35mm Lens Resolution Limiting?

[BigBuzz]

Are the new 11MP and 14MP sensors overkill on resolution? It seems to me that the lenses wll be the limiting factor. I'm not at all sure on this, but here goes. I remember from the past that a very good 35mm kens can resolve about 100 line pairs per mm at the focal plane, which would imply 3600X2400 resolution. This is about 8.6 MP. Or am I mistaken?

 

[Joseph S Wisniewski]

Are the new 11MP and 14MP sensors overkill on resolution? It seems
to me that the lenses wll be the limiting factor. I'm not at all
sure on this, but here goes. I remember from the past that a very
good 35mm kens can resolve about 100 line pairs per mm at the focal
plane, which would imply 3600X2400 resolution. This is about 8.6
MP. Or am I mistaken?

A very good lens (high quality macro, for example) can do 400-600 lpmm at the film plane. A good prime should be doing about 300, and a consumer zoom might be down around 100.

Now, we look at the effect of combining lens and sensor. Overall resolution is roughly reciprocol of the sum or the reciprocols of the resolution of the sensor (or film) and the lens.

Current sensors run around 8um, or 125 lines/mm. This is effectively cut in half by the anti-aliasing (low pass) filter in front of the sensor, so figure around 60 lines/mm for digital. (This doesn't take into consideration sharpening algorithms or other non-linear signall processing).

So the effective resolution is is around:

53 lines/mm with a really good macro
50 lines/mm with a good prime
37 lines/mm with your 100 lines/mm lens.

By comparison, in case anyone is interested, a really epic film like Tech Pan resolves around 160 lines/mm. The film has no antialiasing filter, so it can extract about 120 lines/mm from a really shoar macro, and 61 lines/mm from the consumer zoom.

Ciao!

Joe

 

[Joseph S Wisniewski]

A couple of cool Robert Monaghan articles to help you understand resolution limits for lens and film.

http://people.smu.edu/rmonagha/mf/limits.html
http://people.smu.edu/rmonagha/mf/lenslpm.html

Ciao!

Joe

 

[BigBuzz]

Thanks for the links! These really help with understanding resolution. I guess the 100 lpmm I was remembering is based on measurements from negatives, so the film is the most limiting factor. And there is still room for improvement in digital sensor images.

A couple of cool Robert Monaghan articles to help you understand
resolution limits for lens and film.

http://people.smu.edu/rmonagha/mf/limits.html
http://people.smu.edu/rmonagha/mf/lenslpm.html

Ciao!

Joe

 

[Dick60789]

The analysis on:
http://www.normankoren.com/Tutorials/MTF7.html

indicates that the 1Ds matches the resolution of Provia 100F while the 14n exceeds it by 15%. The photos on Luminous Landscapes certainly support this contention.

This means digital is taking the 35mm format into unchartered territory. How far the format can go until the best lenses are the limiting factor is still not resolved .
--
****
http://www.toskhara.com

 

[Teddy14735]

No problem:
http://www.dpreview.com/forums/read.asp?forum=1014&message=3444636

TeddyBear, Russia
http://digicam.narod.ru editor

Are the new 11MP and 14MP sensors overkill on resolution? It seems
to me that the lenses wll be the limiting factor. I'm not at all
sure on this, but here goes. I remember from the past that a very
good 35mm kens can resolve about 100 line pairs per mm at the focal
plane, which would imply 3600X2400 resolution. This is about 8.6
MP. Or am I mistaken?

 

[Stephen H. Westin]

Are the new 11MP and 14MP sensors overkill on resolution? It seems
to me that the lenses wll be the limiting factor. I'm not at all
sure on this, but here goes. I remember from the past that a very
good 35mm kens can resolve about 100 line pairs per mm at the focal
plane, which would imply 3600X2400 resolution. This is about 8.6
MP. Or am I mistaken?

You are mistaken. It takes two sensels to resolve a line PAIR, so you need
7200 across that 36mm to resolve 100 lp/mm.

But 100/mm is pretty high; see
http://www.photodo.com/nav/prodindex.html
for actual resolution tests.
Their highest-rated lens, the Canon EF 200/1,8L USM,
drops to about 70%-80% contrast at 40 lp/mm, which is the highest
frequency they test. And that's at f:8; wide open it's a bit softer.

Most lenses don't approach this performance.
For example, the Contax Vario-Sonnar T* 35-70/3,4
is down around 30% contrast at the corners at all focal lengths, wide open.

--
-Stephen H. Westin
Any information or opinions in this message are mine: they do not
represent the position of Cornell University or any of its sponsors.

 

[Igor R.]

