« Photography at the Speed of Light | Main | AIPAD Show Marks the Passing of the Years »

Thursday, 17 March 2011

Comments

Mike,

it's getting embarrassing how often I am in full agreement with one of your posts. Will you just quit raiding the time difference between Wisconsin and GMT and sneaking over here to steal my early morning wonderings? ;)

Trying to be positive, I have often struggled with the AF element of my cameras. In the film days (my first AF camera was a Nikon F-601 (AKA N6006 in the US), I rarely knew what the AF was doing, but the results were sort of OK. My first digicam was a Sony P&S, and it "hunted" AF for at least a second before relentlessly making the wrong choice. The was no menu option I could find to turn that feature off. My current Nikon D200 has no less than 4 button controlled AF settings on the back, each of which behave differently in different light levels. The software has about as much intelligence as my daughter's hamster. I have defaulted to one setting which at least I know how to operate, or sometimes just turning off the ruddy AF altogether.

The next technology development I want is a little cursor / thumb control on the back of the camera that I can operate with the thumb of my right hand, to drive a little illuminated circle in the proper glass viewfinder, to a point where I want the focus and exposure to be spot-on. It could be the centre, rule of thirds, somewhere off a gridline, or even right down in the lower left hand corner. I want it to be my choice, and to obey me, and not to hesitate while it gives me a "are you sure?" message.

I suppose what I am asking for is not to be involuntarily subordinated to the "creative amalgamation" of a crowd of youngish Japanese software programmers, none of whom I will ever meet, none of whom will ever understand my photographic mojo, and none of whom care because I am probably a statistical outllier.

Alternatively, a digital FM3 would work for me...

We're also now at the point where the "performance" of a camera is limited by its lens, not the sensor.
Lenses can certainly be made better, but the cost/benefit ratio is becoming prohibitive.

I'm no expert with camera or computer technology, but is it possible that the consumer saturation point for megapixels also relates to the ancillary technology? For example, consumers who sense that their computer and software will not be able to handle the larger files easily or know that their printer will not make larger prints might think twice before jumping to a greater MP camera.

It always seemed to me that the sensor type produced by Foveon was/is superior. Maybe that technology will begin to find its way into more mainstream offerings. On your other point, I remember reading an article years ago that estimated the grain, or "resolution" of the best 35mm slide films to be around 24-26mp equivalent. It seems like that benchmark has been achieved. For the average or even professional 35mm photographer, 24 megapixels seems like enough.

Megapixels may be a cooling battlefield, but are the "better viewfinder wars" starting, or maybe just more "all-purpose autofocus" skirmishes? (And all the lens designers seem to be permanent REMFs)

And the Bayer array? The nails on the coffin are lined up already, but the hammer of "significant difference" has yet to publicly fall. I hope to dance at the funeral in less than 5 years.

Horace Dediu has been exploring similar themes over at asymco.com. His theory is that the entire dynamic of a market can change when the offerings swing from not-good-enough to better-than-needed.

He usually confines himself to mobile telephony/computing, but I'd be very interested to read his take on the digital camera market.

Michael: For me the 'new' battleground for the manufacturers will be dynamic range. I'll call it 'native' high dynamic range; no tricks, no software, no funky ISO settings in the millions, just built-in high dynamic range.
The goal (the 'Holy Grail') is a range equal to, or greater than, the human eye. I realize that the eye does not work like a camera but the range of light that an eye can see is a where we are headed with cameras... I wrote a blog piece about this in the context of talking about HDR and I am more convinced now than when I wrote that piece a few months back.
Marshall

Anybody remember "The Man Who Fell To Earth", with David Bowie? This probably goes with this as well as the posting about whether digital cameras are precision oriented or worth the money. Anyway, in the film, Bowie is an alien trying to get back to his planet, and he starts introducing technology at such a rapid rate, to fund the building of a spacecraft, that eventually everyone is afraid to buy anything for fear that it will be obsolete by the time they get it home. I think Buck Henry, as CEO of the company, gets thrown out the window, by the government, to put an end to the madness.

