Wonderful control for water drop photography

I use the Cognisys StopShot timer, which provides the proper timing, but control needs more than a timer. It also needs something that can be controlled. I discovered a solenoid valve for water drops which actually provides the precision control expected (not all do). It is day and night better than what I was using. The first few minutes with it made the difference be immediately obvious, a real WOW! experience, very memorable and pleasant. Perhaps this article can provide first day hints for those interested in getting into water drops.

I ordered this Shako PU220AR solenoid valve from APEX Industrial Supply, San Diego CA, call Sales at phone: 1-619-661-6200. They were referred to me by Shako Taiwan. This may take two or three weeks if not in stock, but they will order it. Option with 24 volt DC coil and 1/8 inch NPT treads (Stopshot is rated to 40V DC). $32.50 plus $10 shipping.   Shako PU220AR spec sheet. (the NPT it mentions is an American pipe standard, National Pipe Tapered threads.) This solenoid valve actually provides the precision that I imagined was possible.

It will need power too. I used a Mean Well DR-30-24 power supply, 24 volts DC, $33 (surely is great overkill, but it is very well regulated and protected - very much more than just a power cube). But any 24V DC power supply will work, or you could use a 12V solenoid.

The DIY solenoid power cable (cable wiring diagram) used 10 foot of 18 gauge lamp cord, which is larger wire than this needs. One wire is cut near midpoint (maybe at 40%) so StopShot can be the activation switch there. I spliced in a RCA cable from an old TV video cable into it for the StopShot activation - or an audio cable with RCA connector works, it just needs the RCA connector on it. And I made a camera shutter cable from the other end of that old audio/video cable. You probably want to add a disconnect connector near the valve - this one was a Molex connector from an old Y splitter for PC disk drive power. This vertical mounting bar at right is cut from the 3 foot bar offered as a mounting accessory for the power supply (the bar is inexpensive, but it runs up the shipping cost, but there's no other way to mount it). This power supply is small (about 3 inches wide, and a picture is on the cable page above), and it could sit loose, but the only mounting provision is that optional bar.

The nozzle seems pretty important too. The output nozzle used is simply a standard 1/8 inch NPT fitting, barbed to accept 1/8" ID tubing. The threads are 1/8" NPT which fits the valve choices, and the barbed size varies for different tubing sizes. The local stores (and not even Home Depot) don't have much choice in the small 1/8" NPT thread fittings (to fit the valves), but just search for 1/8" NPT Barbed hose fittings. The 1/8" NPT is the thread, and at the other end, I also like the size fitting 1/8" tubing. These are available on Amazon, search for 1/8" NPT barbed fittingm but you may have to buy more than one at Amazon. My original source was about $2 each (search there for 1/8" tubing part number 48755003). Also 3/16" tubing is 48755060, 1/4" tubing is 48755128, and 3/8" tubing is 48755300). These are adapters to barbed tubing sizes from a 1/8" NPT threaded fitting. Most are brass, but polypropylene fittings are less than $1. As nozzles, they only need to be finger tight. The thought occurred to me to try using the value without any nozzle (the valve has its orifice), but I never got around to trying that, I was pleased with the nozzles.

Two barbed sizes (all 1/8" NPT threads) were most useful as nozzles: The 1/8" ID tubing size (nozzle is 0.093 inch ID, about 1/10" ID) used for drop pulse size 7 ms to 30 ms. The 3/16" tubing size (nozzle is 0.125 inches ID) for drop pulse size 25 ms to 55 ms. I normally use the smaller one. Larger nozzles or larger drops become a little more messy (some extra drops, maybe this needs a larger valve orifice??), however a 1/4" tubing size nozzle (0.187 inch ID) can be useful for a 50 to 70 ms drop pulse. At left is a 1/8" fitting that I cut off to be half length, which works the same, and seems of no advantage. I also tried drilling a 1/16 inch hole in a 1/8" pipe plug, but it saw little use. See page two here about messy/clean drops.

