How much better will digital get?

[howard smith]

Bob, I am (was? retired now) an engineer.  My experience with manufacturing was not with lenses.  However, for many mass produced items, the items are made and a sample of product is examined, tested, whatever.  It the sample is OK, then there are statistical reasons to believe that some percentage of the batch have some percentage of "acceptable traits."  Say, 95% confidnet that 95% of the batch is within spec.  The specs can be raised or the sample size increased to get more items in the batch acceptable.

The usual mathod of control is to monitor the line of products and change something that is drifting before there is a problem.  Control charts based on the max, mean, min and standard deviations of the sampled items.

Some process accept (or reject) the whole batch.  This is costly and is usually done when the rejects can be recycled rather than just dumped.

Some processes, small and expensive production runs, sample each item.  Rare in mass production.

Variations in lenses would be caused by a large number of variable, like machine wear, variations in glass, variations in caotings, etc.  How the elements are assembled is a big one.  To be perfect, the center of each perfect element would have to line up with the center of the others, and the axis of each woulf have to be lined up.  Hard to do.  The better job, usually the more expensive the process, and a higher reject rate.

Look at MTFs for lenses.  At the center, the radial and sagittal lines should meet.  They usually don't though.  That difference is most likely due to misalignment of elements.  (Canon's match, but that is because the publish the "theroretical" lines, not actual.  In theory, they are all great.)

Some process do actually "cherry pick" for you.  When things are going good, tehy make their brand.  As the process drifts, they make "non-house" brands.  Or a rejected batch may become Chuck's 28mm/2.8 lens with a really low price.

The bottom line is, the is no free lunch.  If you want to walk into your camera shop and grab any box off the shelf and get a great lens, it will likely cost you.  Yes, there are great Sigma lenses and a few dog Zeiss lens samples.  But they are rare.

[Bobtrips]

OK, I'll play....

(I've been snowbound for two weeks now.  I'm getting a bit 'cabin fever crazy'....  ;o)

1) Much technology from the past will be carried forward.

-35mm form factor -  

Yes and no.  There's no reason for 'standard sizes'.  We aren't all buying boxes of film any more.  We buy our film once per camera.

I'm guessing that cameras will shake out in terms of print size.

35 mm will stick around for a long time because of the vast number of lenses floating around.  

Ten years from now there will be more half-frame lenses in common use than full-framers.

And cameras (sensor sizes) will continue to shrink while maintaining performance for some time to come.  Those 35 mm sized bodies/sensors will be the MF cameras of today.



-optical viewfinder

Most likely it's already about the last twig on that evolutionary branch.  It might extend a bit further with the addition of a 'heads up' display of live histograms, etc.  But I'll bet that within five years EVFs will be acceptable alternatives.

-35mm lens technology

Improvements are due here.  Tighter sensor packing will require it.  Ceramic lenses are going to make some lenses smaller and lighter.  In addition ceramic is reported to have less CA problems.

-metering technology
-auto focus

No idea on these two.  Except to say that most everything improves over time.

-anti shake/stabilization

Manufacturers are going to feel a lot of pressure to include this in their bodies.  Especially if Oly puts it in their dSLRs.

[Ben Rubinstein]

1200mph? were you in a jet fighter?

[howard smith]

DiaAzul, actually the aircraft was a North American A3C.  No idea about the camera.  I could not see the dimples or name on the golf ball, and had it been out in a corn field, it may not haveen identifiable as a golf ball.  It only look like a golf ball because of the proximity to the golfers and on a green.

pom, "Of couse we save in film, but then didn't we charge the film costs to the client anyway?"  I have been told by some who likely know that if you charge for film and processing directly and explicitly (instead of just burying those costs in a bigger "services" bill), the film with the images on it actually belongs to the client.  They bought the film and services seperately.

[BJL]

Ray,

   have you seen tests using best practice conversion of RAW output from the S3? I cannot see any way that the sensor can be indicted as a failure without doing that; the adverse observations on JPEGs could simply reflect limitations of the in-camera JPEG conversion, which in turn could reflect the impossibility of having a "one size fits all" JPEG conversion strategy for high contrast scenes: this probably needs to be done by hand in RAW conversions.


Bob,
   surely you recognize that many high contrast scenes must be taken in a single shot, not by taking two shots and blending. I also suggest that you not judge teh rsluto of a camera by numerological debates about how pixels are to be counted, but instead do it by looking at the results. the S2 already seemed to have noticably higher resolution than other 6MP DSLRs. Using data from 12 million different locations, even if half the measurements  are of lower quality than the others, does potentially give more resolution than data from 6 million photosites alone, though not twice as much.

[Jonathan Wienke]

I don't want to sound disagreeable but, I don't get what your saying. The small sensor compared to a photon is the grand canyon to a grain of sand.

