras.

Now, if a Canon G10 sold for $100,000 then no one would buy one... and that is what one might have cost a decade ago. Similarly, there's no reason a 50/1.4 costs more to make than a 50/2.8, or a 300/2.8 cost more than a 300/4. The cost of materials is within a dollar (probably within a few cents), and the manufacturing costs are about the same.

The manufacturers have to keep offering us something better tomorrow that costs the same as what was 'good enough' today. Otherwise, we won't be buying. And, with the progress computerization has brought to design and manufacturing, there's no reason that a camera line with all of the qualities and capabilities of the Leica M9 and its lenses couldn't be made for 1/10th the cost... and sold for 1/10th the price. However, what something costs to make seldom has much influence on what it costs to buy.

[Mike Johnston]

"I mean if the cameras were really lacking in important functional areas do you think they would even think about giving them wacky colors?"

Well...you may be right of course, but I don't think that's the right way to look at it. I think what Pentax is doing is what most niche players have to do to survive--it is looking at an underserved market and trying to figure out how to appeal to that market. In other words, the marketers started out by noting that most DSLRs are purchased by men, and wondering what "hook" might interest more women. I don't think it has anything to do with questions of functionality, which Pentax obviously believes it addresses as well as any other maker.

It's also about differentiation. Those who have decided to buy an entry-level DSLR anyway have a wealth of entry-level DSLRs to choose from. Pentax's problem in that context is that they too often come in second or worse in the buyer's final decision of what to buy. If they can turn that around--get the buyer to say, "well, these choices are so much alike that it's hard to make a decision, but I love that I can buy a green one of this brand"--then suddenly they move to #1 in many buyers' choice process.

Could work, too. At least until everybody else starts offering colors in entry-level DSLRs.

Mike

[Dan L.]

I would argue that ISO is a more transparent quality than shutter speed, aperture, or maybe even pixel count — at least on cheaper cameras. If the image quality at ISO 200 and 1600 are essentially the same, then you can get away with using slower lenses, which can be either cheaper or smaller. You could also get away with smaller flashes (neither flash size or power have changed for a very long time). The essentials of optics haven't changed much in the digital age, but what we can do with the same optics has been revolutionized. Increasing megapixel count, sensor size, lens speed, shutter speed, flash power, battery life, etc., all require tradeoffs of one kind or another, but high-ISO capabilities seem to me to have relatively fewer technical disadvantages.

Where high-ISO digital cameras will get exciting will be at the level of small dslrs and digicams, where size has been directly proportional to image and imaging quality. What impresses me most about the D3s announcement is that this kind of improvement is even possible, not that Nikon's marketers have decided that there is a (potential) demand for ISO 100,000.

[Paul Amyes]

There comes a point when looking at equipment when you say ok for x amount of dollars I can have a product that meets my needs 90% of the time, spending a multiple of x to buy a feature you'll rarely use becomes a luxury you can't afford or justify.

Case in point is that I didn't upgrade my 5d to the new version. The reason why, 12 Mp gets me a double page spread which is the biggest my shots ever go, why pay another A$5 K for redundant pixels? I don't need movie mode and I haven't hit an ISO ceiling. I would like an upgrade, but I don't need one.

[EZ]

With respect to the (**) - size of sensor. Is there anything that really constrains us to 24x36mm? That's a historical number. Yes, Mike has a point with respect to the lenses, but they already have the ability to "cover" a circle of diameter greater than 43mm. So why not a square sensor, 31x31mm? Or even better, since hexagons will fill a circular sensor fabricator wafer better, an hexagonal sensor that covers 83% of the image circle? That would make them cheaper, right? Just some provocative thoughts...

[bobdales]

I'm sufficient with a GF1 that has a built-in viewfinder and an RF mode the emulates or replicates a split-image RF.
Now how hard is that?

[Miserere]

One problem I see with sufficiency is how camera makers can correctly assess whether it's been reached or not. Let me explain: As ISO performance improves, pixel counts are increasing, at least for some cameras.

The Canon 7D has better high ISO performance than the 50D, and also has more pixels. If I were a Canon shooter, I would have to eat a 3MP increase in order to get higher ISO. So what message does Canon receive from my purchase? That I want more pixels, more ISO, or both? And I'm sure there will be some people buying that same 7D because they want more resolution and couldn't care less for the ISO.

Right now I shoot a 10MP Pentax K10D. If I want better/higher ISO I need to purchase the 14.5MP K20D or K-7. I don't want more pixels, just more ISO, and I am reluctant to spend the money on a new camera when the one I have does everything I need it to...so long as I don't need more than ISO 1600.