Can't tell you how many times I've remembered that movie, not only in the world of digital photography, but computers, especially browsers that are only a couple of 'ticks' old, but can't be upgraded because of your chip, so you have a perfectly good computer that works fine but can't get half the stuff on the net.

Glad to hear the pixel race is over or slowing anyway; I was about to get a few people and start looking for Buck Henry myself. Maybe the silver-lining of this terrible and debilitating economy is that people with limited resources are just going to stop caring about updating and buying that latest widget. What if they invented it and nobody bought?

As for ISO... weeell: Too often, I find myself in situations where I can just eke out a decently-exposed shot at f/1.2, 1/50, ISO 3200 on my Canon 1Ds2 with an 85 f/1.2L hanging off it. Around the dinner table in a social setting, for example. Flash would kill the mood, etc. Motion blur, camera shake and wafer-thin depth of field makes this a bit more exciting than I'd like. And even 85L bokeh grows tiresome when used en masse.

I'd be very happy to be able to get the same shot at f/8 for depth of field and 1/160 or so to eliminate shake and freeze the subject. That'd call for a reasonable-quality ISO of something like, oh, let's call it half a million or so unless my back-of-the-envelope math is wildly out. Though I could accept f/5.6@1/100 if I had to. Means I could bring a zoom lens instead of the stonking big 85, too.

We have a way to go yet.

I'll be satisfied with sensor sensitivity when I can take a picture using ambient light in my living room at night at an aperture sufficiently small to keep an adult's face acceptably in focus from the tip of the nose to the earlobes, with a level of noise that doesn't show up at 25% of full resolution on my computer screen.

Having settled in fairly nicely for the time being at 31MP and 9x12 to 12x16 prints, the pixel race is definitely over for me.

Not so much the low-light sensitivity race, though, as I would love nothing more than being able to accomplish in fractions of a second that which now requires multiples of seconds and fractions of hours with my long-exposure nighttime photography.

I wonder if secondary limits also come into play. Just as with the print size/enlarger issue, I think there is also a pixels/computing power link. Computers are also getiing to the sufficiency level, and many more pixels means needing more power or storage.
Just as it took me a major computer ugrade to even start on digital photography, so it would probably take another to make a major step-up in pixels, and I'm just not ready to do that yet.
Other performance measures (e.g. ISO) do not have that same limit, so maybe the sufficiency point is further along than for pixels.

Throw the Sony A580 into the mix with the D7000 and K5. Same sensor in all, and I hear that the A580 holds it's own in comparison of image quality, esp for the price.

I agree with much of what you said in regards to megapixels, but the Bayer sensor...? I'm not so sure. I remember reading self-proclaimed experts on various digital photography web forums proclaiming that anyone who bought a camera with a Bayer sensor was a fool because the technology was dead as a result of the Foveon sensor. This was when I was researching my first digital camera. In 2002.

In fact, the reason you might be wrong about the Bayer sensor comes directly from the fact your argument for the megapixel wars being over is so correct. Most people have "enough" quality in their cameras already. Megapixels are one way of measuring this, and incidentally one of the reasons why the resolution of current cameras has grown so large is because of the limits of the Bayer sensor architecture. If people don't need extra quality in the form of more megapixels, why would they need extra quality in the form of a superior (but also more expensive) sensor design? Bayer might die, but it is likely to take a very long while or the advent of a new sensor architecture that is both superior (or at least equivalent/satisfactory) and cheaper to manufacture.

Yes, 24MP is sufficiently good for most photographers, even half that amount at 12MP is adequate.

But, I suspect, ISO and DR is a sore point for many photographers. It's also the one area where digital can really distinguish itself from film.

And after that? How about more accurate colour rendition (of WB is part). Perhaps this is where the Bayer array is replaced.

Christian,
Where do you come by your assumption that a superior sensor has to be more expensive?

Mike

We have to stop fixating on size and work on images, My best 2 images of 30 years work are slightly out of focus and not properly exposed but beautifully printed and framed, they are my two wonderful childrens portraits.

Three of the last 4 cameras I bought, I wished I could get a version with fewer megapixels. (The exception is the D700, which, being full frame, devotes enough real-estate to each of its 12 million pixels to do a pretty darned good job.)

