NIkon rumors

Discussion in 'Nikon' started by Gordo, Jul 30, 2007.

  1. Gordo

    ASAAR Guest

    Unlikely, I agree. But if Nikon can figure out how to combine
    sensor IS with lens IS to produce better stabilization than either
    alone would allow, their DSLR bodies wouldn't necessarily slash
    sales of their VR lenses. If this is successful, they could use it
    to solve another problem that has been mentioned, how to implement
    sensor based VR so that a FF sensor could be used instead of a 1.1
    crop/focal length ratio sensor. Just minimize or eliminate
    horizontal correction, so the sensor is primarily moved up and down
    (parallel to the edge of its sensor's shorter dimension) if a VR
    lens is detected. It could have a menu option that disables this
    feature if the user doesn't mind greater vignetting with some
    lenses. My royalties will be modest. :)
    ASAAR, Aug 4, 2007
    1. Advertisements

  2. I think that having both IS systems working at the same time would be a
    nightmare - it would have to be one or the other. Getting them working
    together would be a real achievement, considering the different dynamics.

    I see what you mean about the vertical-only correction and FF sensor.
    That /could/ help with panning shots, perhaps?

    David J Taylor, Aug 4, 2007
    1. Advertisements

  3. Gordo

    ASAAR Guest

    With all types of shots. I'm not sure though. Even if you're
    panning horizontally, your handheld camera will still subject to
    movements in all directions. If you were trying to capture fine
    print on the sides of panned racing cars, you'd want to minimize
    movements in all directions. I haven't thought about it before
    though. Is the reason that it would help with panned shots because
    most people couldn't pan accurately enough, and the side to side
    movement of even a reasonably well panned shot would produce too
    much side to side movement for the IS/VR to be able to deal with?

    There no reason why it would be especially difficult to get two IS
    systems working together. All that really would be needed would be
    a couple of additional contacts on new bodies and lenses, to allow
    communication between the CPUs handling the IS in the body and in
    the lens. For them to work without knowing what each other is doing
    would indeed be a nightmare. Additional contacts might not even be
    needed if currently used contacts have bandwidth to spare, and
    additional IS data signals could be multiplexed.
    ASAAR, Aug 4, 2007
  4. I was simply thinking of the mode in some IS systems where the horizontal
    is switched off to allow panning. When taking such shots (with the IS
    enabled in my Panasonic FZ5) I have been able to get reasonably good
    results, with timing being perhaps the cause of the majority of errors.
    Of course, this was outdoors where the exposures were short in any case.
    I have yet to try this with a DSLR, although results from a recent airshow
    (with less predictable movement) have been encouraging.
    I'm not convinced. Two measurements of vibration, two open-loop servo
    systems? Perhaps one /might/ be a fine tweak on the other, though, to
    correct predicted residual errors. I wouldn't like to be the
    servo-systems designer for that, though!

    David J Taylor, Aug 4, 2007
  5. Gordo

    ASAAR Guest

    Too late for that, it's already been done, and the system designer
    has already been thanked for his contribution to the team effort.
    IIRC, he was invited to the White House where he was told "Brownie,
    ya did a heckuva job." :) I still don't think that there would be
    an insurmountable design problem.

    Instead of having two completely independent systems applying
    corrections and notifying the each other about what they've done,
    try a design like this. Each of the two systems (in-body IS and
    lens IS) communicates its data to a third "master" CPU in the camera
    body, and this central master CPU returns signals to each of the two
    IS controllers, telling them what corrective action they each need
    to take. That should eliminate conflicting corrections, allow
    greater camera deviations to be handled than either of the
    individual IS mechanisms could manage, and simplify the job enough
    to remove it from 'nightmare' territory. These would be, after all,
    two cooperating digital systems, not two independent analog systems
    that are trying to be managed by a digital kluge.
    ASAAR, Aug 4, 2007
  6. Sorry, you've lost me.
    You could, but it goes rather against the KISS principle!

    You raise another question, though. Are the existing IS systems analog or

    David J Taylor, Aug 4, 2007
  7. Gordo

    ASAAR Guest

    Local joke. Bush praising Brownie (head of FEMA) despite
    miserable all around performance just after hurricane Katrina
    devastated New Orleans and many other coastal areas.

