Ad
Ad
Ad
Pages: « 1 2 [3]   Bottom of Page
Print
Author Topic: Leaf backs at 28MP, 35MP  (Read 24172 times)
BJL
Sr. Member
****
Offline Offline

Posts: 5121


« Reply #40 on: January 02, 2006, 07:16:20 PM »
ReplyReply

Quote
With the exception of a few really expensive lenses such as the Canon 200/1.8, you are limited to a maximum resolution at around f8 to f11
[a href=\"index.php?act=findpost&pid=54989\"][{POST_SNAPBACK}][/a]
I believe that many modern lenses have peak resolution at larger apertures than f/8 to f/11; f/4 seems a more reasonable figure for good primes, and even some good zooms (e.g. the Canon 70-200 f/2.8 IS L?) In fact the Canon 17-85 f/4-5.6 EF-S seems sharpest at apertures very close to wide open, and that is only a mid-priced, wide ranging zoom (for a dead-end, low level format!) FourThirds lenses in general seem to be sharp almost all the way to maximum aperture.

It is probably time to revise that old saw that "lenses are sharpest two stops down from wide open", to reflect progress like aspherical lens elements.
« Last Edit: January 02, 2006, 07:18:39 PM by BJL » Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8853


« Reply #41 on: January 02, 2006, 07:57:07 PM »
ReplyReply

Quote
You're deliberately mischaracterizing my remarks, and straying far from the equal-aperture/equal-photon exposure comparison we have been discussing. In any such comparison, the smaller-format camera will (all else being equal) deliver a better result until it is no longer capable of imaging all of the photons without overexposure/clipping. Once the quantity of photons in the exposure exceeds the imaging capacity of the smaller format (and clipping occurs), then the larger format will produce the better result, and will continue to do so in increasingly decisive fashion as the photon count is increased until it's imaging capacity is reached.
[a href=\"index.php?act=findpost&pid=55031\"][{POST_SNAPBACK}][/a]


Not really. I've always conceded that increased dark noise of the larger format is a technological barrier. The difference in our attitudes here is that I seem to be positive about the capacity of technology to solve problems and you seem to be negative, like those who were convinced that heavier-than-air machines could never fly.

I'll also concede that all else being equal the smaller format will do a better job imaging an equal number of photons. But all else is rarely equal, is it?

I've already compared my Canon D60 with the newer technology Sony T1 at equal physical apertures and found the D60 did a marginally better job. I'm fairly convinced the 5D with a sensor about 16x the area of the KM A2 will produce a better image than that camera with the same number of photons.

As I see it, you are making assertions with no concrete evidence to back them up, which is surprising considering how elegantly you espoused the scientific method recently in another thread.

Now if you'd like to provide some factual mathematical evidence along the lines that increased dark noise is not proportional to increased sensor size, but increases exponentially, then I'll eat my words and concede that the fact that current 35mm DSLRs can produce an equal or better image than the smaller sensor with the same number of photons is not relevant to the situation of very large format.
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8853


« Reply #42 on: January 02, 2006, 08:41:43 PM »
ReplyReply

Quote
I believe that many modern lenses have peak resolution at larger apertures than f/8 to f/11; f/4 seems a more reasonable figure for good primes, and even some good zooms (e.g. the Canon 70-200 f/2.8 IS L?) In fact the Canon 17-85 f/4-5.6 EF-S seems sharpest at apertures very close to wide open, and that is only a mid-priced, wide ranging zoom (for a dead-end, low level format!) FourThirds lenses in general seem to be sharp almost all the way to maximum aperture.

It is probably time to revise that old saw that "lenses are sharpest two stops down from wide open", to reflect progress like aspherical lens elements.
[a href=\"index.php?act=findpost&pid=55056\"][{POST_SNAPBACK}][/a]


A point worth making. However, good lenses often have a consistently good performance over a fairly wide range of apertures. That's one of the features you pay for. The Photodo ratings for the Canon 200/1.8 vary from 82 to 90 across the range of f stops from f1.8 to F8. That's roughly a 10% variation.

