Michael Reichmann reasoning for AA filters?

Discussion in 'Digital Photography' started by W, Oct 19, 2007.

  1. W

    W Guest

    Okay, does this make any sense: http://www.luminous-landscape.com/essays/MKIIIs.shtml

    M. Reichmann says that DSLRs have AA filters to save compute power
    needed in demoscaicing Bayer pattern sensors whereas medium format
    backs do not have AA filters because they do not need to demosaic and
    create jpgs in the camera.

    Sampling theory says that ANY sampled data system requires an AA
    filter to keep higher (spatial) frequencies from aliasing into lower
    (spatial) frequencies. I do not understand Reichmann's assertion. It
    seems to igore the basics of sampled data system theory.

    Any comments?
     
    W, Oct 19, 2007
    #1
    1. Advertising

  2. W

    nospam Guest

    In article <>,
    <> wrote:

    > Okay, does this make any sense:
    > http://www.luminous-landscape.com/essays/MKIIIs.shtml
    >
    > M. Reichmann says that DSLRs have AA filters to save compute power
    > needed in demoscaicing Bayer pattern sensors whereas medium format
    > backs do not have AA filters because they do not need to demosaic and
    > create jpgs in the camera.


    from that page:

    They have an AA filter primarily to save battery power and processing
    time when computing a JPG.

    that is absolutely ludicrous.

    > Sampling theory says that ANY sampled data system requires an AA
    > filter to keep higher (spatial) frequencies from aliasing into lower
    > (spatial) frequencies. I do not understand Reichmann's assertion. It
    > seems to igore the basics of sampled data system theory.


    that's correct.
     
    nospam, Oct 19, 2007
    #2
    1. Advertising

  3. W wrote:
    > Okay, does this make any sense:
    > http://www.luminous-landscape.com/essays/MKIIIs.shtml
    >
    > M. Reichmann says that DSLRs have AA filters to save compute power
    > needed in demoscaicing Bayer pattern sensors whereas medium format
    > backs do not have AA filters because they do not need to demosaic and
    > create jpgs in the camera.
    >
    > Sampling theory says that ANY sampled data system requires an AA
    > filter to keep higher (spatial) frequencies from aliasing into lower
    > (spatial) frequencies. I do not understand Reichmann's assertion. It
    > seems to igore the basics of sampled data system theory.
    >
    > Any comments?


    I agree with you. However, the actual AA filter characteristics differ
    between audio and digital cameras. In audio, brick-wall characteristics
    are the aim, whereas in photography certain cameras (Sigma, in particular)
    have a very weak AA filter, or even omit it, "as the results look
    sharper"! Guess which I won't be buying!

    David
     
    David J Taylor, Oct 19, 2007
    #3
  4. W

    Annika1980 Guest

    On Oct 18, 10:57 pm, W <> wrote:
    > Okay, does this make any sense:http://www.luminous-landscape.com/essays/MKIIIs.shtml
    >
    > Sampling theory says that ANY sampled data system requires an AA
    > filter to keep higher (spatial) frequencies from aliasing into lower
    > (spatial) frequencies. I do not understand Reichmann's assertion. It
    > seems to igore the basics of sampled data system theory.
    >
    > Any comments?


    His site has been updated:

    Postscript
    In a version of this review which was online for a few hours on Oct
    18-19, there was a discussion of antialiasing filters and how I felt
    that there would be advantages to the 1Ds MKIII not having one, for a
    variety of reasons. Due to a mix-up, an early version which was not
    intended for publication because of mistakes in my initial analysis,
    found its way online in error.

    I regret any confusion that this may have caused.
     
    Annika1980, Oct 19, 2007
    #4
  5. W

    W Guest

    Wow, he completely removed the whole discussion no AA filters.

    While I understand the need for AA filters (my earlier post). I must
    admit, I do not understand why the medium format digital backs do not
    include an AA filter. Aside from moire issues, I would think that any
    kind of periodic image elements (e.g. a fence) could cause problems if
    there were no AA filter.


