|
de Wijs Dual-Objective Stereo Macro Lens: Tests John Hart Program in Atmospheric and Oceanic Sciences University of Colorado Boulder, CO 80302 nimbus.colorado.edu/hart/science.htm April 10, 2006 In this technote we study the performance of the de Wijs stereoscopic macro lens on a high performance digital single lens reflex (DSLR) camera (17 Megapixel Canon 1Ds-2). The lens we tested is one of a series made by deWijs that offer a range of magnification from approximately 1:3 (magnification 1/3) to 3.6:1 (magnification 3.6). This is based on the width of the object versus the width of sensor (i.e. magnification 3.6 => 10mm wide subject on 32mm wide sensor). Our test lens, model A, had the highest magnification. These lenses are compact stereoscopic imaging devices that place a left/right pair on a single full 35mm frame. Our goal is to use this to visualize particle paths in small volumes of turbulent fluids in three dimensions. However, the test images shown here are simple easily-photographed common macro subjects.
Lens attached to the DLSR.
Front view, showing the pair of small diameter lenses at the front of the deWijs. The aluminum frame allows for positioning and setting the focus without using the viewfinder. Due to the high f-stop (45, 60, 90, increasing with magnification), the viewfinder image is quite dim.
Close-up of the twin lens system. Small diameter lenses allow narrow separation and facilitate an increase in depth of field. Separations vary between units with different magnification. Here are the specifications and links to the deWijs drawings.
Model A. Subject size: 10x14 mm. lens sep.: 6.5 mm. aperture 90.
Each lens paints a stereo (left/right) image onto the sensor. The lenses are designed for full-frame (36mm wide by 24mm high) digital or film cameras. A splitter at the back of the lens keeps the left and right images separate. The splitter does not interfere with the mirror system of the camera.
Because the viewfinder is dim, a focus guide is included with the lens. Its use is illustrated above. Some calibration will be necessary, because you probably want the center focus point to be somewhat in front of the vertical bars. There are screws at the bottom of the lens that allow one to move the guide in and out. de Wijs Macro Lenses for stereo photography.
Sketch of the deWijs macro lens. (Image from the deWijs website)
Technical Data from the deWijs Website (except row A, columns K - S, which were measured by J. Hart). All data in millimeters (mm). * Depends on setting of focus guide, whose distance is variable by about 7mm. K = focus guide width. L = focus guide height. M = usable depth of field. S = working distance (lens face to start-focus point). N = distance past S to center of focus (mid DOF).
The lens we had to test (courtesy of Jon Golden at 3D Concepts) was model A. In some ways this is the most extreme (highest magnification, lowest depth of field). Above is a ruler (in 1/32" units) photographed at 45 degrees. The apparent sharpness spans about 12/32" or so, which converts to about 7mm parallel to the optical axis.
Download full-size, only slightly-compressed version of original image (warning: ~ 2MByte file). The camera records parallel pairs directly onto its sensor. Above is a raw image (reduced in size for the web, of course). It comes out in cross-eye format. There is a narrow fuzzy zone (about 5% of the width of one side of the cross-eye pair), arising from the septum splitter plate at the back of the lens (see back-of-lens picture above).
Color anaglyph of the raw image. You can see the fuzzy zone at the right vertical edge. The image directly out of the camera is well behind the window.
A simple shift puts the subject closer to the window, but of course this costs some sensor real-estate (i.e. some pixels are lost). For the model A lens, we found that the shift costs another 5% horizontally. Thus, between the required shift and the fuzzy zone, you can expect to lose about 10% of the image width.
Here is an image of a resolution test chart (a crude printed one). Made with a flash, for stability. Note that these lenses are fixed aperture and fixed focus. Care was taken to be within the shallow focus zone by illuminating the subject with an intense model light and using the DSLR's viewfinder and a magnifier to critical focus.
