OK that is more reasonable! I must have imagined you speculating about a future low cost 6MP 35mm format sensor option.

a) I have not noticed anyone in this forum arguing that smaller formats wil completely displace 35mm foramt and larger, but I suppose some such hard core "little-endians" do exist. I am personally willing to concede a continued 5% or so of DSLR sale volume to formats bigger than 1.5x!

I suggest that the two best argument for lower size limits are
1) downsizing lenses for smaller formats will naturally be limited to improving resolution (lp/mm) not quite as fast as image size reduces, so a smaller format is inherently limited to somewhat lower ability to record detail, in the sense of "line pairs per picture height".
2) optical design limits on aperture ratios lading to permanent speed disadvantages for too small formats. Maybe there will never be excellent optical performance below about f/2 or even higher.

almost certainly not. Shot noise is an observed feature of the light arriving at the lens; we are stuck with it.

c) almost certainly yes. Yet another possibility with CMOS is sensors with an extremely high frame rate, so that they can be read numerous times during a single exposure, completely eliminating highlight headroom limits. One way or another, I am very confident that highlight headroom can be eliminated as a limitation on dynamic range.

d) absolutely yes! Once highlight headroom is essentially eliminated as a constraint, there is absolutely no DR advantage to a larger format, as the DR is unlimited in either case. That leaves only shadow noise/speed trade-offs, which naturally tend to either a tie or a win for smaller photosites so long as fast enug lenses canb be used.

The question on this point is whether larger sensors with their inherently larger dark noise electron counts can compete, in particular given their naturally higher cost. The lens limitations (speed and resolution) menioned above are the big-endian's best chance, with compact digicam formats already permanently boxed out of the high end I would say.

The joker in the deck is the market.

At this point in time more people seem to want little things than want big things.  If the goal is to produce a camera capable of making a 12" x 18" print with great resolution and great DR and great everything else and you've got two cameras - one big and one little ....

I don't know anyone who would ever argue that little cameras will somehow be more 'capable' than large cameras.  MF film cameras were more capable than 35 mm film cameras. The market went with smaller.

Thing is, small cameras are likely to keep on getting better over time.  Half-frames are replacing 35 mm film, full-frames are replacing MF film.  In a generation or few we may see quarter-frames doing the work that 35 mm film once did and full-fames moving into the LF film world.

Ray, you seem to be severely misunderstsanding what I am saying about eliminating dynamic range limitations. It is NOTHING to do with increasing sensitivity, or the maximum usable ISO; quite the contrary in fact.

Both blending and the other methods already being tried out work by using long enough exposures to handle shadows well, while avoiding blown-out highlights. The plausable goal is totally eliminating the UPPER limit on the amount of light that a highlight photosite can receive during an exposure without problems, not lowering the minimum amount of light needed, which I think we agree is ultimately limited by photon noise. It is about allowing long exposures and very low effective ISO speeds.

The elimination of highlight headroom limits is very far from Star Trekky wishful thinking; let me say something yet again: methods for doing this are already in use, in video surveillance cameras! (And blending is another less convenient such method.)

At heart, all that is required is to check the photosites periodically during an exposure, and read out the ones that are close to full. By noting the different read-out times at different photosites, the correct relative illumination levels can be computed. Thus every photosite can potentially be read out when it has received enough light to have a healthy S/N ratio, if you can wait long enough on the shadow pixels. And motion blur will be minimized in well lit regions, since they are effectively exposed at the highest ISO appropriate to thier light levels!


P. S. you might be more enthusiastic about my argument that there will never be a good quality, general purpose, interchangable lens camera system in a format smaller than the current smallest DSLR format, 4/3. (Ignoring the debate about whether there is or will ever be such as system in 4/3 format!) The lower size limit set by the "law of diminishing returns" has been reached if not exceeded by current DSLR formats: camera and lens costs and sizes can not be reduced much further, and so on.

I suspect we will be able to  divorce shallow DOF and aperture before long.  

