How would lens optics/engineering drive toward an optimal sensor/filmsize?

Discussion in 'Digital SLR' started by Tom, Apr 4, 2006.

  1. Tom

    Tom Guest

    Hi all,
    I have learned a lot as the practitioners and engineers and such (you
    know who you are :) have debated sensor size, physics, angles, and
    limits.

    The related question I have is, if we started with some goals for the
    glass and worked toward a sensor what kind of system would we end up with?

    Is a 50mm 1.8 35mm lens a great lens optically, and for moderate cost,
    because its in the sweet spot for the 'physics of glass' or is just the
    most engineered / largest selling lens and market size reduced the cost?

    For wide angle for a given field of view, is a 24m easier to get 'good'
    (distortion, flair, etc) vs an 18mm lens? So body and sensor wars
    aside, if we could start over and pick a format driven by optics what
    would it be?

    If, optically, bigger is always better (e.g. 200 inch in some fields is
    nice) then is this just a size/quality/cost/preferences trade-off debate?
     
    Tom, Apr 4, 2006
    #1
    1. Advertisements

  2. Tom

    Paul Furman Guest

    I'll take a stab at it. No guarantees though <g>.

    It is a matter of knowing the trade-offs and which to chose for your need.

    Small sensor format has more depth of field and is easier to get long
    focal lengths and a large sensor format has very limited depth of field,
    and is impractical to make long focal lengths for but has room for more
    pixels and larger pixels with less noise but again it's impractical to
    make large sensors because the drive behind technology these days is
    miniaturization. Small sensor formats perform better in low light
    allowing faster shutter speeds but the noise associated with ISO boost
    seems to counteract this advantage so that the high ISO performance of a
    larger sensor holds the advantage up to a point, when you get to big
    view cameras it's completely impractical to improve the already too
    expensive sensor's ISO rating because it's too big and probably near
    quantum limits anyways.

    We'll see if I screwed up anything (probably).
     
    Paul Furman, Apr 4, 2006
    #2
    1. Advertisements

  3. Tom wrote:
    []
    Thanks for your comments, Tom. It triggers a couple of questions.

    I see it as more a sensor noise and camera function trade-off. Taking two
    of the current popular camera styles, so-called P&S (or small sensor) and
    DSLR:

    small sensor:

    - shorter focal lengths, so mechanical tolerances are tighter
    - larger depth of field
    - diffraction more important
    - no need for long back focus
    - less glass so cheaper material costs
    - no need for interchangeable mount

    DSLR

    - longer focal lengths, easier mechanical tolerances
    - smaller depth of field
    - diffraction less important
    - longer back-focus required to clear mirror
    - more glass, so bigger and more expensive material costs
    - interchangeable mount required

    Now extend those comparisons to larger cameras. Of course, the actual
    sensor plays a major part in the resulting performance, not just the
    optics. The 35mm lenses have been developed intensively over the last 50
    years, so you would expect them to have developed to a peak of
    performance.

    Question 1: You see some 12:1 zoom lenses for small-sensor cameras that
    were unmatched in the 35mm world until the recent Nikon 18 - 200mm VR
    zoom. Why is that? Perhaps because the smaller sensor lenses are
    "designed for digital" in the sense that they don't need a very high
    ultimate resolution - they need an MTF curve with a sharper cut-off than a
    long tail.

    Question 2: Why are lenses for 4/3 sensors (about half the linear size of
    35mm film), not half the linear size and an eighth of the weight?

    It seems to me that the sensor size is more important, and that lens
    engineering adapts to the sensor size requirements.

    David
     
    David J Taylor, Apr 4, 2006
    #3
  4. Tom

    Paul Furman Guest

    Why do the small sensor cameras not offer really wide angle? Is that
    just a matter of marketing appeal or a physical limitation? I know you
    can get screw-on adapters but I assume that's a big hit on image
    quality. Large format is usually fairly wide angle though I'm not aware
    of extreme wide lenses for those either, maybe just a matter of so many
    different lenses developed for 35mm because there has been a big market
    with lots of room for innovation.
     
    Paul Furman, Apr 4, 2006
    #4
  5. You have to be careful not to change too many parameters at the same
    time.
    This is not true. Getting DOF from a larger format requires using a
    higher numerical aperture. But that is not magic. Even diffraction doesn't
    get in the way.

    However, for this to be true, the larger format has to be as sensitive
    (per pixel) as the smaller format, and this require a much higher
    ISO sensitivity.