Actual pixels of the small zone around the highest resolution block of the test chart. The optics are limited by diffraction (and perhaps other aberrations). In order to optimize depth of field, these units have a very small numerical aperture (roughly 1/(2 * f-stop). The chart is degraded somewhat from what you would see if you looked at it using a high quality microscope objective having a near unit numerical aperture (shown below). Of course, the microscope view has essentially zero depth of field. An important question: Is there enough resolution to enable good quality presentation of stereo images using standard printing and projection methods? The deWijs can be sharpened dramatically (using USM in Photoshop, for example). In fact this step is probably necessary when trying to make a quality presentation.
The smallest blocks of the printed resolution chart used above, photographed using a high quality microscope.
For presentation in the form of stereocards, or for digital projection, the vertical format (i.e. the so-called "portrait mode"), may not be preferred. If you must crop the DeWijs frame to get to a 4 wide by 3 tall (4:3) aspect ratio typical of digital projectors, a fair amount of image information will be lost. Doing this crop can alternatively be thought of as an increase in the magnification. In this process, defects in the lens, like it's diffraction errors, will become more apparent. EXAMPLE IMAGES: The images below are single sides of stereo pairs. The pairs can be viewed in various formats by clicking on each of the images shown, which activates our stereo image-server. The images have a maximum display resolution of 1024x768 per side. If you want to look at the full camera resolution, you can download some slightly-compressed but fully-sized samples. The full-size pairs are about 4800x1800. All the images below were sharpened using USM 150%, 1.9 pixels. This is fairly strong, but brings out details in these particular shots.
Download full-size slightly-compressed stereo pair, (a, left, portrait mode). Warning: large ~ 2MB file. Download full-size slightly-compressed stereo image (b, right, full frame 4:3 cropped mode) Warning: large ~ 2MB file.
Download full-size slightly-compressed stereo pair (warning: large file). In a projection shootout of these images, my conclusion is that the 4:3 cropped images appear just a tad soft. Square or portrait mode images look OK. Based on this, the latter modes would also make reasonably good stereo cards (where the projector resolution converted to 300dpi prints gives stereo-cards about 3 inches wide, per side), CONCLUSIONS: PROS: Compared with other methods using beam-splitters and such, this system is much smaller, faster to set up, and easier to acquire and process stereo pictures! You get instant stereo images - no inverting images in software or making beam-splitter corrections (keystone, color shift, resizing), etc. The range of available magnifications cover many common situations (short of high-magnification microscopy). Shifting the window and cropping to 4:3 (without the centerline fuzz), results in an image about 2400x1800 in size PER-SIDE (using a 17Mpix camera). Other full frame cameras can be up or downsized according to the square root of the megapixel ratio. For example, a 12 MPix camera would give images about 2016x1512. At 300dpi you can get prints that are 7 x 5, or at 500dpi you get 4 x 3. CONS: Critical focus on Model A was somewhat difficult because of the shallow depth of field. Careful setup and adjustment of the focusing frame is necessary. The splitter at the back of the lens should be protected. Care should be taken not to whack it. Use a back lens cap when the unit of off-camera. Cropping necessary to get to 4:3 aspect ratio costs a lot of pixels (compared with a native 4:3 twin-camera beam-splitter system, for example). The increased magnification in projection (i.e. spreading the cropped image across the full width of the screen, vs. spreading the full raw deWijs image across the full vertical height of the screen - and accepting the resulting black bars on the edges), led to images on-screen that were acceptable but which appeared a little soft. Summary: Recommended. The fact that this is a relatively small single-camera macro system is a major positive. The lower magnification units (e.g. models D - F, say) would appear to be easier to use than model A because they have more depth-of-field and larger fields of view. With model A you have to be quite precise in setting it up for a shot. The main use for these lenses seems to be in quick set-up situations for subjects having motion. Synchronization between left and right sides is automatic. Flash works perfectly. Compared with other systems aperating at similar magnification, these relatively compact, light, and effective deWijs lenses have many advantages. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||
