D30 Replacement

[Trenton]

The one that is in R&D for late 2002 announcement is the more
likely unit I'll go for.

Where did you hear this?

-Trenton

 

[JCDoss]

Although the current ccd sizes being less than film are a problem
as far as multiplier factors(greater than 1) are concerned for
standard 35mm lenses, they will not pose the same problem once
sensors are larger than 35mm film. The reason is simple physical
optics, its possible to distribute or disperse the incoming light
from the external lens by use of a fixed internal lens that adjusts
for the fractional multiplier ( it would be less than 1 for
sensors larger than film when processing light passing through
standard 35mm lenses) Using a technique such as this would ensure
improvements of low signal to noise ratio's, high resolution and
larger pixel area for sensors larger than 35mm film dimensions. I
think the digital cameras of 10 years from now will display
resolution, noise and color fidelity far beyond the capabilities
of any traditional film formats.

Although interesting, this technique seems to make the assumption that sensor technology will not advance beyond what it is currently, which is an awfully dangerous assumption.

First, the more glass you use, the more image degradation. What you're proposing is to put a magnifier between the rear element and the sensor and open up the inside of the camera to allow the magnifier to spread the image circle over a larger-than-35mm frame. This technology virtually already exists... the magnifiers are called teleconverters... and they're used very judiciously by many photographers.

...In my opinion a 24mp digital image would provide
resolution comparable or better than medium format film sizes
leaving 35mm long behind...

Good for US!

Also remember that by increasing the
number of pixels while maintaining the same ccd dimension we reduce
signal to noise per photosite, this makes image processing more
difficult and less likely to provide as noise free an image as
would be possible if the photosite dimensions are not reduced.

That was sort of the point of my original post... ie, how would S/N be maintained while decreasing the pixel size? I suspect some method of on-chip processing will eventually provide that breakthrough... and since on-chip processing is a CMOS thing, then Canon is probably well ahead of everyone else.

JCDoss

 

[David SL]

Although the current ccd sizes being less than film are a problem
as far as multiplier factors(greater than 1) are concerned for
standard 35mm lenses, they will not pose the same problem once
sensors are larger than 35mm film. The reason is simple physical
optics, its possible to distribute or disperse the incoming light
from the external lens by use of a fixed internal lens that adjusts
for the fractional multiplier ( it would be less than 1 for
sensors larger than film when processing light passing through
standard 35mm lenses) Using a technique such as this would ensure
improvements of low signal to noise ratio's, high resolution and
larger pixel area for sensors larger than 35mm film dimensions. I
think the digital cameras of 10 years from now will display
resolution, noise and color fidelity far beyond the capabilities
of any traditional film formats.

Although interesting, this technique seems to make the assumption
that sensor technology will not advance beyond what it is
currently, which is an awfully dangerous assumption.

I don't think I implied that technology would not advance in my previous statement but let me clarify. I am only saying that the area of the CCD can be made larger than film size and with continued subdivision(more photosites) at the larger sizes still gain better signal to noise ratios per photosite(since each one can be larger or the same area of current pro grade ccd's since the ccd area is larger than film) In short there is a limit on the ability to reduce noise given the number of photosites on a ccd but this limit will vary depending on the sizes beyond film that ccd's may (will) be made in the future.

First, the more glass you use, the more image degradation.

That is only true only up to our ability to recognize any distortion, using a fixed lens introduces only optic distortion on one lens element as a factor in image distortion, a single element lens can be ground to astonishing tolerences. The internal lens of which I speak would in fact, because it is a single lens would not need to be ground to the tolerances that external lens groups currently require. Lens groups require precise grinding of the lens on each tranmissive surface and more precision is required for each transmissive surface. A single lens however only needs to be ground so that the inbound image illuminates the surface of the ccd accurately. Any image degredation would be very imperceptible across the image to the human eye under almost all circumstances. Take a look at Fig. 1 to see what I mean:

Fig. 1



The greater than film size CCD is noted as "ccd_Gf" for "greater than film" in area. The film size ccd is "ccd_f" as you can see the green lens on the inside of the development cavity is what I am proposing. It is attached to the internal cavity itself, light from the rear nodal point of the attached lens would then be magnified to fill the area of the greater than film size CCD. The red light rays show the path to the larger ccd. In traditional cameras, this lens is not present and the light focuses on "ccd_f" as show using the purple light rays. As mentioned before, the internal lens allows CCD's to increase in size (up to the size of the development cavity of course) and increase in resolution while improving signal to noise for a fixed MP rating relative to the film size CCD. IE. A 6mp ccd_GF camera would have larger photosites and better noise performance than a 6mp ccd_f camera. For larger MP sizes relative to the film size CCD there will still be an advantage of signal to noise, since 8mp on "ccd_gf" yields larger photosites than 8mp on "ccd_f".