Pressure through an orifice area for a duration produces a certain volume of flow, a drop size. For clean drops, we must find the right combinations of head pressure and duration pulse to be appropriate for our nozzle size. To show this concept of clean and messy drops, and the effect of pressure head and pulse duration, more combination results (different than the link just above) are at:

Shako PU220AR solenoid valve 1.5mm orifice. (valve details Source)

STC 2P 025 solenoid valve 2 mm orifice. (valve details Source)

A Mariotte siphon is a simple device to keep water pressure constant as the container empties. The StopShot Mariotte Siphon needs the 3/8" fitting on valve inlet, for 3/8 inch ID tubing. Home Depot has the larger fitting, but not the smaller ones in 1/8" NPT size. The valve needs two 4 mm metric machine screws to mount it (Home Depot has those, in one of the special flat drawers). The valve's inlet NPT threads must be wrapped with a couple of turns of teflon tape to seal it (output nozzle doesn't need it, it is only finger tight). The little Giottos mini ball head (MH1004, $15 USD) is to allow quick leveling of the sensor, which initially seemed a dumb notion, but it works really well (it has standard UNC 1/4" x 20 thread at both ends, screw and nut). The smaller StopShot sensor (shown) is much more convenient for water drops than their large sensor, but both work. It is mounted with a couple of standard mending brackets from Home Depot (3" T plate to mount sensor, bent). Shown is about 9 inches between valve and sensor. A fall height of 18 to 24 inches (nozzle to water surface) works well. A second bracket allows quick swap to another valve, for example to the original Stopshot solenoid valve (which I did not like its performance. I heard they provide a STC valve now).

My setup uses two drilled 2x4 boards over a 1/2 inch black pipe (see Here), on a 2x2 foot sheet of 1/2 inch plywood (available off-the-shelf pre-sawed at Home Depot). I use a pipe flange screwed to base (about 6 inches from one corner). I used T-nuts below the board for the screws, and four more T-nuts near the corners for screws to level the base (levels the pan). These leveling screws sit in smaller cup washers, but do protect the table under them! The pipe is stable enough against vibrations if you move the StopShot timer off to a different table, so your finger does not shake the pipe (or the water). The 2x4 boards are strong and stable and convenient, except not level, so the ball head makes it be perfect.

Flash setup here. To photograph clear water, you can't light the water directly, you must light what is reflected in the water, so I use a frosted Plexiglas panel as a background. Two Nikon SB-800 flash with color filters are behind that panel, aimed low on the water, more or less reflecting into the lens (best bet is angle of incidence equals angle of reflection). The flash can reach surprisingly far this way, maybe 3 feet from splash for two flashes at 1/64 power (about 1/25000 second duration), to use f/16 ISO 320.

Of course, the speedlight flash is what stops the fast motion. The point of this operation is that very low flash power (1/32 to 1/128 power) is extremely fast to freeze the fast motion of the splash. This is why they are called speedlights. The lower the flash power, the faster it is, which is all the better. See the flash durations in the spec chart at the rear of the Nikon flash manuals. Then the precision timed solenoid valve allows control to make the splash happen, as desired.

So the idea is, now we can just dial in the specific delay required to give the desired result.

Some tweaking is always expected, and to show the control possible, the adjustment of Toff (Time Off, the time the valve is closed between two drops) was tweaked in each picture, from 65 ms to 60 ms, in 1/2 ms steps (reducing the time between the two drops). This simply lowers the second drop down onto the first drop as desired. This is control, awesome control and repeatability, like you would expect precision timing ought to work.

These are eleven sequential pictures, consecutively numbered, not otherwise selected at all.

65 ms Toff. This adjustment simply lowers the 2nd drop onto the column.

64.5 ms Toff

64 ms - No difference, which could be my error, or maybe it is not 100.00% perfect every time.
But this is 1/2 ms, and it is mighty good!!

63.5 ms Toff

63 ms Toff

62.5 ms Toff

62 ms Toff

61.5 ms Toff

61 ms Toff

60.5 ms Toff

60 ms Toff

Following is maybe something you might want to know...

Shutter lag time

Cameras have shutter lag time, the time delay before the shutter actually opens. I use 116 ms for a Nikon D70S and I use 54 ms for a Nikon D300 (StopShot can measure this lag time, and I include a millisecond or two safety factor).

