View Full Version : High power LED drives
01-22-2008, 09:58 PM
I'm working with some Lumiled K2 leds, these can pull up to 1.5A of current and as you might expect are rather bright. I was wondering if anyone had any ideas for driving and switching them, in general I have always just used simple current limited supplies for LEDs based on linear voltage regulators. In this case as well as being able to manually dim them (turn down the current) I'd like to be able to strobe them with the propeller. I'm looking for any ideas really for neat solutions.
Why don't you use a simple tranisitor (one with sufficient current carrying capacity)·switch to turn the LED on/off driven from one of the props I/O pins, and use PWM to control the brightness?
01-23-2008, 12:55 AM
I don't always want to use the lights with the propeller.
01-23-2008, 07:07 AM
·I connected a 12v taillight to the 9v input thru a simple npn transistor as Keli suggests.·I could turn it up and down with PWM. The counter tutorial covers the PWM I think.
······· "What do you mean, it doesn't have any tubes?"
······· "No such thing as a dumb question" unless it's on the internet
Technologically challenged individual, Please have pity.
01-23-2008, 10:49 AM
If you can get a variable resistor in the proper range, here's a very low cost driver. http://led.linear1.org/a-cheap-current-regulated-luxeon-star-driver-design/·· It's not the most efficient.
It should be fairly simple to cook up a MOSFET circuit for the strobing effect.
Phil Pilgrim (PhiPi)
01-23-2008, 11:32 AM
If you want a linear drive, the attached is the classical way to regulate the LED's current. With 5V on the input, the LED will draw about 1.37A. At 0V, the LED will be off. You can strobe the input to strobe the LED at max power, or you can PWM the input to dim the LED. The LED's supply voltage should be kept as low as practical to keep the power transistor cool but must be at least Vfwd(LED)+ Vsat(Transistor) + 0.5V. A high-beta transistor would be the best chioce to keep the op amp from having to source too much current.
01-23-2008, 01:33 PM
FYI: If you want to simulate your LED driver, LTSpice/SwitcherCADIII (Free from www.linear.com/designtools/software/ (http://www.linear.com/designtools/software/)) comes with a couple of Lumiled models, one of which is an LXK2-PW14, which might be close to what you're using. I think I also saw some more Lumiled models in the files area of the third-party moderated LTSpice forum at http://tech.groups.yahoo.com/group/LTspice/.
01-23-2008, 05:25 PM
How many LEDs do you want to controll at once? The TLC5940 led driver chip can drive 16 leds at once, and can be daisy chained if you want to drive more. The chip can only supply fairly low currents by itself (25mA, or thereabouts), but Texas Instruments have written a short article on using MOSFETs to control higher power LEDs. Have a look here (http://focus.ti.com/lit/an/slva280/slva280.pdf). Theres some code somewhere here on the forums for driving the chip with your Prop.
Theres plenty of information on the Texas website here (http://focus.ti.com/docs/prod/folders/print/tlc5940.html#technicaldocuments)
01-23-2008, 07:43 PM
Thinking about it further PWM is out of the question as it will form a lighting system for a high speed video camera, the pwm will show up unless it's a high frequency. So linear is probably what I should go for, I have used the circuit Starman shows for battery charging in the past and was considering it for this app though I'd possibly switch the 317 for a low drop out regulator to reduce heat. The mosfet/transistor part is simple enough I just wasn't sure if there were any clever solutions out there I should consider like a driver chip but if they are using PWM I'd be back to square one. I should have just trusted my instincts and gone with the linear current regulator and transistor/mosfet.
01-23-2008, 08:17 PM
Don't today's high-speed cameras have sync outputs to handle just this type of issue; not only synching PWM, but for synching high-speed, high-power strobes. With the sync output you can simply control the width of the LED driver output pulse in-sync with the HS camera's frame capture.
Have a look at the Linear LT3518 part - it will drive up to 10 LEDs like yours and has a PWM input too. Digikey has the LT3518EUF#PBF type E standard-temperature RoHS part at $5.90 USD qty.-1, $4.21 qty.-25, $3.47 qty.-100.
Note, this is a relatively new part from Linear. As-yet I haven't been able to find their LTSpice model for it. I've posted a request on the LTSpice forum, if nothing comes back from there I'll query Linear's technical support directly about the model. I'd like to simulate a high-current LED driver like this myself.
Link to Linear's LTC3518 Page (http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1003,C1094,C1766,P3 8340)
Link to Digikey's LTC3518 Page (http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=LT3518EUF%23PBF-ND)
01-23-2008, 08:23 PM
There is a project in the January issue of Nuts and Volts magazine for a high power led strobe which may be of interest to you.