What
you're proposing is to put a magnifier between the rear element and
the sensor and open up the inside of the camera to allow the
magnifier to spread the image circle over a larger-than-35mm frame.
This technology virtually already exists... the magnifiers are
called teleconverters... and they're used very judiciously by
many photographers.

Sort of, teleconverters always concentrate or converge the light onto the surface of the film or ccd. The internal lens I am proposing would always disperse incoming light onto the larger than film ccd surface. (So it would do the opposite of a teleconverter.) It would be a (one element) lens for the inside of the development cavity that disperses the incoming light from the rear nodal point of the attached external lens so that it matches the surface area of the greater than film ccd.(As shown in Fig. 1)

...In my opinion a 24mp digital image would provide
resolution comparable or better than medium format film sizes
leaving 35mm long behind...

Good for US!

Also remember that by increasing the
number of pixels while maintaining the same ccd dimension we reduce
signal to noise per photosite, this makes image processing more
difficult and less likely to provide as noise free an image as
would be possible if the photosite dimensions are not reduced.

That was sort of the point of my original post... ie, how would S/N
be maintained while decreasing the pixel size? I suspect some
method of on-chip processing will eventually provide that
breakthrough... and since on-chip processing is a CMOS thing, then
Canon is probably well ahead of everyone else.

I agree, Canon seems to be in another league when it comes to whipping CMOS into shape, having already given us a "pro. level" CMOS camera (D30)

My point is basically that there remain many technological and physical principles waiting to be exploited by digital camera designers.

--DSL

 

[Mohit]

Trenton,

The one that is in R&D for late 2002 announcement is the more
likely unit I'll go for.

Where did you hear this?

Lee doesn't know about any R&D details, I've called him on this in previous threads. He makes claims about what he thinks is going on, then states it as if it is confirmed public knowledge.

The rumors on Canon's announcement next month are bad enough, but to speculate blindly about stuff coming out a year or two from now ...

• Mohit

 

[JCDoss]

I don't think I implied that technology would not advance in my
previous statement but let me clarify. I am only saying that the
area of the CCD can be made larger than film size...

Let me put this another way...

You are proposing a (very good) optical solution to the problem of low resolution for DSLR cameras. Low resolution is caused by large pixels, and represents the trade-off between quality and quantity. DSLRs opt for quality over quantity based on today's technology.

There is no reason to assume that CMOS devices developed in the year 2005 will have twice or four times the resolution per surface area with the same S/N and sensitivity quality of today's sensors. Put another way, it is more likely that the issue of high quality small pixels (ie, high quality, hi res sensors) will be solved at the sensor level in the future. Therefore, such an exotic solution involving extra glass and very large (and undoubtedly exponentially expensive) sensors would ultimately be short-lived and unnecessary for the long term.

First, the more glass you use, the more image degradation.

...Any image degredation would be very imperceptible
across the image to the human eye under almost all circumstances.

The internal lens itself will cause at least some image degradation whether it's perceptible or not. However, one thing any magnifying lens will do is magnify any lens imperfections of the main lens.

For example, this internal lens may or may not cause a degeneration of the image taken with L-quality glass... however slap a Sigma lens on, or a consumer lens with known artefacts or imperfections, and they're blown up to poster size in the images. Chromatic aberrations and edge sharpness should be MUCH more visible when magnified as such.

...As mentioned before, the internal lens
allows CCD's to increase in size (up to the size of the development
cavity of course) and increase in resolution while improving signal
to noise for a fixed MP rating relative to the film size CCD...

The concept is sound, but I have three issues:

1) Is it really possible to get such magnificent, distortion-free magnification from a single element device (ie, very good teleconverters have multiple elements).

2) The cost of a full-frame sensor is currently so high that no full-frame DSLRs are available. In fact, Pentax apparently bailed out of their full frame DSLR project due to cost. Do you really think a "larger than 35mm" sensor is financially feasible?

