| JOHN'S 3D GUIDE | |||||||||||||||||||||||||||||||||||||||||||
|
Back
|
|||||||||||||||||||||||||||||||||||||||||||
|
CHAPTER 3: TWINCAMS (TWO-CAMERA STEREOGRAPHY). SYNCHRONIZING AND MOUNTING |
|||||||||||||||||||||||||||||||||||||||||||
|
Updated Nov. 10, 2004 One must fire the two cameras together. This is called SYNCHRONIZATION. A good support is also needed in twin-camera stereo photography. CAMERA TRIGGERING DEGREE OF SYNC A critical question is: HOW CLOSE TOGETHER MUST THE TWO SHUTTERS FIRE TO RENDER A GOOD STEREO PAIR. A simplistic, but suggestive, analysis is as follows. 1) A good stereo pair will have parts of the image that are separated by about 1/30'th the width of the film. (Example: widest separation for 35mm film ~ 1.2mm). This degree of separation gives the depth from the front to the back of the image. 2) If you want the depths to be accurate to 10%, say, then shutter mis-sync errors should not lead to deviations more than 1/300'th the width of the film. For example, suppose a goose flies across the whole frame in 1 second. In order for the goose to be in the right place (depth-wise), the shutters should fire within 1/300'th second. Another example: Water drops in a waterfall fly around at a speed that would cause them to cross the whole frame in 1/2 second. Even though the may not go that far, we still would require synchronization to about 1/500'th. A last example: Clouds drift across a scene, crossing the frame in 30 seconds. Thus to accurately capture them in 3D, we need the shutters to fire in 30/300 = 1/10'th second apart. The first three examples require careful and accurate synchronization. The last example could be shot by triggering your cameras with your fingers. Any digital pair (pretty much however wired together) will do 1/30'th sync. But the last example also shows that even cloud drift will cause difficulties for photography with a single camera on a slidebar, where the interval between pictures is typically a couple seconds or more. METHODS OF SYNCING CAMERAS Mechanical or double-cable release linkages have been used. Here is one I made for a pair of digital cameras (Sony DSC-U60):
This mechanical trigger routinely caused the cameras to fire together to within 1/40'th second, or better. Electronic Linkages Synchronization can alternatively be obtained by wiring electronic shutter releases together. DETAILS BELOW. Research may be required to make sure you can work this with your particular camera. Certainly do this research before buying another camera body (or a pair of cameras). Single pole (shutter only) releases are easiest to wire, but it is nice to be able to preset the focus and exposure, if only to see what the exposure will be. I 35mm FILM CAMERA SYSTEMS THAT CAN BE ELECTRONICALLY SLAVED TOGETHER:
It is best to do this "Twinning" with identical models. The internal timing for each camera will then be as close as possible, so the probability of success in achieving good synchronization of shutter opening is optimized. DIGITAL CAMERAS THAT CAN BE ELECTRONICALLY SLAVED TOGETHER
Below we give some details that may help those interested in embarking on their own synchronization project. As of Nov. 2004, my favorite system is the SONY V1 (or V3, or F828) in combination with the LANC Shepherd. This system works well and does not require any camera modification.
DETAILS ON WIRING SOME REMOTES FOR PARALLEL TRIGGERING This, at least, should give you an idea what you are in for. Most new cameras are electronic. Make sure your camera body has a remote electronic trigger port (unless you want to get into taking your camera apart). This is a two or three (or more) pronged outlet on the body that a switch trigger can be attached to. Pentax's, Minolta's, Nikon N90, and some Canon's have these fittings and appropriate trigger cables. Digital SLR's and some ConsumerCams (notably Minolta D7) along with other makes and models do also. Ask at a good photo outlet if your camera can be fitted with a remote electronic cable release, and test it out in the store. It is desirable to have a two position switch. The first presets the camera by forcing it to auto-focus and set the exposure (if these options are enabled). The second switch position fires the shutter. To get both cameras to go off simultaneously you have to wire the two cables coming from the remotes of the two cameras into a single release switch. This probably means carefully taking the switch off one of the two releases you will need to buy, and attaching it to the remaining single switch. Alternatively you may be able to buy the parts (switches) at Radio Shack and make your own trigger. This requires a bit of electronic tinkering, but is not rocket science. 1) If the remote port on your camera is wired in parallel to its trigger button, one approach is to simply take the remote cords from each camera and wire them together. The wires in the cords are color coded, so just find a way to solder like-colored wires together and insulate them. With this method, pushing the button on one of the two twin-cam bodies will trigger the other. One disadvantage is that the cameras can only be operated independently if the remote cord joining the two is disconnected. This parallel-wiring method works well with some cameras (e.g. the Minolta Dimage 7 digicam - see below), but will not necessarily work with all cameras because the remote-port and the trigger button on the camera may not be internally hard-wired together. Some people have taken their cameras apart to wire directly into the button's contacts, but this is beyond what most people will want to try. 2) If the remote port operates through some internal camera circuitry, then you may need to insert diodes into the external switch lines. The diodes isolate one camera from the other. This method is illustrated in figures 3.4 - 3.7, and is what I have used with Pentax bodies. 3) If you wish to be able to operate a body independently by pushing its trigger button, you will need the diodes illustrated in figure 3.4.