    Usually, larger formats are used to get more resolution, and more resolution
    comes at the expense of DOF (assuming a constant amount of light).
    It is impractical because it is hard to make large sensors. Miniaturization
    doesn't have much to do with it. Miniaturization does make certain
    things cheaper over time.
    I don't see any reason why small sensor would perform better in low light.
    (well, there are dark current and read-out noise, but I am not sure how they
    scale, and photon shot noise tends to be the dominant source of noise in
    many low light applications).
    The quantum limits are the same for everyone. It is much more likely that
    in many larger formats, read-out electronics is not as optimized as in
    smaller formats.
     
    Philip Homburg, Apr 4, 2006
    #5
  6. You can get 23mm and 24mm equivalent, and the Nikon add-on for the 8400
    gives (IIRC) and 18mm equivalent FOV at high quality. No "big hit on
    image quality".

    I suspect that the real reason is that it ywould not be cost-effective for
    the relatively small number of sales. Who is going to buy a $600 lens for
    a $200 camera?

    David
     
    David J Taylor, Apr 4, 2006
    #6
  7. Tom

    Bill Guest

    A smart photographer...?

    Seriously, my lenses equal or exceed the value of my camera body because
    I know that's where the image is formed.
     
    Bill, Apr 4, 2006
    #7
  8. Whilst this may be true for the SLR enthusiast (and it was true for my own
    SLR outfit), I don't believe it /as/ true for the $200 digital camera
    owner (which is why I set the camera price rather low). Once DSLRs are
    offered at US $200 perhaps it will be true.

    David
     
    David J Taylor, Apr 4, 2006
    #8
  9. Tom

    Bill Guest

    Yes, for P&S, I agree completely.

    Hence the joke...did I forget the smiley?
     
    Bill, Apr 4, 2006
    #9
  10. Probably not - it's still (relatively) early in the day over here!

    David
     
    David J Taylor, Apr 4, 2006
    #10
  11. Tom

    Paul Furman Guest

    Alright, that's not bad (15mm on APS DSLR). That's only on the fancier
    models though, right? Last time I looked the Oly models seemed the only
    reasonable option and they don't go very long.
     
    Paul Furman, Apr 4, 2006
    #11
  12. Tom

    Paul Furman Guest

    This is why smaller formats perform better in low light. Below a certain
    size, quantum mechanics & diffraction start to interfere though. So
    there ought to be a chart that shows where these factors collide & an
    optimal sensor size can be chosen (for a given use).
    Assuming a fixed budget of $3000 I'm thinking it might be possible to
    make a super fast P&S better than 35mm? I guess not though if you look
    at a 4/3 sensor with a honking big expensive lens in that budget I don't
    think that's any faster than APS so that suggests APS is pretty close to
    optimal. Full frame 35mm is faster but you need longer lenses to match &
    that could blow the budget.
     
    Paul Furman, Apr 4, 2006
    #12
  13. Tom

    Ben Brugman Guest

    My guess is that the mirror box is in the way and that the shortest
    distance from the sensor to the back element of the lens is not
    half of that for a 35 mm equivalent.
    That is I assume the most important reason.

    The next reason is that you can not scale down all mechanics, and
    the handling can not be scaled down by a factor of two. (Or you have
    to scale down your hands to operate the lens.)

    (For point and shoots, no mirror box, you can't handle de lens at all,
    so the above constraint aply not or less.)

    ben
     
    Ben Brugman, Apr 4, 2006
    #13
  14. I don't see why this has to be true. Nobody makes 6 Mpixel medium format backs
    in the same process technology as is used for 6 Mpixel P&S sensors.
    Yes, and for $1000 the P&S is likely to win from the APS.

    The original question was about optics. When you spend $3000 on a full frame
    35mm system, you don't have any money left to buy lenses.

    If you remove the sensor from the equation, then $3000 buys you much better
    optics on full frame 35mm than on a tiny P&S format.
     
    Philip Homburg, Apr 5, 2006
    #14
  15. The Nikon 8400 (24mm) was towards the top end of the Nikon range, yes, but
    I would not describe the Kodak V570 as a "fancier model".

    http://www.dpreview.com/reviews/nikoncp8400/

    http://www.dpreview.com/reviews/specs/Kodak/kodak_v570.asp

    I believe there are more cameras now offering 24mm, and doubtless someone
    will give the URL for listing cameras by widest field of view on DP Review
    or wherever!

    David
     
    David J Taylor, Apr 5, 2006
    #15
  16. Tom

    Tom Guest

    The prior posts are mostly about sensors...is that because we are
    sensor-centric or...what?
    Still interested in physics of the "Normal Lens".
    As a D50 owner, a fast 35mm lens will currently cost me 2x - 3x the cost
    of a 1.8 50mm.
    Will we see fast, good, cheap "Normal" lenses for the 1.5X format?
    Same here, just a matter of time for good 18mm glass at 24mm prices...or no?
     
    Tom, Apr 5, 2006
    #16
  17. Yes, I noticed everyone leapt in on the sensor question and rather
    ignored the optics part.

    To some extent the answer to your question is that 50mm on 35mm
    represents the easiest specification for lens design. The angle of view
    - typically 46 degrees diagonal on 35mm film - is about the least
    challenging to perfect. The face that, historically, they tended to be
    the most popular may or may not have resulted in earlier perfection of
    designs, and cheapness through larger production runs, but after several
    decades of sophisticated computer design and of the almost complete
    dominance of zooms in the mass market I would be surprised if this is
    still a major factor.

    Wider lenses tend to suffer from a rapid increase in some design
    problems: vignetting, curvilinear distortion (i.e. barrel or
    pincushion), coma and astigmatism all get worse as FoV increases. The
    solutions include: restricting maximum aperture, using aspherical
    elements and floating elements, and in any event many *more* elements;
    all this results in lenses which are larger than you would expect, and
    much more expensive.

    Longer lenses tend to suffer a different array of problems. Chromatic
    aberration, especially lateral CA, gets worse as relative focal length
    increases. This can be reduced or eliminated by the use of (very
    expensive) fluorite or extra-low dispersion glass, and more complex
    design. Also, temperature effects start to become significant (focus
    point shifts as the lens barrel expands). And of course the size of
    glass elements required to give a decent f-number goes up as f goes up.

    To illustrate this, just look at a few specifications. These are from a
    Canon book as that's what I happen to have, but I have no doubt a
    similar story exists in other makers' lists.

    50mm f/1.8 --- 6 elements in 5 groups
    50mm f/1.4 --- 7 elements in 6 groups

    This shows that even at 50mm you have to use a more complex design to
    get good results at a 0.5 stop wider aperture.

    28mm f/2.8 --- 5 elements in 5 groups
    Similar to the 50/1.8 - but notice we lost 1.5 stops to allow
    simplification

    28mm f/1.8 --- 10 elements in 9 groups
    To get the wide aperture back we have to have a much more complex design

    14mm f/2.8L --- 14 elements in 10 groups (including 1 aspherical
    element)
    About as bad as it gets at the wide end..

    And going the other way:

    100mm f/2.0 --- 8 elements in six groups
    Bit more complex, smaller aperture.

    200mm f/2.8 --- 9 elements in 7 groups
    More complex and a stop down

    400mm f/5.6 --- 7 elements in 6 groups
    Small aperture keeps it fairly simple (though still requires 2 ULD
    elements). But now see:

    400mm f/2.8L --- 17 elements in 13 groups (including 1 fluorite and 2
    ULD elements)
    I was quite surprised, looking this up, to see just how complex the
    design of these lenses are.

    If you look at the price lists you will see a similar (but more extreme)
    story - and again, probably not because of scale of production as the
    50mm fixed focal length has fallen out of favour.

    When I had my first SLR, a 400mm lens typically had 4 or 5 elements -
    no fancy glass - but the performance was very poor, low contrast and low
    resolution, for the reasons I mentioned above. Way below the simple 58mm
    f/2 Helios I had as a standard lens. The 17/13 sophistication (and
    £4000+ price) is required just to get performance back to the same level
    as the £80 Canon 50mm f/1.8 - and still at 1.5 stops less aperture.

    Obviously if you change the film/sensor format then the focal lengths
    change - it is (mostly) the FoV that affects design, apart from
    mechanical things like material strength and size of controls, which as
    someone else pointed out will limit the benefits of a linear scaling
    down. There is no obvious reason why a 35mm f/1.8, or even a 17mm f/1.8,
    should not be available given a format which requires a 40-55 degree FoV
    at that focal length. Indeed, look at cine/video cameras and you can see
    it has been done.

    David
     
    David Littlewood, Apr 5, 2006
    #17
  18. Tom

    Paul Furman Guest

    24 - 85 mm F2.6 - F4.9
    39 - 117 mm F3.9 - F4.4

    Oly C8080:
    http://www.steves-digicams.com/2004_reviews/c8080.html
    28 - 140mm f2.4
     
    Paul Furman, Apr 5, 2006
    #18
  19. Paul,

    You missed the second lens in the V570. From the press release:

    "Using proprietary KODAK RETINA Dual Lens technology, the elegant V570
    camera wraps an ultra-wide angle lens (23 mm) and an optical zoom lens
    (39 - 117 mm) into a small, sleek package less than an inch thin."

    Daivd
     
    David J Taylor, Apr 5, 2006
    #19
  20. Tom

    Paul Furman Guest

    Oh cool, an interchangeable prime built in!
     
    Paul Furman, Apr 5, 2006
    #20
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.