Perhaps 25MP is about "right" in some sense for top-quality slide film (in 35mm); but I promise you, you can make a far far better 24x36 inch print from a 12MP digital image than you can make in a darkroom from that slide film.

And, crossing threads, check out the print sizes, where listed, at the AIPAD show. They certainly do trend to the large end! And several of the small ones listed are decades old, whereas most of the large ones are quite recent.

I have this odd sinking feeling that this discussion is based on a flawed assumption that still photography matters any more. It does, of course, but as the driving force behind sensor development ? I suspect designers are thinking "what can we do to provide better video, faster CDAF" more than they're thinking about how to produce better stills. Hopefully I'm wrong and that the two can be pursued simultaneously.

Anyway, I'm not convinced that we'll reach a high ISO saturation point as fast as providers will run into ceilings, but that could be because I shoot high ISO a lot. I shoot low light a lot, but enough indoor sports to see how it can benefit anyone who shoots action indoors. But maybe that just means that it falls in line with megapixels; the number of people demanding more will shrink.

Speaking of megapixels, going from 6 to 12 was a nice boost for me. I've printed landscapes from 6MP at modest sizes that left me wanting more detail. Looking at 12MP files shot with my primes or my CZ16-80 leaves me with this odd feeling that "just a little more" would hit the spot. Say 14-16. Silly. The difference is trivial. Those numbers just feel right. 18 sounds high. Sillier :) 14MP files from my NEX can be really good, though, and that's with the kit lens. The long-overdue replacement for the A700 is rumored to be 24MP and I can't say I have any great desire to own a 24MP APS-C camera.

Sony's new alternative to the Bayer array is... a Bayer array, rotated 45 degrees. With twice as many photosites.

Their claim of 67% RGB sampling vs. 33% for a Bayer array is because they are are using twice as many photosites, and they round up to get 67% rather than 66%.

The 45 degree rotation does absolutely nothing of value at all. Their claims essentially come down to nothing more than 17.7 megapixels vs. 8.8 megapixels.

As a D700 owner, I'm continually amazed how I get great shots at night in a pub shooting the music that I do. Terrible lighting, moving targets etc etc. But at ISO 1600 and f4 I can get crisp shots at 40-60th of a second.

What would I change? retina focus point selection. I think Canon tried this a while back? It would be so much faster than a joystick (also a good idea) and so much better for moving subjects. No more hunting.

Wishlist
1.1) eye follow focus. whole viewfinder used.
1) 100% clean ISO 6400. Any more is just weird. Hold color gamut well.
2) cheap f2.8 lenses that are not soft / rubbish or massive!
3) Built in Adobe Lightroom (or similar) in camera. ie: common preset/s loaded in camera, wireless sync to iphone / android.
4) size. Bring my d700 down to the size of a leica and i'd be the happiest chap in town


Nick

I think one driver for the next area of attention for improving photos will be in the presentation end. Folks above write about wanting more dynamic range, but I don't see how having cameras with more DR will matter unless external things like displays and printers improve so we can see an increase in displayed DR (as opposed to the DR compression that HDR type effects produce). Don't get me wrong, I'd love an increase in DR, but how one would get there is a mystery to me. How to show mass consumers that a photo isn't showing what they see is a tough job for marketers. Making them care, doubly so!

Me, I think the next wave of attention will be on software-in-camera's capabilities. I agree with Christian that the next thing will catch on because it is less expensive to implement. Remember when CMOS came along, it wasn't for superior performance reasons. It was cheaper. Now, there is only one 35MM Full Frame CCD based system (of which I am aware, the Leica M9), but all MF backs are (AFAIK) all CCD because the image quality demands are higher than what CMOS can provide at any cost.

Patrick

If you want all those cool computational tricks like refocusing the image or changing the point of view after taking the shot, you need many more pixels. Each output pixel (which contains much more than a single colour value) corresponds to (say) 14x14 sensor pixels, so your effective linear resolution is reduced 14 times. For a 24 MP refocusable image you'd need a 4800 MP sensor.

Perhaps this is what you meant by "technical developments" to "obliterate the present context"?