Are the new 11MP and 14MP sensors overkill on resolution? It seems
to me that the lenses wll be the limiting factor. I'm not at all
sure on this, but here goes. I remember from the past that a very
good 35mm kens can resolve about 100 line pairs per mm at the focal
plane, which would imply 3600X2400 resolution. This is about 8.6
MP. Or am I mistaken?

There are some expensive Carl Zeiss fixfocal lenses that has resolution 160 lp/mm, and in your case it is not 3600*2400, it is 7,200*4,800 because it is not lines, it is line pairs, thus multiply milimeters by 2, thus in your case it is limit of 34 megapixels, and much more in my case.

 

[Igor R.]

Current sensors run around 8um, or 125 lines/mm. This is
effectively cut in half by the anti-aliasing (low pass) filter in
front of the sensor, so figure around 60 lines/mm for digital.
(This doesn't take into consideration sharpening algorithms or
other non-linear signall processing).

Yes, it is 60 lp/mm for X# sensor, but it is not 60 for Bayer, it gotta be divided by about 2 or 3 times. Thus you've got 20-30 lp/mm everithing else - interpollation. don't fool yourself!

 

[Dale Bickley]

Current sensors run around 8um, or 125 lines/mm. This is
effectively cut in half by the anti-aliasing (low pass) filter in
front of the sensor, so figure around 60 lines/mm for digital.
(This doesn't take into consideration sharpening algorithms or
other non-linear signall processing).

Yes, it is 60 lp/mm for X# sensor, but it is not 60 for Bayer, it
gotta be divided by about 2 or 3 times. Thus you've got 20-30 lp/mm
everithing else - interpollation. don't fool yourself!

 

[Joseph S Wisniewski]

Current sensors run around 8um, or 125 lines/mm. This is
effectively cut in half by the anti-aliasing (low pass) filter in
front of the sensor, so figure around 60 lines/mm for digital.
(This doesn't take into consideration sharpening algorithms or
other non-linear signall processing).

Yes, it is 60 lp/mm for X# sensor, but it is not 60 for Bayer, it
gotta be divided by about 2 or 3 times. Thus you've got 20-30 lp/mm
everithing else - interpollation. don't fool yourself!

Actually, it's fairly common in CFA systems to use an antialiasing filter at the spacial frequency of the monochrome array, not the bayer cells. You then process the array into a luminence oomponent, and use some form of pixel adjacency algorithm to keep from having color banding due to under-filtering. So it really is closer to 60, for Bayer or for X3 (by the way, you cannot capitalize the "3" in X3, you get "X#").

Ciao!

Joe

 

[Joseph S Wisniewski]

Kodak spec says no AA filter

Yeah, I got a chuckle the first time I read that. Something about the higher resolution of the sensor means the lens will provide all the low pass filtering that the system needs.

Only problem is that the Kodak sensor might be much bigger than previous sensors, but it's still got 8um spacing, no higher resolution than any of the Kodak DCS 600 or 700 models, or the Nikon D100, or the Canon D1s, or anything else. A good sharp lens will need an AA filter.

That's also the look the early X3 pictures on the Foveon web site (pool table, etc.) had. Obviously no AA filter, jaggies on hair and luminance moire in cloth.

Now, if all you had were consumer zoomsm then yeah, the "don't need an AA filter" bit might fly...

Ciao!

Joe

 

[Capital Man]

Are the new 11MP and 14MP sensors overkill on resolution? It seems
to me that the lenses wll be the limiting factor. I'm not at all
sure on this, but here goes. I remember from the past that a very
good 35mm kens can resolve about 100 line pairs per mm at the focal
plane, which would imply 3600X2400 resolution. This is about 8.6
MP. Or am I mistaken?

Based upon the comparison photos at the luminous-landscape.com website, the 11 MP 1Ds is not overkill at all. It makes very sharp images.

 

[TommyT]

Yes, it is 60 lp/mm for X# sensor, but it is not 60 for Bayer, it
gotta be divided by about 2 or 3 times. Thus you've got 20-30 lp/mm
everithing else - interpollation. don't fool yourself!

Please keep in mind the Bayer color filter and interpolation doesn't increase or reduce the orginal resolution of the image produced. They are used to "extract" the missing color information from the neighbor sensors. If your lp/mm resolution is color blind, then I don't think the image resolution is 1/2 or 1/3 of the sensor resolution in any case.

 

[Teddy14735]

For 11Mp (about 4200x2800 pixels) CCD resolution:
4200/36/1.5(eff.pixel size for Bayer) 2 (pairs) = 39 pairs lpmm

For 14Mp CCD (about 4500x3000 pixels) CCD resolution
4500/36/1.5(eff.pixel size for Bayer) 2 (pairs) = 42 pairs lpmm

It's resolution of a budget zooms
Big CCD size mean lower spacial frequency.