    On Oct 19, 7:00 am, Annika1980 <> wrote:
    > On Oct 18, 10:57 pm, W <> wrote:
    >
    > > Okay, does this make any sense:http://www.luminous-landscape.com/essays/MKIIIs.shtml

    >
    > > Sampling theory says that ANY sampled data system requires an AA
    > > filter to keep higher (spatial) frequencies from aliasing into lower
    > > (spatial) frequencies. I do not understand Reichmann's assertion. It
    > > seems to igore the basics of sampled data system theory.

    >
    > > Any comments?

    >
    > His site has been updated:
    >
    > Postscript
    > In a version of this review which was online for a few hours on Oct
    > 18-19, there was a discussion of antialiasing filters and how I felt
    > that there would be advantages to the 1Ds MKIII not having one, for a
    > variety of reasons. Due to a mix-up, an early version which was not
    > intended for publication because of mistakes in my initial analysis,
    > found its way online in error.
    >
    > I regret any confusion that this may have caused.
     
    W, Oct 19, 2007
    #5
  6. "W" <> wrote:
    > Wow, he completely removed the whole discussion no AA filters.


    Somebody he trusts must have clued him in on how far off he was.

    > While I understand the need for AA filters (my earlier post). I must
    > admit, I do not understand why the medium format digital backs do not
    > include an AA filter.


    Bad design<g>? The last time I spent some time lurking on an MF digital
    list, they were just begining to figure out what a disaster it is. If you
    have any fabrics in the scene, many of your images will be unusable. It's
    simply not acceptable for professional work, or any other context in which
    failure isn't acceptable. Going for extra sharpness by leaving out the AA
    filter is amateur in the extreme.

    > Aside from moire issues, I would think that any
    > kind of periodic image elements (e.g. a fence) could cause problems if
    > there were no AA filter.


    You mean like this? (100% crop from 5D, whose AA filter is a tad on the
    inadequate side. Sigh.)

    http://www.pbase.com/davidjl/image/52118414/original

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Oct 19, 2007
    #6
  7. W wrote:
    > Wow, he completely removed the whole discussion no AA filters.
    >
    > While I understand the need for AA filters (my earlier post). I must
    > admit, I do not understand why the medium format digital backs do not
    > include an AA filter. Aside from moire issues, I would think that any
    > kind of periodic image elements (e.g. a fence) could cause problems if
    > there were no AA filter.


    Maybe the lenses used aren't good enough to need AA filters? Too small an
    MTF at half the sampling frequency?

    David
     
    David J Taylor, Oct 19, 2007
    #7
  8. W

    W Guest

    Wow yes, great example with the 5D. I have even seen moire with my
    20D with fabrics.

    Regarding the MF digital backs, I can't believe that the folks
    designing MF sensors are that "stupid". I still don't get it.

    On Oct 19, 9:55 am, "David J. Littleboy" <> wrote:
    > "W" <> wrote:
    > > Wow, he completely removed the whole discussion no AA filters.

    >
    > Somebody he trusts must have clued him in on how far off he was.
    >
    > > While I understand the need for AA filters (my earlier post). I must
    > > admit, I do not understand why the medium format digital backs do not
    > > include an AA filter.

    >
    > Bad design<g>? The last time I spent some time lurking on an MF digital
    > list, they were just begining to figure out what a disaster it is. If you
    > have any fabrics in the scene, many of your images will be unusable. It's
    > simply not acceptable for professional work, or any other context in which
    > failure isn't acceptable. Going for extra sharpness by leaving out the AA
    > filter is amateur in the extreme.
    >
    > > Aside from moire issues, I would think that any
    > > kind of periodic image elements (e.g. a fence) could cause problems if
    > > there were no AA filter.

    >
    > You mean like this? (100% crop from 5D, whose AA filter is a tad on the
    > inadequate side. Sigh.)
    >
    > http://www.pbase.com/davidjl/image/52118414/original
    >
    > David J. Littleboy
    > Tokyo, Japan
     
    W, Oct 19, 2007
    #8
  9. W

    Paul Furman Guest

    David J Taylor wrote:
    > W wrote:
    >
    >>Wow, he completely removed the whole discussion no AA filters.
    >>
    >>While I understand the need for AA filters (my earlier post). I must
    >>admit, I do not understand why the medium format digital backs do not
    >>include an AA filter. Aside from moire issues, I would think that any
    >>kind of periodic image elements (e.g. a fence) could cause problems if
    >>there were no AA filter.