So if we are limited to f2.0 or even f2.8 (as opposed to f1.4 for larger formats) do you think people will buy a larger (and very likely more expensive) system for that additional bit of performance?

(As an aside, I'm assuming that 1/50, f2.8, ISO 100 is approximanetly the same exposure value across formats.)

BJL,
Are you absolutely certain about this   ?

The central part of my argument got lost somewhere. It's this. I agree that the application of much technological wizadry could eventually create a small camera capable of outstanding performance on a par with what we currently get from bigger DSLRs and even what we hope to get in the future from DSLRs, but only if the research dollars go in that direction and cease to flow to further development of the larger format cameras.

The argument that this appears to be what is happening and that large format digital cameras will eventually cease to be economically viable and become dinosaurs, is another issue.

I am merely concentrating on the inherent potential of two hypothetical formats that differ greatly in size. In relation to that argument you are making some controversial statements, the logic of which I find difficult to follow.

Let's get down to some more examples using simple maths. Compare two sensors of equal pixel density but the smaller sensor has 1 micron photosites and the larger sensor has 10 micron photosites. The smaller camera has an f1 to f4 lens. The bigger camera has an f4 to f32 lens. DoF at f1 on the smaller camera is equal to DoF at f10 on the bigger camera. Both sensors have eliminated all dark noise. Photon noise is the only noise.

What you are proposing, if I've understood you, is that we could completely sacrifice the speed advantage of an f1 lens by using the same exposure duration that the larger camera would use at f10. As soon as a photosite is full, the charge is read, transferred, dumped (whatever) but essentially noted. The fact that pixel #1124 was filled 200 times during a 1/100th sec exposure and pixel #13 only once, is recorded and used to construct an image with full dynamic range.

This is a great idea, but suppose the larger camera adopts the same technique?

Here are what I think the issues are.

(1) Nothing happens instaneously (except perhaps in fictitious quantum computers using qantum entanglement). In practice, if we want to get that photosite #13 to fill just once (representing the darkest part of the image) it is going to take a somewhat longer exposure than the larger camera would use at f10. This effectively means the smaller camera would have optimal performance at say ISO 64 as opposed to ISO 100 for the larger camera. So already the larger camera has a speed advantage at equivalent DoF before we've even applied this amazing technology; the opposite of what we're used to.

(2) But the big issue is the significance of the far lower S/N of the bigger sensor. To refresh, 1000 electrons at full well have 33 electrons of photonic noise. 100,000 electrons have only 317 electrons of photonic noise.

If we employ this same technique with the larger camera, we don't need to wait until the well is full. For the same Signal-to-Noise, we can set the camera to transfer the charge (or note the charge) when the well reaches 10,000 electrons (1/10th full well). This effectively means we get the same DR performance of the smaller camera, but at a yet higher ISO setting.

And we haven't even begun to consider the benefits of those diffraction limited F4 lenses on the bigger camera, which I surmise would be ideal for noise-free hand-held shots at ISO 128,000 (or even higher) in the Vatican Museum   .

Let's face it, BJL, these tiny digicams, however sophisticated, are just toys for the consumer mass market   .

Note, I'm jumping in here a second time to correct my maths. Such gross errors are intolerable (no calculator at hand). I figure 33:1000 is equivalent to 317:10,000. The bigger well, 100x the area, needs to be only 1/10th full to match the S/N of the smaller well; ie. we can get the same results at f10 using less than 1/10th the exposure, combining points (1) and (2).

It can be seen if we start using F4 on the larger camera, the ISO benefits (shutter speed) are enormous.

Ray,

  we seem to be close enough to agreement, but this comment is very telling. How do you think the allocation of research efforts is going between formats, maybe grouped as digicam (2/3" and below), the mainstream digital SLR formats (4/3 to 1.5x), 35mm format (including 1.3x?), and bigger than 35mm format. One dominant trend I see is that sensor and camera makers want to reduc costs by getting the best performance possible out of the smallest sensor possible, but with significant inertia in DSLR sensor sizes once a DSLR format has been chosen and lenses made for it.

