Measuring Twin Camera Shutter Synchronization


The Issues: With two digital cameras set up to take stereo photography, how can I measure how closely synchronized the shutters will be? Very little of my photography is still life, and I need to be able to accommodate moving subjects. If the shutters do not fire at the same time, disparity from motion between the stereo images can spoil the stereo effect. A closely related issue to shutter timing with twin cameras is flash photography to supplement existing lighting. And finally, I am also keenly interested in predicting, prior to the exposure, how closely the shutters will fire.

Digital cameras have complex timing issues unique in the field of stereo photography, as previously discussed. However, I've been able to resolve for me many of these issues with my twin cameras simply linking together external wired remote controls, allowing me to power up the cameras together and trigger the shutters electronically.

Power Up Sync Data: Using the remotes, I powered up twin Sony DSC-V1 cameras 25 times, and used the Sync Shepherd to sample the video output signal sync. The measurements show that almost every time (86%) on power up, predicted shutter sync would be within 1/3 msec or about 1/3000 sec. I've noticed that the probability of a power up close sync degrades when the batteries get low, but it either powers up in close sync or really out of sync. Importantly for an individual trying to test power up sync, this power up sync only works on battery power. When the cameras are plugged to an external power supply (the battery charger for example), power up sync is completely random.

Data as HTML.
Data as Excel worksheet.

Run Out Sync Data: Starting with a power up, I again used the Sync Shepherd to sample the video output signal sync. The measurements show that in my twinned cameras, the cycle runs for about 58 minutes from perfect sync to completely out of sync (180 degrees), and a complete cycle would then take about two hours (back into sync). It also suggests that if I power up the cameras together and start with near perfect sync, I can take a stereo picture out as far out as 3 minutes and may still get better than 1/1000 sec sync.

Data as HTML.
Data as Excel worksheet.

Shutter Sync Data: I did some initial measurements with a digital stopwatch (poor man's digital laboratory chronometer, picked up at a local drug store for about $1.99), and the exposure interval certainly was better than 1/100 sec after powering up the cameras together.

So I set about collecting some more detailed data. I powered up the cameras together 10 times, and for each power up acquired about 5 twin exposures (after a focus lock) along with Sync Shepherd measurements, for a total of 50 trials.

The twin exposures are of my CRT computer monitor while it is displaying a graphic with graduations, with the monitor set at a 75Hz refresh rate (a flat panel/LCD monitor would not work). SYNCSCREEN.JPG:1024x768 JPEG white screen graphic with graduations for sync measurements. I displayed this using the built in Microsoft Windows slideshow feature in "My Pictures" of "My Documents". I selected a screen refresh rate of 75Hz in WindowsMe: Right click desktop/Properties/Settings/Advanced/Adapter/Refresh Rate.

The risk of using a computer monitor to measure sync is that a wildly out-of-sync exposure can be accepted as close sync because of aliasing (every displayed screen refresh 75 times a second looks the same). If you read the references above, you might understand that at these close shutter sync times, the expected out-of-sync anomaly related to the unique digital camera timing issues would be 1/30 sec. I chose 75Hz for the computer monitor refresh rate because a 1/30 second addition to the interval between the exposures would produce a 1/2 screen change, a change that would be not detectable at refresh rates that are multiples of 1/30 sec (60Hz and 120Hz). That said, the probability of one of these 1/30 sec "flyers" is still less than 2% anyway over the range of sync intervals that I measured. I guess I would feel better about the data if I had seen a "flyer", but I didn't.

I set the shutter speed on the cameras at 1/500 sec. The measurements were taken at the upper sharper edge of the illuminated portion of the screen (at the point of shutter opening). The display program mentioned above has a zoom feature that allows very close examination of the 5 megapixal images I acquired.

I did have to discard two trials because one of the twin exposure lines was off the screen (As I measure it, 4% of the screen refresh time is used to reset the raster beam from lower right corner back to upper left corner, making the screen graduations actually total 10.4 rather than 10).

Complicated? Yep, but fortunately, the data collection is simple, and only cocktail-napkin calculations are needed to analyze the data. I can hardly believe the data. The longest interval I measure between the start of two exposures is 0.55 msec or 1/1800 sec, and most are much shorter. In addition, the Sync Shepherd predicts very closely the shutter exposure interval.

