EF 100-400 IS USM on a Rebel

[Marty Schultz]

f-stop is not changed with 1.6x crop, because camera engineers
decided this would not be a good thing to happen (its just a
definition of ISO). This is why smaller sensors show more noise for
the same FOV.

No, go back to the definition of f-stop, the focal length divided by the aperture. With a TC you are increasing the focal length but not the aperture, so f-stop is reduced. Cropping the image at the focal plane does nothing to the lens. You said yourself the light density doesn't change (photons per square mm). Smaller sensors usually have more noise because the size of the pixels is smaller, with less light gathering area per pixel you need to increase the amplification of the signal more than you would if you had larger pixels that could gather more light.

DOF is an interesting point. I don't see why it should be different
between 1.0x/TC and 1.6x. Could be an interesting experiment.
Anyone with two DSLRs with different crop factors, a good tele lens
and a TC that exactly matches the cropping factor difference of the
cameras, willing to make the test?

If both cameras have the same number of pixels the DOF will be different because if you make the same size print from both cameras the amount of enlargment is different, the circle of confusion changes with sensor size.

--
Later,
Marty

http://science.widener.edu/~schultz/digipicts.html

 

[AchimS]

f-stop is not changed with 1.6x crop, because camera engineers
decided this would not be a good thing to happen (its just a
definition of ISO). This is why smaller sensors show more noise for
the same FOV.

No, go back to the definition of f-stop, the focal length divided
by the aperture. With a TC you are increasing the focal length but
not the aperture, so f-stop is reduced.

You are right about the definition of f-stop, and your conclusions (you did mean f-stop is increased). My explanation of the TC was not formula based, but rather showing the physics behind them, and is equally valid: the TC is projecting the lens' light onto a larger diameter circle, and the "full"-sized sensor is now smaller than the circle. There is light outside the sensor area which is lost. The light density decreases with the TC. The definition of the f-stop is such that the density is inverse proportional to the square of the f-stop, so with decreasing density, f-stop increases.

Cropping the image at the
focal plane does nothing to the lens. You said yourself the light
density doesn't change (photons per square mm).

Yes, we are in total agreement.

Smaller sensors
usually have more noise because the size of the pixels is smaller,
with less light gathering area per pixel you need to increase the
amplification of the signal more than you would if you had larger
pixels that could gather more light.

OK, now I need some help from formulas. Lets collect the facts we have.
Light density only depends on the f-number, not on the sensor size.
Smaller pixels mean less photons per pixel for the same light density.

The same f-number means less photons per pixel and time for a smaller sensor with the same number of pixels.

Take a full-frame sensor, 100mm/1.8 lens, and say for a given illumination and exposure time there are N photons collected per pixel.

With a 1.4xTC this is effectively a 140mm/2.5 lens, and each pixel has N/1.4 photons collected.
The same lens, on a 1.4x cropped sensor, without TC:

This is effectively a 140mm/1.8 lens, and each pixel has N/1.4 photons collected because their size is 1/1.4 of the original size.

So, although the f-stop is the same as for the full-frame sensor without TC, the cropped sensor has less photons collected, and this has to be compensated by the camera electronics, which will result in higher noise. This is what I wanted to say with the camera engineers and the crop factor, but it turned out to be misleading. In order for the f-stop definition to still be valid, camera electronics has to be adjusted.

In comparison with the TC/full-frame combo, the cropped sensor collects the same amount of light per pixel! No difference in noise here, because for the full-frame sensor you have to double the ISO to have the same exposure time.

DOF is an interesting point. I don't see why it should be different
between 1.0x/TC and 1.6x. Could be an interesting experiment.
Anyone with two DSLRs with different crop factors, a good tele lens
and a TC that exactly matches the cropping factor difference of the
cameras, willing to make the test?

If both cameras have the same number of pixels the DOF will be
different because if you make the same size print from both cameras
the amount of enlargment is different, the circle of confusion
changes with sensor size.

The circle of confusion and DOF are different things.

The first one refers to diffraction effects on the aperture -- they reduce the resolution and get worse for smaller apertures (larger f-numbers). A camera with crop-factor has smaller pixels, and therefore better resolves the circle of confusion as the full-frame sensor does. Cropped sensors do not allow to stop-down as much as full-frame sensor cameras because of this limitation.