The indications from the rate of introduction of new sensors and new camera models are of a bottom heavy pattern, with every step up that size ladder having a lower level of activity. This is almost inevitable given the very bottom heavy distribution of sales and revenues. As far as I can tell, Digital Rebel (or D70) revenues alone greatly exceed those of all 35mm format DSLRs combined.  In fact, as far as I can tell, even the Four Thirds format probably now outgrosses all of 35mm in the digital world. And 1/1.8" probably trounces 35mm.

So my guess is that amongst interchangable lens camera systems, the current mainstream DSLR formats will have the fastest technological development. Just as MF fell of the R&D pace decades ago, maybe 35mm format is next.

Are you sure that's fast enough, BJL? I'm no mathematician or electronics expert, but as I understand a GHz is a thousand, million Hertz.  0.000001 is a millionth. A millionth of a GHz is a mere 1,000 Hertz. Okay, what's the occasional zero between friends .

For sampling to be accurate, it needs to be a 'bare-bones' twice as fast as the activity that's being sampled but preferrably at least 4x as fast or even greater. If we try to mimic with a small photosite the performance of a larger photosite with a full-well capacity of say 100,000 electrons that might fill in 1/100th of a second exposure say, then we'd be looking at a 4x sampling rate of around 40 million Hertz. No problem I guess, but what about the time it takes to discharge and reset a particular photosite and the time it takes to process that data in 16 bit mode (and possibly much higher with this technology) for a good quality RAW image, plus all the other noise reduction techniques that take up real time?

If you shift ground and exclude all the P&S cameras, which there's good reason for doing because there seems to be a big size gap between the 2/3rds format (6.6mmx8.8mm) and the Oly E-1 (13.5x18mm), then all the arguments become more nebulous. Absolute pixel size (or more accurately pixel pitch because manufacturers don't seem to publish the actual pixel size) seems to have less effect than design, noise reduction techniques and such factors as 16 bit processing (or greater) as opposed to 10, 12, or 14 bit processing.

I just did a quick check of the format sizes ranging from the Oly E-1 to the P25. The size differences are fairly evenly spaced with the biggest gap occurring between FF 35mm and the P25 which has about double the area.

It's to be expected that there's going to be a bit of temporary overtaking between one format and the next one up in size.

You must be a bit disappointed that the small brother of the DSLR brigade (the Oly E-1) has not yet overtaken anyone  :(  .

Ray,

  do you really think that I believe that smaller formats are overall superior to larger ones? If so, I need to clarify my mostly centrist position.

a) many claims for natural advantages of one format size over another are greatly exagerated or completely untrue, being based on overlooking some relevant factors (e.g. ISO vs f-stop.)

larger formats tend to have some advantages, at least with current technology:
i) Bigger maximum aperture diameters are possible, allowing greater extremes of high shutter speed and low DOF.
ii) Greater dynamic range when it is possible to get "full exposure" (exposing to the right at minimum ISO, and so almost filling the electron wells at highlights).

c) Smaller formats also have some natural advantages:
i) Less dark current noise (electrons per pixel), potentially improving S/N ratio when a need for both adequate shutter speed and adequate DOF resticts one to less than full exposure (i.e. all electrons wells significantly less than full, even at the brightest highlights).
ii) Size, weight and cost, which can indirectly lead to higher image quality by allowing you to own or carry more or better equipment.

d) With advantages in both directions, it is not the case that either a bigger or a smaller format has an automatic image quality advantage; such simplistic dogma is for partisans of the "digital format size cold war". Only when the specific photographic goals and the specific formats being compared are specified can a useful image quality comparison be made. E.g. is 16x24mm (Nikon) or 645 medium format better for action photography?