Once again, I request modular cameras; at the very least with interchangeable sensors. I'm sure this will happen at some point, but I'm not holding my breath.

[Tom K.]

The camera feature I want to see test your thesis here is "image stabilization". Could advances in that technology render the tripod useless? Hmmmmm.......

[Ulf Ågerup]

I believe the wow-factor of ISO and MP is wearing off. I think the next innovation will be cameras that capture HDR and many focus points at once. This would allow me to choose my exposure and my focus in post. This would be the end result of a long trend. I see it starting with cameras capturing greater dynamic range. The focus merge function is already there in CS4 so I do not think it should be impossible to get this function in-camera at some point.

[Thomas Johns]

All well and true, at least from a photographic perspective. Optical and mechanical performance reached maturity with (insert your favorite analog camera here.) But there's a 'computer' in each of those digital cameras, and that's a different paradigm. Personally, I didn't keep pace with Moore's Law, but I've had several leaps and bounds since that Tandy 1000 in 1982. The pace has slowed lately and I'm sure that will happen in photography with (insert your favorite digital camera here.)

[Craig Arnold]

Some people choose at a feature set, some choose at a price point.

I have found with the move from a 5D to a 5DMkII that my technique is now wanting. The extra pixels are showing me up. Of course since I mostly print at A4 size it doesn't matter too much, I am now operating far inside the envelope of what is possible with my equipment.

I have been operating at a price point until now; a £2,000 camera body every 3 years being what I can comfortably afford. But with the advances being made I am now seriously thinking of starting to operate at a feature set. I think the GF1 is probably good enough for my needs.

In fact I am not quality-limited by 35mm film, with my Zeiss Ikon and Coolscan V I can make the prints I want. If I can just find a way to get a digital equivalent I will be very happy. I think the GF1 may be the digital equivalent. I would love an M9, but the price puts it a bit out of reach emotionally.

[Craig]

The theory of sufficiency assumes there is a trade-off between performance and cost. But with camera sensors, history has shown a steady increase in performance with no increase in cost. Today's 5DII provides significant performance increase over the original iDs, at one third the price.

Perhaps there are two questions. How much ISO would you like? And, how much ISO are you willing to pay for? For me, the answers are not the same. I'll pay for 3200, but I'd sure like 100,000 or more if it gives me comparable results for the same price.

I think we'll continue to see sensors grow in pixel count and performance, and prices will continue to go down.

Maybe we don't need more than 20 mega-pixels for resolution, but why limit our thinking? Why not a sensor with 60 mega-pixels in an HDR array capable of capturing a 20 stop dynamic range? Or, maybe a pseudo-random arrangement of pixels that allows designers to eliminate the anti-aliasing filter.

The camera industry is full of briliant engineers and marketers to develop and sell us all sorts of improvements. I can't wait to see what they think of next.

[Zoltan]

While you may not have written a post bearing the same title, you have used the phrase "points of sufficiency" in your post entitled "Sensor Sizes Part II". You published it on Saturday, 23 February 2008.

[Thom Hogan]

Adam Osborne used to tell me "adequacy is sufficient, all else is superfluous." This was his mantra for computer design. And he was wrong. And right. You are wrong. And right.

At the commodity level, yes, sufficiency is enough. But tech doesn't respect commodities, it has this way of breaking them to establish new ones.

What you're arguing can easily be examined in the personal computer market. Processor speed is a bit like pixels: more is better. Until it isn't for the masses, at which time you establish a commodity baseline and usually a new product line (most recently, netbooks) and the R&D switches to find a new path towards "more."

The underlying decision point is a simple one: is there a bigger profit providing adequate/sufficient products to the masses of premium products to smaller groups? Each time this happened in computers, we got companies pursuing different paths. It even split processor development (e.g. Celeron).

But the smart never bet against the relentless progress of tech. Whether it's a disruptive process (post-Bayer sensors) or just on-going push on the current curve (e.g. the D3s sensor) the name of the game in tech is "price reset." When the market becomes commodity, everything is driven by taking price out of every place you can (mass produced chips, no manuals, no CDs, no straps, no caps, etc.), as pricing in commodity markets is always low and is driven lower and lower. Tech progress always finds a way to reset prices higher. Ultimately, all companies need that price reset, even the commodity ones that have to wait for the new tech to become commmodity.