DOF is the range of distances that are in focus. DOF increases with larger f-numbers. Take again the 100mm/1.8 example, used at full aperture. On the full-frame sensor without TC, we take the portrait of a face that nearly completely covers the frame (Shot 1). Something 1m behind the person is nicely out of focus. Placing the TC, we now have a smaller FOV and have to go further away to take the image frame. Now the distance ratio camera-object/object-background is increased, plus we have a smaller f-number, so the background is more in focus (Shot 2). If instead we take the 1.4x sensor camera without the TC, then FOV is the same as in Shot 2, we take the same distance, but still have f/1.8. So the cropped sensor gives a more shallow DOF, better for people photography? Perhaps for some reason we should not take into account the changed f-number in Shot 2, then Shot 2 and Shot 3 give equal DOF. I do not see why Shot 3 should give a larger DOF as Shot 2.

Your argument of the different amount of enlargement is not valid, because in one case you enlarge from a smaller sensor to the large print, and in the other you already enlarge the lens output in part in the TC, and in another part from the sensor to the print. The amount of enlargement needed from the original lens to the print is the same in both cases.

Which leads me to my original statement, which you still could not disprove, but rather gets more and more proved by calculations:

Using a 1.4x TC on a full-frame sensor has exactly the same effect as using a 1.4x crop factor sensor. (Not taking into account diffraction effects, which is valid for small f-numbers). The only difference is the magnified view in the viewfinder with the TC, for your eyes only, but not for the image.

No difference in FOV, nor in noise level, nor in DOF, nor in perspective. No difference at all. AF-speed might even be in favor of the cropped sensor, because of the smaller f-number.

Thanks for the interesting discussion (for me at least)
Achim

--
Later,
Marty

http://science.widener.edu/~schultz/digipicts.html

 

[Marty Schultz]

Take a full-frame sensor, 100mm/1.8 lens, and say for a given
illumination and exposure time there are N photons collected per
pixel.
With a 1.4xTC this is effectively a 140mm/2.5 lens, and each pixel
has N/1.4 photons collected.

Yes.

The same lens, on a 1.4x cropped sensor, without TC:
This is effectively a 140mm/1.8 lens, and each pixel has N/1.4
photons collected because their size is 1/1.4 of the original size.

Even if both cameras had the same number of pixels you might have even less area per pixel on the smaller sensor, see below on sensor design.

So, although the f-stop is the same as for the full-frame sensor
without TC, the cropped sensor has less photons collected, and this
has to be compensated by the camera electronics, which will result
in higher noise. This is what I wanted to say with the camera
engineers and the crop factor, but it turned out to be misleading.
In order for the f-stop definition to still be valid, camera
electronics has to be adjusted.

I think you mean the ISO definition, not f-stop, the smaller sensor needs more amplification per pixel to get the same ISO. f-stop is only defined by the lens, not the camera.

In comparison with the TC/full-frame combo, the cropped sensor
collects the same amount of light per pixel! No difference in noise
here, because for the full-frame sensor you have to double the ISO
to have the same exposure time.

I theory most of what you said is true, but in reality the pixel sizes are not that different:

The pixels of the 10D are 7.4um
The pixels of the 1Ds are 8.8um

Also as sensors have increased in density noise reduction technology has also improved, offsetting the increase in noise from smaller pixel areas. Just look at the D30, D60, 10D progression, same size chip and the noise keeps going down as technology improves.

There are also other factors that influence how large the photosites are on a chip, the area of the sensor is not all used for collecting photons, there are spaces between the pixels for various parts that are needed to access the pixels. Different designs of CCDs and CMOS make these parts larger or smaller relative to the pixels.

Your argument of the different amount of enlargement is not valid,
because in one case you enlarge from a smaller sensor to the large
print, and in the other you already enlarge the lens output in part
in the TC, and in another part from the sensor to the print. The
amount of enlargement needed from the original lens to the print is
the same in both cases.

If one sensor is larger than the other how can it be the same amount of enlargement? I thought we were comparing a TC that matches the crop factor, so the image capture by both sensors will be the same field of view. In that case the smaller sensor has to be inlarged more if you make the same size print from both.