EXAMPLES: HOW TO PARALLEL WIRE THE REMOTE SWITCH FOR A PENTAX FILM SLR and A MINOLTA DIGITAL CAMERA. Here is the Pentax release switch as shown in the diagram including diode protection.
The cable wires, shown in green in fig. 3.4 are usually color coded (see fig. 3.7). Omit the diodes if you want to be able to fire the cameras by using the trigger button on one or the other bodies. THE DIODE-LESS CIRCUIT MAY NOT WORK ON YOUR CAMERA! For example, on the Pentax ZX5-N, wiring without the diodes causes the two cameras to trigger each other continuously, hands off, while you sit there horrified! On the other hand, for the Minolta Dimage 7 (see below) the diode-less circuit works just fine. Below are some images showing installations on Pentax and Minolta equipment.
TRIGGER TIMING TESTS Once you have completed the cable release, or other connections as suggested above, it may be useful to find out how accurately your two cameras will trigger. Ideally you would like to find this out before building all this stuff, but life isn't always so cooperative. A simple method is to shoot high-shutter-speed images of your TV and observe how far down the screen the refresh has progressed. Fields refresh every 1/60'th second, so the difference between cameras can be measured this way (unless it's more than 1/60'th, whence interpretation is a little harder). Another way to test the trigger speed is to use the pair of cameras to shoot a spinning wheel. The setup for doing this is shown below.
In figure 3.9, for example, the stripe width is 1 inch and the circumference, as measured with a tape, is 60 inches. The stripes are about 4.3 strip-widths apart. Therefore the error (lag or lead) is 4.3 * .5 * 1 / 60 ~ .035 sec ~ 1/30 sec. This is not very good. You would never be able to sync with a single flash triggered by one camera. Shots of water flying around a kayak would never work. In 1/30 of a second the left camera would see a completely different water scene than the right one. A headache for sure. Fortunately, the cameras I use turned out to be pretty accurate. WITH FRESH BATTERIES IN EACH, they routinely fire less than 1/250'th of a second apart. This becomes worse, going down to 1/60 or so, with aging batteries. Therefore if I know I'm going to shoot action, I insert a fresh set and keep track of how many rolls get used on these "action batteries." Usually four or five rolls come out OK. Then I set these batteries aside for more static landscapes. NOTE: You don't need a fancy motor driven wheel to do this test. I happened to have one lying around, so what the heck. Just measure the circumference of a wheel on your bike, then attach a one or two inch piece of white tape onto the black tire. Set the bike upside down and have a friend spin the wheel. Get about 2 turns per second and shoot. You don't have to be super accurate in the turns-per-second deal. What's the difference if you measure a lag of 1/200 seconds and it's really 1/180 or 1/220? All you really want to know is if you are in the 1/200 or 1/500 ballpark, or if you are down in the 1/30 range.
EXAMPLES OF DIGITAL SYNCHRONIZATION TEST CASES THE OLYMPUS C3040 (aka 2020, 4040, etc., of the Camedia series). These are sweet little rangefinder cameras with high resolution and good optics. The Twin-Cam setup is shown below.
There is no hard-wire remote for these, but like many other digitals, they can be popped with an IR remote. Unfortunately, I learned upon test that there is a built in 3 second lag between IR trigger and the shutter release. Presumably this is so you can take a picture of yourself and hide the IR in your pocket before the camera fires. This is not so good for action stuff! With the cameras set to aperture priority, and pre-focused by pressing the silver trigger down a little, the timing was pretty good. Better than 1/60 of a second for most shots! The system was not accurate enough to run with a flash, which requires close synchronization of the shutters so that the left camera is completely open when the flash on the right camera goes off. NOTE: CANON G3 claims IR remote with no lag. TO BE TESTED (Dec. 2002). MINOLTA DIMAGE 7 (See Technote: Triggering the Dimage 7, for detailed results) This is a high resolution 5 Megapixel camera with a true 28mm (equivalent) wide angle lens. The stereo twin-cam setup is shown below.
One problem with the Dimage 7 is that because of the long barrel associated with the lens assembly, it is difficult to close-mount these cameras. You could overcome this by flipping one camera upside down (i.e. rotate the right camera of figure 3.11a by 180 degrees so that the lenses are side by side). This would be OK, except the bracket would be more complicated and it will be an annoyance when it comes to changing parameters in the upside down camera. Viewing the two LCD panels side-by-side in the upside down configuration would actually work well.