[Joseph E. Reid]

Maybe the ISO race ends when we don't need flash anymore for snapshots in dark rooms. In the consumer world who really likes the look of washed out, flat flash images or the necessity of blinding the subjects momentarily? Image stabilization isn't the answer because IS does nothing for subject motion. In the professional world I can imagine situations none of us has yet thought of in which enormously high ISO will be absolutely essential to showing us the world in a way we have not yet seen it (and isn't that one of the prime reasons we have photography?). Faster and faster lenses are not the answer for low-light shooting. Optics work against that -- the larger the aperture, the smaller the depth of focus, the higher the lens aberrations. Shutter speed can only go so low before subject movement obliterates the photographer's intent. High ISO is the answer for shooting in the dark and cameras like the D3s are only beginning to get us there. Add that to software tricks like HDR and multiple focus points and subtract the need for flash and you have something that may give us a new way of seeing.

[Boris Liberman]

Mike, I generally agree with what you said. It makes perfect sense to me, as for example, I forwent upgrade from K10D to K20D and rather reluctantly upgraded to K7 mostly because it became rather cumbersome to share a camera with my daughter.

I think that what has to be taken into account is also impact or effect that marketing speak has on the customer masses. For knowledgeable people like you marketing speak may not work as you see beyond the hype. But great many people don't.

A friend of mine asked for advice as to which camera/lens to buy (DSLR with video). I suggested Canon 500D and Tamron 17-50/2.8 (not knowing about IS version at the time) so as to get him a good starting kit both from price and performance perspectives. He ended up with Canon 17-85/4-5.6 because USM and IS (mostly IS). But almost immediately he admitted that he wanted faster optics. I think it is good example of him preferring to listen to marketing speak and not to me. No blame or anything here, merely pointing out the fact.

I think the story is also about what manufacturers can actually produce and what their marketing department wish to be selling.

[Michael]

Most people feel a Canon Rebel with a kit lens is sufficient.

Megapixels will never stop increasing until Moore's law stops. There's just no downside to adding more megapixels if data storage and computer speed keep increasing.

[Tony Gamble]

So far much of this discussion has been directed to the use of the pro or semi-pro.

Most cameras are bought by amateurs - to take pictures at parties and on holiday. Plus friends and family around the home.

I forecast the most important move will be when a P&S has enough ASA potential to make a flash not necessary.

A great number of amateurs waste their flash shots anyway. See the number of flashes in use at a daylight football match. Or the number in use in a room far larger than the reach of the flash. Or the times you have not used the flash at it would be intrusive.

That will be the time when we (so called) pros can start pointing out to our chums how to make best use of the available light - and start taking pictures rather than snaps.

Forget pixel counting. Pray that the marketeers start to covet speed.

Tony

[Martin Doonan]

The comment by Joseph E Reid made me think this way - maybe all those who need an optical VF also need high ISO. The high ISO negates the flash, removes it from the camera, leaving room for a good VF.
I don't know, but you can never second-guess the other opportunities that get opened by a given technical development.

[Michael Barkowski]

Perhaps there's an element of whimsy in what the great mass of hobbyists are looking for each season, so that the focus is always the most interesting, sexy, or surprising feature. The rules of the race or competition are written by the participants, not the audience. I overheard someone at a camera store looking for a camera with a certain range of megapixels; that was the whole requirement list. The issue of poor DR performance with high pixel density was forefront in the blogosphere at the time. Meanwhile, great advances in other technical areas can be made without much fanfare. How many customers come with a shopping list of: strong built-in flash, best white-balance accuracy, LCD speed & quality, autofocus accuracy, compact lens distortion, power zoom speed, etc? If they were going by their real needs, they would, but most only think about the requirements that marketing has highlighted for them.

[Jon Bloom]

One of the principal things I see driving DSLR purchases by middle-class families is the desire to take photos of their kids' sports. For those people, the point of sufficiency will come when they can get stop-action shutter speeds under typical indoor (gym) lighting conditions. The D3s is pretty much there now with an f/2.8 lens. When consumer-grade cameras with consumer-grade lenses can get 1/500 in an elementary-school gym, the masses will find it sufficient. ISO 12,800 or so should do it.

[Richard Skoonberg]

My eyes have been opened by this thread...ISO is the way to go! Hurray Nikon!

[ctein]

Dear Mike,

Heh, that's funny. When I saw the title I had the same reaction as you, which is "haven't we both written columns about this?" Turns out not. I, like you, have made many comments and side remarks in my articles about sufficiency (I called it "good enough"), but nothing directly devoted to the topic. Interesting. 'Course, I think you've got it exactly right.