Should be easy to calculate how different the 2 setups would be as far as DOF goes, use any of the DOF calculators and plug in the numbers. You would have to use the new f-stop and focal length because of the TC on the full frame camera, and the actual data from the cropped camera. I really don't feel like trying it but I would guess the enlargement is a minor factor. If you adjust the f-stop on the cropped camera to match the TC on the FF camera you would see only the change due to enlargement.

--
Later,
Marty

http://science.widener.edu/~schultz/digipicts.html

 

[AchimS]

Take a full-frame sensor, 100mm/1.8 lens, and say for a given
illumination and exposure time there are N photons collected per
pixel.
With a 1.4xTC this is effectively a 140mm/2.5 lens, and each pixel
has N/1.4 photons collected.

Yes.

The same lens, on a 1.4x cropped sensor, without TC:
This is effectively a 140mm/1.8 lens, and each pixel has N/1.4
photons collected because their size is 1/1.4 of the original size.

Even if both cameras had the same number of pixels you might have
even less area per pixel on the smaller sensor, see below on sensor
design.

Of course this is right, but to the first order we can comfortably ignore this.

So, although the f-stop is the same as for the full-frame sensor
without TC, the cropped sensor has less photons collected, and this
has to be compensated by the camera electronics, which will result
in higher noise. This is what I wanted to say with the camera
engineers and the crop factor, but it turned out to be misleading.
In order for the f-stop definition to still be valid, camera
electronics has to be adjusted.

I think you mean the ISO definition, not f-stop, the smaller sensor
needs more amplification per pixel to get the same ISO. f-stop is
only defined by the lens, not the camera.

Well, this is all just definition, and it's well worth to remember the assumptions on which any definition is based. It's not a law of nature that f-stop is independent from the sensor. It could have been defined for a particular sensor size or whatever. For the common f-stop definition it is the ISO definition that has to be adjusted, but from law of nature it could as well have been defined the other way around. Canon engineers could have invented a "35mm equivalent f-stop", just as the "35mm equivalent focal length", and just use the same sensor amplification for a given ISO and for all sensor sizes. It's just definition, you can do whatever you want as long as you define it well.

BTW, f-fstop is defined under the assumption that the lens transmits 1OO% of the incident light. In reality, different lenses have different transmittance, so two lenses with the same f-stop may actually have different speed.

BTW2, what is a good definition of the focal length of an objectiv made out of several lenses in several lens groups? If you only have one lens then it's easy: for incoming parallel light, the focal length is the distance from the center of the lens to the focal plane. But for a complex system made of several lens groups? Searching the internet you find several definitions: (a) the distance from the rear element to the focal plane, (b) the distance from the front element to the focal plane, and (c) the distance from the optical center of the lens to the focal plane. Take definitions (a) or (b). Say, you have a 500mm lens with a given length. Coupling it to a 2xTC, the lens' length merely increases by some cm, but in no way doubles its length. So, do we have a real 1000mm focal length, or should we better say "effective focal length" based on FOV? How is it possible to build a 15mm lens for a 35mm SLR if due to the mirror size the distance from the lens to the focal plane is far more? Definition (c) seems the only one that makes sense to me, but then, how do you define the optical center of the lens? Let me take a wild guess: you are comparing to a single lens. The only possibility to do that is compare the FOV for a given sensor size. So, in the end the focal length definition is based on the FOV for a given sensor size. I might be wrong with this, but then please give me a better definition.

In comparison with the TC/full-frame combo, the cropped sensor
collects the same amount of light per pixel! No difference in noise
here, because for the full-frame sensor you have to double the ISO
to have the same exposure time.

I theory most of what you said is true, but in reality the pixel
sizes are not that different:

The pixels of the 10D are 7.4um
The pixels of the 1Ds are 8.8um

In the horizontal direction, the 1Ds has 4082 pixels on 35.8mm. This is 4082/35.8=8.77um/Px. On the 10D, there are 3072 pixels on 22.7mm. 3072/22.7=7.39um/Px. These numbers are not taking into account any different sensor design like different space between the pixels. So you are confirming my first-order approach. If both sensors had the same pixel number, then the 10D pixel size would be a factor 1.58 smaller as the 1Ds pixels size per side, and a factor 2.49 smaller in area. Of course you get different images with a TC on the 1Ds and without TC on the 10D, because they have different numbers of pixels.

Also as sensors have increased in density noise reduction
technology has also improved, offsetting the increase in noise from
smaller pixel areas. Just look at the D30, D60, 10D progression,
same size chip and the noise keeps going down as technology
improves.

