How much better will digital get?

[BJL]

I wonder how much NASA paid for their 8mp sensors 10 years ago.... half a million each?

I am not sure that NASA even had them. The sensors in the current explorations of Saturn and Titan, and recent explorations of Mars, are 1MP or less. That is due to a mixture of reasons
- the roughly decade lead time from design to mission completion,
- the requirements of the sensor being proven to be very rugged for its long, sometimes very cold, sometimes very hot, sometimes violent trip,
- the limited transmission bandwidth, and
- for most of their purposes, mosaicing a number of moderate resolution, narrow field of view images is more flexible and cost effective than taking higher resolution wider angle images, or packing a zoom lens.

[Ray]

Geez! Isn't this a long thread! Okay! The problem: Can modern lenses pick up number plate details from outer space? John has suggested intelligent interpolation of detail from other sources. Who can argue with that. I'm sure it happens. But I'm also sure it doesn't happen when it should.

[Lisa Nikodym]

If you didn't have a bunch of Nikon glass, the 1Ds-MkII would be a solid long-term choice; a 1Ds or 1D-MkII would be solid alternatives if you're on a budget, depending on what you shoot.

Geez, Jonathan, it sounds like you're the guy in the Ferrari in Stef's story, advising us to buy a Corvette if we're on a budget. :laugh:

Lisa (whose most expensive camera is a D70; Honda Civic equivalent???)

[John Camp]

For the people who say, "Wait for the D2x" or maybe a FF Nikon, I would agree that if you've got a lot of Nikon equipment, like I do, that wouldn't be a bad idea. The problem is that some of us (like me) are getting a little old. I figure that if I perform according to the actuarial tables for a guy of my age and health, I've got maybe -- maybe, optimistically -- fifteen years of shooting left. If I wait three years for a Nikon that will make a Epson 4000 print as good as a 1DsMII, that's 20 percent of my remaining time. As I said, some of us have reason for wanting the really good stuff now.

I'm very aware of the ZD because I own a Mamiya RZ system; but if you drive long distances to shoot, you sort of want a backup. Two ZDs will probably run $26,000 to $30,000; two 1DsMIIs will run $16,000, and the lens selection and autofocus (important when your eyes get older) is better with Canon. And nobody has ever shot a ZD. As far as I know, they might be building them the same place they're building the full-frame Nikon.

I shot an F5 from the day they came out, and really think no better camera was ever made. The ergonomics were wonderful, the camera was as durable as a hammer. But I hate wet darkrooms. If only Nikon could make a 1DsMII digital equivalent based on an F5; I'd buy two in an instant. Nikon had the pro market to itself for so long that they not only got arrogant, they got stupid. If somebody from Nikon would call me up and say, 'We believe that we will match the 1DsMII next September,' I'd wait. But even with the rampant disputes on the net about when the D2X is coming out (and the D2X won't match the Canon), Nikon hasn't said a word after the original announcement, even when great doubt is being case on its veracity. No reassuring comments, nothing. That's just plain dumb.

As for Kodak, I like some (most?) prints I get from my SLR/n, but the range of acceptable conditions is too narrow (can't really shoot after about three o'clock in the afternoon), and the camera itself is a kludge. Doesn't anybody from Kodak actually take pictures? You can work around the problems (like taking a Dremel tool to the knobs on your shift lens to make it fit the body) but I don't want to. I want the camera to be right. I actually think the suggestion that Kodak might get out of making DSLRs is a good one, because they're not good at making them.

So. I'm keeping an eye on Nikon and Mamiya through the PMA, which I think is in February. If nothing happens then, some of you may have the opportunity to buy some really good Nikon-based equipment really cheap.

JC

[Sfleming]

Why the lens situation?  Because they don't care.  They don't listen to us and they don't have to.  The situation can only get worse.  We sold our souls when we went whole hog after digital.