"The next technology development I want is a little cursor / thumb control on the back of the camera that I can operate with the thumb of my right hand, to drive a little illuminated circle in the proper glass viewfinder, to a point where I want the focus and exposure to be spot-on."

James - This is already completely possible in a lot of cameras, EXCEPT that it's done with an LCD viewfinder/screen. Doing it with a real viewfinder is certainly possible too, but I'm not sure if the extra complexity is worth the amount of sales it would garner. Maybe the new Fuji X100 viewfinder is getting close to that (or is there already?)

Craig - Stephen is right that what Sony is doing there is not really any different than the Bayer pattern. There are other alternatives though. But I think Mike was referring to when Bayer arrays are no longer being made, not when alternatives show up.

Greetings Mike,

Could the real issue be that most people are not printing their pictures anymore? 72dpi images with small pixel counts transmit though emails real fast and websites want low resolution images for fast uploads. The web based public is used to low quality. So for those people: why would they need to spend the money for a costly high pixel count camera?

Most of the digital cameras are disposable and cheap just like their old film equivilants. Ease of use and no chemical processing is the major benefit. But, how many of us are turning into computer-potatoes because of the time we spend in front of them. I resemble that remark and see no way around the situation.

CHEERS...Mathew

I'll be interested to see what the new Sigma is like, with the higher resolution Foveon.
I've got a Sigma DP1, and while it's a flawed camera in many ways, I've hung onto it because there's something about those images that really appeals to me. I prefer my Sigma pics to ones i took with an EOS300D, EOS400D, or Olympus E-P1.

If the new higher resolution sensor (Marketing aside, it's 3x the resolution of the sensor used in the DP1, DP2 and existing Sigma DSLRS) delivers the same quality at the higher resolution, i'd like to hope Foveon will have the last laugh!

Need ... more ... storage!

Pak

Until there are full frame compacts there is simply no advantage of digital over film other than convenience. It's all hype basically; don't believe it!

I agree. Both megapixels and high iso quality have been singlemindedly targeted by unimaginative camera companies to the exclusion of most other elements in recent times.

As of yet though, none of the manstream dslrs have been able to match the colour integrity of the foveon sensor, nor the high dynamic range of the Fuji sensor in the S3 or S5. Both of these are areas I would like to see tackled next in such a singleminded fashion.

Interestingly, I think we're reaching the point where much of the quality is being wrung out of the sensor, and would like to see the big-hitting camera manufacturers explore the usage of multiple sensors within a camera, a la foveon.

APS-sized sensor cameras would appear to be the real testing ground for such experimental approaches, given the oft-quoted lower cost of APS-size sensors. Would Canonikon be brave enough to explore so, away from their flagship full-frame range though.

I would like see high megapixels allied to high iso performance, with tremendous colour depth and integrity, with the dynamic range of negative film.

Lastly, I would also like to see a black and white only sensor camera..

All I want is a wider dynamic range. Couldn't care less if it was 6mp or 60mp.

Actually, my D3s cameras are already good enough most of the time.

Here's what's missing:

- better spread of the AF sensors (at least a couple of cross-AF points at the 1/3 lines)
- there should be a an EOS RT-type of non-mirror-slap mode (less noisy than the "q" position today)
- how about a D700s with two CF (or even two SD) slots. Price could be close to the D3s price-tag. I don't want to save money, just some weight
- well, throw in a D700x+s with 24 MP and I'll have one of those, too.


200 ISO film was made of a bunch or silver particles with different sensitivities. Some were 100 ISO, some 800 ISO. The average of the film was 200 ISO.

The consequence of this is the DYNAMIC of film that is still MUCH higher that what we currently have.

It is time that they expand the dynamic on sensors. Film still makes better work we can't reproduce...

glennbrown: beware, photos-of-kids hits the higher end of subjective. ;)

I err on the side of wanting more megapixels after the recent post here about painting sizes; the simple inclination is that if I'm going to go to the trouble to make a photo (including setting up tripod, etc) then I'm going to get the best out of that image I can, including the options of putting a (potentially) large print on the wall or selling it as large as possible. I'm happy somewhere around 18 megapixels as the starting point - can upscale (ideally via multi-image super-resolving) from there.