Actually, the sensor sites on P&S cameras are getting pretty close to the smallest possible physical size--the wavelength of visible light. Red light has a wavelength of about .7 microns (~700 nanometers) and IIRC the 8MP sensor used on the Canon Pro 1 and Sony 828 has ~2 micron sensor sites. So it's possible for photosite size to be cut in half again, but that's pretty much as far as photosite shrinkage can go.

[Sfleming]

Actually I'd say if one has the money now is the  time to get the 1Ds Mk II as it may very well be all downhill from here.

Pro DSLR Digital tech is akin to rocket science and brain surgery combined.  Canon has been able to pull it off and they now  virtually stand alone.  Nikon limps along by virtue of their history alone.  Watch carefully the press pit at the games, political events and it is hard to find a single Nikon.

(completely  beside  the point but revelatory:  Why hasn't canon made a good extreme WA lens for the 1Ds II?  They could get away with a 3k price tag and most would pay it if there was any  way  they could scrape up the money.  So why?  Because they don't have  to.  Why bother?  They got you where they want  you ... why go out of their way?)

Kodak who started the digital revolution has fallen flat on it's face for two years now and is pulling ever back from the 'finished product' end of things and concentrating on making chips only.  Don't talk to me about toy cameras.  I'm talking about  Pro DSLRs.

If Mamiya hits a home run with their ZD it will be a milestone ... a pivotal event ... because perhaps it will make someone else believe that  Canon can indeed be challenged.  But  think of all the things that MUST come together.  They have to match or exceed Canon on all the points.  From resolution to iso to build quality.  A bigger chip alone will not do it.

I'm afraid  that Canon stands alone and if this  is indeed borne out then quality will go only one way.  Down.

Perhaps something will come out of all the mergers.  I wish  Nikon would merge with a company with some vision and daring.  As it is now they look like a company ready to fold.  Well not a whole company ... just their Pro DSLR division.

I think the true problem is the corporate culture itself.  There just isn't a ton of money to be  made here and corporate cheiftans look at the DSLR division compared to say, in Nikon's case, the eyeglasses and lens division, and ask themselves, "Why the   do we bother  with this crap"? Hellsbells the binocular division probably makes more  money!

Why would any cut-throat capitalist continue to mess with costly (in EVERY way) high high tech endeavors that don't make much money when they can turn an easy  buck in a hundred different ways?

There's really only  room for one  company and  Canon  is  it.  So buy your  Canon  now it's all you're ever gonna  get.

"__ you!," you say "I'll take my 22mp digital back, thank you very much!"  OK.  You and about 1000 others on the entire planet.  The rest of us have to live in the real world and most likely  face divorce even to get the $8000 Canon.  As to the price of 22mp backs coming below $15k and probably more like below $20k in '05 I say ... HA! Dream on.

Digital capture is a lost world to me and I don't even know why I bother any more.

[BJL]

Canon has been able to pull it off and they now  virtually stand alone.  Nikon limps along by virtue of their history alone.  Watch carefully the press pit at the games, political events and it is hard to find a single Nikon.

The press pit shows that Canon is dominant in one aspect of photography; making the fast, highly automated "ultimate point and shoot cameras" needed for journalistic and sports photography. And this domination started before digital, driven by things like lens technology.

I do not think however, that this shows that "resistence is futile" in all aspects of high quality photography. What works well in the press pit is rather irrelevant to those who operate in "frames per minute", if not "minutes per frame", such as most medium format users.

[Ben Rubinstein]

Kevin, can I ask whether your marketing studies show that digital cameras/backs, will reach a plateu as to what is necessary, each format for it's own. For example will manufacturers try to take on 4X5 with DSLR's and will med format backs try to better 8X10 large format, or will the needs of the clients dictate how good the chips need to be so that at some point the manufacturers will say, 'Good enough, better for this format isn't needed by all but a tiny amount of our clients' and at that point when the digital race is over, the prices will stabilise and drop, and we will see new bodies/backs every 5-10 years.

Will the race not be won at some point in the next 10 years, and then having reached a plateu, stop, or at least slow down drastically? Will you continue to invest time and money in developing a med format back which will better 8X10, market it as a solution for all med format shooters and then in 5 years time sell it at the price the p25 is now, only to find out that a huge amount of your clients will say, 'I don't need anything better than the p25 or whatever, I won't need to upgrade again'. at that point the market falls away drastically and the price of 30,000 dollars is no longer justifiable for extra performance that is no longer necessary?