    >
    > Maybe the lenses used aren't good enough to need AA filters? Too small an
    > MTF at half the sampling frequency?


    My guess is these users want extreme detail and I assume there are
    fairly easy ways to process moire patterns out in post. It's more work
    but that kind of photography is not for fast turnover.

    --
    Paul Furman Photography
    http://edgehill.net
    Bay Natives Nursery
    http://www.baynatives.com
     
    Paul Furman, Oct 19, 2007
    #9

  10. >
    >>> Aside from moire issues, I would think that any
    >>> kind of periodic image elements (e.g. a fence) could cause problems if
    >>> there were no AA filter.

    >> You mean like this? (100% crop from 5D, whose AA filter is a tad on the
    >> inadequate side. Sigh.)
    >>
    >> http://www.pbase.com/davidjl/image/52118414/original
    >>
    >> David J. Littleboy
    >> Tokyo, Japan

    >
    >


    For a photo like that, meaning everything can be in focus,
    there is always God's AA filter: diffraction.

    Doug McDonald
     
    Doug McDonald, Oct 19, 2007
    #10
  11. "Doug McDonald" <mcdonald@SnPoAM_scs.uiuc.edu> wrote:
    >>
    >>>> Aside from moire issues, I would think that any
    >>>> kind of periodic image elements (e.g. a fence) could cause problems if
    >>>> there were no AA filter.
    >>> You mean like this? (100% crop from 5D, whose AA filter is a tad on the
    >>> inadequate side. Sigh.)
    >>>
    >>> http://www.pbase.com/davidjl/image/52118414/original

    >
    > For a photo like that, meaning everything can be in focus,
    > there is always God's AA filter: diffraction.


    That doesn't work very well because diffraction causes more loss in the pass
    band than a good AA filter (the MTF curve for diffraction is a straight line
    from 100% at 0 lp/mm to 0% at 1600/(f number), but what you want for an AA
    filter is a step function). Also, with the 5D, you'd have to stop down to
    f/32 or further to eliminate Moire: it's still there at f/22.

    (Actually, it's getting better: the more recent converters are getting
    better at not accentuating Moire as badly as RSP did.)

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Oct 20, 2007
    #11
  12. W

    acl Guest

    On Oct 20, 3:59 am, "David J. Littleboy" <> wrote:
    > "Doug McDonald" <mcdonald@SnPoAM_scs.uiuc.edu> wrote:
    >
    > >>>> Aside from moire issues, I would think that any
    > >>>> kind of periodic image elements (e.g. a fence) could cause problems if
    > >>>> there were no AA filter.
    > >>> You mean like this? (100% crop from 5D, whose AA filter is a tad on the
    > >>> inadequate side. Sigh.)

    >
    > >>>http://www.pbase.com/davidjl/image/52118414/original

    >
    > > For a photo like that, meaning everything can be in focus,
    > > there is always God's AA filter: diffraction.

    >
    > That doesn't work very well because diffraction causes more loss in the pass
    > band than a good AA filter (the MTF curve for diffraction is a straight line
    > from 100% at 0 lp/mm to 0% at 1600/(f number), but what you want for an AA
    > filter is a step function).


    Actually a step function causes ringing (thus power in higher
    frequencies). But the power in the ringing depends on the cutoff
    frequency, and I must admit I'm too tired at the moment to try to
    guess if it'll be important or not for cameras (if one built a step
    function cutting off near eg half the max spatial frequency).

    Anyway, physically I can't see how to construct a step function filter
    for optical transmission (not that you said such a thing is feasible,
    I'm just wondering).

    But this Luminous Landscape thing is excellent. The quoted passage (in
    the OP) is pure poetry!


    > Also, with the 5D, you'd have to stop down to
    > f/32 or further to eliminate Moire: it's still there at f/22.
    >
    > (Actually, it's getting better: the more recent converters are getting
    > better at not accentuating Moire as badly as RSP did.)
    >
    > David J. Littleboy
    > Tokyo, Japan
     
    acl, Oct 20, 2007
    #12
  13. W

    Rich Guest

    On Oct 19, 9:55 am, "David J. Littleboy" <> wrote:
    > "W" <> wrote:
    > > Wow, he completely removed the whole discussion no AA filters.