3) This new camera that you propose will necessarily be much larger than current models, and I suspect most will balk at the idea of a very large camera like this.

My point is basically that there remain many technological and
physical principles waiting to be exploited by digital camera
designers.

I don't mean to dismiss your proposal, because it does sound like a viable solution to solving several problems with DSLRs as they are today... but I think it's (a) too expensive, (b) prone to image degradation, (c) will make the equipment much larger, and (d) will likely be eclipsed by advances in sensor technology before it can be widely adopted (and therefore able to pay for itself).

I enjoy the discussion!
JCDoss

 

[Lee Rothman]

The guys at Pictureline in Salt Lake City were told this at a Canon show late this last year.

The one that is in R&D for late 2002 announcement is the more
likely unit I'll go for.

Where did you hear this?

-Trenton

 

[JCDoss]

Entertaining, enlightening exchange.

What are your thoughts on the topic, Larry? I know you've got an interest in techie things like this...

JCDoss

 

[Larry H. Smith]

Hi JC,

Unfortunately "interest" doesn't necessarily imply education-in-this-field.

You guys left me way-back in the shallow end of the pool.

But keep-on-talkin', I'll wade around picking-up whatever I can.

Larry

Entertaining, enlightening exchange.

What are your thoughts on the topic, Larry? I know you've got an
interest in techie things like this...

JCDoss

 

[David SL]

Excellent response JC,

You are correct in pointing out that cost NOW is a definite issue, however, I don't see this particular solution in the near term but as something to be done in the future when CCD prices are much reduced from their current levels. Perhaps 5 to 10 years into the future. However, sensor advances using CCD's on the size scale of current consumer chips and approaching film size CCD's will still always be restricted by the size of the photosites and the resulting signal to noise ratio from the smaller elements as MP increases. Physics does place lower bounds on the level of subdivision (increase in resolution) that can be achieved given a particular surface area, simultaneously it reduces signal to noise and resolvable dynamic range while doing this. So, though I agree with you that advances will continue to occur in this area they will reach a limit where it becomes increasingly difficult and expensive to keep noise down and maintain or increase dynamic range. (as for example consumer grade CCD's at 3/4" have with 4 and 5mp) The need for greater resolution and better noise performance and the fast reduction in CCD production costs over time will make larger than film CCD solutions(such as the one I propose) just as viable as increasing sensor efficiency on smaller CCD's. As for the issue of image degredation on the internal lens I would have to disagree, any image imperfections that exists as a consequence of the external lens group will impinge upon the CCD surface any way and may or may not be visible. For the larger than film CCD the inbound image light is only distributed across the CCD, any distortion caused by the external lens group will be proportionatly the same as in an equal MP film size CCD. For larger MP greater than film CCD's, the degredation will actually be reduced as it's dispersed across more pixels comprising the image plane. As for your fear of large camera size, I don't think larger cameras will be needed at all, percentage change of between 5 and 20% will yield CCD's physically not much bigger than film. However, these sensors will be capable of providing several additional megapixels of resolution or improved signal to noise in comparison to the same MP on a film size CCD.

I am confident that however it turns out, we will be using relatively "cheap" cameras over the next 5 to 10 years with image quality far beyond the best of what is available now, and for that I am very thankful. I think Moore's law has taken a while to get into the digital market, but as it takes hold and is forced by large demand and increased competition for a larger piece of the bigger pie, the consumer will reap benefits similar to those currently being enjoyed in the personal computer market. (Cheap memory,processors and HDD's) I appreciate your thoughts on the subject, very good discussion indeed.

Regards,

Dsl

My point is basically that there remain many technological and
physical principles waiting to be exploited by digital camera
designers.

I don't mean to dismiss your proposal, because it does sound like a
viable solution to solving several problems with DSLRs as they are
today... but I think it's (a) too expensive, (b) prone to image
degradation, (c) will make the equipment much larger, and (d) will
likely be eclipsed by advances in sensor technology before it can
be widely adopted (and therefore able to pay for itself).

I enjoy the discussion!
JCDoss

--DSL

 

[Chris Fritz]

Assuming that the noise is random, such as in the D30, you can improve your signal to noise ratio in Photoshop.

The equation states that there is a square root of the number of samples reduction in random noise if these samples are averaged together.

