|
A SYNCHRONIZED SONY V1 DIGITAL STEREO TWIN John Hart Program in Atmospheric and Oceanic Sciences University of Colorado Boulder, CO 80302 nimbus.colorado.edu/hart/science.htm Dec. 10, 2004 Updated Nov. 14, 2004
The SONY V1 has arguably the highest image quality of the 5 megapixel "consumer" digital cameras available at the time of this writing. For example, calibrated measurements by Digital Photography Review ( www.dpreview.com ) suggests that the V1 sensor and associated electronics generates about the lowest image noise of all 5MP consumer still digital cameras. Using the highly integrated and unique "LANC REMOTE CONTROLLER" designed and built by Rob Crockett, a stereo twin-camera, useful in a wide range of applications, has been constructed. The twin uses the horizontal format and has a minimum separation of 3.1 inches (78mm). Details of the mount are given, along with results from synchronization tests with the "spinning wheel". 3D examples obtained with the use of an external flash are included. References: Digital Photography Review of the Sony V1: http://www.dpreview.com/reviews/specs/Sony/sony_dscv1.asp http://www.steves-digicams.com/2003_reviews/v1.html Description of Rob Crockett's Controller (contact him if you are interested): http://www.ledametrix.com/lancshep/index.html
Front and rear views of the inverted, horizontal-format camera mount. Separation from 3 to ~6 inches. The controller is contained plastic housing that serves as a trigger grip.
A window gives continuous information on the degree of shutter synchronization of the two cameras. The control buttons are (top to bottom): preset-fire (the shutter), zoom-tele, zoom-wide, power, time-lapse/self-timer. Basically, you power the two cameras up together using the remote by pressing the lower white button. Most of the time they come up very closely "in-phase". The shutters then go off in-sync to within better than 1/1000'th of a second. As time evolves the sync errors grow slowly as two cameras' master oscillators drift slowly apart. Depending on temperature, the sync error reaches about 1/250'th in 5 to 10 minutes (see data below). The drift rate depends on your cameras and their temperatures, with drift slowing (typically) at a lower temperatures like 50F. Occasionally, the power-up leads to only modest sync (like 1/70'th sec). If this is acceptable, go for it. If not, you must hit the power button to turn the cameras off (~4 seconds), then press it again to restart (another 3 seconds or so). Very rarely does it take more than two tries to get 1/1000'th sync levels.
Disassembled parts for the bracket (L). The only critical part is the spacer tube (3/4" ID, 1/4" OD hole, length 2.56 inches). The parts are cut out of angle aluminum. A hack saw and coping saw, and files, can be used, but a bandsaw and a belt sander make it faster to cut out the parts. Clean up the parts with a Scotch-Brite pad. The alignment surfaces are the back edges of the two holding angle bracket bars, opposite the flanges. The top and bottom parts are cut from 2" x 1/8" thick angle aluminum, usually available at True-Value type hardware stores or Home Depots.
In order to get as close separation as possible, a Philmore right-angle sub-mini jack (model #70-093) was modified in order occupy as small a horizontal space between the two cameras as practical (left picture). Taking the shroud off, filing down the pins as low as possible to still permit soldering, attaching flat cable, and then applying a thin coating of epoxy, reduces the jack thickness to less than 1/4 inch. These jacks go into the two sides of the V1's that face each other. Self-aligning slider bars (right picture) made out of 1/8" thick x 1.5" aluminum angle were attached using the tripod socket and a spot of glue to keep the bracket from twisting. The angle is cut down in a band saw and trim sanded. The small overhang preserves alignment on the back (pre-milled) sides of the mounting brackets shown above. The screws are brass flatheads that have been sanded down and set into countersunk tapers in order to sit below the surface and slide smoothly in the slots of the mount plates.
PERFORMANCE - SYNC DRIFT
Shutter lag is displayed in Crockett's panel (pre-production version shown here). What is actually measured is not the shutter opening lag itself, but the lag of the video sync between the two cameras. This is directly related to the phase of the camera operation cycle. Previous measurements, as well as those reported below, have shown that the actual shutter lag closely follows the operation-cycle (or video) synchronization. Direct shutter synchronization tests were performed by shooting a 4 revolutions-per-second spinning wheel. With this it is possible to measure the sync to within about .0005 sec. In my simple tests, I reset the cameras using LANC "power-up" button, then made a plot of indicated sync time (on the LANC's LCD panel), along with the measured sync (actual, as a comparison of the spinning wheel pictures showed), versus the time since the cameras were powered up.