    >
    > Somebody he trusts must have clued him in on how far off he was.
    >
    > > While I understand the need for AA filters (my earlier post). I must
    > > admit, I do not understand why the medium format digital backs do not
    > > include an AA filter.

    >
    > Bad design<g>? The last time I spent some time lurking on an MF digital
    > list, they were just begining to figure out what a disaster it is. If you
    > have any fabrics in the scene, many of your images will be unusable. It's
    > simply not acceptable for professional work, or any other context in which
    > failure isn't acceptable. Going for extra sharpness by leaving out the AA
    > filter is amateur in the extreme.
    >
    > > Aside from moire issues, I would think that any
    > > kind of periodic image elements (e.g. a fence) could cause problems if
    > > there were no AA filter.

    >
    > You mean like this? (100% crop from 5D, whose AA filter is a tad on the
    > inadequate side. Sigh.)
    >
    > http://www.pbase.com/davidjl/image/52118414/original
    >
    > David J. Littleboy
    > Tokyo, Japan


    Why can't they do AA in post-processing, with software? At least it
    would allow the 40% increase in resolution to be used when you don't
    encounter repetitive high-frequency details.
     
    Rich, Oct 20, 2007
    #13
  14. W

    Scott W Guest

    Rich wrote:
    > On Oct 19, 9:55 am, "David J. Littleboy" <> wrote:
    >> "W" <> wrote:
    >>> Wow, he completely removed the whole discussion no AA filters.

    >> Somebody he trusts must have clued him in on how far off he was.
    >>
    >>> While I understand the need for AA filters (my earlier post). I must
    >>> admit, I do not understand why the medium format digital backs do not
    >>> include an AA filter.

    >> Bad design<g>? The last time I spent some time lurking on an MF digital
    >> list, they were just begining to figure out what a disaster it is. If you
    >> have any fabrics in the scene, many of your images will be unusable. It's
    >> simply not acceptable for professional work, or any other context in which
    >> failure isn't acceptable. Going for extra sharpness by leaving out the AA
    >> filter is amateur in the extreme.
    >>
    >>> Aside from moire issues, I would think that any
    >>> kind of periodic image elements (e.g. a fence) could cause problems if
    >>> there were no AA filter.

    >> You mean like this? (100% crop from 5D, whose AA filter is a tad on the
    >> inadequate side. Sigh.)
    >>
    >> http://www.pbase.com/davidjl/image/52118414/original
    >>
    >> David J. Littleboy
    >> Tokyo, Japan

    >
    > Why can't they do AA in post-processing, with software? At least it
    > would allow the 40% increase in resolution to be used when you don't
    > encounter repetitive high-frequency details.
    >

    The whole problem with aliasing is that it take very high frequency
    information and folds it back down into very low frequency information,
    where it is not right on top of real low frequency information. How do
    you know what if from aliasing and what if real image data?

    You can have the processing program guess at when it is seeing aliasing,
    but this will often lead to problems. The filter in the camera is of
    course not software, it is an optical element.

    Scott
     
    Scott W, Oct 20, 2007
    #14
  15. "acl" <> wrote:
    > On Oct 20, 3:59 am, "David J. Littleboy" <> wrote:
    >> "Doug McDonald" <mcdonald@SnPoAM_scs.uiuc.edu> wrote:
    >>
    >> >>>> Aside from moire issues, I would think that any
    >> >>>> kind of periodic image elements (e.g. a fence) could cause problems
    >> >>>> if
    >> >>>> there were no AA filter.
    >> >>> You mean like this? (100% crop from 5D, whose AA filter is a tad on
    >> >>> the
    >> >>> inadequate side. Sigh.)

    >>
    >> >>>http://www.pbase.com/davidjl/image/52118414/original

    >>
    >> > For a photo like that, meaning everything can be in focus,
    >> > there is always God's AA filter: diffraction.

    >>
    >> That doesn't work very well because diffraction causes more loss in the
    >> pass
    >> band than a good AA filter (the MTF curve for diffraction is a straight
    >> line
    >> from 100% at 0 lp/mm to 0% at 1600/(f number), but what you want for an
    >> AA
    >> filter is a step function).