=================

Dear Timprov,

I'll bet you're right because of sufficiency, but not because of real limits. Quality still goes up with pixel count, even with ordinary lenses, until you hit phenomenal pixel counts. But, other than geeks and perfectionists, who cares? I'm not sure. I can make the technical case for 100+ Mpix, but will anyone buy it unless there's truly NO cost penalty? Which leads me to...

================

Dear Winder,

Cart before the horse. Economies of scale have a huge effect on the price of consumer goods. When absolute demand goes down, price goes up. A Noctilux might or might not be marketable at $1,000, but it would still sell in insignificant numbers because hardly anyone will pay $1,000 for a lens and hardly anyone really needs f/nothing. Sure, it'd sell a lot more than at $10K, but that's just moving up from invisible market share. Which leads me to...

=================

Dear Obijohn,

No, faster lenses cost differentially more than you think. Long technical monologue inserted here. Cliff notes version: larger elements, deeper curvatures, need for additional elements to correct residual aberrations better, manufacturing tolerances have to get tighter as roughly the SQUARE of the aperture. For modest apertures, it's still a small part of the total cost, agreed. That's why a 50mm f/1.7 is only a smidge more spendy than f/2, a f/1.4 is visibly pricier than 1.7, and 1.2 is beyond most people's budgets.

Worse... people are irrationally price sensitive. They'll buy a cheaper product even if it costs more in the long run, if the benefit isn't obvious to them. So "sufficiency" kicks in, sales volumes are lower and the economies of scale work against you instead of for you.

pax / Ctein
==========================================
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photo-repair.com
==========================================

[tv]

I'm thinkin along the lines of what ObiJohn and Joseph Reid said..

For one, a decent digital rangefinder (not like the Epson), let alone a full frame one ala the M9 and accompanying lenses for much less than the Leica. I'm sure loads of street shooters are waiting for that one, myself included..

Secondly, if technically possible, enough quality ISO and what not that would allow us to shoot in low light with little worry of noise, lack of sharpness, image quality in general...

[Andreas]

The post and also and especially the comments are very interesting. I know that it is about marketing, but I still don't like that the more we talk about limits of sufficiency we more and more imply that technical improvements equal quality. We leave that notion of quality rather unreflected and somehow absolute.

For example, for me it is not given that I need higher shutter speeds in every condition. I like motion blur and I like that shooting film a certain speed (regardless if 100 or 400, maybe not 3200) poses a limit on my shooting style, forces me to use slower shutter speeds in certain conditions and at the same time liberates me from thinking about ISO setting. This is a quality on its own. So, pictures at dusk get another character than picutres in sunny conditions. Or: having parts more sharp and some parts more blurred by motion is also a new way of seeing things that we don't see in reality.

Of course, this can be replicated with more capable equipment, but I think I am not the only one who doesn't use to apply those limits to himself. Limitations force me to embrace them and this leads to different results than having all options. This is especially true for not so experienced photogs like myself. I somehow envy those who had to learn to shoot with ASA 25 speed film. Because in the future the standard will be clean, sharp, bright, punchy all day and night. Like modern megacities.

One last example of what I want to say: look here at the last image all way to the right. Bad quality is good quality.

[Kerry Glasier]

Hi Team,
I think a lot of people commenting already should be head hunted by manufacturers looking for whatever is going to be the next big thing. All exciting stuff and yes,isnt it always a buzz to get the new gear out of the box? But people,lets not forget, people make photos not cameras, how would you 'improve' say,a Cartier Bresson shot? Whatever you come up with the shot may well be better in some technical way, but would it be any better in the sense of the 'decisive moment' that he gave us? My bet is, no. Minor White used to say, 'camera
was used faithfully', no shutter speeds or aperture, film or even camera make were ever mentioned.
KG. Cornwall.UK

[Semilog]

"Megapixels will never stop increasing until Moore's law stops. There's just no downside to adding more megapixels if data storage and computer speed keep increasing."

That simply is not so. There is no escape from optical physics. How do I know this? From scientific imaging. Smaller pixels do two things.

(1) they get you closer to the limits of the lens, and even a perfect lens is limited by diffraction. Then, the only way to increase resolution is to use shorter-wavelength light (purple, then UV). This is why far-UV light is now being used to etch microcircuitry. This is also why for a typical microscope, there is never a reason to use a pixel smaller than 6 micrometers on a side. Smaller pixels just resolve blur since we're limited by the optics, and the optics are nearly perfect - they can't really be improved unless you play tricks like illuminating the sample with bands of light or using objectives to image the sample from both the top and bottom.