Folks with online photo-galleries: when was the last time you reviewed your work and said a 10MPel image was too small to sell or print? Or 12MPel?

Recently, a Nikon executive stated something on the order that "now that the megapixel race is over, focus will be on who can create the widest zoom range" (paraphrased, the interview was on NikonRumors.com, but I can't find it. Sorry)

That might explain why every manufacturer now seems to think that a 25-500 zoom is not enough for their customers, since that would be only "20x", and the competition offers 24x, 26x, or 30x.

It seems that they have collectively decided that the Average Customer can only understand one figure at a time. An absolutely nonsensical paradigm for product design.

The commenters who wondered about whether ancillary things might be causing the reluctance to more pixels are right, but looking at the wrong culprit.

It's not the computer that's the gating element. Yes, you'd need more memory and storage, but anyone looking at the price/size trends in those two areas will tell you that progress just keeps happening there. We've got one company just announcing a small 6TB drive (actually two drives) at about US$550, after all. You can store an awful lot of 24mp images on that.

No, the culprit is output. Just exactly where are you going to output 24mp to? All those big screens in your home, no matter how big, are basically 2mp devices (1920x1080). My 30" display is a 4mp device (2560x1600). My Epson printer is a 32mp device (at 360 dpi), true, but do I actually push it that far? No.

So exactly where are we going to send our 24mp images? Well, for us pros, to clients. But for the amateurs that constitute most of the market?

On the other hand, this is just a "pause." Technology has this way of coming along and breaking barriers. Right now most people have an output barrier of about 12mp (8x12" print at 360 dpi), and most of their work gets output far smaller than that. Until that changes, you don't need more than a 12mp camera.

I'd echo what the others say about the new battleground being Dynamic Range. Hell, it's minimal cost for the marketing department. Simply substitute Megapixel with Stops for the last 2-5 years of maketing materials.

Actually DR Stops doesn't quite have the same impact as MP (OMG MEGA Pixels!!). Someone should come up with a sexier unit of measurement for DR Stops.

The photographs that I most admire were taken with cameras that were positively primitive by today's standards. How can I imagine that the cameras available right now, or for that matter the 4 year-old camera I currently own is NOT GOOD ENOUGH for me??

I haven't done the tests but have seen several messages along the lines of this one from Iliah Borg suggesting that anything above 12MP, even on sensors of 36x24 mm, are of limited value when shooting handheld SLRs with 100 mm lenses.

Bad news for those of us waiting for D3S prices to plummet. :)

I have to ask the question what is considered "mainstream"?

Sure, there are 24 MP cameras, but are they really "mainstream"? Even pros appear to be resistant to them based on price. Few amateurs seem likely to purchase them given the current price structure. Could it be that part of the Canon 5D MK II's success is that it breaks out of this price structure and so places it within reach of more prospective purchasers, both professional and amateur?

Then there is the matter of where the market is moving. The technical demands to post images on Facebook/Flickr or whatever are simply not that great. Indeed, a higher MP camera is a hinderance in that circumstance because it has to be processed (down) before posting. Web content in general does not demand that much. Just how large are the image files you post on this web site? Others probably use even smaller files.

Will there be a sufficient market for higher MP cameras at the present pricing to warrant much more development other than engineering work to improve process and reduce costs?

Look at Leica. According to all reports, the M9 is a wonderful little camera, but is priced such that it is a niche product, not a mainstream one, but Leica is sufficiently small that generating collector editions periodically keeps them going.

I suspect that there are a lot of people who might like higher resolution cameras, but not at the current pricing. There are, of course, a lot of people who could either care less or might actually find a higher MP camera to be a hinderance. I see the market at the proverbial "fork in the road", except that it may not be a circumstance where the path forward is one or the other. Certain portions of the market may take one or the other, but cost will be determinant factor. As it is, landscape photographers, as an example, are able to capture images that can achieve the desired result whether it is in the image capture or in post processing to scale the image up to the desired size.

Current events in Japan may adversely impact the ability of camera companies' ability to develop products for longer than we think. It may require the relocation of much of their operations. We may see a stagnation of development for some time.

