Stereoscopic 3D macro lens (2x magnification) for Sony FE cameras

prusaprinters

This design combines two 4x microscope lenses together to create a stereoscopic 3D macro lens, with 2x magnification and a 5mm stereo baseline. It has been designed to be shot handheld in the field using a flash for lighting, rather than in the studio on a tripod, so live subjects in their natural environment can be photographed. Video recording is also possible.The field of view at the focal plane is 18x12mm on a full-frame camera, or 12x8mm on a 1.5x crop camera. Larger subjects can't be photographed, as the magnification is fixed at 2x.Two Sony E/FE cameras are required, ideally identical models.The microscopes shoot through a beamsplitter prism so that one camera can be mounted to the top or bottom of the rig, while the other points towards the subject normally.The stereo baseline of 5mm is achieved by offsetting both microscopes from the centreline of the prism by 2.5mm. Each camera is then offset from that position by a further 5mm. This offset is the equivalent of rear-standard shift in a view camera, and causes the field-of-view and focus planes of both cameras to fully coincide. This approach of shifting the cameras instead of tilting the lenses avoids introducing “toe-in” effects that could make the stereoscopic images hard to view, and ensures that the subject is in-focus from edge to edge from both viewpoints (unlike the problems faced by a Greenough-style 3D microscope).This is a cross-section as seen from above; the first object is the prism at top, then the offset microscope, then the further-offset camera at the bottom:I can add support for these mirrorless cameras on request, just leave me a comment if you are planning to build this: Canon RF, Nikon Z, Fujifilm X, Leica L, or Micro Four-ThirdsIntroduction to the systemPrinting instructionsI have only tested eSUN ABS+ Black filament, and I won't be testing or giving instructions for any others. I would strongly recommend ABS or ASA filament unless you want to experiment with settings and tolerances yourself.Print one copy of each of these files:Full-frame - 5mm baseline - Body set.stlFull-frame - 5mm baseline - Camera connectors - 2x magnification - Sony FE.stl4mm apertures for Reakway 4x microscope.stlWith these settings:Layer height: 0.1mmPerimeters: 3Infill: 13% rectilinear (or similar)9 top layers, 8 bottom layers (or similar)Nozzle: 0.4mm diameterElephant foot compensation (first layer XY compensation): DisableFilament shrinkage: (required for ABS-like filaments)In SuperSlicer set shrinkage to 99.5% in the filament settingsIn PrusaSlicer select the models on the buildplate and set the scale factor for X and Y to 100.5%, but keep Z fixed at 100% (unlock the aspect ratio padlock first to allow this). Perimeter generator: Classic (I did not test Arachne and I've had trouble with it in the past)Supports: NoneFilament: eSUN ABS+ Black. Filament must be black. This specific filament is also very matte, which is optically fantastic.Fan speed: I used:30% constant fan speed for the camera connectors (which is my default setting)60% fan speed for the body set (for better overhangs near the bed)100% fan speed for the 4mm apertures… but best settings will depend on your printer and your chamber temperaturePrint one copy of the “diffuser”:White filament, I used white PETG100% infill0.2mm layer height.Parts and toolsThese parts are required:Reakway 4x plan microscope objective x 2 pieces. This specific model is required, no lookalikes, as some have wildly-different internal constructions. See pictures below for which internals are the correct ones.18mm cube beamsplitter prism x 1 pieceM2x6mm cap head screw x 18 piecesM2 heatset threaded insert, 3.5mm diameter, 3mm length x 18 piecesThin (≤1mm) adhesive black foam sheet (optional)These tools are required:Soldering iron with conical tipHammerChopstick, or similar wooden rod1.5mm ball-headed hex wrench (Allen key)Lens cleaning cloth/microfibre cloth or cotton budsLens cleaning fluid or isopropyl alcoholFor use with this camera equipment:2x Sony E/FE mirrorless cameras (ideally identical models). Crop and full-frame models are both compatible.Twin-head snake-arm macro flash (recommended)AssemblyMicroscopesPrepare the two microscopes by unscrewing the silver outer case (right) from the objectives, and discard them:Note that the internal and external threads on this microscope are too sharp to be gripped with bare fingers without cutting your skin, so use a kitchen rubber glove to protect your fingers! They will also cut through cotton t-shirts.Unscrew the black aperture disk (right) from the rear of the objectives. Your rubber glove can help you get enough grip to start untwisting this (but don't hold the objective by its silver threads while you do that, because those are sharp too). Screw in the 3D-printed 4mm apertures in their place.If you have black adhesive foam, you can use it to cover up the shiny silver ring on the tip of the microscope, so that its reflection doesn't get seen by the prism. I used two wad punches to cut it into the correct donut shape. 