Loading ...
Sorry, an error occurred while loading the content.
 

Phil Groce: Light output of fulldome systems

Expand Messages
  • Ed Lantz
    Posted on behalf of Phil Groce: Dear Pat, Ed and Rob: Published lumen ratings of projectors are simply irrelevant and do not take into account the use of
    Message 1 of 3 , Sep 28, 2005
      Posted on behalf of Phil Groce:

      Dear Pat, Ed and Rob:

      Published lumen ratings of projectors are simply irrelevant and do not
      take into account the use of fisheye lenses or the the fulldome lens
      characteristics such as speed(f ratio), as well as specific lens
      transmission and contrast characteristics (see my IPS 2004 Fulldome
      Summit paper). The same can be said for published contrast ratios.

      To standardize brightness of full-dome system, one should use incident
      brightness of white light projected through the center of the lens
      system measured in lux (lumen/meter squared) at a standard distance such
      as one meter from the lens front surface. This lux measurement
      eliminates the dome reflectivity factor, and the image size factor
      (full-dome vs. truncated dome or pixel/degree factor) when evaluating
      the brightness of projection systems and allows theater designers to
      calculate the desired dome size and/or dome reflectivity to achieve a
      certain reflected NIT or foot-lambert. It will allow us to compare
      apples to apples.

      Contrast ratios can be standardized as well by measuring "black level
      "(at the same projector lamp setting) using this same standard lux
      measurement method and determining the ratio of the black and white
      light lux values. As Ed suggests, measuring incident lux of projected
      black and white checkerboard screens could also be used for an even more
      real-world measurement. While none of these methods will give a true
      representation of the image contrast once a dome gets hold of an image,
      it will allow us to compare oranges to oranges. It has been my
      experience that the contrast differences in data projectors as published

      by manufacturers is not as nearly as critical as the the number of lens
      elements, the effectiveness of internal lens coatings and the
      reflectivity of the dome.

      Finally, there are many other image quality factors. When it comes to
      single full-dome lenses, line pair resolution must be considered. As
      chips get smaller with greater pixel density (resolution), line pair
      separation must go up. A single 1400 x 1400 poorly resolved image will
      not be perceived to be as sharp as a well resolved 1024 x 1024 image.
      Then there is the issue of the chromatic aberration of the lens as
      measured in the center of the lens and at the edge of field. All fisheye
      lenses suffer from a certain degree of "coma". All of these factors
      should be weighed in determining image quality and in evaluating
      full-dome systems.

      Give me on any day a seam-free, pixel-sharp, bright, high contrast 1k x
      1k image relatively free of chromatic aberration and coma over a dim.
      low-contrast, fuzzy, and aberrated 4k x 4k image with obvious seams.
      Audiences are not dummies. That is why they don't seem to mind lower
      resolution systems when the image is bright, seamless and well
      resolved.


      Phil


      Philip Groce
      Konica Minolta Planetariums
      619 Orange Street
      Macon, GA 31201

      Tel: 478- 750-7870
      Fax: 478-750-7826
      cell: 478-719-0661

      e-mail: office: hps4075@...
      home: groce@...
    • Ed Lantz
      ... Light measurements are the most mixed up set of units in all the sciences, I think. Luminous flux, luminance, illuminance, foot-Lamberts and nits. next it
      Message 2 of 3 , Oct 4, 2005
        Building on Phil Groce's suggestions:

        >To standardize brightness of full-dome system, one should use incident
        >brightness of white light projected through the center of the lens
        >system measured in lux (lumen/meter squared) at a standard distance such
        >as one meter from the lens front surface. This lux measurement
        >eliminates the dome reflectivity factor, and the image size factor
        >(full-dome vs. truncated dome or pixel/degree factor) when evaluating
        >the brightness of projection systems and allows theater designers to
        >calculate the desired dome size and/or dome reflectivity to achieve a
        >certain reflected NIT or foot-lambert. It will allow us to compare
        >apples to apples.

        Light measurements are the most mixed up set of units in all the
        sciences, I think. Luminous flux, luminance, illuminance, foot-Lamberts
        and nits. next it will be photons per fortnight (just kidding - that
        would actually be radiant flux, not to be confused with luminous flux
        which accounts for eye sensitivity).

        Below is a summary of units taken from
        http://www.ibiblio.org/pub/multimedia/3d/3d-photo/photometry:

        Q light quantity lumen-hour radiant energy
        lumen-second as corrected for
        eye's spectral
        response

        F luminous flux lumen radiant energy flux

        as corrected for
        eye's spectral
        response

        I luminous intensity candle one lumen per
        steradian
        candela one lumen per
        steradian
        candlepower one lumen per
        steradian

        E illumination foot-candle lumen/foot^2
        lux lumen/meter^2
        phot lumen/centimeter^2

        B luminance candle/foot^2 see unit def's.
        above
        foot-Lambert = (1/pi)
        candles/foot^2
        Lambert = (1/pi)
        candles/centimeter^2
        stilb = 1
        candle/centimeter^2
        nit = 1 candle/meter^2


        I personally prefer total lumens actually delivered to the dome screen
        as the IPS standard for projector "brightness" (luminous flux). Display
        engineers commonly use lumens, which are not dependent on the projected
        angle. Not all projection systems are hemispheric, of course.