I wish I could find a good Plaubel Veriwide 100.

http://medfmt.8k.com/mf/veriwide.html

[Bobtrips]

Another little discussed reason why more MP will stop to make sense very soon is hand shake...

I have never seen a serious 1Ds vs 1DsMKII for hand held shooting, but I am 90% sure that there is virtually no resolution advantage for the 1DSMKII when shooting hand held in most situations.

For all the people shooting handheld, the only solution will be to go for larger sensors and MF is then the only option. Here also, we are probably close to the limit though.

The reasoning here, I believe, involves the size of the individual photosites.  

Based on the size of the photosites used in consumer, fixed-lens cameras you could cram zillions of sites on a full-frame sensor and still hand hold the camera.

(Or am I missing something?)

The 'sites per area' issue is most likely going to be limited by noise and ISO issues.  And if those limits can be significantly pushed then half-frame cameras will start to replace full-frames and the MF digital back market really will be eaten up by full-frames.

[didger]

didger, you say again that the only way out of the jaws of too little DOF or too much diffraction is larger formats, even after I refuted it and pointed out that there is NO WAY out of that dilemma once pixel counts get too high.

It's a question of pixel count relative to sensor "real estate", not a matter of absolute pixel count.  Do you think that a 16.7 Mpixel 8x10 back would give you the same diffraction and DOF constraints as a 16.7 Mpixel 35mm sensor?  If that were true, then 8x10 film would be similarly constrained and useless.  There's reasons why people go for the hassle and expense and cumbersome shooting of large format.  The same reasons (higher resolution) apply to digital sensors in just the same way as film.  

Refuting is not the same as proving.  I can refute the theory that the earth is round.  MF sensors will never rival 8x10 either, even if they could ever pack that many pixels into the relatively small size of a medium format sensor.  At this point it's still highly speculative just how high sensor density can get without noise problems and how good lenses can be made.  The lens problem seems to be the biggest bottle neck and I've seen nothing whatsoever to indicate that 35mm lenses (esp. wide) will ever properly support even 16 Mpixel sensors and I similarly see no hope of MF lenses ever being able to support much more than about 25 Mpixel backs adequately.  Forget rivaling 8x10, or even 4x5 at largest print sizes.

Using any given effective aperture diameter and focus distance gives fixed proportions between image size, circle of confusion sizes and diffraction spot size, regardless of focal length. So on prints with the subject appearing at a given size, you get the same degree of blurring from both diffraction and OOF effects.)

Yeah, but if you have low to moderate sensor element density, then number of sensor sites is the resolution limiting factor more often and you have relatively more perceived DOF.  If you have a huge sensor site density then the optics are the resolution limiting factor, whether it's basic lens resolving power or DOF constraints (leaving you a much smaller area in your print that's up to the resolution level that this ultra dense sensor is capable of).  Same thing for diffraction.  The lenses diffract the same at a given aperture regardless, but a denser sensor reveals a smaller amount of diffraction blurring than a less packed sensor.

I expect that electronic sensors can improve to the point that lenses are the dominant determinant of image sharpness, contrast and such

No need to wait.  The 1ds sensor already outperforms a lot of lenses (even very expensive ultrawide zooms) and 1dsMKII outperforms almost all lenses available for the camera.

[Jonathan Wienke]

Seems to me there's some powerful resolving going on in the Pro I's glass.  Is there something about smaller lenses that makes them better?  (Or am I lost in space?)

Lenses that cover a small image circle are much more easily made to resolve more lp/mm. The bigger the image circle, the harder it is to maintain a decent MTF at higher lp/mm. The Pro/1's lens resolves more lp/mm at 50% MTF than the EF 135/2L, but it would vignette like crap in front of a 1Ds sensor.

[Jonathan Wienke]

Well, just a few minutes ago Jonathan's 24x36 print that he made for me arrived and it's incredibly close to what the monitor shows and also incredibly sharp.  The sharpness and absolute absence of any artifacts from pretty massive processing really amazed me.  Of course, I had to do a lot of cleanup before I sent the file to Jonathan, but the whole procedure from shooting through printing really worked.