Example: if there is a 2x horizontal and 2x vertical reduction in resolution by averaging, there is a 2x (sqrt[4]) decrease in the noise.

It is still better to have a sensor with bigger pixels due to the increase in sensitivity.

Chris

My guess it that the S/N would be improved over using 32 Mpixels,
but wouldn't be as good as using 8 Mpixels of larger size.

And I'm not sure that just a "downsizing" in photoshop would give
that S/N improvement. It might have to be done in-camera, or with
some other external program other than Photoshop.

Hopefully some of the people here more familiar with this can give
some insight.

If what you're saying is right, then Yippeeee!

The beauty should be that downsizing a 32MP image should provide
the same signal/noise ratio as an 8MP image taken with larger
pixels. That's by today's standards, though.

--
The Unofficial Photographer of The Wilkinsons
http://thewilkinsons.crosswinds.net

 

[Andrew Grant27981]

I don't agree. I don't think photosites have to get that much smaller to fully utilize the image quality of most lenses. Once sensors have reached 35mm size, I doubt resolution will go much over 10-12MP in 35mm bodies which is plenty for most uses. Those who need to go over 11x14 on a regualr basis will probably just buy a medium format digital as they do now.

You are correct in pointing out that cost NOW is a definite issue,
however, I don't see this particular solution in the near term but
as something to be done in the future when CCD prices are much
reduced from their current levels. Perhaps 5 to 10 years into the
future. However, sensor advances using CCD's on the size scale of
current consumer chips and approaching film size CCD's will still
always be restricted by the size of the photosites and the
resulting signal to noise ratio from the smaller elements as MP
increases. Physics does place lower bounds on the level of
subdivision (increase in resolution) that can be achieved given a
particular surface area, simultaneously it reduces signal to noise
and resolvable dynamic range while doing this. So, though I agree
with you that advances will continue to occur in this area they
will reach a limit where it becomes increasingly difficult and
expensive to keep noise down and maintain or increase dynamic
range. (as for example consumer grade CCD's at 3/4" have with 4 and
5mp) The need for greater resolution and better noise performance
and the fast reduction in CCD production costs over time will make
larger than film CCD solutions(such as the one I propose) just as
viable as increasing sensor efficiency on smaller CCD's. As for
the issue of image degredation on the internal lens I would have to
disagree, any image imperfections that exists as a consequence of
the external lens group will impinge upon the CCD surface any way
and may or may not be visible. For the larger than film CCD the
inbound image light is only distributed across the CCD, any
distortion caused by the external lens group will be proportionatly
the same as in an equal MP film size CCD. For larger MP greater
than film CCD's, the degredation will actually be reduced as it's
dispersed across more pixels comprising the image plane. As for
your fear of large camera size, I don't think larger cameras will
be needed at all, percentage change of between 5 and 20% will yield
CCD's physically not much bigger than film. However, these sensors
will be capable of providing several additional megapixels of
resolution or improved signal to noise in comparison to the same
MP on a film size CCD.

I am confident that however it turns out, we will be using
relatively "cheap" cameras over the next 5 to 10 years with image
quality far beyond the best of what is available now, and for that
I am very thankful. I think Moore's law has taken a while to get
into the digital market, but as it takes hold and is forced by
large demand and increased competition for a larger piece of the
bigger pie, the consumer will reap benefits similar to those
currently being enjoyed in the personal computer market. (Cheap
memory,processors and HDD's) I appreciate your thoughts on the
subject, very good discussion indeed.

Regards,

Dsl

My point is basically that there remain many technological and
physical principles waiting to be exploited by digital camera
designers.

I don't mean to dismiss your proposal, because it does sound like a
viable solution to solving several problems with DSLRs as they are
today... but I think it's (a) too expensive, (b) prone to image
degradation, (c) will make the equipment much larger, and (d) will
likely be eclipsed by advances in sensor technology before it can
be widely adopted (and therefore able to pay for itself).

I enjoy the discussion!
JCDoss

--
DSL

 

[JCDoss]

I don't agree. I don't think photosites have to get that much
smaller to fully utilize the image quality of most lenses. Once
sensors have reached 35mm size, I doubt resolution will go much
over 10-12MP in 35mm bodies which is plenty for most uses. Those
who need to go over 11x14 on a regualr basis will probably just buy
a medium format digital as they do now.