        Using lumens as a measure, a truncated hemispheric projection will
        deliver more lumens because of the greater utilization of the video
        frame (thus better light utilization), but the field-of-view of the lens
        is immaterial (except for optical losses in the different lenses, which
        should already be accounted for in the spec). The luminance in
        foot-Lamberts (or nits, if you must) can be calculated by dividing the
        luminous flux in lumens by the area of the screen in square feet (or
        square meters for nits), then multiplying by the screen reflectivity.
        This assumes a lambertian screen that scatters light equally in all
        directions, which most domes are designed to do.

        Regarding peak brightness, with all due respect to D'nardo, CRT
        projectors often use this measure and it is useful for them. When doing
        a shootout between a 7,000 lumen (ok, 4130 IPS lumen) fisheye LCoS
        projector (a D'nardo's lens, in fact) and 6-projector CRT system, I was
        surprised to see the CRT grids and stars appearing brighter than the
        LCoS. That's because, although the ANSI brightness for the CRT
        projectors were 250 lumens each, they had over 1200 peak lumens, and
        stars/grids permit peak operation. When projecting the entire earth,
        however, the LCoS projector won out, since the CRT went into electron
        beam current limiting mode. In another decade or so this will be moot
        since all CRT projectors will have died by then and you will need to
        visit a museum to see one.

        >Contrast ratios can be standardized as well by measuring "black level"
        >(at the same projector lamp setting) using this same standard lux
        >measurement method and determining the ratio of the black and white
        >light lux values. As Ed suggests, measuring incident lux of projected
        >black and white checkerboard screens could also be used for an even more
        >real-world measurement. While none of these methods will give a true
        >representation of the image contrast once a dome gets hold of an image,
        >it will allow us to compare oranges to oranges. It has been my
        >experience that the contrast differences in data projectors as published
        >by manufacturers is not as nearly as critical as the the number of lens
        >elements, the effectiveness of internal lens coatings and the
        >reflectivity of the dome.

        Measuring contrast is tricky - fortunately it is a simple ratio, so
        absolute measurements are not required (one can work in luminance,
        illumination, or luminous flux). A projected checkerboard (ANSI
        Contrast - ratio of white checkerboard square to black one) actually
        measures the theater scattering performance, not the display system
        performance. That's because the contrast limit with a checkerboard in a
        dome is cross-dome scattered light, which in the best of cases yields a
        10:1 contrast ratio for a 0.3-ish reflectivity dome screen. A 0.45
        reflectivity screen yields a 7:1 contrast ratio as I recall. This
        figure also depends on the brightness of fixtures, furnishings and
        finishings in the theater. White chairs, for instance, scatter more
        light and wash out the screen even worse. "Checkerboard contrast" is a
        useful measure - but for the theater, not the display system.

        One IPS standard for displays should be sequential contrast, the ratio
        of an all-white frame to an all-black frame. Problem is, the black
        level is often out of range of common light meters... I therefore
        measure sequential contrast using a white reflective (paper) target
        placed a meter or so from the projector. Make sure that the white level
        is not saturating the meter - this recently caused me some grief. I've
        also pointed a spot photometer directly into the lens of a CRT and
        measured the ratio of black/white on the phosphor. It's fine as long as
        you don't change anything between measurements except for the video
        level (a calibrated ND filter may be required to extend the range of the
        meter). Sequential contrast, more than any other measure, tells you how
        well the projection system will look when the image fades to black or
        attempts to project a minimal image against black such as stars.

        Another IPS standard that we need is measuring the contrast limit of the
        projector itself due to internal scattering. This cannot be measured
        without projecting an image. In my display standards paper:
        http://extranet.spitzinc.com/reference/IPS2004/Fulldome%20Display%20System%20Specifications_%20Proposal_Lantz.pdf
        and
        http://extranet.spitzinc.com/reference/IPS2004/Fulldome%20Display%20System%20Proposal_lantz.ppt
        I suggest that we adopt a test pattern
        consisting of a single 12 degree diameter white disc with a 3 degree
        diameter black hole cut out in the center of it for this measurement.
        The hope is that light scattered into the center of the disc will be
        dominated by internal scattering within the projector and not cross-dome
        scatter. I don't know what the limits of this measurement are, but it's
        sure better than a checkerboard. The best way to measure projector
        contrast is, of course, to remove it from the theater and place it into
        a controlled space without a reflective dome hanging above.