With my cheap-ish flatbed I can get a good 25MP scan from 6x6 medium format film. I might shoot a roll or less of that per month, and aim to scan no more than half the roll, but I can still barely handle that amount of data. I'd be utterly overwhelmed if I had the ability to shoot hundreds, or even tens, of 25MP images at a go.

Frankly, the 7MP or so I can get from 35mm film is enough for my needs and I'd be entirely content with a 6MP digital for the "throwaway" stuff. Seriously. 14+ MP is a non-selling point for me.

But then I am a bit odd.

"Actually DR Stops doesn't quite have the same impact as MP (OMG MEGA Pixels!!). Someone should come up with a sexier unit of measurement for DR Stops."

Agreed that "stops" sound boring, but even more than that, most people don't know what it means. I'm assuming they'll come up with a new term, or maybe just substitute "light range" or something. And then focus on the difference that stops imply. For instance, say you had a 3-stop improvement; The ad could read "Camera X can capture an incredible 8 TIMES the light range of Camera Y".

If you have a 24Mpixel Bayer camera you can use 2x2 pixels to get a foveon effect. Resulting in a 6 Mpixel picture :-).

I would just like to say that I think a black and white only camera would be a step backward. There is already some evidence that using only black and white pixels does not increase resolution much, if at all, and you would lose all the spectacular post processing abilities. It would be back to filters which would steal light away from the sensor, etc., etc.

Reading reviews, it seems ~9 EV is typical sensor dynamic range. that is 2^9 or 512X. Fuji, I believe, tried different sized sensor cells - hexagonal no less - to get better. One might imagine a CMOS sensor that gangs cells in dark areas, trading off resolution for DR. That is software, which is where I believe most progress will be made. Maybe we talk about K-EV?
But what about lens designs? Optical limitations cause many current lenses to lose resolution because of diffraction at F/8 or so and beyond. Stop down for DOF and lose resolution. That is hard to fix.

Dear Bahi,

I think that's the wrong message to come away with.

All things being equal and **perfect**, a 24 MP sensor would resolve 40% more line pairs than a 12 MP sensor.

Iliah got 20% more in a bench test. That's very reasonable, considering that he couldn't have a perfect lens, assure absolutely perfect focus, etc.

The message you should be taking from this is that it is just as silly to equate the *working* resolution of a digital camera with the sensor resolution as it would be to equate the resolution of a film camera with that of the film.

It's a nonsensical position to take, contrary to both reason and reality. In fact, using the term of the week, I would call it an OCD that way too many photogeeks suffer from.

Sharpness is not limited by just the sensor. It is not limited by just the lens. It is not limited by just the mechanical tolerances and accuracy.

pax / Ctein

Hello, Ctein.

What I took from Iliah's note is really his point about camera support being an important limiting factor; because I shoot almost entirely hand-held and usually in less-than-ideal light, I'd expect to see relatively little consistent return from using a higher resolution SLR.

I completely agree with your point about the system rather than just the sensor. In my case, since I'm not an obsessive lens tester and generally use phase-detect focus in preference to manual LiveView or CD, I'm generally not in the running for (I've just looked it up) 61 lp/mm, which is what DXOMark reports for the Nikkor nifty fifty on a D3x. In my case, a camera's noise characteristics and speed of operation are much more important, which is why the D3S, despite being twice as bulky as I'd like, holds a special place. :)

I disagree, I think we'll enter a new race for 2x2 binned resolution on Bayer sensors. Or at least I think we should.
The advantage of that is you sum out all the loss in resolution from the low pass filter, and get a very sharp image right off the chip, without algorithms guessing how to remove the blur of the LPF.
The 5D mkII already offers 5MP binned raws, and from what I've read those images are very good.
So if 12MP is enough now, we should see a new race to 50MP, with special ultra-high resolution cameras going all the way to 100MP.
Just my 2 cents.

Hynee,
I think you're right. The question is whether those cameras will be viable products that will sell well--i.e., that will drive efficiencies of scale and further development.

Mike

Dear Hynee,

And then there's my take on it:

"Why 80 Megapixels Just Won't Be Enough..."

http://tinyurl.com/c4dj5n

pax / Ctein

The comments to this entry are closed.