0.5mm of space has been allocated for this, and my 1mm adhesive foam successfully compressed to this thickness.You could probably paint it black instead, or just leave it virgin, with some loss of contrast.Main body heatsetsUse your soldering iron to install heat-sets in the indicated positions only (the 4 brass coloured inserts shown below, not the other holes). I set my iron to the same temperature that I extrude my filament at:Repeat that on the bottom side too:Install the 5 inserts shown into both copes of this lens mount piece, then screw a screw into each one of them to push out any excess melted plastic from the rear. Then remove the screws and cut off those excess plastic dags:Installing the prismThe prism has 3 transparent sides and 3 opaque black-painted sides. The transparent sides will face the microscopes and the outside world, while the black-painted ones will rest against plastic. If you wear cotton gloves while handling the prism you can avoid putting fingerprints on it, which makes cleanup much easier. Push the prism (red object in the cross-section below) into the square channel of the main body from the rear, so that its split line is oriented like so, to send half of the light to one camera and half to the other:You can stick your finger in through the bottom hole to help lift and align the prism (it will be better guided in if it is pressed upwards, in the image shown above).This prism is retained and aligned by being squeezed by the print itself (see where the channel narrows in the diagram above, near the front aperture of the main body), so it needs some “mechanical persuasion” to be installed. I stuffed a microfibre cloth into the square installation hole to protect the prism's surface, and then used a wooden stick as a punch with my hammer to hammer the prism home. Verify that the prism sits flush with the front aperture all the way around.Clean up any exposed faces of the prism with a lens/microfibre cloth or cotton-bud, dampened with lens cleaning solution or isopropyl alcohol. Avoid spraying the cleaner into the assembly directly, because droplets can get trapped somewhere unwanted and leave behind residue when they evaporate.Install the microscopesScrew a microscope into each of the lens mount pieces in the orientation shown below:Use your kitchen rubber glove to give you better grip to fully tighten the objective into place (remember that those threads are sharp!). Clean the front element of the microscope now (if needed) since it will be inaccessible later.Slide each completed mount+microscope into the main body, and add 4 screws (red) to secure them into place. Tighten the screws gradually in a circular pattern so the mount doesn't tilt or deform as it tightens:Now you can join the camera adapter tubes to the main body. There are two kinds of tubes, ‘A’ and ‘B’ (this is labelled on the rim of the camera mount), and they differ only by which way up the camera will face on the mount. The labelling on the rim of the tube will face the bottom of the camera.The suggested orientation puts a “B” tube on top and an “A” tube on the rear, as shown below:The example camera model used in this image is by guileman on Grabcad.If you want to have a bottom-mounted second camera (pointing upwards) instead of a top-camera as shown above, swap the A and B tubes so that the main camera doesn't end up upside-down.Put one of the thin “gasket” prints between each tube and the main body, then screw down the tube using 5 screws and your ball-ended hex key (it must be ball-ended, as you'll be screwing them in on an angle).The full assembly looks like this in cross section:The “gasket” part allows the focus distance of both microscopes to be matched with each other by varying the thickness of the gasket between them, to ensure that the focal plane of both microscopes coincides. In practice I haven't needed to do this thus far.Slide the ridge on the white printed diffuser sideways into the small keyed gap in the front face of the main body, highlighted blue:Shooting with the 3D macro lensThe aperture is quite dark in order to give an extended depth of field, so you'll want to use a flash for illumination. A twin macro flash where the heads are mounted on flexible arms is ideal, because you can mount this to the hotshoe on the main camera to relieve stress on the mount, and reposition the flashheads to the front of the 3D lens.The flash that I'm using is the Canon MT-24EX on custom arms, but these days there are dramatically-cheaper noodle-armed macro twin flashes