D30-sized pixels on a full frame sensor would yield in the neighborhood of 8MP, which isn't too far from your 10-12MP projection of "plenty." That amounts to a reduction in pixel size from 10x10um to around 9x9um... not much of a reduction.

How small do pixels need to be to accommodate the finest 35mm lenses (afterall, that should be our goal, right?). I don't know what the maximum resolvable resolution of a that lens is, but I do know that 40lp/mm is a standard that many lenses are measured at. Therefore, let's assume that we're gunning for 50lp/mm (it makes the math easy).

50lp/mm = 50 line pairs per millimeter = 100 pixels per millimeter =
0.1 pixels per micron = 10 microns per pixel

10um/pixel is the maximum sampling rate of this lens. However, the maximum sampling rate of the sensor must equal the Nyquist frequency, or 1/2*lens sampling rate, in order to capture all data without aliasing or producing other errors. Therefore...

1/2*10um/pixel = 5um/pixel

...the dimensions of the pixels we're after are 5x5um, producing an image on a full frame sensor of 34.5MP (!).

Correct?
JCDoss

 

[David SL]

Andrew,

I thought of stating that there will be a limit on the high side as to how large sensor MP's will need to get but I left it out. I agree with your assessment below. 10Mp would be good enough for 17 x 14 inch prints at 204.5 ppi, which is about as sharp as an 8 x 10 3mp print today. There are some who have performed tests to show that the D30, a 3mp camera is able to exceed 35mm film in terms of resolving power. For this reason alone I am curious to see how the D30 replacement will perform.

Regards,

Dsl

Andrew Grant wrote:
I don't agree. I don't think photosites have to get that much
smaller to fully utilize the image quality of most lenses. Once
sensors have reached 35mm size, I doubt resolution will go much
over 10-12MP in 35mm bodies which is plenty for most uses. Those
who need to go over 11x14 on a regualr basis will probably just buy
a medium format digital as they do now.

You are correct in pointing out that cost NOW is a definite issue,
however, I don't see this particular solution in the near term but
as something to be done in the future when CCD prices are much
reduced from their current levels. Perhaps 5 to 10 years into the
future. However, sensor advances using CCD's on the size scale of
current consumer chips and approaching film size CCD's will still
always be restricted by the size of the photosites and the
resulting signal to noise ratio from the smaller elements as MP
increases. Physics does place lower bounds on the level of
subdivision (increase in resolution) that can be achieved given a
particular surface area, simultaneously it reduces signal to noise
and resolvable dynamic range while doing this. So, though I agree
with you that advances will continue to occur in this area they
will reach a limit where it becomes increasingly difficult and
expensive to keep noise down and maintain or increase dynamic
range. (as for example consumer grade CCD's at 3/4" have with 4 and
5mp) The need for greater resolution and better noise performance
and the fast reduction in CCD production costs over time will make
larger than film CCD solutions(such as the one I propose) just as
viable as increasing sensor efficiency on smaller CCD's. As for
the issue of image degredation on the internal lens I would have to
disagree, any image imperfections that exists as a consequence of
the external lens group will impinge upon the CCD surface any way
and may or may not be visible. For the larger than film CCD the
inbound image light is only distributed across the CCD, any
distortion caused by the external lens group will be proportionatly
the same as in an equal MP film size CCD. For larger MP greater
than film CCD's, the degredation will actually be reduced as it's
dispersed across more pixels comprising the image plane. As for
your fear of large camera size, I don't think larger cameras will
be needed at all, percentage change of between 5 and 20% will yield
CCD's physically not much bigger than film. However, these sensors
will be capable of providing several additional megapixels of
resolution or improved signal to noise in comparison to the same
MP on a film size CCD.

I am confident that however it turns out, we will be using
relatively "cheap" cameras over the next 5 to 10 years with image
quality far beyond the best of what is available now, and for that
I am very thankful. I think Moore's law has taken a while to get
into the digital market, but as it takes hold and is forced by
large demand and increased competition for a larger piece of the
bigger pie, the consumer will reap benefits similar to those
currently being enjoyed in the personal computer market. (Cheap
memory,processors and HDD's) I appreciate your thoughts on the
subject, very good discussion indeed.

Regards,

Dsl

My point is basically that there remain many technological and
physical principles waiting to be exploited by digital camera
designers.

