Are IS lenses doomed ?

Discussion in 'Digital SLR' started by VC, Jan 11, 2007.

  1. VC

    VC Guest

    The release of Sony Alpha with the image stabilization in camera ( although
    this is not new) highlighted the fundamental problem with Canon.
    Canon have had IS lenses long ago as it would be very difficult to do
    in-camera stabilization in film cameras. The digital cameras had to support
    older lenses including the ones with IS. If Canon developed a camera with
    in-body stabilization it would hurt Canon sales and reputation.
    So I guess Canon will continue with its nonstabilized bodies and when Sony
    or someone else will achieve the same image sensor quality Canon will find
    itself in a very difficult situation.
    There is a very small advantage in having IS in the lens but it is not
    significant enough to grant double and triple cost of the same quality
    What do you guys think ?
    VC, Jan 11, 2007
    1. Advertisements

  2. VC

    Mark² Guest

    Until someone comes up with a sensor-based IS that is as effective as Canon
    and Nikon IS/VR at all focal lengths, they have nothing to worry about save
    for Sony's less-than-honest marketing tactics.
    Mark², Jan 11, 2007
    1. Advertisements

  3. VC

    Lionel Guest

    How on earth would in-camera IS hurt Canon's reputation? Sales, maybe,
    because in-camera IS would /eventually/ kill off their market for IS
    lenses, but a huge number of photographers (including myself) would
    consider in-camera IS on Canon DSLRs to be a godsend, & would be
    saving their pennies to buy one ASAP.
    Lionel, Jan 11, 2007
  4. VC

    ForrestPhoto Guest

    because in-camera IS would /eventually/ kill off their market for IS
    There might even be a little to be gained by using an IS lens on an IS
    chip. Probably not much, but possibly something.

    Also, I'm thinking IS in a 300+ mm lens is going to do a lot better
    than on a chip.
    ForrestPhoto, Jan 11, 2007
  5. AIUI, the problem with in-camera IS is that the range of movement
    required in the chip is *much* greater than the movement required in an
    optical correction element in the middle of the lens.

    Thus, other things being equal, an in-lens system will always have the
    advantage, and be capable of being smaller, lighter, faster acting, or
    more effective (more f-stops effective benefit) or all of the above. The
    corollary is of course that you will need one in every lens, instead of
    just one in the body, but at least that one in each lens will be
    optimised for that lens, not constrained to some generic compromise

    So it is probably a choice of cheapness versus maximum effectiveness. I
    know which side I come down on, YMMV.

    David Littlewood, Jan 11, 2007
  6. VC

    Bill Guest

    There is no problem, that's just marketing hype.
    Where do you get that idea? Canon has a good reputation as it stands, so
    how would adding another feature to the dozens of current features hurt
    their market share?

    Was Canons rep hurt when they introduced a sensor cleaner in the XTi?

    While we know the sensor cleaners are mostly hype, it doesn't seem to
    hurt image quality or camera performance, so how is it detrimental to
    I think you're an easy target for marketing campaigns.

    Bill, Jan 11, 2007
  7. VC

    Lionel Guest

    Dunno. It'd be awfully hard to prevent the two IS mechanisms from
    confusing each other. If Canon do ever introduce an in-camera IS
    system, I'd bet that they configure the camera firmaware to turn off
    the body IS when an IS lens is mounted.
    On that, I agree with you. In-camera IS would need a huge amount of
    travel to compensate for the amount of shake you get with big tele
    Lionel, Jan 11, 2007
  8. What actually happens with in-body IS is that the travel is the same in
    extent, it's just the velocity (speed of travel) which increases with
    longer focal lengths. Sony (Minolta) use angular momentum sensors, since
    it is an angular shake which counts, so shake is not considered in mms
    it's in degrees (or small fractions of a degree) plus velocity. When a
    lens covers 8 degrees a quarter degree of shake is substantial, when a
    lens covers 80 degrees it's not so much.

    However, you are 'safe' to shoot at 1/30th without IS on an 18mm lens,
    but you have to use 1/500th to be safe on a 300mm lens (both APS-C
    examples). The long lens magnifies the shake in effect, so that the
    image moves as far in 1/500th with the 300mm, as it moves in 1/30th with
    the 18mm.

    The long lens only gets into 'huge amount of travel' if you try to
    hand-hold 1/30th at 300mm. With anti-shake, you can do 1/8th and maybe
    1/4 at 18mm. You can do 1/250 or 1/125 at 300mm. In each case, the
    sensor is travelling about the same amount, but it is having travel
    faster for the 300mm.

    Shake does not just keep going in one direction, anway. It tends to be
    tremor or vibration-like when it is not a brief, fixed jerk caused by
    pressing the shutter. Sony's SSS will cope with tremors between 1Hz
    (swaying gently back and forth once a second - heartbeat, breathing) and
    60Hz (someone just plugged you into a wall socket by mistake). Most
    shake is apparently around 10Hz, a typical frequency of human body
    tremor. So the system, whether in the lens or the body, has to respond
    to acceleration, fixed velocity, vector (direction) including rapid
    changes of all three.

    Both in-lens and in-body IS appears to function equally well over a wide
    range of conditions. It's not possible to state that in-lens IS is
    definitely superior at long focal lengths, on in-body superior with
    extreme wides and hand-held 1/4s. In practice I have found my KM and
    Sony bodies very similar to Canon IS with 100-300mmm/70-300mm lenses
    (the KM 100-300mm is much smaller and lighter than our early Canon IS
    70-300mm, but I don't think this improves the efficiency - if anything
    the large Canon lens is a bit easier to hand-hold steadily).

    What I forget - and I suspect many others forget - is that you really
    should not be able to use 1/30th with either system, if the lens is at
    300mm. I do so regularly, and the result is nearly always perfectly
    sharp. That's 4-5 stops of stabilisation, not the claimed 1-2 for the
    older Canon lens, or 2-3 for the KM/Sony systems. Yet both, with a
    little care, will give a high success rate.

    David Kilpatrick, Jan 11, 2007
  9. VC

    Skip Guest

    The lenses are doomed, the companies that make them are doomed, photography
    as we know it is doomed, we are all doomed.
    Skip, Jan 11, 2007
  10. VC

    Skip Guest

    I keep seeing this bandied about as the premium for IS/VR, but nowhere do I
    see it in actual practice. It is about a $400-500 increase in price over
    the non IS version, if such does exist in the lineup. The only times this
    has occurred is with the old 75-300, a cheap lens with a gimmick, as far as
    I am concerned, the 70-200 f2.8L ($1100 vs $1600) and the current 70-200 f4L
    ($600+ vs. $1100). Not exactly triple the price. When you get into the
    long teles, the price premium becomes such an insignificant part of the
    whole as to drop out of consideration, like with the 600mm f4L at more than
    So far, newer Canon IS lenses maintain their lead over sensor based IS, the
    24-105 f4L and 70-200 f4L IS lenses give a minimum of 4 stops of correction.
    Skip, Jan 11, 2007
  11. No, you haven't worked out your math. Your 1/500 on 300 mm is
    above the 1/fl guide, so IS is not needed. Newest IS on canon
    claims 4 stops, so if 1/300 is the guide, then 4 stops
    is 1/20 second. At 18mm, the guide is 1/20 second, and
    4 stops is 1 second. The IS feedback loops don't work well
    at 1 second.

    Image shake of ~1 arc-minute per 1/20 second, and tracking needs
    to be at least ~10 times that level (you can't keep your shake to
    1 arc-minute pointing accuracy). Thus a range of 10 arc-minutes
    is needed at a rate of 20 arc-minutes/second. For a 300 mm lens
    that works out to a range of 0.9 mm (900 microns), and a rate of
    1800 microns/second, with a required accuracy of ~ 2 microns
    (pretty difficult.

    Now try that with a 600 mm lens, then a 600 mm + 2x TC.
    Rate goes up in proportion to focal length, but so does the
    range, because you can't accurately point to the 1-arc-minute
    limit, you need a larger limit, and that is at best constant
    and in reality worsens as you hold more weight.
    Thus 1200mm would require 3.6 mm range at a rate of 7200 microns/sec
    with an accuracy of ~2 microns. It can be done in a lens because
    you can tune the power of the optical element being moved
    to give image movement within the range and accuracy of
    the device doing the movement.

    Also, in my experience, the 1/fl guide falls apart as the
    lens size goes up, and the longer you hold the lens.
    Try holding an 8-pound 500 mm f/4 lens for a while.

    Thus, for telephoto work, in lens IS is the only reasonable
    engineering solution.

    Here are example of high magnification hand help in lens IS:

    For this image, a friend was shooting with a 300 f/2.8 non IS lens
    and got no good images, despite using faster shutter speeds
    and lower magnification. My image was 1/1600 sec at 500 mm:

    This one even with IS, about half the images I took were
    sharp, the other slightly blurry:

    Roger N. Clark (change username to rnclark), Jan 11, 2007
  12. VC

    -hh Guest

    Canon's 8x25 IS Binoculars retail for $200 after rebate.

    While implementation complexity (& cost) will obviously vary, this does
    suggest at least that a low performance IS system has to cost less than

    In general, I think that the "+$400-ish" rule of thumb is probably
    around the mark (retail), which means that anyone claiming "double",
    "triple", etc are only doing their comparisons based on one or two
    cheap lenses, rather than considering a broader range of IS lenses.

    -hh, Jan 11, 2007
  13. I think I must have missed some vital point here, David. How does an
    angular movement of the camera, which moves the image in linear fashion
    along the sensor, require a change in the angular orientation of the
    sensor. ISTM that it must require a linear movement of the sensor to
    correct it.

    OTOH, it is quite feasible to change the location of an image on the
    sensor by a small angular displacement of the image-forming beam
    somewhere in the lens, and quite credible that this would require much
    smaller movements.

    If I have misunderstood this (and many others have too, including Canon,
    who I see give a similar rationale for favouring lens-based IS), is
    there a suitable web site which explains the position? I freely
    acknowledge a lack of specific expertise here, and accept the situation
    may be more complicated than my simple thoughts here.
    I guess this is the ultimate test - if in-camera works as well as
    in-lens, then it is as good. Would be good to see some rigorous tests

    David Littlewood, Jan 11, 2007
  14. VC

    l v Guest

    I can't find the article but I remember reading about the comparison
    between in-camera IS and in-lens IS. It stated that the in-camera IS
    would prevent a full sized sensor from working with the in-camera IS.
    It stated that the camera would be very large to allow for full sensor
    movement and the light from the lens could fall off the edges of the
    sensor, therefore even a larger then full sensor was needed, thus making
    the camera even larger.

    Am I blowing smoke or does anyone else remember this?
    l v, Jan 11, 2007
  15. VC

    RichA Guest

    The bad of lens-based IS: COST!
    The good. If one lens goes down, it doesn't take the whole camera with
    it, compared to if IS in the camera goes down.
    RichA, Jan 11, 2007
  16. It's quite obvious if you think about it. The lens throws a circular,
    fixed-size image into the camera. The size of this image circle is designed
    to cover the film frame (or sensor, obviously) and not a smidgeon more than
    necessary- or, if the image circle size IS significantly bigger than the
    film frame you will have to carry around a lens that is correspondingly
    larger, heavier and more expensive than would otherwise be the case.

    Now, with in-camera IS the sensor has to move around to "chase" the image as
    it wobbles around the sensor plane as your hands shake. The image circle is
    of course constant and stays in place, so the sensor has a hard limit on how
    far it can move in any direction before parts of it actually pokes out into
    the darkness outside the image circle.

    So, in-body IS is a good idea only as long as the image circle is
    sufficiently larger than the sensor. A full-frame sensor on a full-frame
    lens is a particularly bad candidate for in-body IS, an APS-sized sensor on
    a "digital" APS-sized lens with reduced image circle size ain't too hot a
    proposition either. Full-frame lens and APS-sized sensor gives you the most

    Compare this to an IS lens, which de facto lets the image circle chase the
    wobbling image so that it stays in the same place relative to the sensor.
    Ståle Sannerud, Jan 11, 2007
  17. VC

    John Francis Guest

    It's not at all obvious. It very much depends on focal length, of course.
    But for medium telephoto and onwards (say 200mm focal length and upwards)
    the size of the image circle has very little effect on lens size and weight.
    A 300mm/f4 for a 6x7 camera is just about identical in all dimensions to a
    300mm/f4 for a 35mm body, and costs just about the same. The determining
    factor for size (and weight, and cost) is the fromt element.

    Mind you, a 50mm lens for a 6x7 system *will* be larger/heavier/costlier
    than a 50mm lens for a 35mm camera.
    John Francis, Jan 11, 2007
  18. I didn't mean the sensor changes angle in any way - it correct while
    remaining strictly plane-parallel in its film plane. The calculation as
    to how much linear movement to give is based on research into camera
    shake, which shows that linear movement of the entire camera is rare and
    causes very little visible shake; angular (so-called rotational) shake
    is universally the main cause of unsharpness, but it has to be corrected
    by a linear movement of the sensor corresponding to the image movement.

    This subject has been discussed to death on the Sony forum on dPreview
    as a result of a misinformed ideas by a physics graduate (it's very easy
    to get it wrong if you think camera shake is a displacement of the lens
    axis, rather than a dip/swing/tilt of the lens axis).

    My conclusion, after some very useful posts from others including one
    which emphasised that long lenses don't need greater sensor movemen if
    you stick to the rules about shutter speeds (with the IS factor
    included), was that the programming is pretty complex and relies on
    'typical' shake detection. There is no way the sensors in the Sony can
    really tell whether displacement, rotation, etc is causing them to
    detect motion. Somewhere in the system (same goes for IS in lens) the
    designers have said 'if the signal from the sensors is THIS, the
    required correction will be amost certainy be THIS'. It is not actually
    a direct link.

    David Kilpatrick, Jan 11, 2007
  19. In fact the displacement in IS lenses does cause loss of illumination
    (asymmetrical vignetting) as well as localised aberrations. One of the
    main arguments against using IS lenses is that at the best (IS turned
    off) they may not entirely match the performance of similar non-IS
    designs, and at the worst (moving element or group at maximum
    displacement) optical performance will be visibly compromised.

    One slightly ironic point is that so far all the sensor-stabilised
    cameras have been 1.5X APS factor. In practice, even lenses like the
    11-18mm and 10-20mm wide angles provide 2mm or more coverage beyond this
    at their shortest setting - the Tamron 11-18mm will cover full frame at
    13-14mm. Yet Canon uses 1.6X sensors, which would be even better (more
    room to move round!) with sensor-based IS than the 1.5X format.

    The only lens I know which causes problems with sensor-based IS on the
    KM/Sony 1.5X bodies is the Sigma 30mm f1.4 - it was, I think, really
    designed to work on Canon 1.6X or Sigma 1.7X sensors, and it's already
    pushing its coverage on 1.5X when wide open.

    David Kilpatrick, Jan 11, 2007
  20. VC

    King Sardon Guest

    Camera shake shakes the image, not the image circle, which is fixed by
    the optical axis (with IS is turned off). With IS now on, the
    corrective element moves the image circle around so that the image is
    placed correctly.

    So yes, IS lets the image circle chase the wobbling image so that it
    (the IMAGE) stays in the same place relative to the sensor. The
    relationship between sensor and image circle is the same as for
    in-camera IS. The difference is that with lens-based IS, the image
    circle moves, whereas with in-camera, the sensor moves.

    The image circle demands are the same for the two situations. However,
    not having to have an IS group in the lens allows for a simpler
    optical design (read: cheaper and maybe better performance).

    King Sardon, Jan 11, 2007
    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.