The fact is, when a lens is diffraction limited at the apertures being used, resolution increases inversely in proportion to the f stop number. With the 200/1.8 there's a sweet spot at f4. The lens is marginally sharper at f4 than at f8. If it was diffraction limited at f4, resolution would be double that at f8. This is the difference I'm talking about. Large format allows one to use the lens at diffraction limited apertures (or at least closer to the diffraction limit) whatever your DoF requirements are. Using a lens at a diffraction limit could be thought of as using that lens with 100% efficiency.

If we consider the performance of the wide angle lens designed by Paul Weissman for the Gigapxl project, a lens at f22 that is diffraction limited will have a maximum resolution of 66 lp/mm at 9% MTF. The MTF chart for this lens shows an MTF of 50% in the centre at 30 lp/mm, falling off to only 30% in the corners if the lens is used with the smaller format 8x10. I don't know what the resolution would be at 9%, but it seems a reasonable assumption it would be close to 66 lp/mm, at least in the centre. This lens seems pretty close to being diffraction limited at f22, wouldn't you say?
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8853


« Reply #43 on: January 03, 2006, 07:57:35 PM »
ReplyReply

Quote
But in order to make use of this increased resolution potential, large-format cameras must image more photons per exposure than their small-format counterparts. And that means that their practicality decreases as format size increases.
[{POST_SNAPBACK}][/a]

Image more photons or use technology to remove dark/thermal noise?

The great fallacy in your argument, Jonathan, is this. Whilst it's true that the smaller camera with the smaller sensor has an inherent advantage imaging a small number of photons, to fully exploit and capitalize on this inherent advantage requires the application of state-of-the-art noise reduction technology, including microlenses to increase sensitivity, dark frame subtraction, on-sensor signal processing and super cooling with micro-miniature refrigeration devices to eliminate almost all dark noise.

Such a small format camera would have no peer amongst the larger formats when it came to taking reasonable resolution images in low light without flash and with short exposures. But such a camera would no longer be small, light and cheap. It would have a high ISO setting with remarkably low noise, but still only be capable of a small dynamic range because of its small pixels and relatively low resolution because of the small number of pixels. It would be a specialised camera suitable for certain scientific applications, and probably as big and heavy as a 1Ds.

People expect large format cameras to be big and heavy. I notice that Sony now has a notebook weighing only 1.2Kg which includes a 60GB HD and (I believe a first in this category) a built-in dual layer DVD recorder. (Ideal for photographers downloading and burning images in the field).

When a camera weighs say 4Kgs, an extra O.5Kg for a powerful computer to help reduce noise is acceptable. When a camera weighs 10oz it's over the top. The same applies to cooling systems.

A browse on the net leads me to believe there are a number of lightweight and efficient micro refrigeration systems currently available, not including those in the process of development such as the following;

Quote
The devices, called "micro-channel heat sinks," circulate coolant through numerous channels about three times the width of a human hair. Such devices might be attached directly to electronic components in military lasers, microwave radar and weapons systems, as well as in future computers that will generate more heat than present computers, said Issam Mudawar, a professor of mechanical engineering who is leading the research.

The researchers are adapting refrigeration systems by using the micro-channel heat sinks to replace conventional "evaporators" components in household refrigerators that contain a labyrinth of tubing. As coolant circulates through the tubing, heat is removed from the refrigerator to cool the food inside.

"We are substituting these conventional evaporators which might be well over a meter long in the typical refrigerator with a heat sink that's only about 1 inch square," Mudawar said. "The challenge is how to unplug this large evaporator and put in its place this tiny heat sink and make the whole system work."

[a href=\"http://www.physorg.com/news3725.html]http://www.physorg.com/news3725.html[/url]


Again, a browse on the net leads me to believe that lowering the temperature of a sensor (and its environment presumably) by just 20 degrees C can result in as much as a 10 fold decrease in dark current noise. There are small, efficient cryogenic devices already available which can lower the temperature to around minus 200 degrees C in less than 15 minutes.

Imagine something the size of the average 8x10 field camera made of super lightweight carbon-fibre materials and containing inside a double vacuum thermal insulation. The refrigeration system could be initially powered from the car battery to bring the temperature down to say minus 50 degrees C (in about 15 minutes). From there on the gradual slight rising of temperature could be countered by use  a standard lithium battery back.

Taking an average of the lowest and highest rates of dark noise reduction I've seen on the net, I arrive at a figure of around 7x per 20 degree reduction in temperature. From +30 degrees (in Australia) to minus 50 should result in a dark noise reduction of over 2,000 fold. Do you think that would be sufficient?
Logged
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #44 on: January 03, 2006, 09:01:26 PM »
ReplyReply

Any thermal noise reduction technique usable on large-format sensors will work equally well on small-format ones, so the net result is a wash. If you can cool an 8x10, you can cool a 1Ds-MkII. A Peltier chip/heatsink for a 1Ds is going to be a lot smaller/lighter than one for an 8x10 sensor, so its weight will be increased much less. In both cases, the additional cost/bulk will be proportional to the imaging area, so the size/weight/cost/quality ratio between them will be relatively unchanged. The same goes for software noise reduction, which is better done in post with a real computer than in-camera anyway. Neat image can improve a 35mm image just as well as an 8x10 image. You've proved nothing here, except your propensity for wishful thinking.
Logged

Ray
Sr. Member
****
Offline Offline

Posts: 8853


« Reply #45 on: January 03, 2006, 09:21:13 PM »
ReplyReply

Quote
Any thermal noise reduction technique usable on large-format sensors will work equally well on small-format ones, so the net result is a wash. [a href=\"index.php?act=findpost&pid=55143\"][{POST_SNAPBACK}][/a]


Well that's essentially what I just said in my previous post.... except it's not a wash. The larger format sensor is capable of higher resolution and greater dynamic range. We have an instrument which is not so specialised and therefore more desirable for those wanting the ultimate in image quality. Ie. a camera that can deliver a stunningly detailed 8ft x 10ft print but also a remarkably sharp and noise free image in poor lighting conditions at high shutter speed.

Design a small format camera to produce a better image with the same number of photons, increases the weight, size and cost and leaves you with a relatively low resolution image with relatively poor dynamic range but superb signal-to noise at high ISO's. That's not a wash.

Did it occur to you that at certain sub zero temperatures there may be no further  noise reduction of practical consequence in the smaller format. The wash is in the potentially equal level of dark noise reduction, with the larger format requiring a greater reduction of temperature to achieve that. Having achieved it, the diffraction limited large format lenses provide the extra resolution with equal noise per pixel, but less noise per image because of the greater number of pixels.
« Last Edit: January 03, 2006, 09:35:25 PM by Ray » Logged
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #46 on: January 03, 2006, 10:34:54 PM »
ReplyReply

Except that it is a wash at best. Given theoretically perfect, zero noise sensors, diffraction-limited lenses, equal aperture diameters, and equal-photon exposure, net image quality will be the same. Given any deviation from theoretical perfection in the sensor department, the smaller format wins. Larger formats can beat smaller ones, but only when they image more photons than the smaller format. Period.
Logged

Ray
Sr. Member
****
Offline Offline

Posts: 8853


« Reply #47 on: January 03, 2006, 11:11:06 PM »
ReplyReply

Quote
Except that it is a wash at best. Given theoretically perfect, zero noise sensors, diffraction-limited lenses, equal aperture diameters, and equal-photon exposure, net image quality will be the same. [a href=\"index.php?act=findpost&pid=55150\"][{POST_SNAPBACK}][/a]

I'll go along with that. Except you seem to have already forgotten that diffraction limited lenses at really small f stops do not exist. If I recall, BJL has stated that around f5.6 is as wide as we can expect for diffraction limitation. Maybe f4. However, if you are talking about the small format, you've still got dynamic range limitations.

I'd agree in principle that if you were to apply all the noise reduction technology to the small sensor plus the 'still under research' frequent sampling and charge emtying of the photodiode as it fills, then there might be little point in developing a large format sensor.

However, you've still got the reducing photodiode efficiency factor as pixel sizes approach the dimenson of light waves, as you pointed out to me some years ago. You can't fit 100 million photo-diodes on a small 2/3rds sensor without each photo  site being too small and inefficient for optimal results.

I would still maintain that, after all the noise reduction technology has been applied, there is nothing the smaller format can do in respect of image quality that the larger format cannot, except be smaller, lighter and cheaper. But that of course is an advantage in its own right. If all you need is a 6mp image, say, under adverse lighting conditions, then you might as well use a relatively light and cheap camera that's been designed for the job (say a $20,000 3Kg camera as opposed to a $100,000  10Kg camera).

However, if you think you can use a $1000 P&S to capture images at fast shutter speeds that the large format camera with all its technology can't manage, then that's wishful thinking.

I'm prepared to let this rest now, having had my say.  
« Last Edit: January 04, 2006, 03:25:18 AM by Ray » Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5121


« Reply #48 on: January 05, 2006, 12:51:30 PM »
ReplyReply

Quote
If I recall, BJL has stated that around f5.6 is as wide as we can expect for diffraction limitation. Maybe f4.
[a href=\"index.php?act=findpost&pid=55151\"][{POST_SNAPBACK}][/a]
Actually I said rather the opposite; current good lenses can go to at least f/5.6 or f/4 before aberrations take over from diffraction as the main limitation on resolution. I am very open to the possibility than in the future, further development of technologies like two-sided aspherical lenses might achieve diffraction limited resolution down to f/2.8 or even f/2. Or at least, it is no less likely that your imaginings of 8"x10", 800MP sensors, which is about what would be required for diffraction to be the resolution limit at f/16.

You argument seems to be based an assuming vast progress in some areas (mostly do do with bigger sensors), while assuming that in other aspects, there is little room for further improvement (35mm format lens performance, small photodiodes). Oh, and pretending that, while diffraction spots are relevant to resolution, circles of confusion are not. Ignoring circle of confusion (DOF,OOF effects) at best makes sense only for purely planar subjects, like when copying documents.

By the way, "100 million photo-diodes on a small 2/3rds sensor" is a red-herring in a discussion that was, up to that point, all about DSLR's, not compact digicams. It is a fallacy to argue that the weaknesses of one extreme shows the superiority of an opposite extreme when the comparison at hand is to other intermediate options. (Like criticizing Stalinism to argue in favor of Fascism, or vice versa, both of which have been done of course.)
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8853


« Reply #49 on: January 05, 2006, 09:06:33 PM »
ReplyReply

Quote
I am very open to the possibility than in the future, further development of technologies like two-sided aspherical lenses might achieve diffraction limited resolution down to f/2.8 or even f/2.

So am I. When I wrote that you had said diffraction limitation at f5.6 to f4 is the most we can expect, I meant of course expect at the present time. Unless technological progress comes to an end, we can always expect future improvements in all areas.

Quote
You argument seems to be based an assuming vast progress in some areas (mostly do do with bigger sensors), while assuming that in other aspects, there is little room for further improvement (35mm format lens performance, small photodiodes).

I deliberately side-stepped the issue of whether it would ever be practical to manufacture an 8x10" sensor. I simply don't know, but never is a long time. However, I can't help wondering what happens to all those rejected sensor chips that are supposed to keep the cost of full frame 35mm DSLRs so high. A flaw that makes a single sensor unsuitable in a 35mm camera does not necessarily make it unusable in a stitched array of around 50 or so. Also, a supercooled environment allows for the possibility of much more powerful computers using super conductors. Whatever seems impossible or impractical today can become achievable tomorrow with a jump in computer power.

Quote
Oh, and pretending that, while diffraction spots are relevant to resolution, circles of confusion are not. Ignoring circle of confusion (DOF,OOF effects) at best makes sense only for purely planar subjects, like when copying documents.

I've pretended no such thing   . I've already addressed this issue but you seem to have forgotten. The whole idea of supercooling the large format sensor is so that it can provide at least the same quality image with the same number of photons as the smaller camera. If the smaller camera is not cooled, then I would maintain the larger format that is cooled would definitely deliver a better quality image with the same DoF.

To recap, 35mm f8 has equal DoF to 8x12" f64. Traditionally, f64 on a large format field camera requires a long exposure. However, if the LF sensor (and preferrably the whole interior of the camera) is super-cooled, then we can effectively use ISO 6400 which gives us the same shutter speed, same number of photons and same DoF as 35mm f8 at ISO 100.

The 'wash' that Jonathan was referring to seems to be based on an assumption that after all the techology has been applied to both the small and large format, virtually eliminating dark noise, significantly reducing read noise etc so all we are left with is photon noise, then the two systems will perform equally but only in that very narrow range where total photon count , equal shutter speed and DoF, is the same.  Increase the photon count and the larger format will have the advantage, admitted by Jonathan.

Maybe so, but I would maintain the larger format can contain more technology simply because it's larger. Is Jonathan trying to say that a notebook or laptop can ultimately be as powerful as a desk top computer? Imagine the possibilities for on-sensor processing with such a large 8x10" CMOS board.

Call me impractical if you like. I'm just trying to work out from basic principles where the limitations are.
« Last Edit: January 05, 2006, 09:08:35 PM by Ray » Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8853


« Reply #50 on: January 05, 2006, 10:04:32 PM »
ReplyReply

I think I'd be prepared to accept that if you were to design a small format and large format camera incorporating equal amounts of computing power, cooling, radiation insulation etc, so that each camera was roughly equally heavy and bulky, (the smaller format being just a little lighter) then the small format would win in that narrow range of usage imaging an equal number of photons.

But who would want such a limited camera? As I said, it would only be bought for specialised scientific purposes.

To flesh out this concept a little more I'll have to refine my argument. I've used 35mm f8 and 8x10 f64 as being equivalent because of their popular usage. In fact they are not equivalent. There's no doubt at all that LF f64 is a diffraction limited aperture. That's not the case for 35mm f8 which is more of a sweet spot. Photodo have provided a reason for not testing 35mm lenses at f11. They're all equally bad at that aperture. There's no doubt there's a slight exaggeration in that statement. The really cheap lenses would probably still show some spherical and other aberration at f11, but there would be a sameness of performance amongst all good lenses.

I think a more accurate equivalence between 35mm and 8x12 would be f11 and f90, so let's start off from that point with our heavy, technology laden cameras.

The two super-cooled cameras are giving approximately the same image quality respectively at f11, ISO 100 and f90 ISO 6400. Let's say it's a wash in this respect. As I stop down with both cameras, DoF increases equally and image quality deteriorates equally due to increased diffraction.

But what happens when I stop up? With 35mm there's a sweet spot at f8, say an extra 10% in resolution, maybe less. If the lens is really expensive, like the Canon 200/1.8, there's another sweet spot at f4, say another 10% increase in resolution. At shallower DoFs than f4 there's a deterioration. So, by stopping up from diffraction limited f11, the maximum resolution increase I can get with 35mm is about 24%.

Now let's look at what happens when I stop up from diffraction limited f90 with 8x10" format. If we take the example of the Gigapxl wide-angle lens designed for 9x18" format, we see that such a lens is close to diffraction at f22 and almost certainly at its diffraction limit at f32. Stopping down from f90 to f45 doubles resolution. Stopping down from f45 to f32 (still in the domain of diffraction limitation) increase resolution by another 50%. Stopping down further to the sweet spot at f22 (not quite diffraction limited) gives us say another 20% increase.

I make that a total cumulative increase in resolution with the large format (and all at equivalent DoFs) of around 350%, as opposed to 24% for 35mm.

Now let's get into perspective what a 350% increase in resolution means.

The 5D delivers 100% more resolution than the D30 (ie. double the pixel count in each dimension or 4x the number of pixels). A future 24MP FF 35mm camera with double the pixels of the 5D would potentially give us a further 40% increase in resolution, provided the lenses were up to the job, which it seems clear they aren't. That's a cumulative increase from the D30 to a 24MP FF 35mm of 280%.

So, having demonstrated that current LF lenses are up to the job at the present time, the potential is already there for LF lenses to deliver spades more resolution than 35mm lenses. This not marginal stuff we're talking about. We're talking about a 300% increase in resolution over and above what the 1Ds2 can currently deliver at F4 with a good prime, but equal resolution at greater DoFs than f8 all the way to the smallest aperture the lens supports.

I hope my maths is correct. If it isn't I'm sure you'll correct me   .
« Last Edit: January 06, 2006, 12:43:10 AM by Ray » Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8853


« Reply #51 on: January 06, 2006, 08:38:44 PM »
ReplyReply

Perhaps the confusion results because you think I'm saying that this dramatic increase in resolution as one stops up from f90 with the LF camera is also achievable at the same shutter speed as the 35mm format at equivalen apertures (same physical aperture).

No, not at all. Having designed the LF camera so that use of ISO 6400 produces about the same noise and resolution as the 35mm format at ISO 100, allowing use of equal shutter speeds for equal image quality in all respects, then these conditions apply across all equivalent f stops.

The dramatically higher resolution potential of the LF at f32 (compared with 35mm f4) can only be realised by increasing the photon count, ie lengthening the exposure. This of course not only results in a much higher resolution image, but much higher dynamic range. If a faster shutter speed is required at f32, to capture the same movement of subject that might necessitate the use of f4 with 35mm, then there is probably no advantage using the LF camera. On the other hand, there is no disadvantage, which is really all I've been trying to say in these last few hundred words.
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5121


« Reply #52 on: January 16, 2006, 08:27:10 PM »
ReplyReply

Quote
Increase the photon count and the larger format will have the advantage, admitted by Jonathan.
[a href=\"index.php?act=findpost&pid=55329\"][{POST_SNAPBACK}][/a]
And as I said before, getting as larger photon count with the larger sensor than is possible with the smaller sensor requires at least one of two things: (1) larger absolute aperture, meaning less DOF (2) longer exposure time.

If we are not ignoring the need for adequate DOF, this mostly means the longer exposure times idea that the main advantage of a larger formats like 8"x10" or even medium format is long exposures of stationary subjects.

Another thing I have said many times: you seem to assume that photosites of a given size will always have a maximum photon count that increases with size, whereas there are numerous technological possibilities for completely eliminating upper limits on photon counts. What substantial advantage would that leave for those far bigger, more expensive sensors requiring super-cooling?
« Last Edit: January 16, 2006, 08:27:42 PM by BJL » Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8853


« Reply #53 on: January 16, 2006, 09:35:09 PM »
ReplyReply

Quote
Another thing I have said many times: you seem to assume that photosites of a given size will always have a maximum photon count that increases with size, whereas there are numerous technological possibilities for completely eliminating upper limits on photon counts. What substantial advantage would that leave for those far bigger, more expensive sensors requiring super-cooling?
[a href=\"index.php?act=findpost&pid=56094\"][{POST_SNAPBACK}][/a]


Huge spades of extra resolution. The rapid filling and emptying of small sensors during one long exposure, increases the dynamic range but does nothing for absolute resolution.

In any case, my concession to the smaller format that it can perform at least as well as the larger format with equal photon count and equal DoF, is probably too generous.

Based on what I've observed with my own DSLRs, resolution does not fall off in inverse proportion to ISO. In fact it seems to be maintained pretty well up to ISO 1600, falling only marginally with increased noise.

Getting back to this hypothetical situation of an 8x12" format at f90 and ISO 6400 producing the same resolution, same DoF and same noise as 35mm at f11 and ISO 100, both employing the same shutter speed, I have assumed that the pixels of the larger format will be binned to reduce read noise, say in groups of 16, ie. the larger format is delivering just one quarter of the full resolution it was designed for.

As we stop up to f8 for 35mm and f64 for LF, keeping everything else the same except shutter speed which varies with f stop, the same conditions apply, ie, same DoF, same noise, same photon count, same resolution, EXCEPT there is the option of binning fewer pixels with the large format to achieve the significantly greater resolution that f64 provides over f90 (40% greater). This extra resolution may well come at the expenses of greater noise, but the option simply isn't there with the 35mm format. F8 can only deliver a very marginal increase in resolution compared with f11, with most if not all 35mm lenses.

I'm sorry BJL. I consider myself to be a reasonable bloke but I haven't seen any convincing arguments from you yet to support any innate technological advantage of the smaller format   . (I guess I'll have to add, innate technological advantage with respect to any aspect of image quality. There's no doubt that something lighter and cheaper has its own uses.)
« Last Edit: January 16, 2006, 09:41:16 PM by Ray » Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5121


« Reply #54 on: January 17, 2006, 04:19:03 PM »
ReplyReply

Quote
Huge spades of extra resolution.
...
I'm sorry BJL. I consider myself to be a reasonable bloke but I haven't seen any convincing arguments from you yet to support any innate technological advantage of the smaller format   . (I guess I'll have to add, innate technological advantage with respect to any aspect of image quality. There's no doubt that something lighter and cheaper has its own uses.)
[a href=\"index.php?act=findpost&pid=56095\"][{POST_SNAPBACK}][/a]
I'm sorry Ray, but as I have argues many times before, it seems quite clear that
1) sensor resolution can easily go to the point where overall resolution will soon enough be limited almost entirely by lens resolution
2) the main optical limitations relate to angular resolution (related lines per picture height) and this does not increase much with format size.

The obvious advantages of smaller formats in cost, size and weight are of value in themselves to all but those with infinite funds and infinite cargo capacity. Even Ansel Adams used medium format (Hasselblad) and miniature format (Contax) at times!. So the fundamental physical advantage of less total dark noise is icing on the cake; similar physics applies to film too, and is probably part of the reason for the long-standing preference for 35mm over larger formats in action photography. Given these advantages to a smaller format, the onus is surely on the advocates of a larger format to show if and when that larger format has sufficient compensating advantages.

And of course these advantages do exist in certain situations: for example, all current DSLR formats (4/3" and up) all have a clear advantage over current compact "digicam" format (2/3" and down) for high speed//low light/low noise, since what any DSLR can do at f/2 would require a lens faster than f/1 in any digicam format, probably resulting on quite noticably worse aberrations.

But such arguments have to be made in comparison of specific formats as applied to particular purposes; the desire to show that "bigger/smaller is always better" seems quixotically futile.
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8853


« Reply #55 on: January 17, 2006, 06:41:57 PM »
ReplyReply

Quote
But such arguments have to be made in comparison of specific formats as applied to particular purposes; the desire to show that "bigger/smaller is always better" seems quixotically futile.
[a href=\"index.php?act=findpost&pid=56152\"][{POST_SNAPBACK}][/a]

Agreed! But I think I'm covered on that account since I've already acknowledged earlier in the thread (somewhere) that one should always try to use the best tool for the job. This exercise for me has been an attempt to try and get a handle on the fundamental issues regarding image quality alone, as they apply to different formats.

Jonathan made the point that the only 'ultimate' advantage of the larger format is greater dynamic range as a result of the collection of a greater number of photons through longer exposures.

If that point is correct, then all advantages of the larger format will eventually be wiped out with the introduction of that new technology you've referred to, whereby small photodetectors are repeatedly filled and discharged during the course of one 'longer than usual' exposure.

I'm uneasy with such an explanation because; if we look at lens performance of the most popular (historically) of all formats, the 35mm format, we see that it's difficult to find a lens that's truly diffraction limited at f8, never mind f5.6 or f4.

If we go down the scale from large format to very small format at equivalent diffraction limited apertures, each potentially capable of producing the same resolution and DoF, provided the sensors in all cases have sufficient pixels, we get something like the following:

8x10.....f90;  4x5...f45;  6cmx7cm...f22;  35mm...f11;  APS-C and 4/3rds...f6.3;  2/3rds P&S....F3.

The above are of course approximate figures, but I don't believe they are far out. The essential principle is, as I understand it, each of those apertures will potentially provide the same 'picture' resolution and DoF provided the lenses are diffraction limited at those apertures.

Now the question I'm asking is this. Which of those formats currently have lenses available that are truly diffraction limited at those apertures? My guess is, only the first 4 (8x10 to 35mm).

Supposing we stop up by 2 stops in order to double resolution whilst still maintaining equivalent DoF across the formats. We get:

8x10... f45;  4x5...f22;  6cmx7cm...f11;  35mm...f5.6;  APS-C & 4/3rds...f3.2;  2/3rds P&S...f1.5.

Now I think already we've eliminated all the small formats from competing in terms of 'picture' resolution because there aren't any lenses for those formats that are diffraction limited at those apertures. We are left with just large and medium format.
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5121


« Reply #56 on: January 19, 2006, 09:40:59 PM »
ReplyReply

Ray, if you are trying to assess the image detail limits of imagined future cameras, with sensors far beyond what is currently possible (like 8"x10"!), it does not make sense to me to assume that the lenses will have roughly the performance limits of current 35mm format lenses designed within the far more modest resolution needs of keeping up with film. Especially when your estimate of those performance limits seem rather pessimistic, or at least are not based on the very best quality prime lenses, as would presumably be used in this pursuit of ultimate image detail.

So at very least, I suggest that you allow that future 35mm format lenses could avoid significant aberration limitation of resolution down to somewhere between f/2 and f/4, not f/11. And even f/4 gets one to the point that DOF would be inadequate for almost all extreme high resolution images, given the high degree of enlargement (or extremely close viewing) needed to show the great detail provided by that high resolution.
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8853


« Reply #57 on: January 19, 2006, 11:01:27 PM »
ReplyReply

Quote
So at very least, I suggest that you allow that future 35mm format lenses could avoid significant aberration limitation of resolution down to somewhere between f/2 and f/4, not f/11. And even f/4 gets one to the point that DOF would be inadequate for almost all extreme high resolution images, given the high degree of enlargement (or extremely close viewing) needed to show the great detail provided by that high resolution.
[a href=\"index.php?act=findpost&pid=56367\"][{POST_SNAPBACK}][/a]

BJL,
F4 is a very usable f stop on 35mm. I can't of course predict the nature of future developments in lens design. It would seem a reasonable assumption that 8x10 and 4x5 is likely to become increasingly irrelevant in the digital age, but maybe a full frame 6x7cm sensor is a possibility. In fact, it might well be more easily technologically feasible to produce say a 100MB 6x7cm sensor with a 90mm lens truly diffraction limited at f8, than a 100MB 35mm sensor with 50mm lens truly diffraction limited at f4.

I mean, if a 50mm lens for 35mm format were diffraction limited at f4, we'd be looking at an MTF response of around 70% at 80 lp/mm, would we not? Is such performance really feasible?
Logged
BJL
Sr. Member
****
Offline Offline

Posts: 5121


« Reply #58 on: January 20, 2006, 06:45:37 PM »
ReplyReply

Quote
F4 is a very usable f stop on 35mm.
[a href=\"index.php?act=findpost&pid=56375\"][{POST_SNAPBACK}][/a]
With normal print size/viewing combinations yes, but if we for example quadruple the linear resolution (16 times the pixel count) and so enlarge four times more so as to be able see that extra detail, the perceived DOF is reduced four-fold, so it looks like f/1. I would think that very, very few of such highly detailed images would work at "f/1 equivalent" DOF. Most such "high detail" work is already done at f/8, f/11 and beyond even with poor old 35mm film.

Anyway, I will try to sign of discussing hypotheticals far beyond any current or forseeable technological progress. My expectation though is that lenses will be the dominant limit or resolution and such, far more than sensor capabilities.
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8853


« Reply #59 on: January 21, 2006, 02:31:43 AM »
ReplyReply

Quote
With normal print size/viewing combinations yes, but if we for example quadruple the linear resolution (16 times the pixel count) and so enlarge four times more so as to be able see that extra detail, the perceived DOF is reduced four-fold, so it looks like f/1. I would think that very, very few of such highly detailed images would work at "f/1 equivalent" DOF. Most such "high detail" work is already done at f/8, f/11 and beyond even with poor old 35mm film.
[a href=\"index.php?act=findpost&pid=56461\"][{POST_SNAPBACK}][/a]


BJL,
That's a cop out. Okay! So you're getting tired of this thread because of the hypotheticals. Let's bring it back to your assertion that current f1 DoF would look like f4 Dof if we had sufficient resolution (on 35mm).

My gut feeling is, this is not true. There's a tendency in that direction, but it's not linear or proportional.

Perhaps you'd like to start another thread and share your insights with us   .
Logged
Pages: « 1 2 [3]   Top of Page
Print
Jump to:  

Ad
Ad
Ad