[BJL]

Perhaps the evil effects of diffraction are being over-estimated here. Diffraction is a significant factor in sharpness from about f/22 in 35mm format, but it is far from being the sole determining factor until far smaller apertures: "pin-hole" territory. The resolving power of lenses and emulsions also play a significant role at apertures like f/22 or f/32, so it is not at all surprising to me that MF can give sharper images than 35mm at the same print size, or equal sharpness in larger print sizes even when the MF is used about one stop smaller to get the same DOF.

Good 35mm lenses often have optimal sharpness at about f/11 or even below, with some measurable degree of diffraction limitation beyond that, but this does not stop knowlegable photographers from using f/22 and beyond when appropriate for the sake of extreme DOF, and stil getting some impressivly sharp images.

The 9 micron pixel pitch of current 22MP backs seems to start feeling modest diffraction limitation at somewhere around f/16 to f/22 and should not be "diffraction crippled" at all even at f/32; these numbers give DOF comparable to f/11, f/16 and f/22 in 35mm format, which does not sound like a painful DOF restriction at all, especially with normal to wide FOV and/or distant subjects.


P. S. From my small experience, it seems to be that 8x10 is mostly used at normal to wide FOV, meaning with focal lengths mostly 300mm or less. Ansel Adams seemed to use a 10" (250mm) lens more than any other. This greatly helps in getting adequate DOF.

[howard smith]

Don't underestimate Chapter 11.

"Since I've talked to Fatali and David Muench AND Bev Doolittle I'm almost famous already just by association."  I've never seen a photo of all three of you at once.  Are you really Bev Doolittle?

Gosh, I shook the hand that shook the hand of Bill Clinton.  I'm almost famous and didn't even know it.  My sister had dinner with Richard Nixon, but I'm not a crook.  I had a date with FDR's granddaughter.  And, IMUO, Fatali is a jerk.  So there.

[didger]

Are you saying that large photosites can generate a stronger signal because their greater surface area will be struck by more photons than a small photo site? Or are you saying that large photosites simply have more 'capacity'?  

Hopefully someone here is an engineer that can really answer this in the detail you want.  Basically, whether it's camera sensors or computer chips, the smaller the components, the more likelihood of errors due to random events that can't be eliminated and only mitigated with ever more challenging measures the smaller the components become. Eventually you reach a point of diminishing returns and an unacceptable gains vs cost ratio and then finally an absolute wall for a given type of technology.  Altogether new kinds of technology are not predictable.  These random events are of a low order of magnitude and get absorbed with no effect in a larger component, but may trigger a false signal output in a smaller component.  Consider it like the difference between putting a gram of arsenic in a quart of water or a big storage tank full of water.  The one situation will kill you with a sip and the other won't hurt you if you drink a quart.

[Bobtrips]

Let's say the camera takes 3-6 quick exposures to get the 'blend' shots.  

Isn't this the problem? No quick exposure can capture detail in the shadows. If you want to blend images, you need one long exposure, one medium exposure and one quick exposure. Total time is one very long exposure.

Yes, this may be the downfall of that approach.  But let me try one other idea.  

Shadow areas don't throw off a lot of photons.  The problem becomes separating the signal derived from those few photons from the inherent system noise.

If you had multiple frames to compare it might be possible to do a pixel by pixel analysis to determine which were the 'real' signals by their appearance across multiple frames.


(And, yes, this is all going to take a lot of processing power.  But it doesn't need to be done in-camera.  The camera can create a quick Jpeg for review.)

[Ray]

Given the proper processing workflow, would discerning print buyers, at normal viewing distance, perceive any significant difference between a 16x20 print of a Digital Rebel image and 16x20 print of a high-end digital camera image? If so, where would the difference lie?

I think it's clear the differences would lie in the amount of fine detail and texture that would be visible close up. However, I suspect these differences would not be discernible from a viewing distance appropriate for taking in and appreciating the whole scene.

I can't comment on how this would affect sales except to say, if I was offered a choice of two 16x20 prints of the same scene, but one was sharp close up, as though it had been taken by a 1Dsll, and the other was slightly blurred close up but identical in all other respects, I'd choose the sharp one, if the price was right.

[Quentin]

I understand that no less a name than Leica has cooperated with Sigma (yes Sigma) and that some Leica lenses have exactly the same lens specs on paper as their Sigma counterparts.  Can't verify if this is right, but it comes from a reliable source.  The difference is that Leica operate stricter quality control and probably  manufacture to closer tolerances.

So if the market (and profit) is there, Sigma could produce a super-EX range to the same exacting standards as Leica or Zeiss.  In fact even if the story is baloney, it still stands to reason that the likes of Sigma Tamron, and Canon Nikon etc are in the business for profit and could do it anyway.  But is the market for premium lenses big enough to justfy the hassle and expense?