        Regarding resolution, the simulator standard is line-pairs/arcminute,
        with eye-limited resolution at one line-pair per arcminute or so (varies
        from person-to-person and also varies with brightness!). I've suggested
        pixels per degree as an IPS standard measure of resolution. Here's why.
        To start with, modern digital projectors have discrete pixels and
        typically have a high MTF (a measure of how much one pixel blurs into
        adjacent pixels), so pixels per degree, while not a "pure" measure of
        resolution, does the trick when comparing most systems. Should a
        non-digital system come along (i.e. CRT, scanned laser, etc) we can
        force them to an "effective pixels" measurement with a 50% MTF
        requirement. This means that two parallel lines define a pixel pitch or
        spacing when they are so close together that the dark space between them
        is 50% white. In other words, when the two lines are so close that they
        bleed onto one another and fill in the space between them to 50% of
        their brightness - we arbitrarily call that the effective pixel width.
        Of course, this depends on electron beam focus (for CRT) or Gaussian
        beam width (for scanned laser). Additional specification would be
        required to fully describe these non-digital systems: the total number
        of addressable pixels. For instance, 4000x4000 pixels may be
        addressable (according to the A/D sample rate and analog bandwidth), but
        only 2000x2000 may be resolvable (based on 50% NTF criterion).

        Confused yet?

        Phil continues:

        >Finally, there are many other image quality factors. When it comes to
        >single full-dome lenses, line pair resolution must be considered. As
        >chips get smaller with greater pixel density (resolution), line pair
        >separation must go up. A single 1400 x 1400 poorly resolved image will
        >not be perceived to be as sharp as a well resolved 1024 x 1024 image.
        >Then there is the issue of the chromatic aberration of the lens as
        >measured in the center of the lens and at the edge of field. All fisheye
        >lenses suffer from a certain degree of "coma". All of these factors
        >should be weighed in determining image quality and in evaluating
        >full-dome systems.

        >Give me on any day a seam-free, pixel-sharp, bright, high contrast
        >1kx1k image relatively free of chromatic aberration and coma over a dim.
        >low-contrast, fuzzy, and aberrated 4k x 4k image with obvious seams.
        >Audiences are not dummies. That is why they don't seem to mind lower
        >resolution systems when the image is bright, seamless and well
        >resolved.

        Coma and chromatic aberration are more of an issue with fisheye systems,
        while edge-blend artifacts and color balance are the bane of
        edge-blended systems. My paper presents a method for measuring
        edge-blend uniformity and color balance. Test patterns can help with
        coma and chromatic issues, especially if you are looking for a pass-fail
        test. But you'll not likely see manufacturers voluntarily quantify
        these factors. That is, after all, the reason for these specs. Please
        also note that these are fulldome SYSTEM specifications, not individual
        PROJECTOR specifications. The idea here is to define what will be
        actually seen on a given dome screen so there are no customer surprises.


        Thoughts?

        Ed
        ed@...
      • Philip Groce
        Ed: Thanks for the thoughtful response. I only have a few additional comments: 1) With incident lux value of both black level and white light measurements at a
        Message 3 of 3 , Oct 5, 2005
          Ed:

          Thanks for the thoughtful response.

          I only have a few additional comments:
          1) With incident lux value of both black level and white light
          measurements at a standard distance we can easily standardize the method
          of stating the brightness and contrast ratios of projection systems
          since dome size and reflectivity are generally unique to each facility.
          A prospective buyer only needs to apply the inverse square law and the
          percentage dome reflectivity to calculate the resulting theoretical
          foot-lamberts. For the record, most Konica Minolta meters are sensitive
          enough to accurately measure black levels and are used my many projector
          manufacturers for this purpose. The other reason to use incident lux is
          that there is over a half century history of using the incident lux
          value of the starfield as a general way of communicating the brightness
          of a projected starfield. This allows us in one measurable way to
          compare analog and digital starfields. Such knowledge can only benefit
          the prospective planetarium buyer and help digital systems reach for the
          much coveted optical-mechanical standard of starfield brightness,
          contrast and resolution.

          2) It doesn't matter what the pixel density of the system is, if the
          optical system is incapable of resolving those pixels. You have thoughfully
          and accurately described the problem of "what is a resolved pixel?" It
          also brings up the issue of the pixel and the relative size of the
          separation of pixels. The better the LCD or DLP projector chip, the
          smaller (and darker) the separation and the less "pixelated"(I'm not
          sure that this is a word) the apparent image for any given resolution.
          To me the ultimate line pair resolution of a lens is one that resolves
          the spaces between pixels. Resolve that separation and you have a very
          sharp image. Most optics sold today for most data projectors may be
          capable of resolving the pixel, few resolve the space between.

          Phil
        Your message has been successfully submitted and would be delivered to recipients shortly.