    >
    > Actually a step function causes ringing (thus power in higher
    > frequencies). But the power in the ringing depends on the cutoff
    > frequency, and I must admit I'm too tired at the moment to try to
    > guess if it'll be important or not for cameras (if one built a step
    > function cutting off near eg half the max spatial frequency).


    Isn't ringing what you get when you try to implement a step but fail? (I.e.
    don't have enough terms in the Fourier series?)

    > Anyway, physically I can't see how to construct a step function filter
    > for optical transmission (not that you said such a thing is feasible,
    > I'm just wondering).


    I wonder what the response of the AA filters actually used looks like. I'm
    under the vague impression that they consist of two (or more) layers of
    birefringent materials each of which shifts the image a bit in one
    direction. That is, that they are really quite gross kludges. (Hmm. Does the
    effect of the AA filter change if you use a non-circular polarizer???)

    > But this Luminous Landscape thing is excellent. The quoted passage (in
    > the OP) is pure poetry!


    But it's been removed. (Actually, that was a bit of a surprise. MR is one of
    the more stubborn blokes in this world, and usually responds to criticism by
    becoming more obstreperous. I wonder who got through to him, and how they
    did it.)

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Oct 20, 2007
    #15
  16. In article <>, David J.
    Littleboy <> writes
    >
    >
    >That doesn't work very well because diffraction causes more loss in the pass
    >band than a good AA filter (the MTF curve for diffraction is a straight line
    >from 100% at 0 lp/mm to 0% at 1600/(f number), but what you want for an AA
    >filter is a step function).


    Actually, the diffraction MTF for a circular aperture is not quite a
    straight line, it flattens off significantly towards the cut-off. The
    diffraction MTF is the autocorrelation of the pupil.
    --
    Kennedy
    Yes, Socrates himself is particularly missed;
    A lovely little thinker, but a bugger when he's pissed.
    Python Philosophers (replace 'nospam' with 'kennedym' when replying)
     
    Kennedy McEwen, Oct 20, 2007
    #16
  17. In article <>, acl
    <> writes
    >
    >Actually a step function causes ringing (thus power in higher
    >frequencies).


    It causes ringing in the fourier domain (time in audio, space in
    imaging) not in the frequency domain. There is NO power at higher
    frequencies - that is the point of the stop filter.
    >
    >Anyway, physically I can't see how to construct a step function filter
    >for optical transmission (not that you said such a thing is feasible,
    >I'm just wondering).
    >

    You can't, and it is fairly obvious why. If you look at the fourier
    terms required to create a brick wall type filter, the even terms are
    positive, but the odd terms are negative. That's easy enough to
    implement in an electronic signal, such as an audio signal. However
    there is no such thing as negative light intensity, black (ie. zero
    intensity) is as low as it gets. Consequently it is impossible to
    create an optical brick wall spatial filter, ie. an ideal anti-aliasing
    filter and all practical optical AA filters lose some MTF below the
    critical Nyquist frequency, whilst having significant residual MTF above
    it. They are, however, better than no AA filter at all.
    --
    Kennedy
    Yes, Socrates himself is particularly missed;
    A lovely little thinker, but a bugger when he's pissed.
    Python Philosophers (replace 'nospam' with 'kennedym' when replying)
     
    Kennedy McEwen, Oct 20, 2007
    #17
  18. In article <>, David J.
    Littleboy <> writes
    >
    >I wonder what the response of the AA filters actually used looks like. I'm
    >under the vague impression that they consist of two (or more) layers of
    >birefringent materials each of which shifts the image a bit in one
    >direction. That is, that they are really quite gross kludges.


    Yes, they are layers of birefringent crystals cut at the correct angle
    and thickness (0.825mm on the Canon 20D & 30D, 1.00mm on the Canon 5D)
    to displace the image in one polarisation axis relative to the other by
    a pre-determined distance. The result is a modulus cosine frequency
    response, with the first zero at 1/x where x is the relative
    displacement between the two images. Of course, this means that of its
    own, the filter is not an AA filter at all, merely one component of it.
    Another component of the AA filter is the pixel itself, which has a
    modulus sin(x)/x response with its first zero at 1/w where w is the
    width of the nominally square pixel. By positioning the zeros from the
    two filters at the right point, the MTF of the combination can be kept
    reasonably high below the Nyquist sampling frequency yet fairly low
    above it for several "cycles". Diffractive limits of the lens ensure it
    stays low, rather than increasing at very high spatial frequencies.