on AliExpress which look suitable.Configure the cameras like so:Manual exposure modeMatching white balance, shutter speed, and ISO settingsDisable the orientation sensor (if possible) so that the downwards-facing camera doesn't erratically rotate images depending on what angle it tilts at.Single shot modeRAWWhen holding the setup, it's best to support the 3D lens with your left hand underneath it, to carry some or all of the weight of the top-mounted camera.Synchronising the firing of the shuttersFor Sony cameras with Multiport connectors, there is a third-party cable available with a multiport connector on both ends which causes the shutter button on one camera to fire the other camera (the cable is directional, so you can swap ends to choose the master). The official Sony version of this cable is called VMC-MM2, but I've only tested the third-party one.If both cameras are identical, then I found that with my Sony and Canon cameras, I could take two shutter release to phono jack cables, and join tip-to-tip, ring-to-ring and sleeve-to-sleeve. This effectively puts both camera shutter buttons in parallel, which makes it so that pressing the shutter button on one camera will also fire the other at the same time.If both cameras are not identical then this approach might not be safe, as the cameras may use different voltages for the internal pull-ups on their focus and shutter pins. In this case you need to first measure the tip-to-sleeve and ring-to-sleeve voltages with a multimeter (e.g. to confirm they're both 3.3V, and not a mixture of 3.3V and 5V). Because the cameras are not expecting an external voltage to be applied on this port, a voltage mismatch here could blow up one of the cameras.The other issue with non-identical cameras is that they could have a different shutter-opening delay, so they won't open their shutters in sync even though the shutter buttons are synchronised. e.g. my Sony A7's shutter delay is 19ms while my Sony A7R's is 160ms! If one camera has a mechanical first-curtain shutter like my A7R, while the other uses an electronic one, see if you can configure the other camera to also use a mechanical first-curtain shutter, to better match the shutter delays.Note that joining the cameras like this causes them to have a shared electrical ground. This could have potential consequences (pun intended) if both cameras were connected to different mains-powered equipment like studio strobes, or PCs over USB.It's difficult to conceive of a fault scenario for solely battery-powered equipment.Editing imagesBecause the top/bottom camera is effectively shooting through a mirror, you'll need to flip the images from that camera (not just rotate them).In Lightroom I mark the photos from one camera Red (by pressing 6) and the photos from the other camera Yellow (by pressing 7) so I can tell them apart. Then I use the stacking feature to group together matching photos from both cameras.I've included a zip file with lens correction profiles for Adobe Lightroom (.lcp) and Darktable (.xml) which correct for the off-centre vignetting caused by the lenses being mounted 5mm off-axis. If you're using crop-sensored cameras then you won't really need to use this profile, because vignetting is minimal there.Unpack the zip, then in Lightroom click "File > Import Develop Profiles and Presets…" and select the two LCP files. New profiles will now become available to be picked manually in the “lens corrections” panel in Develop mode:You need to pick the correct profile (camera A or camera B) depending on which tube the photo was taken on, or else the vignetting will be offset to the wrong side.In the Library view, you can select both of the left and right photos and press N to enter the survey view. If the window is wide enough, this will show both photos side-by-side, so you can cross your eyes to view them in 3D (see the next section for instructions). You may need to collapse the side panels for the window to be wide enough. If the images are shown in the wrong order for cross-eyed viewing, their depth will appear to be backwards. You can drag and drop them within the stack to swap their display order to fix this.Viewing imagesYou can show the images on-screen in either a “left - right” or “right - left” order, which requires them to be viewed using parallel view or cross-eyed view techniques respectively.Cross-eyed viewing allows much larger display sizes, so that's how I've arranged my sample photos above.If you hold your finger up in front of your face and look at it, your eyes converge (cross) in order to both point at it. If you now move your finger towards you and away from you, in the background you will see the two side-by-side photos cross over each other more and less, respectively. Move your finger until both images in the background fully fuse together to