Data as HTML.
Data as Excel worksheet.

Flash Sync: Flash sync is shutter sync with a slight twist. The flash occurs at the start of an exposure just after the shutter opens. To prove this to myself, I put a motorcycle on a stand, put a thin strip of reflective tape at intervals around the tire, and had my daughter spin the wheel clockwise while I tried to take a picture of the tire at 1/30 sec while blocking half from the flash with a card. The photo shows the sharp flash reflection at the very origin of the more vague blurred remaining exposure. From measurements of flash and camera video cycle sync using exposures with the Sync Shepherd connected, the flash seems to occur with these cameras about 0.35 msec or 1/2800 sec after the shutter opens.

Both cameras have to be set up the same way, so that exposures are identical, but one of the flashes must be blocked, so that neither exposure gets two flashes. The word on the street is that if one were to tape over the blocked flash, the flash could be damaged by the excess heat. I have blocked the flash with my finger, and I can attest to the impressive heat deposited by the flash, so I made a flash diverter, a little black card that I tape diagonally in front of the flash. For this discussion, the "Master" flash is the one not blocked, and the "Slave" flash is the one blocked. So the key to getting the master flash to appear on the slave exposure is to put the beginning of the master camera exposure somewhere within slave camera exposure. As the exposure sync time lengthens, the camera with the slower governing internal electronics must be the master with the flash (so that the first part of the slave exposure will occur before the first part of the slower master flash exposure).

I trained my twin cameras on my computer screen once again, this time at a vertical refresh multiple close (60Hz or 120Hz) to my cameras NTSC video rate (29.94 Hz), and the camera shutters set at 1/100 sec or faster. (I could have used a television also). The computer is slightly faster than the cameras, so the illuminated band rolls downward on the screen. Observing the illuminated band on each camera LCD monitor shows that the illuminated bands slowly diverge. Here however, the camera with band moving downward relative to the other is the slower camera, the master, and the one that needs to have the flash.

Flash sync with the Sony DSC-V1 cameras is complicated a little by a "pre-flash" that allows the camera to fine tune exposure with the main flash--each main flash is preceeded by a mini flash and the brightness of the miniflash image is examined by the camera. The pre-flash sync window (where the master pre-flash brightness is examined by slave camera sensors) seems to be quite wide, and is lost only slightly (0.1 msec exposure sync) before the main flash sync is lost. The improper adjustment made as a result is to the intensity of the slave flash that is blocked, but the camera may not take an exposure at all when it gets this invalid reading. If it does take an exposure, the slave flash ramps up to its maximum, producing an impressive "POP" as the slave flash pumps out photons for all it is worth into the flat-black card I taped diagonally in front of the flash.

Anyway, the flash sync behaves otherwise just as one would expect with this camera rig. If the camera video/interrupt sync time is less than 0.35 msec, the master flash always shows up on the slave exposure, the 0.35msec being the time from the master shutter opening to the master flash (see above). To considerably extend the flash sync window, one can make the camera with the slower internal electronics the master with the flash, and the flash sync window extends out almost to the length of exposure (at 1/500 sec exposure, the camera video/interrupt cycle can be out of sync as long as 1.2 msec or 1/800 sec and still get perfect flash on the slave exposure. One would think that longer exposures (say 1/30 sec) would just make the video/interrupt sync even less of an issue--it does so by making flash sync possible while further out of video/interrupt sync--but one will start to see more shutter sync "flyers" as the video/interrupt sync gets further out, and flash sync will become less reliable as the video/interrupt sync time increases. With a synchronized power up and both cameras set up the same, shutter sync is very reliable for several minutes if the open flash is on the camera with the slower internal electronics.

Data as HTML.
Data as Excel worksheet.

Crossed View . . . Parallel View. . . Images open in a new window. 110k.

Crossed View . . . Parallel View . . . Images open in a new window. 75k.

Good luck,
Rob Crockett
Copyright 2003 - 2008. All Rights Reserved.

Digital Stereo Photography Home