(aborted?) Sony 10.66 MP APS (D100, etc.) sized sensor

[Ron Parr]

I just noticed this one, although there are references to this dating back to 2001. According to acrobat, the document was created in July 2001:

http://www.adelco.it/online/company/Sony/pdf/ICX263AL.pdf

• The 1.6 FPS rate is discouraging and would seem to limit how high upmarket this sensor could move.

• It is capable of 9FPS by skipping 1 out of every 6 vertical pixels for a total of 441 lines. This seems a bit awkward since it's too slow for video and too low res. to be useful for high quality stills.

One has to wonder if Sony is continuing to develop this product, perhaps waiting for the technology to improve the frame rate, or if it was completely shelved in favor of the 6MP sensor that found its way into the D100 and *ist.

Is this the prototype for the next generation of low to mid priced digital SLRs, or just a relic?

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Joseph S Wisniewski]

I just noticed this one, although there are references to this
dating back to 2001. According to acrobat, the document was
created in July 2001:

http://www.adelco.it/online/company/Sony/pdf/ICX263AL.pdf

• The 1.6 FPS rate is discouraging and would seem to limit how high

upmarket this sensor could move.

I wouldn't worry about that. It's running at 20 MHz. Most Sony CCDs were clocking about that speed, back in 2001. As were the cameras of that time period, such as Nikon D100 or Canon D60, both of which clocked their sensor at about 18 MHz.

By 2004, clock speeds of 35 MHz are not uncommon. That will get you to 3 fps at 10.6mp.

• It is capable of 9FPS by skipping 1 out of every 6 vertical

pixels for a total of 441 lines. This seems a bit awkward since
it's too slow for video and too low res. to be useful for high
quality stills.

But it is useful as an electronic viewfinder, when the camera is in mirror lockup mode. 9 frames/sec is a lot better than what you see now, which is nothing at all.

And, if you raise the clock from 20 Mhz to 35, you go from 9 fps to 15.

One has to wonder if Sony is continuing to develop this product,
perhaps waiting for the technology to improve the frame rate, or if
it was completely shelved in favor of the 6MP sensor that found its
way into the D100 and *ist.

Is this the prototype for the next generation of low to mid priced
digital SLRs, or just a relic?

Both. Neither.

I think it's more than Sony could handle, back in 2001. Look at the ICX413AQ sensor used in the D100, *ist, Minolta 7D. It's a bit less ambitious than the ICX263AL. Bigger pixels, and less of them. And from what I've heard, it was supposed to have electronic shutter from day one, but it took Sony almost 2 years longer to get that working right, just in time for D70.

Don't be surprised if they clean it up and it appears in one or more cameras at Photokina this fall. In a slightly different varient.

I really wonder what market Sony had intended this chip for. It's got an "L" suffix. That means monochrome. There's nothing online about an ICX683AQ. Or even an "AK".

--
A cyberstalker told me not to post anymore...
So I'm posting even more!

Ciao!

Joe

http://www.swissarmyfork.com

 

[Anastigmat]

I just noticed this one, although there are references to this
dating back to 2001. According to acrobat, the document was
created in July 2001:

http://www.adelco.it/online/company/Sony/pdf/ICX263AL.pdf

• The 1.6 FPS rate is discouraging and would seem to limit how high

upmarket this sensor could move.

• It is capable of 9FPS by skipping 1 out of every 6 vertical

pixels for a total of 441 lines. This seems a bit awkward since
it's too slow for video and too low res. to be useful for high
quality stills.

One has to wonder if Sony is continuing to develop this product,
perhaps waiting for the technology to improve the frame rate, or if
it was completely shelved in favor of the 6MP sensor that found its
way into the D100 and *ist.

Is this the prototype for the next generation of low to mid priced
digital SLRs, or just a relic?

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

Sony may well be working on a higher resolution chip that is the same size as the current 6mp chip found in the D70 and *ist D, but such a chip would not be the next generation of DSLR sensor, since such a chip would almost certainly be noisier, given its small size, than the current chip. The next generation of DSLR chip is going to be 1.2-1.3x format. The first affordable camera that will carry such a chip will most likely come from Canon. Sony may need to be dragged kicking and screaming into making a bigger photo sensor, but there is no doubt that it will do so, unless it wants out of the DSLR chip market.

 

[Ron Parr]

Sony may well be working on a higher resolution chip that is the
same size as the current 6mp chip found in the D70 and *ist D, but
such a chip would not be the next generation of DSLR sensor, since
such a chip would almost certainly be noisier, given its small
size, than the current chip. The next generation of DSLR chip is
going to be 1.2-1.3x format. The first affordable camera that will
carry such a chip will most likely come from Canon. Sony may need
to be dragged kicking and screaming into making a bigger photo
sensor, but there is no doubt that it will do so, unless it wants
out of the DSLR chip market.

I remember when people were upset that the D60 had same size sensor and 2X the pixels as the D30. They predicted more noise. Well...

http://www.dpreview.com/reviews/canoneosd60/page16.asp

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Daniel (NL)]

Sony may well be working on a higher resolution chip that is the
same size as the current 6mp chip found in the D70 and *ist D, but
such a chip would not be the next generation of DSLR sensor, since
such a chip would almost certainly be noisier, given its small
size, than the current chip.

Noise can't be that bad, compared the APS size to smaller CCD's as used in the fixed lens digicams. doubling the pixel count should be possible I think, without much problems with noise, the pixel size will still be much larger than for instance a 3 MP small ccd.

Daniel

 

[Anastigmat]

Do you expect the same to keep repeating itself ad infinitum? Sooner or later it becomes impractical and undesirable to add more pixels and keep the sensor the same size. The current crop of 8mp prosumer fixed lens models are hitting that ceiling. The Minolta Dimage A2 and Nikon 8700 for example, are notorious for their high noise levels compared to earlier 5mp versions of the same cameras.

Smaller pixels also have lower angular response, so that vignetting becomes a more noticeable problem. I can see the advantage to a chip maker for maintaining small chip sizes in order to maximize the number of sensors that will fit on a single wafer. Photographers, however, want bigger chips. Canon's full frame EOS-1DS for example is a sought after camera even at its lofty price. An $8,000, 11mp, 1.6x format Canon camera would be much less popular because of higher noise levels, even though you may think that it is possible for Canon to fit that many pixels on a chip and reduce noise levels at the same time. The 1.5x and 1.6x formats have probably reached the limit of how many pixels can be squeezed into them without major gains in technology.

Sony may well be working on a higher resolution chip that is the
same size as the current 6mp chip found in the D70 and *ist D, but
such a chip would not be the next generation of DSLR sensor, since
such a chip would almost certainly be noisier, given its small
size, than the current chip. The next generation of DSLR chip is
going to be 1.2-1.3x format. The first affordable camera that will
carry such a chip will most likely come from Canon. Sony may need
to be dragged kicking and screaming into making a bigger photo
sensor, but there is no doubt that it will do so, unless it wants
out of the DSLR chip market.

I remember when people were upset that the D60 had same size sensor
and 2X the pixels as the D30. They predicted more noise. Well...

http://www.dpreview.com/reviews/canoneosd60/page16.asp

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Ron Parr]

Do you expect the same to keep repeating itself ad infinitum?
Sooner or later it becomes impractical and undesirable to add more
pixels and keep the sensor the same size. The current crop of 8mp
prosumer fixed lens models are hitting that ceiling.

If that's the ceiling, then there's plenty of headroom for APS sized sensors given their currently huge pixel size in comparison to the 8MP 2/3" CCDs.

Of course one cannot expect things to continue ad infinitum, but there's scant evidence that current digital SLR sensor technology is pushing up against a hard pixel size limit.

Perhaps the best argument against smaller pixels is diffraction, but this is overplayed. In practice, the effects of diffraction are less than what many think and smaller pixels don't make diffraction worse. The diffraction is always there and smaller pixels simply mean that you don't gain as much resolution as you might expect for smaller apertures because the image projected by the lens isn't as sharp as you might have thought.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[Thom Hogan]

Do you expect the same to keep repeating itself ad infinitum?

Yes. There's no evidence that we've "hit the wall" in sensor technologies at the larger sizes. And that doesn't even take into account add-on NR or in-camera post processing techniques.

Smaller pixels also have lower angular response, so that vignetting
becomes a more noticeable problem.

You seem to be guilty of the same thing that everyone else is, equating photosite size to light reception area size. Most current sensor designs have very small light reception areas compared to the photosite size. It's advances like those of FillFactory or Nikon with LBCAST that are making changes to that equation. In the near term I believe that any loss of photosite size is likely made up by improvements in light retention wells, microlenses, and improved dark current conditions. Long term, it's harder to see what the improvements will be, but I've heard enough to convince me they'll be there, at least at APS and full frame sizes.

Photographers, however, want bigger chips.

Not sure I agree. Real photographers are interested in results, not technology.

Canon's full frame
EOS-1DS for example is a sought after camera even at its lofty
price.

Your argument seems to be that it's sought after because it's full frame. No. It's sought after because it is 11mp, low noise, high quality, in a pro body. None of those things REQUIRE full frame.

An $8,000, 11mp, 1.6x format Canon camera would be much
less popular because of higher noise levels, even though you may
think that it is possible for Canon to fit that many pixels on a
chip and reduce noise levels at the same time.

Your contention that it would have to have higher noise levels is your opinion, just as the opinion of others is that an APS sensor size can be produced with higher resolution and low noise. If your opinion is to be taken more seriously than those of others, you're going to have to share you technical background and indicate what access you have to current research and development.

The 1.5x and 1.6x
formats have probably reached the limit of how many pixels can be
squeezed into them without major gains in technology.

So has full frame. What's your point? The thing is, as consumers we're seeing technology about two to four years behind what is actually possible. I'm not aware of any sensor manufacturer indicating that they've hit the limit to improvements they can make in their technology (though I'll also point out that third-parties trying to implement the Sony 8mp consumer sensor in cameras point out that they think they may be seeing quantum limitations at that particular sensor size).

The question that the market is headed towards is very similar to the one facing the computer makers right now. How many gigahertz do processors have to run at? Is your 3.2Ghz Pentium really that much faster from a practical standpoint than your 2.4Ghz Pentium? In digital cameras the issue is how many megapixels do sensors need to have? We can argue whether that's 4mp, 6mp, 8mp, or 11mp, but I'm hard pressed to see where the advantages are above that, at least for the print sizes that 99% of the market makes. Remember, more pixels mean more processing horsepower and memory, too, and while those things come down in price, the expectation in DSLRs is that they will go down in price with higher quality, not stay at the same price at higher quality. Thus, I'm guessing that APS sensor size can satisfy 99% of the market.

Note that this doesn't mean that higher resolution sensors won't get produced and used by a few, much like MF and large format film is today.

--
Thom Hogan
author, Nikon Field Guide & Nikon Flash Guide
author, Complete Guides to the Nikon D70, D100, D1 series, and Fujifilm S2 Pro
http://www.bythom.com

 

[MrX]

Noise can't be that bad, compared the APS size to smaller CCD's as
used in the fixed lens digicams. doubling the pixel count should be
possible I think, without much problems with noise, the pixel size
will still be much larger than for instance a 3 MP small ccd.

Noise isn't just relative to the individual pixel size. The larger the sensor the noisier it becomes as the signal degrades accross it's surface. The Contax N Digital had huge pixels, but the noise wasn't very good.

So, as pointed out in another reply, the number of pixels that can be added will hit a certain limit just like there is a limited with CPU technology. What we will have to wait for is new technology that allows us to overcome todays problems in CCD design.

--
------
AdrianX - BSc. MSc. Director.
http://www.AdrianJudd.com

 

[MrX]

I remember when people were upset that the D60 had same size sensor
and 2X the pixels as the D30. They predicted more noise. Well...

We did lose a stop of dynamic range though, but most people will agree that it was woth it set against a doubling of pixels.

--
------
AdrianX - BSc. MSc. Director.
http://www.AdrianJudd.com

 

[Olivia T. BvG]

Thank you Thom! For myself, I am rather hoping that they would continue developing the APS sized sensors, and make them perform better then they are now.

Especially in areas such as dynamic range. I believe that technologies will improve such that noise issues won't even be much of an issue in the very near future for such sensors.

This will enable the manufacturers to keep lens and maybe, camera sizes smaller, and since I have small hands and am rather smallish in stature, I would be more interrested in such technology rather then the FF-sensor based equiptment.

I do like to see more developments in the electronic viewfinders, and love to see them eventually overtake the mirror/shutter system, presently used in dSLR cameras. Thanks again, Thom! I have been reading your contributions with much interrest.

regards,
Olivia

You seem to be guilty of the same thing that everyone else is,
equating photosite size to light reception area size. Most current
sensor designs have very small light reception areas compared to
the photosite size. It's advances like those of FillFactory or
Nikon with LBCAST that are making changes to that equation. In the
near term I believe that any loss of photosite size is likely made
up by improvements in light retention wells, microlenses, and
improved dark current conditions. Long term, it's harder to see
what the improvements will be, but I've heard enough to convince me
they'll be there, at least at APS and full frame sizes.

Canon's full frame
EOS-1DS for example is a sought after camera even at its lofty
price.

Your argument seems to be that it's sought after because it's full
frame. No. It's sought after because it is 11mp, low noise, high
quality, in a pro body. None of those things REQUIRE full frame.

Your contention that it would have to have higher noise levels is
your opinion, just as the opinion of others is that an APS sensor
size can be produced with higher resolution and low noise. If your
opinion is to be taken more seriously than those of others, you're
going to have to share you technical background and indicate what
access you have to current research and development.

So has full frame. What's your point? The thing is, as consumers
we're seeing technology about two to four years behind what is
actually possible. I'm not aware of any sensor manufacturer
indicating that they've hit the limit to improvements they can make
in their technology (though I'll also point out that third-parties
trying to implement the Sony 8mp consumer sensor in cameras point
out that they think they may be seeing quantum limitations at that
particular sensor size).

The question that the market is headed towards is very similar to
the one facing the computer makers right now. How many gigahertz do
processors have to run at? Is your 3.2Ghz Pentium really that much
faster from a practical standpoint than your 2.4Ghz Pentium? In
digital cameras the issue is how many megapixels do sensors need to
have? We can argue whether that's 4mp, 6mp, 8mp, or 11mp, but I'm
hard pressed to see where the advantages are above that, at least
for the print sizes that 99% of the market makes. Remember, more
pixels mean more processing horsepower and memory, too, and while
those things come down in price, the expectation in DSLRs is that
they will go down in price with higher quality, not stay at the
same price at higher quality. Thus, I'm guessing that APS sensor
size can satisfy 99% of the market.

Note that this doesn't mean that higher resolution sensors won't
get produced and used by a few, much like MF and large format film
is today.

--
Thom Hogan
author, Nikon Field Guide & Nikon Flash Guide
author, Complete Guides to the Nikon D70, D100, D1 series, and
Fujifilm S2 Pro
http://www.bythom.com

--
Olivia
http://public.fotki.com/NatureSpirits/
http://photos.yahoo.com/whispersfromspirit

 

[Anastigmat]

Do you expect the same to keep repeating itself ad infinitum?

Yes. There's no evidence that we've "hit the wall" in sensor
technologies at the larger sizes. And that doesn't even take into
account add-on NR or in-camera post processing techniques.

Pixel size cannot keep shrinking. There are two sets of limits to how small a pixel can get. First, there is the wavelength of visible light. Make a pixel smaller than the wavelength of light, then it would not be able to detect the light. Second, photosensors are made of silicon, and there is a limit on the light sensitivity of silicon. Too small a pixel makes it insensitive to light. A small pixel also requires a lot of signal amplification, another source of noise. According to the Extreme Tech article "Anatomy of a Digital Camera Image Sensors"

"The smallest CCD pixel feasible has been demonstrated at 2.5 microns
square. But that's for digital video and not still digital cameras. Note that
the dynamic range of an image is affected by the pixel size and design. A
digital still camera requires at least 10-bit dynamic range, meaning it needs
slightly larger pixels of present designs. Currently, the smallest pixel size
for digital cameras is about 3.3 microns in a CCD and about 4 microns for

CMOS. The popular Sony 3.3MP sensor has a 3.45-micron pixel. Many other consumer digital cameras have pixels averaging around 5 microns."

According to these figures, pixel size in existing cameras are close to the lower limit.

Smaller pixels also have lower angular response, so that vignetting
becomes a more noticeable problem.

You seem to be guilty of the same thing that everyone else is,
equating photosite size to light reception area size. Most current
sensor designs have very small light reception areas compared to
the photosite size. It's advances like those of FillFactory or
Nikon with LBCAST that are making changes to that equation. In the
near term I believe that any loss of photosite size is likely made
up by improvements in light retention wells, microlenses, and
improved dark current conditions. Long term, it's harder to see
what the improvements will be, but I've heard enough to convince me
they'll be there, at least at APS and full frame sizes.

I am aware of the difference between the light sensitive area of a pixel and the electronic part. Even if you can make the photodiode fit the entire pixel, there is still a limit on how small a pixel can get. So pixels simply cannot keep shrinking. No matter how small a pixel can shrink, a full frame can pack more of them and give higher resolution, while allowing the use of wide angle lenses designed for 35mm. Further, the smaller a pixel, the less sensitive it is too light coming in at an angle, which means vignetting will be worse for cameras with small photodiodes.

Photographers, however, want bigger chips.

Not sure I agree. Real photographers are interested in results, not
technology.

Bigger chips give them better results at any stage in the development of sensor technology. One photographer is comparing the EOS1DS images to medium format film and claim that resolution is close enough and noise is similarly small. Can't get the same results with an APS sized sensor yet. A photographer with a full frame and a 17mm f/3.5 lens can go out and get results that a D70 owner cannot. The same photographer can get results with a 35mm shift lens on a full frame that is impossible for a D70 owner. No one makes 24mm DX shift lenses.

 

[Anastigmat]

Anastigmat wrote:
Canon's full frame
EOS-1DS for example is a sought after camera even at its lofty
price.

Your argument seems to be that it's sought after because it's full
frame. No. It's sought after because it is 11mp, low noise, high
quality, in a pro body. None of those things REQUIRE full frame.

The full frame sensor, given its large size, allows larger photodiodes. That means this sensor can allow 11mp and low noise at the same time. The so called pro body is overrated, since it is available in the 1D MkII as well. The Canon 10D is not a filmsy camera either. The biggest difference between these 3 cameras is sensor size. Despite your disagreement, large sensors are a highly sought after feature that consumers are willing to pay megabucks for. Of course, not that many consumers can afford the lofty prices. Unaffordability does not diminish the desirability of large sensors.

An $8,000, 11mp, 1.6x format Canon camera would be much
less popular because of higher noise levels, even though you may
think that it is possible for Canon to fit that many pixels on a
chip and reduce noise levels at the same time.

Your contention that it would have to have higher noise levels is
your opinion, just as the opinion of others is that an APS sensor
size can be produced with higher resolution and low noise. If your
opinion is to be taken more seriously than those of others, you're
going to have to share you technical background and indicate what
access you have to current research and development.

My opinion is based on the known fact that smaller photodiodes are less sensitive to light and thus require more amplification to achieve the same ISO. Amplification introduces noise. That is a well known fact of life for engineers. Take the night vision cameras used in the military for example, they are monochromatic and they are grainy. That is because at the light levels they are used, very few photons actually reach the light sensors. It does not take a physics major to figure out that a lot fewer photons will reach the front element of a 50mm f/3.5 lens versus the number that will reach the front element of a 50mm f/1.4 lens. The f/1.4 lens has a much larger surface area than the f/3.5 lens. Larger photodiodes have the same advantage as the faster lens because of a larger surface area. I am surprised that you are not aware of this simple fact.

The 1.5x and 1.6x
formats have probably reached the limit of how many pixels can be
squeezed into them without major gains in technology.

So has full frame. What's your point? The thing is, as consumers
we're seeing technology about two to four years behind what is
actually possible. I'm not aware of any sensor manufacturer
indicating that they've hit the limit to improvements they can make
in their technology (though I'll also point out that third-parties
trying to implement the Sony 8mp consumer sensor in cameras point
out that they think they may be seeing quantum limitations at that
particular sensor size).

Tests have shown that these 8mp cameras are noisy. Therefore we are unlikely to see the same sized sensors with even larger numbers of pixels. So, in order to make cameras with higher resolution, Sony would have to make a bigger sensor. That is quite simple. The same is true. Sony will sooner or later run out of real estate on the APS sized sensor and it will have to go to a larger chip. That is why what Nikon says publicly does not make sense. No one can committ oneself to the APS sized sensor and hope to compete with those who show no similar committment. Canon certainly isn't committed to the 1.6x format.

The question that the market is headed towards is very similar to
the one facing the computer makers right now. How many gigahertz do
processors have to run at? Is your 3.2Ghz Pentium really that much
faster from a practical standpoint than your 2.4Ghz Pentium?

Yes it is, if you are using appliactions (such as digital movies) that require the most processing power.

In
digital cameras the issue is how many megapixels do sensors need to
have? We can argue whether that's 4mp, 6mp, 8mp, or 11mp, but I'm
hard pressed to see where the advantages are above that, at least
for the print sizes that 99% of the market makes.

As many as possible. No one may ever need to make a 50x75 inch print on a regular basis, but given a high enough resolution sensor with low noise, one can crop the image taken with a 50mm lens, and get the same quality image one would have gotten from a 400mm lens. Wouldn't that be an advantage that is practical and welcomed?

Remember, more
pixels mean more processing horsepower and memory, too, and while
those things come down in price, the expectation in DSLRs is that
they will go down in price with higher quality, not stay at the
same price at higher quality. Thus, I'm guessing that APS sensor
size can satisfy 99% of the market.

Yes, bigger files are one of the reason for impracticality of large, high resolution sensors. Nevertheless, one can go full frame while limiting resolution to a manageable 11 megapixels. The megapixel count can grow when improvements in write speed, buffer size and photodiode technology make even higher resolution feasible in the future.

You are not guessing. Practically 99% of the market are buying 1.5x or 1.6x format digital cameras. But that does not mean that the consumers are necessarily thrilled with the need to get digital only lenses just to get the equivalent of a 28mm lens on their new digital SLR cameras. If 1.3x format cameras are selling for the same price as 1.5x or 1.6x format cameras, then 99% of them would buy the 1.3x format camera.

 

[HAE]

This could actually be a version of the sensor in the D1x/h. Around a year ago Bjorn Rorslett published a macor photo of the D1h sensor on this site clearly showing four microlenses for each pixel, indicating each pixel actually is made up of four photo sites. The D2X has odd rectangular pixels and is double the D1H in pixel count (2 photo sites per pixel?). If you do that math the sensor in the D1X/H should have 10.6 MP. I think it was Thom Hogan that suggested that that The D1X/H and D70 sensors are Nikon modifications of Sony designs. So this could be the D1X/H sensor with information released to the public after Nikon's execlusive on it expired ???
Just a thought.

I just noticed this one, although there are references to this
dating back to 2001. According to acrobat, the document was
created in July 2001:

http://www.adelco.it/online/company/Sony/pdf/ICX263AL.pdf

• The 1.6 FPS rate is discouraging and would seem to limit how high

upmarket this sensor could move.

• It is capable of 9FPS by skipping 1 out of every 6 vertical

pixels for a total of 441 lines. This seems a bit awkward since
it's too slow for video and too low res. to be useful for high
quality stills.

One has to wonder if Sony is continuing to develop this product,
perhaps waiting for the technology to improve the frame rate, or if
it was completely shelved in favor of the 6MP sensor that found its
way into the D100 and *ist.

Is this the prototype for the next generation of low to mid priced
digital SLRs, or just a relic?

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/

 

[DavidMillier]

Do you expect the same to keep repeating itself ad infinitum?

Yes. There's no evidence that we've "hit the wall" in sensor
technologies at the larger sizes. And that doesn't even take into
account add-on NR or in-camera post processing techniques.

Pixel size cannot keep shrinking. There are two sets of limits to
how small a pixel can get. First, there is the wavelength of
visible light. Make a pixel smaller than the wavelength of light,
then it would not be able to detect the light. Second,
photosensors are made of silicon, and there is a limit on the light
sensitivity of silicon. Too small a pixel makes it insensitive to
light. A small pixel also requires a lot of signal amplification,
another source of noise. According to the Extreme Tech article
"Anatomy of a Digital Camera Image Sensors"

"The smallest CCD pixel feasible has been demonstrated at 2.5 microns
square. But that's for digital video and not still digital cameras.

The Sony 8MP chip is 2.9 microns isn't it? And despite the noted issues, Michael Reichmann (a 1Ds user) is pretty favourable towards the cameras using it.

Note that
the dynamic range of an image is affected by the pixel size and
design. A
digital still camera requires at least 10-bit dynamic range,
meaning it needs
slightly larger pixels of present designs. Currently, the smallest
pixel size
for digital cameras is about 3.3 microns in a CCD and about 4
microns for
CMOS. The popular Sony 3.3MP sensor has a 3.45-micron pixel. Many
other consumer digital cameras have pixels averaging around 5
microns."

According to these figures, pixel size in existing cameras are
close to the lower limit.

But 1/2 frame sensors are much bigger than 2/3 sensors so shrinking the current 8 micron pixels down to say 4 or 5 micron shouldn't be too difficult even with current technology. How many pixels will fit 1/2 frame at this pixel pitch? I would be quite happy to sacrifice ISO speeds above 400 (even 200 for that matter - I'm using a 14n and this full frame monster has rather embarassing noise at ISO 200...) to get high res in a smaller package. Current pro SLR form factors are just too big for my tastes.

Smaller pixels also have lower angular response, so that vignetting
becomes a more noticeable problem.

It is noticeable on the full frame 14n too, despite software corrections.

You seem to be guilty of the same thing that everyone else is,
equating photosite size to light reception area size. Most current
sensor designs have very small light reception areas compared to
the photosite size. It's advances like those of FillFactory or
Nikon with LBCAST that are making changes to that equation. In the
near term I believe that any loss of photosite size is likely made
up by improvements in light retention wells, microlenses, and
improved dark current conditions. Long term, it's harder to see
what the improvements will be, but I've heard enough to convince me
they'll be there, at least at APS and full frame sizes.

I am aware of the difference between the light sensitive area of a
pixel and the electronic part. Even if you can make the photodiode
fit the entire pixel, there is still a limit on how small a pixel
can get. So pixels simply cannot keep shrinking. No matter how
small a pixel can shrink, a full frame can pack more of them and
give higher resolution, while allowing the use of wide angle lenses
designed for 35mm.

Maybe not keep shrinking them, but you could certainly shrink a little bit - enough to get the 10MP or so that is all you really need for the common Super A3 printers.

Further, the smaller a pixel, the less
sensitive it is too light coming in at an angle, which means
vignetting will be worse for cameras with small photodiodes.

Photographers, however, want bigger chips.

Not sure I agree. Real photographers are interested in results, not
technology.

Bigger chips give them better results at any stage in the
development of sensor technology. One photographer is comparing
the EOS1DS images to medium format film and claim that resolution
is close enough and noise is similarly small.
Can't get the same
results with an APS sized sensor yet.

The problem with the 1/2 frame is the pixel count isn't quite high enough. But shrinking pixels to allow 10MP should be OK as long as you are prepared to trade a bit of sensitivity.

A photographer with a full
frame and a 17mm f/3.5 lens can go out and get results that a D70
owner cannot.

10.5MPDX, 12-24Dx, 12-24 Sigma??

The same photographer can get results with a 35mm
shift lens on a full frame that is impossible for a D70 owner. No
one makes 24mm DX shift lenses.

 

[Anastigmat]

The Sony 8MP chip is 2.9 microns isn't it? And despite the noted
issues, Michael Reichmann (a 1Ds user) is pretty favourable towards
the cameras using it.

I am sure that there are people who like these cameras but they are unmistakably more noisy.

ISO 800 film is more noisy than ISo 100 film but the faster film has its share of users. But does that mean that ISO800 film is better in terms of image quality? Not by a long shot.

But 1/2 frame sensors are much bigger than 2/3 sensors so shrinking
the current 8 micron pixels down to say 4 or 5 micron shouldn't be
too difficult even with current technology. How many pixels will
fit 1/2 frame at this pixel pitch? I would be quite happy to
sacrifice ISO speeds above 400 (even 200 for that matter - I'm
using a 14n and this full frame monster has rather embarassing
noise at ISO 200...) to get high res in a smaller package. Current
pro SLR form factors are just too big for my tastes.

I am afraid, however, that you are sacrificing ISO speed without getting a smaller package. Digital camera bodies haven't shrunk to fit the small ensor. The 10D for example, is bigger than any film camera I have ever owned, including the Nikon cameras I have owned.

The problem with the 1/2 frame is the pixel count isn't quite high
enough. But shrinking pixels to allow 10MP should be OK as long as
you are prepared to trade a bit of sensitivity.

But why? Why shrink the pixel to get more resolution when you can grow the sensor without increasing camera size or having to redesign the lens mount to get a larger image circle. 35mm lenses are already capapble of covering the full frame sensor, so why keep shrinking the individual pixel if it is not even necessary from the point of view of the photographer? I know the chip maker can make more money by making small sensors, but why let the bean counters at chip making companies dictate to us photographers how big of a sensor we are allowed to have?

 

[DavidMillier]

The Sony 8MP chip is 2.9 microns isn't it? And despite the noted
issues, Michael Reichmann (a 1Ds user) is pretty favourable towards
the cameras using it.

I am sure that there are people who like these cameras but they are
unmistakably more noisy.

However, are they too noisy? MR obviously thinks not as he was happy to compare his A2 against the 1D Mk2 and found that in actual prints up to A3 there was really not that much in it as far as image quality is concerned. And to a certain extent I can confirm that - I shot a couple of test shots with an A2 from the door of my local branch of Jessops and printed them to 15 x 10 inches. I find the prints remarkably good. Are they worse than similar sized prints from my 14n - difficult to tell for sure.

ISO 800 film is more noisy than ISo 100 film but the faster film
has its share of users. But does that mean that ISO800 film is
better in terms of image quality? Not by a long shot.

But 1/2 frame sensors are much bigger than 2/3 sensors so shrinking
the current 8 micron pixels down to say 4 or 5 micron shouldn't be
too difficult even with current technology. How many pixels will
fit 1/2 frame at this pixel pitch? I would be quite happy to
sacrifice ISO speeds above 400 (even 200 for that matter - I'm
using a 14n and this full frame monster has rather embarassing
noise at ISO 200...) to get high res in a smaller package. Current
pro SLR form factors are just too big for my tastes.

I am afraid, however, that you are sacrificing ISO speed without
getting a smaller package. Digital camera bodies haven't shrunk to
fit the small ensor. The 10D for example, is bigger than any film
camera I have ever owned, including the Nikon cameras I have owned.

Well my D100 and my 14n are both based around the F80 body. The F80 is hardly small (say compared to the Pentax ME Super of yesteryear) , the D100 is a little larger but the 14n is huge in comparison to the D100.

The oly E1 may point the way: whilst not a small camera it does aim to be a pro camera and it is a tad smaller that a D1x...

The problem with the 1/2 frame is the pixel count isn't quite high
enough. But shrinking pixels to allow 10MP should be OK as long as
you are prepared to trade a bit of sensitivity.

But why? Why shrink the pixel to get more resolution when you can
grow the sensor without increasing camera size or having to
redesign the lens mount to get a larger image circle. 35mm lenses
are already capapble of covering the full frame sensor, so why keep
shrinking the individual pixel if it is not even necessary from the
point of view of the photographer? I know the chip maker can make
more money by making small sensors, but why let the bean counters
at chip making companies dictate to us photographers how big of a
sensor we are allowed to have?

Two main reasons: much lower production costs will mean lower prices in the shops once price competition really starts to bite in the DSLR market and very importantly, there is no reason why DSLRs shouldn't shrink dramatically as the technologies mature. The Pentax DSLR is pretty small already actually, but I like to eventually see 10+MP DSLRs the size of the Fuji s5000 (which looks the part but isn't). As much as I appreciate what my 14n can do image wise, it is a tank.

 

[Joseph S Wisniewski]

But 1/2 frame sensors are much bigger than 2/3 sensors so shrinking
the current 8 micron pixels down to say 4 or 5 micron shouldn't be
too difficult even with current technology. How many pixels will
fit 1/2 frame at this pixel pitch?

For a 1.5x crop factor like Nikon, 24mp at 4um, 15.3mp at 5um.

Well, you asked.

I would be quite happy to
sacrifice ISO speeds above 400 (even 200 for that matter - I'm
using a 14n and this full frame monster has rather embarassing
noise at ISO 200...) to get high res in a smaller package. Current
pro SLR form factors are just too big for my tastes.

Then again, noise decreases with downsampling, so you could have that 24mp at ISO 100 or 200 in a studio, or for architecture or landscape, and sample down to 6MP for ISO 800 on up.

I won't get into 14n weirdness...

You haven't been trying to change it's focusing screen, have you?

Smaller pixels also have lower angular response, so that vignetting
becomes a more noticeable problem.

It is noticeable on the full frame 14n too, despite software
corrections.

Yeah, worse than any other camera on the market, 1.3x, 1.5x, 1.6x, or even the 2x Oly. It's more about sensor design than sensor size.

I am aware of the difference between the light sensitive area of a
pixel and the electronic part. Even if you can make the photodiode
fit the entire pixel, there is still a limit on how small a pixel
can get. So pixels simply cannot keep shrinking. No matter how
small a pixel can shrink, a full frame can pack more of them and
give higher resolution, while allowing the use of wide angle lenses
designed for 35mm.

Maybe not keep shrinking them, but you could certainly shrink a
little bit - enough to get the 10MP or so that is all you really
need for the common Super A3 printers.

You're making too much sense, talking about what peopel actually need.

Further, the smaller a pixel, the less
sensitive it is too light coming in at an angle, which means
vignetting will be worse for cameras with small photodiodes.

Photographers, however, want bigger chips.

Not sure I agree. Real photographers are interested in results, not
technology.

Bigger chips give them better results at any stage in the
development of sensor technology. One photographer is comparing
the EOS1DS images to medium format film and claim that resolution
is close enough and noise is similarly small.
Can't get the same
results with an APS sized sensor yet.

The problem with the 1/2 frame is the pixel count isn't quite high
enough. But shrinking pixels to allow 10MP should be OK as long as
you are prepared to trade a bit of sensitivity.

A photographer with a full
frame and a 17mm f/3.5 lens can go out and get results that a D70
owner cannot.

10.5MPDX, 12-24Dx, 12-24 Sigma??

There you go with that "reality" stuff again.

--
A cyberstalker told me not to post anymore...
So I'm posting even more!

Ciao!

Joe

http://www.swissarmyfork.com

 

[DavidMillier]

But 1/2 frame sensors are much bigger than 2/3 sensors so shrinking
the current 8 micron pixels down to say 4 or 5 micron shouldn't be
too difficult even with current technology. How many pixels will
fit 1/2 frame at this pixel pitch?

For a 1.5x crop factor like Nikon, 24mp at 4um, 15.3mp at 5um.

Well, you asked.

Welll, in that case 12MP would obviously fit into 1/2 frame witjhout shrinking too far - if it's good enough for the 1ds to max out a printer at super A3, it's good enough for me even if I'm limited to ISO 50. My tripod doesn't care whether it's ISO 50 or 500!

I would be quite happy to
sacrifice ISO speeds above 400 (even 200 for that matter - I'm
using a 14n and this full frame monster has rather embarassing
noise at ISO 200...) to get high res in a smaller package. Current
pro SLR form factors are just too big for my tastes.

Then again, noise decreases with downsampling, so you could have
that 24mp at ISO 100 or 200 in a studio, or for architecture or
landscape, and sample down to 6MP for ISO 800 on up.

I won't get into 14n weirdness...

You haven't been trying to change it's focusing screen, have you?

ROFL whilst clutching recently emptied wallet...

Smaller pixels also have lower angular response, so that vignetting
becomes a more noticeable problem.

It is noticeable on the full frame 14n too, despite software
corrections.

Yeah, worse than any other camera on the market, 1.3x, 1.5x, 1.6x,
or even the 2x Oly. It's more about sensor design than sensor size.

I am aware of the difference between the light sensitive area of a
pixel and the electronic part. Even if you can make the photodiode
fit the entire pixel, there is still a limit on how small a pixel
can get. So pixels simply cannot keep shrinking. No matter how
small a pixel can shrink, a full frame can pack more of them and
give higher resolution, while allowing the use of wide angle lenses
designed for 35mm.

Maybe not keep shrinking them, but you could certainly shrink a
little bit - enough to get the 10MP or so that is all you really
need for the common Super A3 printers.

You're making too much sense, talking about what peopel actually need.

Further, the smaller a pixel, the less
sensitive it is too light coming in at an angle, which means
vignetting will be worse for cameras with small photodiodes.

Photographers, however, want bigger chips.

Not sure I agree. Real photographers are interested in results, not
technology.

Bigger chips give them better results at any stage in the
development of sensor technology. One photographer is comparing
the EOS1DS images to medium format film and claim that resolution
is close enough and noise is similarly small.
Can't get the same
results with an APS sized sensor yet.

The problem with the 1/2 frame is the pixel count isn't quite high
enough. But shrinking pixels to allow 10MP should be OK as long as
you are prepared to trade a bit of sensitivity.

A photographer with a full
frame and a 17mm f/3.5 lens can go out and get results that a D70
owner cannot.

10.5MPDX, 12-24Dx, 12-24 Sigma??

There you go with that "reality" stuff again.

--
A cyberstalker told me not to post anymore...
So I'm posting even more!

Ciao!

Joe

http://www.swissarmyfork.com

 

[Ron Parr]

Pixel size cannot keep shrinking.

Nobody is arguing this position, so you can stop going after this straw man.

There are two sets of limits to
how small a pixel can get. First, there is the wavelength of
visible light. Make a pixel smaller than the wavelength of light,
then it would not be able to detect the light. Second,
photosensors are made of silicon, and there is a limit on the light
sensitivity of silicon. Too small a pixel makes it insensitive to
light. A small pixel also requires a lot of signal amplification,
another source of noise. According to the Extreme Tech article
"Anatomy of a Digital Camera Image Sensors"

"The smallest CCD pixel feasible has been demonstrated at 2.5 microns
square. But that's for digital video and not still digital cameras.
Note that
the dynamic range of an image is affected by the pixel size and
design. A
digital still camera requires at least 10-bit dynamic range,
meaning it needs
slightly larger pixels of present designs. Currently, the smallest
pixel size
for digital cameras is about 3.3 microns in a CCD and about 4
microns for
CMOS. The popular Sony 3.3MP sensor has a 3.45-micron pixel. Many
other consumer digital cameras have pixels averaging around 5
microns."

According to these figures, pixel size in existing cameras are
close to the lower limit.

Several things come to mind here:

1. You appear to have confused a statement about the current smallest pixels with a statement about the a fundamental limit.

2. You seem confused about pixel sizes since the sensor discussed in the datasheet that started this thread is 10.66 MP with 5.9 micron pixels. You began by criticizing this, but now it seems to fall well within the limits you are describing.

3. The recently introduced P150 seems to have pixels that are around 2.3 microns, which is already smaller than the supposed limit you have indicated above. We don't have the exact dimensions of the new Panasonic DMC-FX7, but it is likely to have pixels smaller than 2.5 microns given the pixel cound and 1/2.5" sensor size.

I am aware of the difference between the light sensitive area of a
pixel and the electronic part. Even if you can make the photodiode
fit the entire pixel, there is still a limit on how small a pixel
can get. So pixels simply cannot keep shrinking. No matter how
small a pixel can shrink, a full frame can pack more of them and
give higher resolution, while allowing the use of wide angle lenses
designed for 35mm. Further, the smaller a pixel, the less
sensitive it is too light coming in at an angle, which means
vignetting will be worse for cameras with small photodiodes.

Again, nobody is arguing for infinitely small pixels so your comments are addressing a straw man that has nothing to do with the discsussion.

You began by arguing, in reference to a 5.9 micron pixel design, "such a chip would not be the next generation of DSLR sensor, since such a chip would almost certainly be noisier, given its small size, than the current chip."

When challenged, you have offered no evidence to support that this engineering problem cannot be solved and have, instead, only attacked straw man positions that none of the discussants hold.

Not very convincing.

--
Ron Parr
FAQ: http://www.cs.duke.edu/~parr/photography/faq.html
Gallery: http://www.pbase.com/parr/