I don't mean to dismiss your proposal, because it does sound like a
viable solution to solving several problems with DSLRs as they are
today... but I think it's (a) too expensive, (b) prone to image
degradation, (c) will make the equipment much larger, and (d) will
likely be eclipsed by advances in sensor technology before it can
be widely adopted (and therefore able to pay for itself).

I enjoy the discussion!
JCDoss

--
DSL

--DSL

 

[Michael Martin]

I don't agree. I don't think photosites have to get that much
smaller to fully utilize the image quality of most lenses. Once
sensors have reached 35mm size, I doubt resolution will go much
over 10-12MP in 35mm bodies which is plenty for most uses. Those
who need to go over 11x14 on a regualr basis will probably just buy
a medium format digital as they do now.

D30-sized pixels on a full frame sensor would yield in the
neighborhood of 8MP, which isn't too far from your 10-12MP
projection of "plenty." That amounts to a reduction in pixel size
from 10x10um to around 9x9um... not much of a reduction.

How small do pixels need to be to accommodate the finest 35mm
lenses (afterall, that should be our goal, right?). I don't know
what the maximum resolvable resolution of a that lens is, but I do
know that 40lp/mm is a standard that many lenses are measured at.
Therefore, let's assume that we're gunning for 50lp/mm (it makes
the math easy).

50lp/mm = 50 line pairs per millimeter = 100 pixels per millimeter =
0.1 pixels per micron = 10 microns per pixel

10um/pixel is the maximum sampling rate of this lens. However, the
maximum sampling rate of the sensor must equal the Nyquist
frequency, or 1/2*lens sampling rate, in order to capture all data
without aliasing or producing other errors. Therefore...

1/2*10um/pixel = 5um/pixel

...the dimensions of the pixels we're after are 5x5um, producing an
image on a full frame sensor of 34.5MP (!).

Correct?
JCDoss

I'd rather have less pixels at stay at the same size as my EOS 1D 11.5 microns.

 

[Andrew Grant27981]

I don't agree. I don't think photosites have to get that much
smaller to fully utilize the image quality of most lenses. Once
sensors have reached 35mm size, I doubt resolution will go much
over 10-12MP in 35mm bodies which is plenty for most uses. Those
who need to go over 11x14 on a regualr basis will probably just buy
a medium format digital as they do now.

D30-sized pixels on a full frame sensor would yield in the
neighborhood of 8MP, which isn't too far from your 10-12MP
projection of "plenty." That amounts to a reduction in pixel size
from 10x10um to around 9x9um... not much of a reduction.

How small do pixels need to be to accommodate the finest 35mm
lenses (afterall, that should be our goal, right?). I don't know
what the maximum resolvable resolution of a that lens is, but I do
know that 40lp/mm is a standard that many lenses are measured at.
Therefore, let's assume that we're gunning for 50lp/mm (it makes
the math easy).

50lp/mm = 50 line pairs per millimeter = 100 pixels per millimeter =
0.1 pixels per micron = 10 microns per pixel

10um/pixel is the maximum sampling rate of this lens. However, the
maximum sampling rate of the sensor must equal the Nyquist
frequency, or 1/2*lens sampling rate, in order to capture all data
without aliasing or producing other errors. Therefore...

1/2*10um/pixel = 5um/pixel

...the dimensions of the pixels we're after are 5x5um, producing an
image on a full frame sensor of 34.5MP (!).

Correct?
JCDoss

I'd rather have less pixels at stay at the same size as my EOS 1D
11.5 microns.

I don't think they need to be that large. The D1x has pixels about half the size of the D30's but that odd rectangular shape. I have heard though, that at least one Nikon lens (the 14/2.8) has some problems resolving enough detail for the D1x pixels.

JC, I don't dispute your math, I would say though that lenses tend to fall off in sharpness towards the edges, particularly wide angle and zoom lenses. Therefore I think something a lttle closer to the 9um would probably be close enough.

 

[Edmund Ronald]

I am not very happy with this reasoning, The noise of each small pixel will be a random, and we will average over the number of small pixels. Noise dispersion (variance) will diminish, COMPARED TO THAT FROM A SINGLE SMALL PIXEL. But this does not tell us how the Noisesmall/Noiselarge ratio looks, as this depends on process electronics, temperature curves etc.