    From where I'm standing, that appears to be the application of the
    KISS principal. But (reading ahead), that's based on my assumption
    of both systems being digital. Coordinating two independent analog
    systems would be much more difficult, I think.

    From where I'm standing I can't tell. It's too hard to peer into
    either of those black box systems. Oh, wait, I think I see what's
    going on. Canon appears to be using a dumb analog IS system and
    Nikon has implemented a smarter digital VR system. :)
    ASAAR, Aug 4, 2007
  8. Ah, naturally! <G>

    David J Taylor, Aug 4, 2007
  9. Gordo

    Alan Browne Guest

    Sorry, I didn't read it that way...
    Alan Browne, Aug 4, 2007
  10. Gordo

    Alan Browne Guest

    If the sensors in camera and in lens are modeled correctly and then
    controlled correctly full integration is possible. However, I believe
    most of the benefit originates with the in lens VR.

    BUT: if the sensor were controlled only in rotation about the lens axis
    (z) (as Oly will be doing [as well as translation in x,y]), then the
    lens corrections for pitch and yaw; and the sensor correction for roll
    would bring the whole to an optimum.

    Alan Browne, Aug 4, 2007
  11. Gordo

    Alan Browne Guest

    Not at all. Consider a "6 axis flight simulator". 6 axis? What the hell?

    3 of the "axis" are rotations in pitch, roll and yaw.
    3 of the "axis" are translation in x,y and z.

    All of these have to be measured and actuated separately.

    For the case of a lens VR (pitch, yaw) and sensor movement: (translation
    in x,y and roll about the z axis, you would end up with 5 degrees of
    freedom (3 rotation and 2 translation).

    Note that the missing translation axis is potentially focus, but the
    degree of movement would too fine to correct with the relatively coarse
    AF in all SLR's.

    Measuring, computing and actuating for such cases is the stuff of motion
    control engineers. I know a couple and their ability to control
    machinery in turbulence and dynamic to fine degrees of precision (given
    the adequate measurements) in real time (250 Hz or so) is impressive.

    Inertial measurement systems in aircraft often integrate in the 1000 to
    2000 Hz range (possibly higher now, it's been a few years...)

    Alan Browne, Aug 4, 2007
  12. Gordo

    Alan Browne Guest

    Not at all. See my other post (today). Explained differently (and I
    believe more simply).
    Alan Browne, Aug 4, 2007
  13. But in your example, six separate and distinct quantities are being
    measured. In the case of two IS systems the three angular quantities
    would be measured twice.

    What makes you think that in a sensor-shift A-S system it is linear motion
    which is measured? Surely it is the pitch and yaw which is measured?
    Yes, the result is applied as a linear shift of the sensor, but so it is
    in an IS/VR system where a lens element (or group) is shifted orthogonally
    to the optical axis.

    How much computing power is required for the degree of control you
    mention? I'm still wondering if the current systems are analog or

    I do like your suggestion of IS/VR in the lens, but image rotation
    correction in the body. Nice one!

    David J Taylor, Aug 4, 2007
  14. Gordo

    Alan Browne Guest

    The lens IS measures pitch and yaw. The body A-S measures translation
    (x,y (and in the case of Oly soon to come) rotation around z (roll)).

    And don't call it "two IS systems". It would best be one, with one
    algorithm using all the measurements and making all of the actuations.
    The measurement positions may be different.

    I admit that I'm not considering that the lens IS/VR data might not be
    communicated to the body. I'm really talking about "what _could_ be
    done" rather than what _can_ be done with today's IS/VR systems.
    I believe in the Maxxum cameras it is up/down - left/right measurement,
    but I may be wrong. (and see below)
    See prior discussions. Yes, you can measure rotation rates and based on
    some non-0 length arm, translate that into translations. But with
    varying lens types, lengths and CofG it gets more approximate due to
    "known unknowns".
    I assume digital as it's usually the easiest way to engineer things.
    This does not imply only CPU's with s/w, but could be implemented in FPGA's.
    I occasionally stumble on something smart... and I don't doubt that
    others have imagined this too. (But do mark the date, just in case!)

    Alan Browne, Aug 4, 2007
  15. Alan,

    I'm sure you are wrong about sensor-shift A-S being based on linear rather
    than angular acceleration measurements. Consider an object at infinity -
    linear motion of the camera doesn't matter, but angular motion certainly

    Yes, I agree that if you have both, it needs to be considered as a single
    system, but I do think it would be a rather more complex system.

    David J Taylor, Aug 4, 2007
  16. Gordo

    Paul Furman Guest

    If VR does work better in long lenses, it might make sense. I thought I
    heard in-body VR wouldn't be that useful at longer focal lengths. Also,
    I believe in-lens VR is more powerful. I think VR would be great for
    wide angle or normal shooting but it's unlikely they'll bother making
    those with VR in the lens.

    There is the 24-120 but that's not their best lens. The 100 VR macro
    might be a stupid one too, so that's just two lenses that might become
    irrelevant. I suppose people would be more tempted to get a Sigma 70-200
    which lacks VR. Nikon could get an edge on Canon with this feature,
    especially considering full compatibility with old lenses. I would like
    to see in-body VR even if a few lenses became somewhat irrelevant, I
    don't think it's a total deal breaker, just a few awkward points.
    Paul Furman, Aug 4, 2007
  17. Gordo

    Alan Browne Guest

    W/o the facts in my hands I won't argue it. I just don't know if they
    use angular or linear accelerometers.
    For motion control it reduces to sets of equations considering the
    sensor capabilities, lags, errors, loop time, biases, etc. Adding more
    is more complex (by definition) but it is not overwhelming by any
    stretch ... just more development time & dollars.
    Alan Browne, Aug 4, 2007
  18. Gordo

    ASAAR Guest

    I don't know if it's true, but I've read that while IS can help a
    great deal in reducing movement, it also slightly reduces overall
    sharpness. If so, then a greater correction due to the IS lens
    element might produce a greater the loss of detail. Consider that
    even if the only camera motion detectors are located in the lens,
    the amount of necessary IS lens excursion could be reduced if the
    sensor was shifted enough to compensate for the reduced IS lens
    movements. This *might* reduce loss of resolution.

    But the body would still have to have its own motion detection
    system, if only because not all lenses used would have IS sensors.

    Another thing that a dual IS system could do, although I don't
    know if it has the potential to produce a significant improvement,
    would be to have each system deal with a reduced range of movements.
    The lens IS system could correct higher frequency jitter, and the
    body IS could attempt to correct only the greater amplitude, lower
    frequency camera movements. Which system does what could depend on
    whether the sensor or the IS lens has the greater mass, and this may
    well depend on the size of the lens.
    ASAAR, Aug 4, 2007
  19. Gordo

    Alan Browne Guest

    IS lenses resolve (MTF tests) _slightly_ less sharp than their non-IS
    equivalent. Emphasis on slightly. So a tripod mounted IS lens (with IS
    off) and a non IS equivalent will produce almost exactly the same
    resolution image as the very slight difference would not be noticeable
    in most images.

    Did I mention that the resolution difference is very slight? Well, it is.

    The benefit for _any_ hand held shot will go absolutely to the IS lens
    for several stops slower than "rule of thumb" speed.

    Alan Browne, Aug 4, 2007
  20. Gordo

    ASAAR Guest

    Your mention of tripod mounted lenses only supports what I was
    considering. I didn't assume that the IS lens's optics were
    inferior in any way, just that when the IS lens element is moved to
    introduce a correction, it would probably reduce the resolution by a
    slight amount. If the degree was too small to notice, then people
    would have nothing to notice. But they have noticed, so the slight
    difference probably wouldn't be described as vanishingly small.

    Even if it is *very* *very* slight, you don't think that a $4,000
    stabilized lens would want to employ every possible technique
    (within reason) to enhance resolution or acuity? I also didn't say
    that I know that it would definitely produce a better result :
    ASAAR, Aug 4, 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.