If I seem to favor smaller formats, it is mostly because I am so often responding to exagerated or untrue claims in favor of larger ones: partisans of "bigger is better" are far more common in discussion forums than partisans of "small is beautiful".

I tend to look at camera selection in terms of the percentage of possible shots that one will not be able to take with a given camera.

There's possibly a camera out there which will 'do it all'.  Maybe a special built MF jobbie with incredibly fast burst rates and capable of using very long lenses.  Something in the, say, $75k+ range that weighs in a 25+ pounds.

OK, none of us (except Michael ;o) is likely to go that route.  So whatever we choose means that we write off a certain percentage of possible shots from the entire picture universe.

As soon as you decide that full-frame 35 mm digital is the answer you've limited your shooting portion of the universe.  But due to cost and weight decisions you may choose to pass on the non-achievable shots.  

The same holds with other camera choices.  I find it uncomfortable to walk for 8-12 hours a day, day after day, carrying even a dSLR and a couple of lenses.  

I'm interested in seeing small sensor lens cameras improve in ability so that I can reduce the (even now currently small) number of shots that I can't get with current monster zoom cameras.  

I pass on the low ambient motion shots, I have to struggle to get some indoor shots without flash, and I have to use PS to create shallow DOF.  But I get the 95%+ shots that I want with a one pound camera.

Get me an 8 meg 12x camera with Fuji's high ISO ability and some software that does a better job of shallow DOF and there's little that I won't be able to do with a tiny monster.

Ray,

  I believe that you are one of many serious photographers who did this with film: a 6x7cm format camera as well as a 35mm, right? And this somewhat common approach of using both 35mm and medium format usually has the pattern I have suggested:
- the smaller format is better for "limited light" situations where the larger format would need to be used at uncomfortably high ISO speed, with long telephoto being a common case
- the larger format is better in situations where you can get all the light you need to work at low ISO: tripods and stationary subjects, flash or studio lighting usable, and so on.

What I envision is that for a sufficiently serious photographer it could similarly make sense to use several digital cameras in formats at similarly large jumps in size, about a factor of two to four in sensor area, or "one to two stops". Smaller image format changes than that can probably be handled adequately by cropping down to the desired size, so using the next available format up from what is "ideal" for the situation.

Of course our host Michael does this at about four levels! Mainly 36x48mm ("near 645") and 24x36mm ("35mm"), but also with 15x22.5mm ("1.6x"), and one or two digicams at 6.6x8.8mm (2/3") and smaller.


P. S. If and when you add a 24x36mm format DSLR to your kit, I think it will be more likely that you prefer the 15x22.5mm format 20D to it on some of those long walks!

P. P. S. I like your idea of redefining "digital medium format": the most common DSLR formats from 13.5x18mm (4/3") to 16x24mm (DX) do seem to sit in the middle: about three time bigger than the dominant "digital miniature formats" like the 4.3x5.8mm of 1/2.5", with the "digital large formats" of about 36x48mm being about three times bigger again.

So, three levels of 'bodies'.  

Full frame for highest resolution and very wide angle work.  A replacement for MF film.

Half frame/high pixel count for long lens maximization with high resolution.

Half frame/moderate pixel count for a "walk around", more affordable camera.  (This can be the small bodied version a la 350D.)

And the lenses largely interchangeable across the three bodies.

There might be some super small/light lenses for walking.

The "64 megapixel question" is this:
how good can the MTF be at about 50lp/mm and distances of up to 14mm from the center of the frame be for (a) current good Canon 35mm format lenses, and ( future professional level EF-S lenses, if and when Canon makes any such.

My guess: if Canon chooses to do it, they will be able to make lenses that can handle the job.


By the way, I read recently of an analogy between some DSLR developments and what Oskar Barnack did almost a century ago: creating a new high quality camera in a far smaller format than ever before, exploting the increasing resolution of the light sensitive media available, and relying on designing new smaller format lenses of distinctly higher resolution. That was the essence of the Leica revolution.