Extremely important: Auto Focus must be Off, and then this lag timing assumes the rear LCD is Off, and the viewfinder is On. Otherwise the lag time is longer and variable, and you will miss the shot (a black frame). A shutter button half press will put them into this Ready state (until viewfinder timeout). You could instead set the camera menu for rear LCD always Off and viewfinder always On, but I choose to use the LCD video output to the television to monitor the shots. Mirror rise time is not a factor, it occurs during shutter lag time (locking mirror up does not improve shutter lag time). I do not bother with mirror lockup, since the purpose of the fast flash is to freeze the motion, both splash and camera shake.

One simple way to do high speed photography is to open the camera shutter in Bulb setting and hold it open, and then start the action, which triggers the flash to capture the motion, and then the shutter is closed manually. The shutter will be open a second or two, so background needs to be dark (the fast flash is what stops the motion). This Bulb shutter method seems necessary for sound action like breaking water balloons or bullets, since the trigger does not happen until the action happens. Shutter lag time makes anything but Bulb shutter (already open) impossible. So for that, we seek near zero delay from the sound trigger. The speed of sound can be a small delay, moving microphone slightly farther from action. Sound travels roughly about 1.1 feet per millisecond, which is a trigger delay.

But water drops allow positioning a sensor under the nozzle, to trigger the action long before the drop splashes on the surface. Bulb shutter can still be used, but StopShot can control the camera shutter, and if carefully set up, we can use 1/100 second shutter speed instead of Bulb. This allows working in a bright room, instead of a dark room. For these pictures, the camera was f/16 1/50 second (Stopshot can control the shutter open time in Bulb, or I simply used the camera's 1/50 second setting). Start slower though, and work up, because the 1/50 second shutter is a 20 ms window, following lag time, during which the flash must trigger.

For StopShot in Sequential mode, there must be sufficient spacing between valve and sensor so that all drops can exit the nozzle before the first drop enters the sensor - else a long drop sequence can miss the first at the sensor, and trigger on the second drop. My sensor started at only 5 inches below the nozzle, which could be a problem with longer pulses and longer intervals, so I increased it to 9 inches (orifice to sensor). That will just handle 2 drops of 30 ms and Toff of 210 ms, and thus most cases of two drops.

My flash is two Nikon SB-800, at 1/64 power, maybe 1/25,000 second duration to stop the motion. Lens is a Nikon 105mm micro, on a D70S body. An AC power supply for the camera is a good thing for this. So the camera has three cords from it, the AC power supply, the shutter cable, and the television output cable. The flash is instead connected to the timer.

StopShot has three timed outputs

I use output 1 to control the solenoid valve (Manual mode), so that it is triggered when I push a button. I must ready the camera lag time sync first, manually, with shutter half press. So first one hand resets the camera, and then my finger button triggers the timed drops from the valve (in the next several seconds before viewfinder timeout). The first drop eventually triggers the sensor under the valve. I call this setting M1 (Output 1 in Manual mode).

I use output 2 to control the camera shutter (Trigger mode), following X ms delay after the sensor trigger is detected. This is a shutter head start, to allow for shutter lag. Technically the picture must still wait for the flash, but practically, this X ms adjustment is what is used to adjust the phase of progress of the first drop splash. I call this setting T2.

I use output 3 to control the flash (Delay mode, with Sync Yes), triggering it after a Delay from the camera shutter, to provide sync after shutter lag. This is a constant. For the D70S, this is a fixed 116 ms to wait for shutter lag. For the D300, it is 54 ms. I call this setting D3. This flash trigger must occur during the window (duration) of shutter open time (so if you add to this lag delay, you may need a longer shutter speed with it).

The water drops are created by M1. There is a sub-menu for number of drops, and the drop's duration called Pulse, and the drop spacing called Toff (time off when valve is shut between drop pulses). The delay time from sensor to flash is T2 plus D3. T2 is the adjustment to modify fall time of first drop, and thus the splash progress. The camera shutter lag begins at T2. D3 is a constant for the camera shutter lag to trigger the flash after the lag. So first drop timing is adjusted by T2, but actual flash delay is sum of T2 and D3.

The setting for this series was:

M1 - pulse 24 ms, 2 drops, Toff around 60 ms (see pictures). A shorter pulse will create a thinner column, usually preferable. BTW, if the drop were dyed green (this one is not), the early splash crown will not be green, but the top of the column will be green. So the valve was open 24 ms, closed 60 ms, and then open again for 24 ms. The closed "Toff" time controls the spacing between drops. A shorter interval lowers the top drop (less delay), which controls the state of progress of the actual collision (see pictures at left).

T2 - 137 ms. The first touch of the drop on surface was at 58 ms (experimentally seen), but the suitable rebound column was 137 ms (again, adjusted for effect desired), which (here) was 79 ms after the first touch. So the interval between drops should be such that the second drop arrives here about then too. T2 is the delay after the drop in the sensor was detected, until the shutter was clicked and started. The flash follows the shutter, so this adjusts the flash relative to the first drop (controls the phase or stage of the first drop as seen in the picture).

D3 - 116 ms - constant for D70S shutter lag, time the flash is delayed after the shutter. D300 is 54 ms. Sync YES in this step.

So first surface touch was T2 + D3 = 58 + 116 = 174 ms, and appropriate column rebound was T2 + D3 = 137 + 116 = 253 ms, or 79 ms longer than first touch. See math at bottom, but Toff for a collision will be a bit less than this (subtracting fall time to surface), around 60 ms here.

This will all be confusing the first day, but it quickly becomes obvious and easy. Above is hints for first day. The Nikon D300 will do 54 ms shutter lag, and it also outputs HD to TV via HDMI, but I've been stingy with its shutter count while just playing. :) To swap cameras, just change D3 for appropriate shutter lag, and change T2 accordingly, to have the same T2+D3 total.

With such a setup, you can get splash after splash after splash, just like you want. Extremely consistent timing, except the splashes do vary slightly, no two alike I guess. Now and then one will miss and fail, but it's relatively rare. More often, I screw up readying the sync, and then get a black frame. After setup, success rate is probably better than 95% (virtually perfect, with a rare exception or two). All are good pictures, but keeper rate becomes vastly lower as your standards increase to those few which "stand out". But it is trivial to create collisions, and repeat it many times. No more need for luck to get anything at all. Frees your time to be working on the lighting and such.

... if you have a good valve.

Nozzle philosophy

I am bewildered about nozzle design. It is very clear however that this valve and nozzle are much superior to other choices. In this case, a large drop remains in the nozzle at all times, and it falls out as you unscrew the nozzle. There must be a mild vacuum created in there as one drop exits, holding it in. The first one drop can vary a little if it has sat idle awhile. But this valve has a very strong "one drop in and one drop out" situation, actually related to the valve timing.

Some math to estimate time between drops for collisions (different numbers than the case shown):

Galileo's Two New Sciences (1638) said falling bodies fall this way:

Distance = 1/2 at²

We have perhaps 1/2 PSI head pressure (roughly 1/2 PSI per foot of head), so our initial exit velocity is not zero, but pretty close.

A nozzle that is 26 inches high (26/12 = 2.17 feet) computes square root of (2 * 2.17 / 32) = 368 ms to fall 26 inches. The sensor at that time was only 5 inches below the nozzle, which computes 161 ms to fall 5 inches. So the sensor to surface time is 368 - 161 = 207 ms. This is ballpark - the nozzle will slow it, and pressure will increase exit velocity to speed it.

But it is a close starting point, and suppose we experiment, and discover (photographic results) that the first drop just touches the water surface at say 210 ms from sensor (to the flash). It will take the second drop this long too. Same ballpark.

Then we experiment to see that maybe an additional 86 ms (296 ms total, sensor to flash, T2+D3) gives a strong rebound column like we want for collisions (this depends on things, fall distance, water depth, surface tension, viscosity, etc). Therefore, we specify two drops, and then 86ms Toff between drops will cause a second one to be right there at that time. One drop rebounding up for 86 ms, and the second drop 86 ms behind the first drop will collide then. Almost... Technically, the drop interval is Toff + Pulse. And the second would fall to the surface in the same time (like the first did), so maybe about 10 ms less than this to account for that column's height, will raise the second drop to where it should be. So maybe 76 ms Toff, based on the 86 ms column rise time from first touch. Presto! Just ballpark, you can experiment instead, but a close starting point, and it will be repeatable.


Copyright © 2010-2015 by Wayne Fulton - All rights are reserved.

Previous Menu Next