And on this Canon v Medium format thing, in the days before digital, Canon did not compete with medium format, but still did perfectly well, so I don't see why they should be at all concerned if Medium format re-establishes its supremacy in the digital future.  The bigger threat to Canon, and Nikon, may instead come from "new" formats that are "good enough" for photojournalism, sport and consumer use - possibly 4/3, or maybe something completely different, leaving 35mm size sensors caught between two stools.

Quentin

[Ray]

Since the added photosites are only used to measure bright light and only need to cover a range of a few stops, they can be far smaller than the regular photosites. With Fuji's roughly octagonal photosites, the small highlight photosites seem to fit almost entirely into gaps that were there between the bigger photosites anyway.

The first implementation of the SR concept had both detectors under the one microlens. It can be clearly seen in the old and new diagrams of the photosite arrangement that the size of the S sensor was severely compromised. The old diagram shows it as being about half the size it could be, if the small R sensor wasn't there. I always believed this was a serious flaw in Fuji's concept.

They've addressed this issue in the new design by squeezing the R pixel, now with its own microlens, in between the main pixels. But why is there still not a major improvement in DR?

People are always talking about pixels getting so small they are beyond the resolving capabilities of the lenses used. I would say that the R pixel in Fuji's system is such a pixel. It's smaller than the resolving capabilities of the Nikon lenses and is therefore of limited usefulness.

In other words, the distances between the tiny R pixels are relatively huge and these spaces are very much withing the resolving capabilities of the lenses. In any high DR situation, most of the light which would normally result in a blown highlight, is not being captured by the R pixels. It's falling everywhere else, much of it presumably overspill on the S pixels, and is essentially being wasted.

[Jonathan Wienke]

FOr an interesting DR improvement concept, look at this thread.

[John Camp]

Most web photo forums have frequent discussions of whether to buy now or wait for better equipment, lower prices, etc. We then get the usual inane "buy now and prepare top spend big bucks to upgrade because that's the way the world is gonna work" or "relax and enjoy what you've got, it's good enough, and besides, a really good photographer can take great photos with 1mp Kodaks and a stick like Ansel Adams did."

But there are reasons (I have a couple, think of your own) for owning really good equipment right now; and also reasons for waiting, if the wait is short (less than a year.) So here's the question: Are the top line affordable cameras -- say the Canon 1DsMII -- good enough in producing 13x19 prints, that it's going to be a long time before newer cameras can produce better 13x19 prints? I don't want a discussion of technique, etc., so let's just say, prints of a photographic target under a variety of lighting conditions. Will there be massive improvements in dynamic range, or color quality, or light sensitivity, or noise control? Or are we now talking about small incremental changes? I read Reichmann's piece comparing the 1DsMII to the P25, and I wonder, if you are working with an Epson 4000, are you going to see a practical difference at 13x19? Will you ever? I understand that if you built a billion megapixal chip you could make a billboard that would be sharp under a loupe, but what will people finally settle upon as a practical optimum, where the critical factors become lens quality and ergonomics rather than bigger chips? Are we there now? Will be there soon?

I want to buy a camera now that I can work with, and settle on, for a long time, without giving up much in the way of print quality. Like everybody else, I think about price, but that's not as  critical as the other considerations. I own a ton of Nikon equipment, but I'm getting very tired of waiting for Nikon to produce a camera that will match other camera's finest 13x19s. Right now I'm working with 2 Kodaks, and the good photos are fine, but the cameras have limits, especially in low light, and frankly, the ergonomics drive me crazy. I'm willing to make the switch to Canon, but I'd be less willing if I were convinced that the 1DsMII is just another one-year temporary waypost on the way to a real camera.

JC

[didger]

I'm surprised that in this discussion and other similar ones the Mimiya ZD never gets mentioned.  This is a camera that weighs a little LESS than 1ds/1ds2 and inexpensive relatively light world class lenses are available, often used.  The price could come in at not too much above 1dsMKII and you have 22Mpixel medium format (exactly double the 1ds/1ds2 sensor size).  If Mamiya's fabulous quality record holds for this camera, it will blow away everything else there is.  1dsMKII?  What's that?  Why go Corvette if maybe the Ferrari is financially in reach?

[Sfleming]

No, I'm not talking about  chicken feed.  I'm not averse to spending money.  I rather enjoy it actually. ;]  

I'm just  standing for the  position that these huge expenditures  for cameras that are outdated every two years  is  not logical unless you  are in a business  that is  paying for  them within their obsolescence period.

I don't think this position gets aired all that much on photography websites and it should.

Digital is not 'HERE' yet  for  fine art shooting.  It is not sensical.  All we ever hear however is the 'sold-out-to-digital' viewpoint.  

There are lots of guys selling lots  of prints and not many of them are doing it with digital capture.  It's too bad we can't get the numbers but I think they are something like 8 to 1.  It's just that the 1 bunch is the web bunch for the  most part.