I assume from this that you found the print satisfactory?


Did you take a look at the CD included with the print?

[BJL]

About well capacity and signal to noise ratios. The short answer is that with current sensor technologies, larger photosite areas give greater dynamic range: a greater range from the high luminosity level beyond which highlights are blown out because the electron well is full, to the lowest level where signal to noise ratio is tolerable in the shadows.

For barely acceptable appearance in shadows a local S/N ratio at individual shadow pixels of about ten is needed according to Kodak, and as Ray indicates, this needs at least 100 electrons counted just to overcome photon shot noise. (Other noise sources push the practical lower limit up). If your well capacity is about 100,000 (as for current medium format back sensors), that optimistically gives a range of about 100 to 100,000, or 1,000:1, or about ten stops from highlights to darkest decently reproduced shadow areas.

Bigger photosites gather more light at the same ISO, and hence have better S/N. But to get the same shutter speed, same DOF, same resolution needs focal length increased in proportion to pixel size and aperture ratio in increased in proportion to focal length, and so a higher ISO. The total rate of illumination per pxel is then the same (larger pixel, but lower intensity due to higher f-stop). So, same FOV, DOF and shutter speed give the same photon count and same S/N ratio at each pixel in the image, but the bigger sensor has more highlight headroom.

So, increasng pixel and sensor size gives you choices like
(a) less visible noise, but less DOF and/or longer exposure time ("same highlight range, better shadows, more OOF or motion blur"), and
( same DOF and shutter speed, same visible noise levels, more highlight headroom ("same shadows, more highlight range").

[BobMcCarthy]

smaller sensors; smaller, lighter, cheaper, and easier to build lenses.

'Fraid not. That would only be true if the resolution didn't need to be increased to match the smaller sensor sites.  

I don't think its quite that simple. Lots of lens defect are less applicable with smaller field to cover. I know a large format lens fight the old curvature of the focus plane. Stopping down for more depth of field is almost a must. Fast LF lenses are for achieving a bright image to compose and do preliminary focus with a touchup upon stopping down.

Another practical issue is how to best use the massive installed base of lens' in the market, and the current higher end manufacturing capacity largely devoted to 35mm. because the typical lens is sharpest in the middle, what format best uses the quality of the installed base. I always thought changing the 35mm digital from 1x1.5 to 1x1.25 would be very useful. It fits the old 8x10 or 16x20 paper standard. Even a small reduction in size should help.

What I do know is there are a number of very bright people working on this as we speak. Change happens fast now days.

Nice thread. BJL, I appreciate your posting. I find this a very interesting topic.

[didger]

And those 8 megs of Pro 1 photo sites sure are tighter packed....

(Now Digger, you ought to be clamoring for things to move in the direction of smaller sensors/smaller lenses.  Imagine sticking a 22 meg camera in your shirt pocket on a future trip high into the mountains.)

Well, we're not talking about 8 megs here, we're talking about going beyond 11, going beyond 16.  I see nothing on the horizon in terms of sensor technology or lens technology, especially not lens technology, that promises much in the way of going beyond 11 megs for 1.6x sensors or 16 megs for full size 35mm sensors.  I AM an SF fan, but I know the difference between science and science fiction.  Nanotechnology is a nice buzz word and means so many different things in different contexts, but  I have not been informed of anything on the way for lenses that has anything to do with any definition of nanotechnology.  Sure, I'd like a 50 Mpixel shirt pocket camera with a 12 to 500 zoom lens and perfect image stabilization and 15 stop dynamic range, but clamoring for that?  That's wasted clamor.  I do expect to own a 22 Mpixel MF outfit within a year or so, however, and that will be as backpackable as my present 1ds kit.  It relies on no clamoring for SF to become reality.  The technology is here already and it only needs to be packaged more conveniently and get a little cheaper.  ZD may be that thing, but if not, something else will be here soon.  Didger says so.