    The filters themselves are compound devices with horizontal and vertical
    displacement stages implemented in different layers with quarter wave
    plates in between.

    http://www.sensor-film.com/filter.html

    >(Hmm. Does the
    >effect of the AA filter change if you use a non-circular polarizer???)
    >

    Yes - the first (horizontal) displacement stage doesn't work reliably at
    all.
    --
    Kennedy
    Yes, Socrates himself is particularly missed;
    A lovely little thinker, but a bugger when he's pissed.
    Python Philosophers (replace 'nospam' with 'kennedym' when replying)
     
    Kennedy McEwen, Oct 20, 2007
    #18
  19. "Kennedy McEwen" wrote:
    > David J. Littleboy <> writes
    >>


    <Snip>

    Thanks! Much appreciated.

    > Diffractive limits of the lens ensure it stays low, rather than increasing
    > at very high spatial frequencies.


    Right; and if it's not diffraction, other lens infelicities reduce contrast
    at higher frequencies, so the AA filter only needs to work in octave above
    the Nyquist frequency.

    > http://www.sensor-film.com/filter.html
    >
    >>(Hmm. Does the
    >>effect of the AA filter change if you use a non-circular polarizer???)
    >>

    > Yes - the first (horizontal) displacement stage doesn't work reliably at
    > all.


    From that reference, it looks as though the IR filtering is so tied in with
    the AA filter that the IR-enabling surgery that one can have done probably
    loses the AA filter as well. Oops.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Oct 20, 2007
    #19
  20. W

    acl Guest

    On Oct 20, 1:47 pm, Kennedy McEwen <> wrote:
    > In article <>, acl
    > <> writes
    >
    >
    >
    > >Actually a step function causes ringing (thus power in higher
    > >frequencies).

    >
    > It causes ringing in the fourier domain (time in audio, space in
    > imaging) not in the frequency domain. There is NO power at higher
    > frequencies - that is the point of the stop filter.


    Yes of course the ringing will be in the spatial domain. Otherwise it
    wouldn't be a step function in the frequency domain, would it? I have
    no idea why I wrote "thus power in the higher frequencies", or what I
    was trying to say. Hmmm...

    Anyway, I still don't know whether this ringing will be a problem or
    not.

    >
    > >Anyway, physically I can't see how to construct a step function filter
    > >for optical transmission (not that you said such a thing is feasible,
    > >I'm just wondering).

    >
    > You can't, and it is fairly obvious why. If you look at the fourier
    > terms required to create a brick wall type filter, the even terms are
    > positive, but the odd terms are negative.
    > That's easy enough to
    > implement in an electronic signal, such as an audio signal. However
    > there is no such thing as negative light intensity, black (ie. zero
    > intensity) is as low as it gets. Consequently it is impossible to
    > create an optical brick wall spatial filter, ie. an ideal anti-aliasing
    > filter and all practical optical AA filters lose some MTF below the
    > critical Nyquist frequency, whilst having significant residual MTF above
    > it. They are, however, better than no AA filter at all.


    Ah excellent, thanks!
     
    acl, Oct 20, 2007
    #20
    1. Advertising

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

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. Marc
    Replies:
    8
    Views:
    429
    trout
    May 6, 2005
  2. Bill Mcdonald
    Replies:
    8
    Views:
    527
    Michael J. Moline
    Jan 19, 2004
  3. deryck  lant
    Replies:
    18
    Views:
    488
    Neil Ellwood
    Sep 4, 2005
  4. Annika1980

    Reichmann vs. Rockwell: "It's the equipment, Stupid!"

    Annika1980, Dec 20, 2006, in forum: Digital Photography
    Replies:
    8
    Views:
    325
    Kinon O'Cann
    Dec 21, 2006
  5. Giuen
    Replies:
    0
    Views:
    1,418
    Giuen
    Sep 12, 2008
Loading...

Share This Page