Sync-wheel images from the two cameras. Exposure 1/1000'th, ISO 800, VGA 640x480 resolution. The wheel spins at 4rps, and the interval between the lines is 1/64 of the circumference (or since the wheel turns at 4rps, an interval is 1/256 = .0039 sec). Measuring the difference in the position of the markers gives the sync error. Here it is about .1 division (+ or minus .1, or .2, or so). Thus, the shutters have fired within about 1/2560 second of each other.
Test results from time of camera power-up using the LANC pulse-reset button. The horizontal axis shows time after reset, while the vertical axis show shutter lag (actual = blue, LANC reading = red). Very high performance is found over the first few minutes. Even after 10 minutes (600 sec) the sync error is only a little over 1/200'th of a second. In actual operation, after about 3 minutes, I shut down and restart the cameras using the power button (unless they turn themselves off if not being used, whence all you need to do is restart them). These data were taken at T ~ 80F. At a temperature of 50F the drift rate slows to roughly half that shown above (i.e. ~20 minutes to get to error of ~.006 sec). THE SYNC DRIFT ONE GETS DEPENDS ON THE INDIVIDUAL CAMERAS Given that the actual sync tracks the LANC output fairly well, as shown above, it is easy to check the sync drift just by reading the LANC. I have done this with six Sony V1's, labeled 1 through 6. The above graph comparing actual to LANC-reading synchronization was made with older (Nov. 2003) cameras, numbers 1 and 2. For these inter-camera tests (made in Nov. 2004) the room temperature was 65F and the cameras were set to the A mode. Here are the results:
Cameras 3 and 4 and the pair 5 and 6 hardly drift at all! I was so surprised I made several readings and averaged the results. The serial numbers of cameras 3 and 4 are not close (497xxx vs. 474xxx). The results are approximately commutative (as expected). For example, since cameras 1 and 3 and 1 and 4 don't drift quickly, neither does 3 and 4. If you like, camera 2 might be loosely considered to be a "lemon" w.r.t. 1, 3 and 4, but works well with 5 and 6. This is a small sample of cameras. I don't think this reliably says if you buy 3 you'll get two that match, but...... There is a loose suggestion from the data that the cameras come in two flavors (i.e. 1,3,4 vs. 2,5,6 ). If this is rigorously true (??), if you did buy three your chances of getting two that match well is 100%. This is a remarkable conclusion! FLASH Previously it has been somewhat difficult to run flash with twinned stereo cameras (either film or digital). In fact, reliable external flash with digital twins has not been reported previously (to my knowledge). With some care the Sony V1 Twin-Camera functions quite well with an external flash.* For external flash, both cameras are set to "hot shoe on" and "flash enabled". A single flash is attached to the hot shoe through a cable (which is better than mounting a heavy flash on the camera itself and risking bending the alignment). The external flash cable should be attached to the M camera. The M camera is, by definition, attached to the controller cable coming out nearest the row of buttons. But which camera do you use? The "M camera" (i.e. the one connected to the cable coming out of the LANC closest to the button row) is chosen to be the slower (i.e. the "laggard") of the pair. That is, connect the cables so that the sync errors are denoted by M on the controller panel. For example, the display should initially show something like M 0.1 msec , then grow slowly with time to M 0.2 msec , and so on. Again, both cameras need to be "Set-Up" with hot-shoe on (this disables the internal flash). ALL FUNCTIONS ON BOTH CAMERAS SHOULD BE IDENTICAL (mode, focus type, zoom, flash on, ISO, WB, etc. etc.). For more information on flash, see Rob Crocketts LANC USER MANUAL.
Unit with external Vivitar 283 flash attached. For sync errors (as read out on the LANC controller screen) from ~ 0 to ~5 msec or so (i.e. for about ten minutes of shooting without a new power-up reset) the flash can be used. Eventually errors get large enough that the slave camera (the one without the flash cable) doesn't see the flash at all. The A (aperture preferred) mode seems to be most reliable, but I had success with S (shutter preferred) and M (manual) modes as well, although for these latter modes there would sometimes be a mis-fire (black field in the camera not attached to the flash). For some strange reason the actual shutter lag (not the sync, but the time between pressing the trigger and the whole rig actually firing), seemed to be noticably longer in the S and M modes, compared to the situation with no flash at all. A guestimate of a delay of about .3 to .5 seconds for M and S modes suggests some possible issues in critically timed flash shots. Note: I don't particularly care for the internal flash. It's pretty wimpy and the TTL - pre-exposure routine dramatically slows the shutter response (i.e. increases the shutter lag between trigger and shutter release release). EXAMPLES (click on thumbnail for 3D view along with a description of camera settings)
A few not-so-great flash shots. But they show that flash works.
Now this is a little neater. New England boiled dinner. CONCLUSION: This system probably represents the potential for digital-3D still photography at its current best: High-quality images, wide operating range, excellent synchronization, external flash, ........
|