form a third image in the middle of your field of view.The trick is now that you need to keep your eyes at the same convergence (still crossed to point at your finger) but you need to refocus to look at the monitor. This takes some practice, but once you get good at it you can cross your eyes at will without using a finger. The fused third central image will now appear to be 3D.Shooting videoUse settings like this for video recording:Manual exposure modeMatching white balance, shutter speed, and ISO settingsMatching frame rate, and shoot at as high a frame rate as possible to improve synchronisationAudio onStart both cameras recording, and clap or make some other noise to serve as a synchronisation point.Editing video (Davinci Resolve 18 Free Version)The free version of Davinci Resolve doesn't have support for 3D video, but we can work around that!In a new project, drop your source clips into the Media pool so that the project picks up the correct frame rate settings. Go to File > Project Settings and change the Timeline Resolution so that the width is double the width of your source clips (i.e. for 1920x1080 source clips, the Timeline Resolution should be 3840x1080).Switch over to the Edit tab, and drag both of your media clips into the timeline, one above the other in separate tracks. Grab the lines at the vertical midpoints of the audio tracks, and drag them upwards to boost the volume of both tracks loud enough that you can easily see the peaks for syncing the tracks:Find your clapper sound in the audio tracks, and use this to help you trim the start off one of the two videos until the video/audio tracks both line up:The alignment error here will max out at half a frame interval, which is why shooting at high FPS is valuable even if you deliver at a lower frame rate.Select one camera, and in the Transform menu offset the X position by one whole frame width (e.g. 1920px), and flip the image:Select the other camera, and offset the X position by -1920 instead. Don't flip this one. Now your two cameras should display side-by-side. For compatibility with YouTube, you want them to be arranged for parallel viewing (left camera on the left, right camera in the right), so if you got them the wrong way around, swap them over now in the Transform pane. If you're not targeting YouTube, you'll probably want them in Right-Left cross-eyed order instead for cross-eyed free-viewing.Now I used the razor tool cmd+B to trim my clips as needed, making sure that both the left and right cameras were selected at the same time so they were edited together and maintained their sync.Next, head to the Color tab. Select all the clips from V1, right click them and click “add to a new group”, give it a name like “Left”. Do the same for V2, but with a different group name.Select a clip from V1. In the Node editor, select “Group Post-Clip” from the dropdown. Right click and add a new Corrector node after the default node:We'll use the first node to make global corrections to match this camera's appearance to the other camera, and the second node to make corrections for vignetting.Select the first node and use the Gamma/Lift/etc controls as needed to match the appearance with the other camera in the centre of the image, but try to ignore the asymmetric vignette for now.Then select the second node. In the Window toolbar at the bottom of the page, enable the circular window. Enable the “invert” icon (the first icon on the right side here):Position it like this, vertically centred, but horizontally offset to one side of the middle of the frame. This mimics the off-centre mounted lens:Now you can use the Gain controls to brighten up the vignette until it isn't very visible any more. I found it easier to see the impact of the vignette if I zoomed out the preview window.Switch over to a clip from V2 and repeat this process. The vignette settings will be much the same, except offset to the opposite side from the centre this time. Head over to the Render tab and render out the video, you don't need any particular settings for this. I rendered to mov format.Finally, we need to mark the video with some metadata so that YouTube knows it's a 3D video. You'll need to install ffmpeg on your computer. Then you can use a command like this for the conversion:ffmpeg -i Source.mov -c copy -metadata:s:v stereo_mode=left_right Output.mkvThis will cause the video to show in cyan-magenta anaglyph on YouTube (for anaglyph 3D glasses), and prompt the user to view it in VR if they have goggles. However, since 2023 it seems that this isn't sufficient for the video to be detected as 3D on YouTube on the Desktop. To fix this, use the YouTube Video Editor to make a small edit (e.g. trim 1 second off the end of the video), and after processing finishes, it'll fix it!Click to view this example video on YouTube directly: