Note: Some of the information below is anecdotal, since my measurements use a Canon 1D Mark III dSLR, rather than specific light-sensing instrumentation
I've just started work on a strobe softbox for my own product photography studio use, especially motion capture (multiple exposure of moving parts in a single frame): I will be using RGB LED strips to get color control.
Couple of things to note:
- If you have 12 Volt or 24 Volt LED strips, those already incorporate the current limiting resistors in each cut-able piece, so you do not need current limiting - Just provide the rated voltage.
- A 3 millisecond strobe every 500 ms is trivial to achieve with most LED power supplies and LED strips: I have been experimenting with a 40-Watt LED power supply, using MOSFETs to switch power to the R, G and B lines, and I am able to achieve strobe duration reliably down to 2 strobes in 1/8000 second (0.125 milliseconds).
My results get muddy below around 30 microseconds per strobe. My camera does not support a shorter shutter time than 1/8000, so I cannot capture more precise data.
Lessons learnt:
- A 24 Watt power supply used on a 24 Watt LED strip leads to spurious pulses during the time my shutter is open, but a supply with at least double the LED strip's rated power works fine.
- Minimum clean strobe duration varies widely between LED strips of different manufacturers, and slightly between different LED strips from a single manufacturer.
- Pulse edges get softer due to capacitance along the conductors, as the length of the LED strip increases, so connecting lots of the "cuttable" cut pieces in parallel works better.
- The green intensity needs to be reduced a lot compared to the other two, the red a bit lower than the blue, else a color cast appears.
- The green needs to have duration reduced by about 0.25-0.5 microseconds compared to the other two colors, which may be due to secondary phosphor afterglow: See this answer for a detailed discussion on the "long tail" subject.
The last two points are not applicable if white LEDs are used. However, white LEDs will have a slight color cast when using very short duration (say less than 0.1 millisecond, exact limit depends on specific LED used) strobes, due to the yellow-emitting secondary phosphor continuing to glow after the blue / UV junction emission stops at each pulse. At 3 millisecond strobes this color cast will not be discernible.
Recommendation:
Experiment with the specific LED strip you intend to use, to determine the point at which any shorter strobe duration causes a color cast. The figures should be far shorter than the 3ms specified in the question.
People typically use JST-3 connectors to connect these strips because that is typically how they come from the factory. You can get more of these connectors from AdaFruit or Amazon or lots of other places.
If you are going to be soldering a lot of these connectors on, then I recommend getting a solder pot. You can then dip the tips of the wires into the pot, then the end of the strip into the pot, and then just hold the tip of each wire to its pad on the strip and touch with a soldering iron for a second. It goes quick.
I'd try to organize your cable runs so that maybe you hit a whole row of boxes horizontally with each logical string. This is really just about physically making the cable easy to deal with, so do it however works best.
For power, I personally would run a bus of of ~14 gauge black/red wire along with each physical string of strips. This cable should be able to support at least 1500 pixels if connected to a suitable power supply. About once every 100 pixels or so, I'd tie the power lines of the strips into the power bus using tap connectors. If you want to whole thing to be completely disassemblable, you could use Molex power connects between the boxes (or groups of boxes), just make sure the connectors can handle the full amperage you will be pushing though them.
I'd prefer to run each physical string into its own power supply rather than having one big one, again just to keep things physically manageable and to be able to use smaller supplies and keep your wires cooler. :) I've had very good luck with the Meanwell 5V DC supplies, but you can get very cheap supplies on amazon and alibaba.
For driving the data to the whole display, you could use something like an arduino but most of the code out there can only drive a single string so you'd have to daisy chain all of your strips into one long logical string. Yuo will also have to jump though hoops to drive that many pixels with an arduino's tiny amount of RAM unless your patterns are very simple. I'd probably use a beaglebone black running this fork of the popular ledScape code...
https://github.com/Yona-Appletree/LEDscape
This can drive up to 48 strings simultaneously and has lots of convenient ways get you pixels into it, including being able to use Processing for very fancy stuff.
If the data cables from the last string end up very close to the beagle bone, you can connect them directly to the bone's IO pins. Even though the bone is sending 3.3 volts and the neopixels want at least 3.8 volts, I've found that it almost always works fine. If the run between the bone and the farthest string is going to be more than a foot or two or you see glitching on the display, then you can make a very simple level converter from a couple of transistors and a resistor or you can buy nice (but expensive) ones here..
http://rgb-123.com/shop/
Note that they also sell the PixelPusher which should work for you, but is also a little expensive for my tastes.
Note that when you are connecting everything up, try really hard to always get the pixels powered up before sending a data signal into them otherwise you can blow the 1st pixel in the string.
Report back with your results!
UPDATE: Here are some photos of 1500 pixel panels I build using some of these techniques...
https://www.flickr.com/photos/bigjosh/sets/72157655896681772/with/19099344714/
Note that using a Beagle Bone also lets you use a Wifi to control the display from your cellphone!
Best Answer
750W of LEDs is a lot. You'd probably need a welding mask to do anything in that room. Bear that in mind - perhaps consider setting the software to never get anywhere near maximum brightness.
It's not a good idea to connect 2 power supplies in parallel to the same rail. The voltages they output won't be exactly the same and, depending on the exact topology of the power supplies, unwanted and potentially damaging things could happen.
I assume that you're proposing using the pipe as the main conductor, much like the cable track lighting shown on that page? You should break the pipe and provide an independent piece of pipe for each power supply you have. The 0V can (and should) be in common.
Monitoring of the power use could be done in a number of ways. The simplest would be a fuse or circuit breaker. It would also be the crudest method. Current monitoring is typically done using a current shunt monitor - essentially a low value resistor and a method of monitoring the voltage drop over it. As the current increases, the voltage drop increases until it's above a threshold, at which point something is done. You can get ICs which do most of it for you, or you can build a current limiting circuit with not much more than an opamp, a handful of resistors and a mosfet. I assume you're actually controlling the LEDs with a microcontroller or similar? An alternative option would be to feed the output of a current monitor into the microcontroller and to take action when the value is over a specified amount - perhaps by uniformly dimming the LEDs.
Without knowing your specific application it's hard to make recommendations. If you can reduce your requirement to a few hundred watts (which is still a huge amount of light to have indoors) then life will be a lot easier. Remember that just because an LED strip can consume a certain amount of power at full brightness doesn't mean you have to command it to.