Stereo Digital Camera Sync Shepherd


The transition of stereo photography from film to digital has been hampered somewhat by the difficulty synchronizing twin digital cameras. This circuit detailed below samples the video output from each camera, and indicates video synchronization, allowing the user to better predict the degree of twin digital camera shutter synchronization.

Discussion:

Most consumer digital cameras cannot ordinarily be consistently synchronized in a twin stereo rig to better than 1/30 sec. The reason is a little difficult to explain without a lot of charts and hand waving, but a quick experiment with your TV will let you know if this is an issue with your digital camera also.

With your camera set to a shutter speed of 1/125 or faster, take a couple pictures of a television screen, and see if the bright band in the picture is in exactly the same position on the TV screen for each picture. Since computer screens use a different refresh rate than the NTSC standard, you must use a television screen instead of your computer. If your camera does not show shutter speed, try to make it show the bright/dark bands by using "sport" or "action" modes or a bright background.

NTSC video is refreshed at about 30 frames per second. Frames are interlaced though, so the refresh rate looks like 60/sec, since I can't tell the odd fields from the even fields. If I use a 1/125 shutter speed, about one half of the screen shows illuminated raster lines like you would expect:

60 fields/sec * 1/125 sec = 1/2 field

The weird part is that, picture after picture with my Sony DCS-S50, the EXACT same part of the field is illuminated and does not roll. If I restart the camera, the photos show the illuminated portion in a different location, that then stays the same again picture after picture. If I took pictures with a film camera, the parts illuminated would NEVER be exactly the same. Even if I tried, I could never take a film picture twice of the exact same part of the television refresh cycle. This can only mean that, despite variations in flash, focus, and exposure, my digital camera makes the calculations and adjustments, then waits for its current internal video frame to finish, before finally taking a picture.

Since one digital camera video frame will never finish at the same time as that of another camera (same NTSC or PAL cycle rate, different time of cycle end), the best twin digital camera sync I could ordinarily reliably obtain would be 1/30 sec (0.033 sec) if the synchronization of the cameras is not known.

If you do the math, you will see that when you try to take a picture with a stereo digital camera rig, there are two possible sync times (time between the two exposures), each with a certain probability that depends on the degree of synchronization (degree of phase difference between the two video cycles) and on exactly when during the video cycle the request for a picture comes. A complete video cycle with the two interlaced frames takes about 1/30 sec or 0.033 sec. Phase sychronization varies from 0 to 360 degrees as the cameras cycle slowly in and out of sync.

Event Probability Sync Time
1 (360-Phase)/360 (Phase/360)*0.033 sec
2 Phase/360 ((360-Phase)/360)*0.033 sec

Unless you are a math wiz, the equations may not be easy to visualize, so here is a graph to demonstrate. Time or sync phase are on the horizontal axis, as the sync will slowly change over time since the timing reference crystals in the two cameras are not exactly the same. The vertical axis is the difference in time between each of the twin camera exposures. If you keep track of which camera is which (one master, the other slave), and plot (sync time)=(master-slave), the result is the second graph, which agrees well with experiment (John Hart's twin Nikon digital camera sync data) . The slope direction (positive vs negative) in the second graph depends on which of the two camera crystals is faster.

So . . . clearly synchronization of many twin digital cameras can be significantly improved by knowing and allowing for synchronization of the twin cameras' video cycles. You can expect to have a much higher probability that the camera shutters will be sychronized if the camera video cycles are also sychronized. I designed the circuit to show the degree of video cycle signal synchronization.

How it works:

National Semiconductor makes a handy chip, the LM1881 video sync separator, that can pull the sync components from a composite video signal. I use the odd/even frame output from two of these chips to generate 5v 30Hz square waves with a 50% duty cycle for each of the two cameras. A quad NAND gate chip configured as an exclusive OR gate (XOR) combines these square waves to a 5v 60Hz square wave, but now with a duty cycle that is inversely proportional to the degree of synchronization. The duty cycle varies from 0% (fully in sync) to 100% (fully out of sync).

Heavy filtering with R3 and C5 then converts this pulse width modulation to an analog signal, a voltage that varies from about 0.2 volts to about 4.9 volts, again inversely proportional to the degree of sync. Finally, the LM3914 displays this voltage on a dot scale, with the R5 and R6 divider network setting the top end and ground as the low end of the voltage scale. To keep the comparators of the LM3914 from going off the bottom of the volt scale, the R4 pull-up resistor boosts the bottom voltage by about 0.2 volts to around 0.4 volts.

R7 sets the current through the LED's of the display. With R7 of 620 ohms, the illuminated display LED draws about 20 mA, which is about maximum for this display, probably needed for viewing the display outside. The LED display uses the vast majority of the power for this unit, and a with standard 9v alkaline battery, you can expect about 20 hours of operation. If you plan to use the unit inside in lower light, you could preserve the battery by using double the resistance (1.2K ohms) for R7, for 10 mA through the illuminated LED and about 40 hours battery lifetime.

With the Fairchild Dot display, the top green indicates the lowest voltage and the best sync.

Parts List:

Part Value Digikey Number Price Description
R1-2 680K 680KEBK-ND $0.28/5 1/8 Watt Carbon Film 5% Resistor
R3 100K 100KEBK-ND $0.28/5
R4 1.0M 1.0MEBK-ND $0.028/5
R5 680 ohms 680EBK-ND $0.28/5
R6 2.2K 2.2KEBK-ND $0.28/5
R7 620 ohms 620EBK-ND $0.28/5
C1-4,C7,C8 0.1uF 399-2127-ND $0.16 Ceramic Capacitor 50V 20%
C5 1uF P2105-ND $0.42 Tantalum Capacitor, 16v
C6 100uF P904-ND $0.42 Radial Electrolytic Capacitor, 6.3V
IC1-2 LM1881N-ND $2.78 Video Sync Separator
IC3 296-1603-5-ND $0.52 TTL Quad 2-Input NAND Gate
IC4 LM3914N-1-ND $2.91 Dot/Bar Display Driver
IC5 160-1066-ND $2.10 10 Element Red Bargraph Dsply
IC6 5V LM2931Z-5.0-ND $0.86 5v 100mA regulator, TO-92, 0.3 volt drop out
T1-2 CP-1403-ND $0.61 PCB RCA Jack Horizontal Mount
T3 Voltmeter Terminal/Wire (optional)
T4 2238K-ND $0.49 9 volt battery clip
S1 EG1906-ND $0.71 PCB Right Angle Switch SPDP

Construction Notes:

Attach power to the board in the right lower corner. The circuit runs at 5v and the LM3914 needs at least 5.3 volts to regulate at 5v, so you can use the 9v battery clip and 9v battery that will last forever. (The prototype in the pictures used a surface mount voltage regulator instead of the TO-92 specified above). To cut a little weight, you can use a smaller 6v camera battery (Energizer A544 alkaline or L544 lithium), or an even smaller 12v garage door opener battery (Energizer A23)--you can solder directly to the terminals of the battery after sanding the terminals a little, or put a glob of solder on the coiled end of the wires and tape the wire over the ends of the batteries (much easier to change in the field). If Digikey is out stock of one of the resisitors, use a 1/ 4 watt instead, with part number ___QBK-ND instead of ___EBK-ND. If you want the Bargraph display to be Red-Yellow-Green instead of the all red unit in the parts list above, you can order a Red-Yellow-Green Bar/Dot Display from Mouser Electronics (no minimum order) part number 512-MV5A164 (Fairchild MV5A164) for $1.39.

For the video input cables, you can pick up a couple of cables from RadioShack similar to those provided with your camera, cut, strip, and solder them on, or use the RCA PCB terminals in the parts list above connected with your original camera cables. My camera uses a standard 3.5mm stereo jack for video with mono audio at the tip, video in the middle, and ground/shield at the base, but you can test yours with a continuity tester. I used a Dremel motortool to route out two slots along the left end of the board, and used a zip-tie to secure the cables to the board for the prototype boards.