(2) Smaller pixels catch fewer photons. This is a real consideration. The ideal pixel size for typical optical microscopy applications is 6 micrometers on a side, as mentioned above. Yet the most sensitive (EMCCD) cameras currently available (e.g., this one or this one) have pixels that are 10-16 microns on a side, or even bigger. This is to maximize both the light-gathering area of the pixels, and the charge-carrying capacity of each pixel, which gives you much better dynamic range. You can choose: small pixels, or deep ones. Not both. Sorry.

[Semilog]

"...how would you 'improve' say, a Cartier-Bresson shot?"

By having Josef Koudelka take it? (Ducks.)

[ctein]

Dear Semilog,

Physics has its limits, but we are not at them.

You are correct to point out that increasing pixel counts has its downsides. It also has its upsides.

Resolving blur is, in fact, an extremely useful thing to be able to do. Besides, while I cannot speak for microscope optics, good 35mm camera optics, used wisely, can deliver 250-300 linepair/mm in the "film plane." If you allot only 1 pixel per line (an inadequate choice for optimal quality), you're still talking about 1.5-2u pixels. And that's just to achieve parity. There are still visible gains in quality when sensor resolution exceeds lens resolution (folks who disagree should reread some of my back columns on blur/diffraction/etc. before arguing).

Dynamic range is not a phyicss limit problem, yet. It's an engineering problem: increasing charge density while controlling leakage. The focus is on better dielectrics, better barrier layer design, new well geometries, continuous readout, and so on. Hard engineering but not a brick wall.

Counting statistics *may* be the one area where we're approaching physical limits. But I've not been able to find a QUANTITATIVE analysis from first principles. I may be forced to do it myself, if I can get one more number I need, but I'm basically lazy-- if you can point me at an existing paper on this, I'd be thrilled.

It's OK to be aware that physics imposes limits. Unless you know with some precision the magnitude of those limits, it doesn't make much more sense to invoke them than to argue that there's no point in building a faster train because physics says it'll never be able to exceed c.

Entirely true! But I don't think rail designers lose much sleep over it.

pax / Ctein

[Zoltan]

250-300 lp/mm? Wow! Can you cite a few examples, from any manufacturer?

[Semilog]

Ctein,

In oversimplified terms, for microscope optics resolution is limited by the wavelenghth of light, lambda, and the objective's numerical aperture NA, which is itself limited by the refractive index of the media that we are looking through. For realizable wide field optics the highest attainable NA is about 1.5. The smallest resolvable features (line pair or point source pair) are ~200 nm with blue-green light, so with a 63 x objective this means that we get Nyquist sampling at ~6 micron pixel size, and with a 100 x objective we get Nyquist at ~10 micron pixel size.

A more technical (but accessible) discussion can be found here

[ctein]

Dear Semilog,

Thanks! Yeah, for your purposes, unless you need to do a kernel deconvolution to extract maximum detail, 6 micron is likely adequate. I'm suspicious of any systems trying to run right at the Nyquist limit (as I'm sure you are, as well), but that's a quibble.

Camera systems, as is apparent, have different limiting criteria.

pax / Ctein

[ctein]

Dear Zoltan,

Hmmm, probably the easiest and cheapest one for people to play with would be the SMC Pentax-M 50mm f/1.7 prime that was standard with the Pentax ME and MX cameras. They can be found extraordinarily cheaply online. They'll diaphragm couple, so you can use them with aperture priority exposure systems; that's about it. Any camera you can adapt a K mount lens to should work fine with it.

f/4.8 is the optimum aperture, I think, at which it should easily surpass 250 lp/mm in the center of the field.

No idea about the edges of the field. It's incredibly hard to measure resolutions off of center; I don't even try.

pax / Ctein

[Tim]

Maybe I'm just sensible, but I still trundle along with a D200 that's beginning to show its age. The main reason I've not considered the immediate successors (D300 or D700) is imply that I don't need high-ISO "performance" - if by that word one means the ability to render scenes of darkened smoky rooms as the same effective brightness as a daylight scene with no noise. (I'm a landscapie, not a gig-photographer.) In any case, if I'm out taking a photo of a field in moonlight, if a 30s exposure is sufficient to get me my midtone then it's still not a realistic photograph until the processing algorithm simulates the Purkinje shift too.