TEST RESULTS FOR THE DIMAGE 7: a) Manual Exposure Mode M is the most consistent. Lag can be as low as 1/250 or as high as 1/20'ish. See the technote Triggering the Dimage 7 for details. b) Aperture Priority. Lags similar to a), but oftentimes much longer because of differences in setup times. d) Auto Mode. Lag < 1/140 on shots that both cameras fired. Many times sync was non-existent. Conclusion: 1) Provided both cameras lock in their exposure and focus, the sync can be good, but is only guaranteed good in Manual mode M, and then the sync is 1/20 or better. 2) Exposure seek and lock in can be problematic on a moving subject. M mode is most reliable. 3) Flash with the internal strobe of one camera never worked, regardless of exposure mode. Apart from cost, and difficulty in close-mounting these cameras, a major problem with the D7 is battery life. In all my testing over two days, I never got more than 25 shots out of a set of fully charged 1600 mAh NIMH batteries. This means carrying around multiple sets of 8 heavy batteries in the field.......Ugh. NIKON COOLPIX 5000 (See Technote: Triggering the CP5000, for detailed results) The Nikon Coolpix 5000 is a 5 megaPixel camera introduced in December 2001. It has several good qualities for 3D: small size, 28mm equivalent wide angle, good battery life, good noise reduction, recallable user settings, and generally beautiful images. There was much anticipation of this camera, and some people were disappointed by the image quality in initial postings. There are by now many reviews out of the general photographic character of the CP5000 (see www.dpreview.com, for example), and these, as well as my own experiences, suggest this to be a high quality camera capable of generating really nice 16 x 20 prints. For 3D applications we again are most interested in: What is the shutter sync? How can the cameras be mounted?
Figure 3.14 shows the setup for two CP5000's. These cameras have a remote port that uses a serial protocol. The remote controller contains a microcomputer that sends serial commands to the cameras in parallel. We use a Digisnap controller (about $150 from www.harbortronics.com ). The output from this controller is split and goes out through two serial cables to the two cameras. One can easily make a compact cable to do this, but for convenience we used a StereoSnap adapter (also about $150 from www.harbortronics.com). The digisnap controller has the nice feature that the cameras can be placed in the "preset" mode where the focus and exposure is frozen. Then when the preset button is released the cameras fire (because the firing serial code is sent simultaneously to both cameras). How well does this work?
Summary of results:
This strange behavior is described in more detail in the technical report Triggering the CP5000 . In a nutshell, the camera can be very accurate or only moderately in sync depending on when the input serial port registers are polled by the CPU in the cameras. This polling varies with startup phase, and with overall time because the master oscillators in the cameras will always be slightly different. This camera system is close to being workable for 3D, but is not quite there. The digital camera field is evolving rapidly. Can I recommend this combination for stereo now? Yes and no. If you are doing landscapes, the sync is probably good enough to stop leaf and cloud motion (at the wide angle setting). If you want to do macro with flash, and action scenes that might require freezing the motion of something like balls or water drops then the CP5000's will work only some of the time. A WORKABLE DIGITAL 3D (Aug. 28, 2003) Rob Crockett discovered that you can parallel wire the switches on a Sony LANC remote controller to effectively synchronize the Sony 5Mpix V1 camera. Following this, I successfully re-wired a pair of 2MP waterproof Sony DSC U60's. Ninety percent of the time, after powering up the cameras simulataneously using a parallel wired ON button, the cameras trigger within 1/500'th second or better. Here is a picture of the rig:
A Detailed description of the wiring project, along with quantitative test results, is given in the Technote "DSC-U60 Twinning for 3D Photography".
CAMERA SUPPORT NOTE: THIS MATERIAL LARGELY SUPERCEDED BY THE NEWER MORE EFFECTIVE LANC-SONY DIGITAL TWINCAM.
Although it is possible to buy a twin bar, because of the flexibility I wanted, I decided to make my own. I got a piece of square aluminum channel from the local hardware store and milled slots in the back. The channel was cut to 18 inches, to permit a fairly wide stereo base separation. Actually I have made several of these, from little short ones to a five footer. A tripod quick release mount is attached to the bottom of the bar. so that a trigger grip or a tripod can be easily attached. The cameras are bolted to angle brackets as described in figure 1.4 ff. Vertical slots are cut in the center of the angle bracket flange that is not attached to the camera. It is important that the angle brackets are as parallel as possible to the film plane, so that the cameras will be pointed perpendicular to the bar. I used a small amount of silicone rubber cement on the bracket, then I mounted the bracket with the glue on the camera and screwed it down lightly. Now put on telephoto lenses (135mm say, though 50mm will work). The cameras are placed on the bar and twisted slightly until they point towards the same infinity object. Use the focus grid or viewfinder reticule to check for parallelism with the cameras mounted on the bar. After the glue dries take the cameras off the bar and tighten the mounting screw securely. If you don't like the idea of glue, use double sided sticky tape, but hold the camera off the tape until the infinity point is matched up properly in the viewfinder. Attach the cameras to the bar using large brass wing-nuts soldered to 5/16 thread rod for easy adjustment. The slider arrangement permits quick removal or repositioning of the cameras.
|
|||||||||||||||||||||||||||||||||||||||||||
| Return to 3D GUIDE |
