larger sensors

[eronald]

BJL

I wasn't "condescending", I was seriously suggesting, and do so again, that you take a look at the yield curves - your note re. Sony underlines yet again that this is an issue of economics, not of technical fesability.  An apparently small increase in surface area can dramatically impact the yield and thus the economics of the process. The mu is built into the yield equations, but these are far from linear. Process shrink does not necessarily entail advanatges AFAIK because we are counting photons and thus well capacity is important ...

This discussion is going nowhere. Summary: I say sensors will get even bigger. You say no they won't. We sound like a 5 year olds

Edmund

No need to be condescending: there is nothing in your post that I did not already know. Of course there will be some downward trend in the cost of sensors of any given size due to improving "mu" value, but that is a relatively modest factor in chip cost reduction, compared to the method that is vastly more popular in the digital imaging industry: reducing the size of the photo-sites (and thus of the sensors) needed to achieve a given level of image quality.

P. S. It is perhaps noteworthy that the team of Sony and Nikon has not even bothered to go to sensors at the maximum size possible with a single exposure, which is reportedly something around Canon 1D size (reticle size about 26x33mm according to Canon, so sensor active area a bit smaller.) Nikon is a major maker of IC fab. equipment, so there should not be a problem of access to suitable tools.
[a href=\"index.php?act=findpost&pid=92673\"][{POST_SNAPBACK}][/a]

[Gigi]

BJL


This discussion is going nowhere. Summary: I say sensors will get even bigger. You say no they won't. We sound like a 5 year olds

Edmund
[a href=\"index.php?act=findpost&pid=92694\"][{POST_SNAPBACK}][/a]

But cute ones at that

[Ray]

This discussion is going nowhere. Summary: I say sensors will get even bigger. You say no they won't. We sound like a 5 year olds
[a href=\"index.php?act=findpost&pid=92694\"][{POST_SNAPBACK}][/a]

But 5 year olds with rather high IQs   .

[BJL]

this is an issue of economics, not of technical fesability.
[a href=\"index.php?act=findpost&pid=92694\"][{POST_SNAPBACK}][/a]

I completely agree that it is matter of economics. (I am fairly sure that Sony could be making 24x36mm sensors and Nikon using them, if they saw an adequately profitable market for them.)

We disagree only on the tougher question of where the cost/benefit trade-offs will take us. My reading of recent trends is that the main efforts now and in the future are on improving performance at the various sensor sizes that have established themselves, up to but probably not beyond Kodak's current 37x49mm.

I can see some small chance of sensors going all the way to "645" (42.5x56mm), but almost none of going beyond that, since the medium format industry had mostly abandoned formats larger than 645 already in the late days of the film era.

[Ray]

I can see some small chance of sensors going all the way to "645" (42.5x56mm), but almost none of going beyond that, since the medium format industry had mostly abandoned formats larger than 645 already in the late days of the film era.
[a href=\"index.php?act=findpost&pid=92820\"][{POST_SNAPBACK}][/a]

There does seem likely to be an optimal middle ground where a compromise between maximum, ultimate quality and the practicality of convenience and cost will stabilise.

On the other hand, this balance can be tipped either way with new possibilities resulting from technological innovation. It wasn't long ago was it, that most digital backs on MF cameras had to be tethered to a computer. The fact that this is no longer a requirement must have given a huge boost to the popularity of digital backs for the cropped MF format.

But MF still has DoF disadvantages, as witnessed in some of Bernard Languillier's recent images from Japan using his ZD. As you continue to increase the size of the sensor, the DoF limitations become greater, shift movements become necessary, which don't always have the desired effect, and the whole system becomes cumbersome and heavy and of course very expensive.

For such disadvantages, one wants a substantial increase in image quality (at least I do   ).

BJL seems to have settled on the Olympus 4/3rds format as being ideal for his purposes, producing sufficient quality at the print sizes he's interested in. I'll probably settle on full frame 35mm as being adequate for my purposes. Canon have managed to reduce the gap between pixels on their new 400D, so that the photodiodes are not smaller than those on the 30D, resulting apparently in no increase in noise compared with the 30D. A full frame 35mm sensor with the same pixel density as the 400D would be something like a 26 or 27mp sensor. I think that should be sufficient for me, but I'm not a professional photographer striving to get an edge in the market place, or produce the sharpest billboards ever   .

[Ray]

Well, I just nipped over to the dpreview site to check my facts on the Canon 400D and of course didn't find any direct comparison between that camera and the 30D; different price category.

However, checking the two separate reviews (of the 30D and 400D), it seems that dynamic range for both cameras is the same, ie. 8.4EV up to and including ISO 800.

At ISO 1600, the 30D seems to have a slight advantage, dropping to 8EV as opposed to the 400D's 7.8EV.

I suppose John Sheehy would disagree with these figures.  

[John Camp]

I don't have a PhD in physics, but I do have a degree in history, and I know that larger, "higher resolution," cameras were once ditched by almost everybody (in the 1940s and 50s) for 35mm machines because the smaller ones were faster, cheaper, lighter and "good enough."

After it was established, 35 held on because it had a huge base of both users and equipment, to the point that some digital shooters now demand "full frame" as though the 35mm frame size were annointed by God as the only right one; that's the power of an installed base.

The point being that much larger sensors may indeed be possible, but who will take the risk of creating a new camera system around them -- especially when, in terms of quality, you could get in any high-end magazine with nothing more than a Canon, Nikon or Leica? Big sensors (after a point) mean bigger lenses, bigger files, more weight, more processing power...for what? So your photograph will look better in People, which is printed on toilet paper? They'd be like that huge (24-inch perhaps?) Polaroid camera that used to tour around the US; or maybe there were two of them. Sure, you could do it, but why?

The above note, by the way, applies to "life time" developments. I don't doubt that a hundred and fifty years from now, things will be different -- but I wouldn't be surprised if the technology 50 years from now is more or less a refinement of what we are already using...say the difference between 40s film and 90s film.

JC

[Ray]

After it was established, 35 held on because it had a huge base of both users and equipment, to the point that some digital shooters now demand "full frame" as though the 35mm frame size were annointed by God as the only right one; that's the power of an installed base.
[a href=\"index.php?act=findpost&pid=92853\"][{POST_SNAPBACK}][/a]

John,
I understand what you are getting at, but I would suggest that the reasons why the demand for the 35mm frame size has held on are similar to the reasons why the format was created in the first instance. The film already had a huge base of users and equipment in the movie industry.

It was a brilliant idea at the time but had one serious drawback for both professionals and amateur enthusiasts, the image quality was not up to scratch for prints larger (or much larger) than 8x10". Nevertheless, because millions of happy shooters seemed quite satisfied with (or at least familiar with) the image quality of 35mm film, it was easy to introduce a cropped digital format that could compete quality-wise on an 8x10(12)" print and even slightly larger.

The lure of full frame 35mm is due to the fact that, with the existing 35mm lens base, everyone can enjoy the convenience, automation and portability they remember from the 35mm film days, but also get that increased image quality that used to be only available with expensive, more cumbersome and less automated MF systems.

Now, I know you can claim that cropped format cameras such as the 12mp D2X produce virtually the same image quality as the 5D, but they are pushing it. Ultimately, a FF 35mm sensor can hold more than twice the number of pixels of a D2X size sensor and over 4x the number of pixels of an Olympus 4/3rds sensor.

To put it another way, when the pixel count race has ended, you'll be able to make a print from FF 35mm that has more than twice the area of a print from a D2X size sensor, whislt keeping the same apparent sharpness from the same viewing distance, assuming further improvement in the quality of 35mm lenses.

[John Sheehy]

Well, I just nipped over to the dpreview site to check my facts on the Canon 400D and of course didn't find any direct comparison between that camera and the 30D; different price category.

However, checking the two separate reviews (of the 30D and 400D), it seems that dynamic range for both cameras is the same, ie. 8.4EV up to and including ISO 800.

At ISO 1600, the 30D seems to have a slight advantage, dropping to 8EV as opposed to the 400D's 7.8EV.

I suppose John Sheehy would disagree with these figures. 
[a href=\"index.php?act=findpost&pid=92850\"][{POST_SNAPBACK}][/a]

Firstly, dynamic range is not a monolithic concept.  There are different standards for dynamic range.  The one I concern myself with mainly is the one that compares maximum RAW signal to the black frame noise floor.  IOW, how far the 1:1 SNR is below the clipping point.  For the 30D, saturation is 3943 ADU above black, and the black noise is about 4.7 ADU.  For the XTi/400D, the figures are 3800 and 7.25.  3943/4.7 = 839; 3800/7.25 = 524.  839:1 is 9.71 stops, and 524:1 is 9.03 stops; a difference of 0.68 stops.

What Phil seems to be measuring is how single RAW conversions from the two cameras compare, but you don't necessarily get the full DR with that approach; the very approach tends to equalize the results.  The standard conversion of the 400D is to set whitepoint 1/2 stop lower than the 30D, so with defaults, the 400D will get more highlights clipped by a converter.

[bjanes]

Firstly, dynamic range is not a monolithic concept.  There are different standards for dynamic range.  The one I concern myself with mainly is the one that compares maximum RAW signal to the black frame noise floor.  IOW, how far the 1:1 SNR is below the clipping point.  For the 30D, saturation is 3943 ADU above black, and the black noise is about 4.7 ADU.  For the XTi/400D, the figures are 3800 and 7.25.  3943/4.7 = 839; 3800/7.25 = 524.  839:1 is 9.71 stops, and 524:1 is 9.03 stops; a difference of 0.68 stops.

What Phil seems to be measuring is how single RAW conversions from the two cameras compare, but you don't necessarily get the full DR with that approach; the very approach tends to equalize the results.  The standard conversion of the 400D is to set whitepoint 1/2 stop lower than the 30D, so with defaults, the 400D will get more highlights clipped by a converter.
[a href=\"index.php?act=findpost&pid=92955\"][{POST_SNAPBACK}][/a]

Electronics engineers define dynamic range as the ratio of full well capacity to read noise, both expressed in electrons. Since the sensor response is linear, and the ADU number is proportional to electons, this corresponds to what John is measuring.

Phil determines DR by photographing a Stouffer (or similar step wedge), apparently exposed for a mid gray tone. Apparently, he uses classical mid-tone based exposure, rather than exposing to the right. The quote below is taken from his D40 review:

"Shadow range is more complicated, in our test we stop measuring values below middle gray as soon as the luminance value drops below our defined 'black point' (about 2% luminance) or the signal-to-noise ratio drops below a predefined value (where shadow detail would be swamped by noise), whichever comes first."

He uses either the in camera JPEG conversion or ACR conversion from a raw file rather than examining the raw values directly. In this situation, the black point would be affected by the tone curve in the first instance and whatever S/N cutoff is chosen for the point where shadow noise is swamped by noise in the second case. The 2% luminance value does not make sense to me since 100%/2% = 50:1 = 5.64 stops. I think Phil needs to refine his testing criteria. Perhaps John can comment.

Bill

[John Sheehy]

I think Phil needs to refine his testing criteria. Perhaps John can comment.
[a href=\"index.php?act=findpost&pid=92978\"][{POST_SNAPBACK}][/a]

I agree.

The "whichever comes first" part is the one that bothers me most, and of course, his methodology relies too heavily on a RAW converter to maintain highlights.  The 400D has 1/2 stop more than the older Canons, relative to average grey metering.

[bjanes]

Now, I know you can claim that cropped format cameras such as the 12mp D2X produce virtually the same image quality as the 5D, but they are pushing it. Ultimately, a FF 35mm sensor can hold more than twice the number of pixels of a D2X size sensor and over 4x the number of pixels of an Olympus 4/3rds sensor.

To put it another way, when the pixel count race has ended, you'll be able to make a print from FF 35mm that has more than twice the area of a print from a D2X size sensor, whislt keeping the same apparent sharpness from the same viewing distance, assuming further improvement in the quality of 35mm lenses.
[a href=\"index.php?act=findpost&pid=92951\"][{POST_SNAPBACK}][/a]

12 MP is 12 MP in terms of resolution whether one is using a full 35 mm frame or a cropped frame in the D2X. As Michael explained earlier, you would really have to double the pixel count to 24 MP to see a real difference in resolution. At base ISO the D2X does well, but at high ISO it has a distinct disadvantage to the 5D in terms of noise and dynamic range.

If you had a 24 MP camera, you would probably need to use a tripod to capture the full resolution of the camera, although image stabilized lenses can extend the limits of hand holding. The necessity to use a tripod obviates much of the appeal of 35 mm photography, at least to many amateurs.

Bill

[John Camp]

John,
I understand what you are getting at, but I would suggest that the reasons why the demand for the 35mm frame size has held on are similar to the reasons why the format was created in the first instance. [a href=\"index.php?act=findpost&pid=92951\"][{POST_SNAPBACK}][/a]

Ray, I don't disagree. I was making the minor point that many people regard the legacy 35mm size as somehow perfect, when, in fact, I find it somewhat awkwardly shaped, as do other follks. I would be happier if Nikon, say, got to 22mp with something more on the lines of a 4x5 or 6x7 aspect ratio, rather than 3x2, if that can be done using legacy gear (lenses) That is, I don't think there is anything holy about the **particular** size and shape of the 35mm frame.

My overall point was to argue that we are now getting to the place where practicality will begin to rule; that is, 98 percent of the professional/serious amateur market will be satisfied with the next generation of DSLRs (~22 mp) essentially forever, and will see no real benefit in upgrading until a camera no longer functions. We will be in a market like the film market in 1995. Another 1.9 percent will want somewhat larger MF chips, but those, too, are reaching the point where the cost/benefit equation will slow development of ever-larger chips.

I don't believe there is any technical reason that we couldn't have a 4x5 inch chip...and it might be bought by several people per year. How many companies really want a market comprising several people?  

JC

[ErikKaffehr]

Hi!

It's not only about chips and sizes. It's also about lenses. For denser chips we need better lenses. There must be a limit set by lenses, go beyond that and there will be diminishing returns from increasing pixel size.

As things stand now we have essentially around four formats.

1) 4/3
2) APS-C
3) 135 Full frame
4) 645 reduced frame

2-4 use lenses designed for film cameras. As far as I understand 135 full frame
(as practised by Canon) have some serious issues with wide angle lenses. This problem is aggravated by the preference for zooms in 135 photography.

On medium format it is a bit easier. Sensors are still smaller than nominal format, so the outermost part of the image circle is not used. Also MF photographers don't seem to demand the extensive zooming capability normally accessible to the 135 folks.

There are many other things affecting image quality than noise and pixel size.

a) Depth of field
 Precision of focusing
c) Camera shake
d) Vibration introduced by mirror
e) Vibration introduced by shutter
f) Precision of camera assembly, including heat expansion
g) Diffraction effects
h) Quality of optics

We need to control all of the above to achieve optimal image quality.

It may be that we need to kinds of camera systems:

For action: Big pixels, so we can use high ISO

For tripod based photography: Dense pixels, we can use low ISO so noise is not that much an issue

Best regards

Erik

 

Ray, I don't disagree. I was making the minor point that many people regard the legacy 35mm size as somehow perfect, when, in fact, I find it somewhat awkwardly shaped, as do other follks. I would be happier if Nikon, say, got to 22mp with something more on the lines of a 4x5 or 6x7 aspect ratio, rather than 3x2, if that can be done using legacy gear (lenses) That is, I don't think there is anything holy about the **particular** size and shape of the 35mm frame.

My overall point was to argue that we are now getting to the place where practicality will begin to rule; that is, 98 percent of the professional/serious amateur market will be satisfied with the next generation of DSLRs (~22 mp) essentially forever, and will see no real benefit in upgrading until a camera no longer functions. We will be in a market like the film market in 1995. Another 1.9 percent will want somewhat larger MF chips, but those, too, are reaching the point where the cost/benefit equation will slow development of ever-larger chips.

I don't believe there is any technical reason that we couldn't have a 4x5 inch chip...and it might be bought by several people per year. How many companies really want a market comprising several people?   

JC
[a href=\"index.php?act=findpost&pid=92999\"][{POST_SNAPBACK}][/a]

[eronald]

For those interested, here is a  Translation Key to the jargon in some of the above postings.

Now to mix it some more:
I once shot a D200 to 1Ds2 comparison, photographing a guy in a Paris café, across the room. Holding the cameras with elbows braced on a table. In this low light the shake and noise and focus problems cumulatively *wrecked* the D200 shots while the 1ds2 was still running strong. I mean wrecked, not just deteriorated. Outdoors, little difference would have been visible.

Moral of the story, I would like the following added to the sophisticated discussion of ADUs and noise:

Camera shake *in prints* depend on the enlargement factor.
Noise *in prints* depends on the enlargement factor.

and then ...

Focus Aliasing depends on the resolution of the autofocus *actuators*, as much as on the AF sensor system. A cropped format with a legacy AF lens implies larger atomic focus steps.

Edmund

[david o]

Ray, I don't disagree. I was making the minor point that many people regard the legacy 35mm size as somehow perfect, when, in fact, I find it somewhat awkwardly shaped, as do other follks. I would be happier if Nikon, say, got to 22mp with something more on the lines of a 4x5 or 6x7 aspect ratio, rather than 3x2, if that can be done using legacy gear (lenses) That is, I don't think there is anything holy about the **particular** size and shape of the 35mm frame.

My overall point was to argue that we are now getting to the place where practicality will begin to rule; that is, 98 percent of the professional/serious amateur market will be satisfied with the next generation of DSLRs (~22 mp) essentially forever, and will see no real benefit in upgrading until a camera no longer functions. We will be in a market like the film market in 1995. Another 1.9 percent will want somewhat larger MF chips, but those, too, are reaching the point where the cost/benefit equation will slow development of ever-larger chips.

I don't believe there is any technical reason that we couldn't have a 4x5 inch chip...and it might be bought by several people per year. How many companies really want a market comprising several people?   

JC
[a href=\"index.php?act=findpost&pid=92999\"][{POST_SNAPBACK}][/a]

100% agree on that, as I never really liked the 35mm proportion.
6x7 to me offers the bests proportions, don't get me wrong, it could be smaller, but same ratio.

[sjprg]

I came across an article about a 100MP Dalsa sensor for either the military or NASA. awhile back, so I suspect that larger sensors are coming when the economics can be justified.
Here is the Dalsa page. I'm not sure if the article is in here or not. Maybe I saw it on a NASA site.

http://www.dalsa.com/news/news.asp?itemID=165

[bjanes]

For those interested, here is a  Translation Key to the jargon in some of the above postings.

Edmund
[{POST_SNAPBACK}][/a]

Edmund,

The link to Jerry Lodriguss's paper is very useful, but he does perpetuate one misconception that has confused me in the past and may confuse some forum members as well. He states that since human vision is logarithmic, it is also necessary that the digital file also should be log so as to comply with the nature of human vision.

Actually, the gamma correction is applied to the image for coding efficiency. In the resulting log transformation, a proportional change in intensity (say by a factor of 1.01 implied by the Weber-Fechner law) results in a proportional increment in the recorded number. In a linear scale, there is a much greater proportional increment between pixel values going from 1 to 2 than from 254 to 255. The gamma function is a power equation: y = x^(1/2.2). In a linear file, the brightest f/stop contains half the data, which is not efficient for coding the shadows.

When the image is displayed on the monitor,  an inverse function is applied so that the resulting image brightness on the monitor corresponds more less 1:1 to the values in the scene. (y = x^2.2). In order to display the greater dynamic of most scenes on a monitor with less DR, a tone curve is generally applied to the data in order to compress the shadows and highlights.

This is explained in [a href=\"http://www.poynton.com/notes/colour_and_gamma/GammaFAQ.html#desktop] Charles Poynton's[/url] Gamma FAQ.

According to the Rec 709 transfer function there is a linear segment towards the shadows, so that the actual power is 2.5.

[Ray]

12 MP is 12 MP in terms of resolution whether one is using a full 35 mm frame or a cropped frame in the D2X. As Michael explained earlier, you would really have to double the pixel count to 24 MP to see a real difference in resolution.[a href=\"index.php?act=findpost&pid=92980\"][{POST_SNAPBACK}][/a]

Bt that's exactly what I've implied, Bill. At the end of the day, when the pixel count race is over and the minimum size photodiode for useful quality has been reached (and I believe Michael R has said this as around 5 microns), a FF 35mm sensor will hold more than double the number of pixels of a D2X size sensor and over 4x the number of pixels of an Olympus 4/3rds sensor.

[Ray]

Ray, I don't disagree. I was making the minor point that many people regard the legacy 35mm size as somehow perfect, when, in fact, I find it somewhat awkwardly shaped, as do other follks. I would be happier if Nikon, say, got to 22mp with something more on the lines of a 4x5 or 6x7 aspect ratio, rather than 3x2, if that can be done using legacy gear (lenses) That is, I don't think there is anything holy about the **particular** size and shape of the 35mm frame.

[a href=\"index.php?act=findpost&pid=92999\"][{POST_SNAPBACK}][/a]

John,
This is a complete red herring. Aspect ratios vary to suit the composition. They can vary from square to 6x17cm. If you are talking about aspect ratios that are in general more suitable for, say portraits, then 4/3rds or 6x7 would probably be more appropriate than the 35mm 3:2.

If we are talking about a general purpose aspect ratio for all types of subjects and compositions, then the arguments in favour of 35mm are at least as strong as the arguments in favour of any other aspect ratio.

Consider the effect a square aspect ratio would have using your widest 35mm lens. It simply wouldn't be as wide in either the horizantal or vertical plane.

The only practical solution to the subjective preferences for different aspect ratios is a circular sensor which matches the image circle of the lens as close as possible without significant peripheral light fall-off. The end users, through use of zoom lenses, could then create any aspect ratio they thought appropriate for the composition, without feeling they were compromising either maximum field of view or maximum image quality.