[Peter McLennan]

Judging by the number and length of posts on this topic, I'd guess the shooting condistions are poor out there.

Here's what I want the boys in the lab to work on.

1) 25 MP imaging, 35mm form-factor sensor.  I can use legacy lenses and that's enough data for a 24X36 inch print at resolution better than my eyes can see.

2) Lots more exposure latitude.  Three more stops should do it.  I want to put all my grads on eBay.  

3) Get rid of that stupid mirror and curtain shutter stuff.  It's complex, noisy, uses power, wears out and makes my pictures bad at long exposures.  Use electronics to make the whole thing solid state, silent, programmable.

4) While you're getting rid of that stupid mirror stuff, make the viewfinder MUCH image bigger and brighter.  If you ever have the chance, peer through an Arriflex or Panavision movie camera.  Now THAT'S a viewfinder.

6) Perfect the IS process.  Put it in the body.  When it works, it's a huge benefit.  Unfortunately, currently, it's not constistent.  My 200mm Nikkor VR has made motion-blurred images at 1/250th.

7) Why not three chips, one each for RGB?  No self-respecting cameraman uses a single-chip camcorder, why not us stills users?  Bye-bye Bayer.  BJL: what resolution gains could we expect if we ditched the Bayer algorithms?

Ruggedizing/weatherproofing  My D70 won't tolerate much abuse.  I crave abuse.

9) In-camera noise reduction improvements.  Noise is always with us.  The earlier in the process we eliminate it, the better.

10) Foolproof sensor dust elimination.  Whatever it takes, boys.

[howard smith]

"Instead of an RRP of $1000 for each and every lens, a lens could come in 3 flavours; $2000 for the few that are spot on; $1000 for the bulk of them that are average, and $500 for those near the lower end of the QC range. Total revenue the same."

No so fast my friend.  This assuems perhaps that the number of $500 lenses and the number of $2000 lenses balance to $1000.  It also assumes that the market for these lenses would not change.  If fact, no one or a fewer number of people would want the $500 lens, and maybe a lot, or no one would want to pay twice as much for the lens they may have gotten for $1000, especially if they cherry picked.  Then there would be $2000 lenses that were only a tad better than the best $1000 lens.  "I payed twice as much as you did, and I can't see any difference.  What a rip off."

MAybe everyone woulf be happier if they just assumed they would get the worst example, and anything better is a steal.

[BJL]

"Expose to the right" is histogram-speak for "expose for the highlights", classical advice for users of sensors such as reversal film that have distinctly limited latitude for overexposure.

"Expose to the left" is overdoing it if taken literally, but it sounds something like "expose for the shadows", equally classical advice for negative film usage, where overexposed highlights are less of a worry becasue there is more latitude for burning them in. That is, choose exposure level to ensure sufficient exposure of all relevant shadow regions, but not very much more, because the cost of further increased exposure is longer exposure times. If future sensor technology gives us a nice long, gentle "shoulder" on the response curve, this could make sense.

[didger]

If people would only let 35mm lenses be 35mm lenses....

Hey, hey, I'm definitely among the people that are open to that, as long as camera manufacturers don't insist on trying to squeeze MF or LF performance into a 35mm sensor.  I find that even though a 1ds is double the resolution of typical 35mm film, my assortment of lenses is equal to the tasks demanded by that sensor resolution.  However, even those Zeiss distagon 18 and 28 lenses are only just good enough and the best Canon wider than 35mm are also only just good enough at best.  The way to go beyond 16 MPixels is with larger sensors so that we can let 35mm lenses be 35mm lenses and not MF and LF lenses.

As for 18x12 prints being just fine with whatever kind of lens, sure, but what if I want to print really sharp looking 30x40 even for a reasonable viewing distance?  I'll need to go beyond anything like present day 35mm lenses (especially ultrawide).