I would call your analysis erroneous.

Edmund

The equation states that there is a square root of the number of
samples reduction in random noise if these samples are averaged
together.

Example: if there is a 2x horizontal and 2x vertical reduction in
resolution by averaging, there is a 2x (sqrt[4]) decrease in the
noise.

It is still better to have a sensor with bigger pixels due to the
increase in sensitivity.

Chris

My guess it that the S/N would be improved over using 32 Mpixels,
but wouldn't be as good as using 8 Mpixels of larger size.

And I'm not sure that just a "downsizing" in photoshop would give
that S/N improvement. It might have to be done in-camera, or with
some other external program other than Photoshop.

Hopefully some of the people here more familiar with this can give
some insight.

If what you're saying is right, then Yippeeee!

The beauty should be that downsizing a 32MP image should provide
the same signal/noise ratio as an 8MP image taken with larger
pixels. That's by today's standards, though.

--
The Unofficial Photographer of The Wilkinsons
http://thewilkinsons.crosswinds.net

--No film, no hassles !

 

[brian]

I don't think they need to be that large. The D1x has pixels about
half the size of the D30's but that odd rectangular shape. I have
heard though, that at least one Nikon lens (the 14/2.8) has some
problems resolving enough detail for the D1x pixels.

JC, I don't dispute your math, I would say though that lenses tend
to fall off in sharpness towards the edges, particularly wide angle
and zoom lenses. Therefore I think something a lttle closer to the
9um would probably be close enough.

Hi Andrew:

I use a D1x, and have tested it extensively with a wide range of lenses at all apertures. The horizontal and vertical resolution of this camera is 84cycles/mm and 42cycles/mm, respectively. I've photographed test charts that clearly show contrast reversal at precisely those frequencies in both the center and all four corners of the image field. This means that the ultimate limits of the lens was not yet reached, even at 84 cycles/mm. Admittedly, some lenses are better than others, but the conclusion is inescapable: 5 to 6 micron pixels CAN be realistically utilized by 35mm SLR lenses. That means that a 30-35mp 24x36mm sensor is a reasonable goal.

Brian

 

[brian]

David:

What you are essentially arguing for is a true medium format digital camera. This will work better than a 35mm digital camera just as a medium format film camera works better than a 35mm film camera. Your idea for modifying normal 35mm lenses to cover a larger format is similar to the old Speed Magny made by Nikon back in the 1960's to convert F-mount lenses for use on 2.25x3.25 or 4x5 film. The Speed Magny used a relay lens to magnify the 24x36mm image up to the desired size.

A couple of points:

1) In your proposed system the internal lens would need to have negative power, just like a teleconverter. You show a positive lens.

2) A single element teleconverter will produce an absolutely horrendous image, no matter how well you grind and polish the surfaces. There is a very good reason that high performance lenses have multiple elements.

3) Teleconverters disperse light, they don't concentrate it. Thats why they cause a loss of system speed: a small image gets dimmer when you magnify it. Front converters are afocal, so they don't change the system speed.

4) A better solution would be to develop a large sensor for existing medium format cameras. Kodak has already done this with their ProBack, which has an area of 35x35mm and has 16megapixel resolution. Existing medium format lenses will work better than modified 35mm lenses.

Brian

 

[Tom Colling]

Ewe no, sum thymes eye Ken bee sew stupid

Didn't you mean "stew pit"? (...or is it stoop-Ed"?)

Monica says, "Close, ...but no cigar!"

Larry

Your write. (I can't believe I couldn't come up with that! And I worked so hard on the other ones!)

 

[Andrew Grant27981]

Hi Andrew:
I use a D1x, and have tested it extensively with a wide range of
lenses at all apertures. The horizontal and vertical resolution of
this camera is 84cycles/mm and 42cycles/mm, respectively. I've
photographed test charts that clearly show contrast reversal at
precisely those frequencies in both the center and all four corners
of the image field. This means that the ultimate limits of the
lens was not yet reached, even at 84 cycles/mm. Admittedly, some
lenses are better than others, but the conclusion is inescapable:
5 to 6 micron pixels CAN be realistically utilized by 35mm SLR
lenses. That means that a 30-35mp 24x36mm sensor is a reasonable
goal.

Well thats good new Brian, if they can sort the noise issue out, pixels that size will completely kill film, bothe medium format and 35mm.