TeddyBear, Russia
http://digicam.narod.ru editor

Are the new 11MP and 14MP sensors overkill on resolution? It seems
to me that the lenses wll be the limiting factor. I'm not at all
sure on this, but here goes. I remember from the past that a very
good 35mm kens can resolve about 100 line pairs per mm at the focal
plane, which would imply 3600X2400 resolution. This is about 8.6
MP. Or am I mistaken?

 

[Igor R.]

For 11Mp (about 4200x2800 pixels) CCD resolution:
4200/36/1.5(eff.pixel size for Bayer) 2 (pairs) = 39 pairs lpmm
For 14Mp CCD (about 4500x3000 pixels) CCD resolution
4500/36/1.5(eff.pixel size for Bayer) 2 (pairs) = 42 pairs lpmm
It's resolution of a budget zooms
Big CCD size mean lower spacial frequency.

Heh, totally agree!
What about comapring to 6*4,5 - it makes me laugh, thats it!

 

[Derrick Siu]

Scaning a 35mm negative or slide using a 4000dpi film scanner such as Nikon 4000ED will has a resolution of 3,762 x 5,646 which is about 21 megapixels. Does it mean that Nikon 4000ED is just a kids toy if you own a Contax or Leica 35mm film camera which use Carl Zeiss lens?

Are the new 11MP and 14MP sensors overkill on resolution? It seems
to me that the lenses wll be the limiting factor. I'm not at all
sure on this, but here goes. I remember from the past that a very
good 35mm kens can resolve about 100 line pairs per mm at the focal
plane, which would imply 3600X2400 resolution. This is about 8.6
MP. Or am I mistaken?

There are some expensive Carl Zeiss fixfocal lenses that has
resolution 160 lp/mm, and in your case it is not 3600*2400, it is
7,200*4,800 because it is not lines, it is line pairs, thus
multiply milimeters by 2, thus in your case it is limit of 34
megapixels, and much more in my case.

 

[Igor R.]

Scaning a 35mm negative or slide using a 4000dpi film scanner such
as Nikon 4000ED will has a resolution of 3,762 x 5,646 which is
about 21 megapixels. Does it mean that Nikon 4000ED is just a kids
toy if you own a Contax or Leica 35mm film camera which use Carl
Zeiss lens?

Well, 21 real megapixels, not really a toy, but if you use low ISO film and tripod with your Leica, then you'll get better results with drum scanner.

 

[Nancy Ann]

Great links. This goes a long way to explain why focusing is so awkward on the D100 with AF lenses.

A couple of cool Robert Monaghan articles to help you understand
resolution limits for lens and film.

http://people.smu.edu/rmonagha/mf/limits.html
http://people.smu.edu/rmonagha/mf/lenslpm.html

Ciao!

Joe

 

[Rinus Borgsteede]

A very good lens (high quality macro, for example) can do 400-600
lpmm at the film plane. A good prime should be doing about 300, and
a consumer zoom might be down around 100.

I do not know what planet you are living on but I have never seen a lens on a consumer camera do even 150 lppm

Now, we look at the effect of combining lens and sensor. Overall
resolution is roughly reciprocol of the sum or the reciprocols of
the resolution of the sensor (or film) and the lens.

Combining is not a real test to go by. The only test is the CCD itsef. The CCD photosites and their relative distance determine resolution only as long as the lens exceeds this requirement. I have to say most lenses cannot cut it with the 6MP 1.5x sensors.

Current sensors run around 8um, or 125 lines/mm.

I suspect you are talking about 1.5x CCD sensors. The newer full size sensors have a lot less problems (in regards to sharpness) coping with existing lenses.
This is

effectively cut in half by the anti-aliasing (low pass) filter in
front of the sensor,

The filter is not needed for the new full size CCD or the Foveon.
so figure around 60 lines/mm for digital.

(This doesn't take into consideration sharpening algorithms or
other non-linear signall processing).

So the effective resolution is is around:

53 lines/mm with a really good macro
50 lines/mm with a good prime
37 lines/mm with your 100 lines/mm lens.

By comparison, in case anyone is interested, a really epic film
like Tech Pan resolves around 160 lines/mm.

A really good ortho micro film Agepe from Agfa can do 300 lppm but it made our Leica lenses look sick.
The film has no

antialiasing filter, so it can extract about 120 lines/mm from a
really shoar macro, and 61 lines/mm from the consumer zoom.

Ciao!

Joe