All this is true, but I am comparing the behaivour of a full-frame xyz-MP sensor with TC with a cropped xyz-MP sensor without TC, being xyz the same amount of pixels for both sensors. I am not talking about a 5 yrs old full frame and a recent technology cropped sensor or vice versa. When comparing two things, you should take other parameters constant otherwise you'll mix things up. Obviously the actual noise level for a given ISO and photons per pixel is not a nature's law, but depends on the technology in use.

There are also other factors that influence how large the
photosites are on a chip, the area of the sensor is not all used
for collecting photons, there are spaces between the pixels for
various parts that are needed to access the pixels. Different
designs of CCDs and CMOS make these parts larger or smaller
relative to the pixels.

This is true, but only a minor effect and we do not in general know if for a smaller sensor the space lost for the photons is relatively larger or smaller. Also, most sensors use a lenslet array in front of them in which one dedicated lens per pixel focuses the light into its pixel's center. In this case the actual design of the sensor does not matter, but that of the lenslet array, and these simply scale in size.

Your argument of the different amount of enlargement is not valid,
because in one case you enlarge from a smaller sensor to the large
print, and in the other you already enlarge the lens output in part
in the TC, and in another part from the sensor to the print. The
amount of enlargement needed from the original lens to the print is
the same in both cases.

If one sensor is larger than the other how can it be the same
amount of enlargement? I thought we were comparing a TC that
matches the crop factor, so the image capture by both sensors will
be the same field of view. In that case the smaller sensor has to
be inlarged more if you make the same size print from both.

Take a seat and try to understand my initial argument. In both cases you start with the same lens without TC. This lens produces an image on the image plane with a given size.

In one case you first magnify this image in the image plane optically with the TC, and in a second step you magnify it to your print.

In the other case you do not magnify in the image plane, but purely magnify to the print.

The magnification factor from the original lens image to the final print image is the same in both cases. The characteristics of the original lens are magnified by the same amount. It's just two different ways to achieve exactly the same thing.

Should be easy to calculate how different the 2 setups would be as
far as DOF goes, use any of the DOF calculators and plug in the
numbers. You would have to use the new f-stop and focal length
because of the TC on the full frame camera, and the actual data
from the cropped camera. I really don't feel like trying it but I
would guess the enlargement is a minor factor. If you adjust the
f-stop on the cropped camera to match the TC on the FF camera you
would see only the change due to enlargement.

Well, I for my part will go back to real life and let this exercise for others.

Best regards
Achim

--
Later,
Marty

http://science.widener.edu/~schultz/digipicts.html

 

[Marty Schultz]

Focal ratio was defined hundreds of years ago, f-stop began when the iris was inserted into the lens at the beginings of photography. Focal ratio is based on the focal length of the primary objective or mirror. For multi-element lenses it is not as straight forward, and then you add the iris to stop down farther. You seem to be thinking of how much light gets through the lens, f-stop does not indicate how well the lens(es) are coated, or if they are uncoated, it does not give you an absolute amout of light, it only tells you the ratio of focal length to aperture. There is no standard listing for how efficient the coatings are, Fujinon uses some amazing coatings on their binoculars and claim more than 95% transmission for the entire light path, but it's up to the buyer to believe them.

If you are trying to say lenses made by different companies with the same focal length and f-stop are not going to give the exact same light output and field of view, then that would be true since the basic definitions don't account for coatings and zoom lenses with moving element groups. But that has nothing to do with crop factors or camera design.

--
Later,
Marty

http://science.widener.edu/~schultz/digipicts.html

 

[AchimS]

I completely agree. Many things we included into our discussion have
nothing to do with our starting point, but were interesting anyway.

As far as the starting point concerns -- in his post, tdhg566 claims: "The risk of being imprecise about this is to leave people with the mistaken impression that they can obtain more zoom/magnification/enlargement, call it whatever you want, if they get a camera with a higher crop factor." And he goes on with an example of a guy who supposedly was so terrible wrong wanting to buy a 10D so he could get rid of his TC.

My claim was that the impression is not so mistaken. If both sensors have the same pixel number, then you could use a smaller sensor with the same results as if you were using a full frame sensor with TC. I guess I finally convinced you, or are you just thinking that these crazy arguments should be better left uncommented?