User instructions:

Connect the video output from each camera to the RCA terminals, and turn on the cameras and the unit. Cycle the power on one of your twin cameras until the sync is close enough--each power cycle has a one in ten chance of being in the best sync zone on the display (at the top of the board). Depending on the tolerances of the resistors, the very best sync may be off the top of the scale with no LED illuminated.

If the 10-element display is not enough detail, you can solder a connector or wire to the three pin terminal at T3. The top pin is the signal lead that varies from about 0.35v (best sync) to 4.97v (worst sync), the bottom pin is ground, and the middle pin is +5v if you need to power the meter from the unit's battery. If you follow the voltage, you will see that the LED display is not linear--above about 3.8 volts, the bottom LED stays on, so that the in-sync part of the display is higher resolution. If you have a voltmeter that can display millivolts, you can considerably improve the resolution of your sync observations, to better than 1 part in 1000. RadioShack has a pocket tester, part 22-802, that has millivolt resolution for about $25. Also, you can easily see if the sync is getting better or getting worse-usually the frame rate of each camera is not exactly the same and the sync will very slowly cycle in and out.

Epilog:

This circuit does not in itself sychronize the cameras--it merely shows the degree of synchronization. However, without disassembling the cameras and re-engineering the cameras, this circuit is the next best thing for consumer level digital cameras that have this issue. Ideally, one would select cameras that do not have this issue--the video cycle is not present, or the cycle is appropriately interrupted for the least and most reproducible shutter lag time. For me, the goals of this project have been met--learn about video signals, improve my stereo photography, and generate a little winter evening entertainment.

Update:

With the pair of Sony DSC-V1 cameras, the power up sync is usually (but not always) very close. To show this very close sync on power up, and help me time the absolute best sync, I narrowed the displayed sync range quite a bit by changing R4 to 20M (two 10M resistors in series), R5 to 4.7k, and R6 to 220 ohms. This brings the bar graph display to a very narrow range, covering perfect sync (top of scale green) out to about 5% out of sync (bottom of scale red), for volts ranging from 0.026 volts to 0.25 volts at the T3 terminal. Here is a table showing the relationship of the bar dot scale to sync times with these new resistors installed:

Data as HTML.
Data as Excel worksheet.

One More Update:

To prevent parasitic capacitive coupling, break off pins 1,3, and 5 of each the LM1881N chips before soldering them in (or cut them off after they are soldered in). These pins are not connected to anything anyway, and breaking them off or cutting them keeps them out of trouble. I also increased the size of the power supply capacitor C6 and added a decoupler capacitor C8 to quiet things down and improve reliability. Please note the polarity of the C5 and C6 capacitors in the parts diagram. For the ceramic capacitors (the rest of the capacitors), polarity does not matter.

Yet Another Update: Making a Circuit Board

For the uninitiated, making a circuit board can be trying. If you are interested in making the circuit board yourself (as I did using a photo etch technique for the prototype above), I have some slightly dated general notes about circuit board design and production here Making Printed Circuit Boards.

However, I think most hobbyists are using a local lab or using online services to have the board made to order. For a local lab, one can draw up a computer design using a program such as CadSoft Computer's Eagle Software. For this small project, the Freeware version is more than adequate. To get you started, Manuel Dejonghe sent to me the Eagle Files for his Sync Shepherd Project. "Right click" this link, "Save Target As" to a place on your computer, then "right click" the saved ZIP file and select "Extract All" to uncompress the file set to a folder, then open the files within the Eagle program. The files include a schematic drawing and a circuit board CAD drawing.

Still another way is to design a board using an online service's proprietary software, then have the service manufacture the board based on the board design you create. I have been using ExpressPCB; their software is easy to use, and their circuit board quality is excellent.

Good Luck,
Rob Crockett
Copyright © 2003 - 2011. All rights reserved.

Digital Stereo Photography Home