[howard smith]

"I agree that most of the price difference between Leica and more mainsteam but good quality lenses is related to economies of scale, plus perhaps the labor costs of satisfying customers who will only accept a lens that has been hand assembled and inspected in Europe or the USA rather than say Asia (Japan is off-limits for many lens snobs, let alone China.)"

This is driving Leica toward a niche market.  The price of a niche product is driven by different factors than the mainstream.  Some people will buy a Leica (as an example only) simply because it is a Leica, not whether it is a good value.  The niche market is far from a perfect economy.  "You get what yo pay for" doesn't work, unless part of what you want is the name on the box.

Total profit, as opposed to profit per unit, can be increased by increasing cost, if that additional cost allows for more units to be sold.  If I buy some glass for a higher price, the cost per unit goes up.  But if the extra glass will allow more units to be made and sold, profit can rise.

Total profit can be increased by increasing price, even if fewer units are sold.

Reducing cost will increase per unit profit at the same price.  What that really gives a company is operating margin.  They can make more profit or still compete by reducing price to amintain or increase market share.

In my opinion, Canon may be at a disavantage in the bigger than 35mm camera market.  They don't make (that I know of) lenses for bigger cameras and getting into that market could be costly.  Maybe someone like Kodak will make a larger format digital that uses whatever available lenses, like their current DSLRs.  The merger of Imacon and Hasselblad suggests a digital back for Hasselblad (?).  Mamiya is going to larger format digital, but they have the MF platform to start from.

[howard smith]

The Cassini-Huygens mission to Saturn was launched from Cape Canaveral on 15 October 1997.  It took 6.7 years to just get there.  So the camera could be no better than what was available 7 years or more ago, plus all the additional time suggested by JBL to design and build the craft.

[DiaAzul]

What if you were to take x number of your satellite's highest-res photos and then applied some kind of sampling software to come up with a most probable licence number? If they managed to distinguish three grapefruits on a contrasting background, then a license plate might not be entirely out of the question. License plate numbers and letters are almost by definition set up to be easily distinguishable, and each letter/number is a significant fraction of the visual size of a grapefruit.

You would also have non-photographic data to add to your observation: the car is in California and the letters and numbers seen to be right for California; so you scan the California license data base on the most probably numbers and find that of your fifty best guesses, only one is on a readily-distinguishable Hummer...  

JC

Wouldn't it just be easier to paint their car's license number in big letters on the roof of the vehicle? If the authorities really wanted to track people from space there are far easier methods than high resolution optics and expensive satellites.

[didger]

If someone 5 years ago had told me that there would be any sort of market at all for an $8000 35mm camera I'd have figured they were crazy.  If they had told me that I'd be spending way over $10,000  for a complete outfit I'd have known they were totally around the bend.  I reckon I'm not going to try to predict limits for what people will spend money for and how much.  I do feel confident that small formats will improve and get cheaper and that larger formats will also happen in a serious way and the technology and prices will get better partly from larger format development in itself and partly by incorporating new technologies and improvements made in the much busier small format sector.  If a relatively poor semi-pro like me can buy such expensive stuff (it's all a matter of priorities), how small can the potential high end larger format market be?

it occurred to me to wonder exactly what I'm hoping/expecting/demanding to get from better lenses and sensors...Maybe the ability to make loupe-quality 3-foot-by-five-foot shots?

I can't comment on your hopes and expectations;  you know best.  However, my reasons for going for more and more resolution have nothing to do with "loupe-quality 3-foot-by-five-foot shots".  A 24x36 print viewed from several feet away will look better with 1ds than 35mm film or small sensor cameras and look much better with 4x5 than 1ds.  Just check out a Michael Fatali or David Muench exhibition some time and compare the quality with Galen Rowell large prints.  These are all highly competent and inspired photographs, but resolution DOES matter if you want the best looking large prints.  You DON'T need a loupe to see the difference, even if you are happy with the stitches in that quilt.  Just look at some large prints from large format media.  All will be clear.