If you are specialized in tele-photos, then there is nothing wrong with smaller sensors with smaller pixels. The downside is wide-angle photography, but for tele it is really the same as using a TC. In his example, the guy wanted to change a 1Ds with a 10D, in order to get rid of his TC. In this case, he would be better off just cropping his images in post-processing.

Just one final remark. Even definitions that were defined hundreds of years ago are just that - definitions. We as a society or even you as individuum are completely free to make our own definitions and don't care about the way other people explained the world. Sometimes for real progress you have to throw old views (definitions) over board and start thinking from scratch. I am not saying we should re-define f-stops, its just a general remark on your history facts, which have nothing to do with the usability of these definitions at present day and with the question if we should not just agree upon a new set of definitions. Anyhow, this is not a philosophy forum.

Focal ratio was defined hundreds of years ago, f-stop began when
the iris was inserted into the lens at the beginings of
photography. Focal ratio is based on the focal length of the
primary objective or mirror. For multi-element lenses it is not as
straight forward, and then you add the iris to stop down farther.
You seem to be thinking of how much light gets through the lens,
f-stop does not indicate how well the lens(es) are coated, or if
they are uncoated, it does not give you an absolute amout of light,
it only tells you the ratio of focal length to aperture. There is
no standard listing for how efficient the coatings are, Fujinon
uses some amazing coatings on their binoculars and claim more than
95% transmission for the entire light path, but it's up to the
buyer to believe them.

If you are trying to say lenses made by different companies with
the same focal length and f-stop are not going to give the exact
same light output and field of view, then that would be true since
the basic definitions don't account for coatings and zoom lenses
with moving element groups. But that has nothing to do with crop
factors or camera design.

Oh, but one of your arguments was that a 2xTC doubles the focal length of a lens, while a smaller sensor does nothing to the focal length. So we need a valid definition of focal length in order to know what we are talking about. Remember that the original post was about a zoom lens.

--
Later,
Marty

http://science.widener.edu/~schultz/digipicts.html

 

[Marty Schultz]

My claim was that the impression is not so mistaken. If both
sensors have the same pixel number, then you could use a smaller
sensor with the same results as if you were using a full frame
sensor with TC. I guess I finally convinced you, or are you just
thinking that these crazy arguments should be better left
uncommented?

I don't think I ever disagreed with the concept, a 1.6x TC on a full frame camera will give the same field of view as a 1.6x smaller sensor with the same lens and no TC. The quality of the TC and the 2 cameras would determine if you could get the same image quality out of both, and that's something that requires real world testing.

As for definitions, in science and technology it kinda helps if they don't change with the whims of society. Hard to build stuff if the definition of a meter were to change every few years. It's bad enough we have metric and english units that have caused problems due to people not converting between them correctly.

--
Later,
Marty

http://science.widener.edu/~schultz/digipicts.html

 

[Jason Hutchinson]

My claim was that the impression is not so mistaken. If both
sensors have the same pixel number, then you could use a smaller
sensor with the same results as if you were using a full frame
sensor with TC. I guess I finally convinced you, or are you just
thinking that these crazy arguments should be better left
uncommented?

I think the same pixel count stipulation isn't all that great because:

1) this means you're associating pixel density with magnification, which is ok, but is not the same as "crop factor"

2) I believe the assumption that things will go along with small sensors having the same pixel count as big sensors is wrong. It's not true today, and I don't believe it will be true tomorrow.

Jason

 

[AchimS]

I don't think I ever disagreed with the concept, a 1.6x TC on a
full frame camera will give the same field of view as a 1.6x
smaller sensor with the same lens and no TC. The quality of the TC
and the 2 cameras would determine if you could get the same image
quality out of both, and that's something that requires real world
testing.

Couldn't agree more.

As for definitions, in science and technology it kinda helps if
they don't change with the whims of society. Hard to build stuff
if the definition of a meter were to change every few years.
It's bad enough we have metric and english units that have caused
problems due to people not converting between them correctly.

Again, couldn't agree more with the concept. Note that the definition of the meter did change every now and then, as technology progressed. See http://www.mel.nist.gov/div821/museum/timeline.htm

• one ten-millionth part of the quadrant of the earth
• the distance between the two graduation lines at 0 °C of the International Prototype Meter
• 1,650,763.73 vacuum wavelengths of light resulting from unperturbed atomic energy level transition 2p10 5d5 of the krypton isotope having an atomic weight of 86
• the length of the path traveled by light in vacuum during a time interval of 1/299,792,458 of a second

In all of these changes in definition, the goal was not only to improve the precision of the definition, but also to change its actual length as little as possible.

Greetings
Achim

 

[AchimS]

My claim was that the impression is not so mistaken. If both
sensors have the same pixel number, then you could use a smaller
sensor with the same results as if you were using a full frame
sensor with TC. I guess I finally convinced you, or are you just
thinking that these crazy arguments should be better left
uncommented?

I think the same pixel count stipulation isn't all that great because:

1) this means you're associating pixel density with magnification,
which is ok, but is not the same as "crop factor"

First principle in science: if you want to see the effect of one parameter, change this one parameter and try to keep other conditions as constant as possible. In order to compare the effect of a TC and the effect of a crop factor on a final image, then the final image should have the same pixel count.

Do you want to compare (a) a 1Ds full-frame image taken with a TC with (b) a cropped image (in post-processing, from the original full-frame image) without TC, both (a) and (b) showing the same FOV? With a decent TC, obviously image (a) will have better resolution.

2) I believe the assumption that things will go along with small
sensors having the same pixel count as big sensors is wrong. It's
not true today, and I don't believe it will be true tomorrow.

I think I showed in my previous posts that for any associated effect we came up with, the sensor size did not matter, but only the pixel count. This is true today and will be true tomorrow. You will say, hey, but smaller sensors show higher noise. The answer is, higher noise at the same ISO level, but the same noise at the same "effective" ISO level. Its all in the definition of f-stop. A f/2.8 lens for a small sensor collects less light than an equivalent (in FOV) f/2.8 lens for a larger sensor. Because of the definitions we have, the aperture values cannot be compared 1:1 between different sensor sizes. Use lenses with the same effective speed, and you get the same noise independently from the sensor size. There's nothing wrong with the four/thirds system, for example. Use adopted lenses, and you get the same images as with 35mm sensors.

In the classic world, large format cameras have better resolution as 35mm cameras, because in both you use the same emulsion. But nobody wants to use large format cameras for action shots with big teles. In the digital world, it may be cheaper to pack a given amount of pixels into a smaller area. Then you can do this without sacrifying anything.

Achim

 

[Jason Hutchinson]

First principle in science: if you want to see the effect of one
parameter, change this one parameter and try to keep other
conditions as constant as possible. In order to compare the effect
of a TC and the effect of a crop factor on a final image, then the
final image should have the same pixel count.

Of course you've just broken your own rule, and changed the pixel size.

Do you want to compare (a) a 1Ds full-frame image taken with a TC
with (b) a cropped image (in post-processing, from the original
full-frame image) without TC, both (a) and (b) showing the same
FOV? With a decent TC, obviously image (a) will have better
resolution.

Given that the first assumption of all above this isn't true currently (and I believe never will) I'll take B. In fact I've done this already with the 1DII:

http://66.180.118.166/lenstests/crop_factors_compared2/

I've seen this done with the 14c too. It performs very well. Even the back of the pack 1Ds does well enough to just say what the heck and go with B.

2) I believe the assumption that things will go along with small
sensors having the same pixel count as big sensors is wrong. It's
not true today, and I don't believe it will be true tomorrow.

I think I showed in my previous posts that for any associated
effect we came up with, the sensor size did not matter, but only
the pixel count.

No, it would not only be pixel count. It'd be pixel density. Density is the only way to take both the size of the sensor and pixel count together. In fact it is the more direct approach to the solution.

This is true today and will be true tomorrow. You
will say, hey, but smaller sensors show higher noise. The answer
is, higher noise at the same ISO level, but the same noise at the
same "effective" ISO level. Its all in the definition of f-stop. A
f/2.8 lens for a small sensor collects less light than an
equivalent (in FOV) f/2.8 lens for a larger sensor. Because of the
definitions we have, the aperture values cannot be compared 1:1
between different sensor sizes. Use lenses with the same effective
speed, and you get the same noise independently from the sensor
size. There's nothing wrong with the four/thirds system, for
example. Use adopted lenses, and you get the same images as with
35mm sensors.

Of course there are no adopted zooms. They're all slowed down considerably. That 2/3rds stop lossed really helps their "smaller lens" compaign when really they have no inherent advantage on many lenses.

In the classic world, large format cameras have better resolution
as 35mm cameras, because in both you use the same emulsion. But
nobody wants to use large format cameras for action shots with big
teles. In the digital world, it may be cheaper to pack a given
amount of pixels into a smaller area. Then you can do this without
sacrifying anything.

You can't do this without sacrificing anything. If that were the case the E1 would have adopted a 2/3" sensor to make life easy on themselves.

Jason

Achim

 

[AchimS]

First principle in science: if you want to see the effect of one
parameter, change this one parameter and try to keep other
conditions as constant as possible. In order to compare the effect
of a TC and the effect of a crop factor on a final image, then the
final image should have the same pixel count.

Of course you've just broken your own rule, and changed the pixel
size.

When I downscale the sensor size, I obviously have to downscale the pixel size in order to have the pixel number constant. My main assumption was that I compare two images, obtained with different concepts, but with the same pixel number. By this I avoid influence from a varying pixel number, which was not my interest, and can test the effects of changing sensor size.

Do you want to compare (a) a 1Ds full-frame image taken with a TC
with (b) a cropped image (in post-processing, from the original
full-frame image) without TC, both (a) and (b) showing the same
FOV? With a decent TC, obviously image (a) will have better
resolution.

Given that the first assumption of all above this isn't true
currently (and I believe never will) I'll take B. In fact I've
done this already with the 1DII:

I don't understand this sentence. My point was to clarify that comparing case A with case B is a different thing to the comparison I was interested in.

http://66.180.118.166/lenstests/crop_factors_compared2/

I've seen this done with the 14c too. It performs very well. Even
the back of the pack 1Ds does well enough to just say what the heck
and go with B.

If you want to say that just cropping the image instead of using a TC is sufficient for most images and print sizes, then I will agree with you.

2) I believe the assumption that things will go along with small
sensors having the same pixel count as big sensors is wrong. It's
not true today, and I don't believe it will be true tomorrow.

I think I showed in my previous posts that for any associated
effect we came up with, the sensor size did not matter, but only
the pixel count.

No, it would not only be pixel count. It'd be pixel density.
Density is the only way to take both the size of the sensor and
pixel count together. In fact it is the more direct approach to
the solution.

Pixel density and pixel count seem to me very related things when comparing different sensor sizes, so what's the point? In the output image its the pixel count that matters, not the pixel size.

This is true today and will be true tomorrow. You
will say, hey, but smaller sensors show higher noise. The answer
is, higher noise at the same ISO level, but the same noise at the
same "effective" ISO level. Its all in the definition of f-stop. A
f/2.8 lens for a small sensor collects less light than an
equivalent (in FOV) f/2.8 lens for a larger sensor. Because of the
definitions we have, the aperture values cannot be compared 1:1
between different sensor sizes. Use lenses with the same effective
speed, and you get the same noise independently from the sensor
size. There's nothing wrong with the four/thirds system, for
example. Use adopted lenses, and you get the same images as with
35mm sensors.

Of course there are no adopted zooms. They're all slowed down
considerably. That 2/3rds stop lossed really helps their "smaller
lens" compaign when really they have no inherent advantage on many
lenses.

Of course they can use it as a marketing campaign: look, we have f/2.8 lenses, when in reality these cannot be compared to 35mm f/2.8 lenses with the same FOV. But, comparing effectively the same lenses, say a 4/3 lens with f/1.8 and a 35mm lens with f/2.8 (I have not calculated well to give correct numbers), then still the 4/3 lens will be smaller and lighter, because it only needs to produce a smaller circle of light. Of course the lens has to be fabricated with better optical resolution, so it will cost about the same.

In the classic world, large format cameras have better resolution
as 35mm cameras, because in both you use the same emulsion. But
nobody wants to use large format cameras for action shots with big
teles. In the digital world, it may be cheaper to pack a given
amount of pixels into a smaller area. Then you can do this without
sacrifying anything.

You can't do this without sacrificing anything. If that were the
case the E1 would have adopted a 2/3" sensor to make life easy on
themselves.

4/3 sensor has diagonal size of 22.3mm, a 2/3" sensor of 11mm.

Perhaps there are present limits in pixel density and in the optical quality that can be achieved in the lenses?

Jason

Achim