01-23-2008, 11:03 PM
The frame rate will be 4000fps, to be honest I want to avoid using any special camera features as it is something we borrow and I won't have time to test things out. Simplicity is best.
I'll check out those other parts
thanks the the heads up Gerry
01-24-2008, 12:57 AM
Gerry Keely said...
There is a project in the January issue of Nuts and Volts magazine for a high power led strobe which may be of interest to you.
Could someone point me in the right direction to get that. I'm not sure if I can fin the magazine around here.
Lil Brother LLC (SMT Assembly Services) (http://www.lilbro.mosquito.net.nz/index.html)
01-24-2008, 01:13 AM
I just subscribed and its an interesting article. Although most of the time I want to use the LEDs CW I would like to pulse and from what he says switching regulators don't really like doing this, not are the required pulse rates/lengths. He is using a histeretic switched drive chip I can't get easily so I think I am back to a linear regulator or a circuit like Phil's and a mosfet.
Also found Osrams 75W pulsed laser diodes (30ns pulse length) which might be interesting for my applications.
01-24-2008, 01:26 AM
So you've got "borrowed" access to a "high-end" high-speed camera. Something from the likes of http://nacinc.com/. These cameras cost a LOT of money but can record at very high frame rates even in regular indoor-light environments (anyone seen MythBusters on TV?, do a google search for "Memrecam").
So you just want to do a simple analog programmable high-brightness ligh-source for use with the likes of the cameras mentioned above. Just do a MOSFET or IGBJT or similar linear interface (no PWM) on the Luxion/Lumileds parts you have. I think we've come full-circle here on that... BUT:
You might also want to look at InfraRed illuminators. Sometimes these high speed cameras are far more sensitive at infrared than at visible light spectrum, and there are some pretty impressive and cheap IR illuminator modules out there from the security camera vendors. Take a look at the specifications for the camera you're borrowing. Black & White cameras have substantially higher sensitivity especially at IR, but you need the specs (many dB advantage) if you pick the right wavelength "white" or IR illumination diodes (various wavelengths, again check the IR "sweet-spot" specs for your HS cam).
At your desired FPS speed for room light 1 Lux is minimum, if you want very high performance ambient light night vision, then 0.01 Lux sensitivity is needed, these numbers are possible. These numbers seem to hold regardless of the FPS and can vary with respect to number of lines of resolution. I can buy a camera that gives down to 0.005 Lux B&W that outputs more than 400 lines of resolution for less than $200 if it outputs real-time NTSC in real-time. I think I would have to spend something like $20K USD to get even close to this at thousands of frames per second (remember high-speed requires buffer memory cost compared to real-time).
Beware when trolling the "security camera" options on the Web for IR illuminators, these Security Camera Vendors are purveyors of much SPAM and unwanted physical mail. - Really a bunch of crooks, but they're selling to crooks and Desperate Housewives in the first place. Security camera sites are seond only to Porn sites if you want to be infected by a Trojan.
01-24-2008, 03:04 AM
For power LED's up to 3W (1A) i am using the ZXLD1360.
I have also made a small PCB for that IC. For more information see:
www.systech-gmbh.ch -> electronics (Elektronik) IREGLER-Z. (as an idea)
The newest version is now even smaller see -> attachment.
The good thing on the ZXLD chips are the easy way for dimming. There are two ways:
a) with a low frequency PWM (100-300Hz)
b) with a higher frequency of > 10 kHz.
Phil Pilgrim (PhiPi)
01-24-2008, 03:42 AM
One advantage that a switching regulator like Joerg's has over a plain PWM is that the output ripple should be pretty insignificant. This is especially true as the switching frequency and/or inductance increases, since the inductor will continue providing current between switching pulses. The amount of ripple will also depend on the hysteresis in the switching mechanism, which may or may be be a controllable parameter.
Joerg, if you were to place a fast-response light sensor in front of your LEDs (optically filtered so it doesn't saturate), what kind of ripple do you observe?
01-24-2008, 03:58 AM
I use a 1W power LED with this circuit for illuminating my door of the office (so i find where i have to put my key!) an i do not see any flickering since the ripple is 15mV/RS which is 15mV/.33Ohm = 45mApp. So i think this will be hard to see.
01-24-2008, 06:10 PM
Attached is a first-stab at simulating the circuit provided earlier by Phil. This is a linear LED driver. It is surely not optimal but illustrates that what can be thrown together quickly works fairly well. The LED is a Lumileds LXK2-PW14, which it seems from the it's model to operate at 1.6A.
This circuit (see Schem.jpg) is designed to take a linear 0V-3.3VDC in and provide a constant linearly corresponding LED current of 0-1.65Amps. At 1.65 amps selected (see Plot1.jpg), the LED current is constant with a LED input voltage range of 9V-12V. With just 5mA LED current selected (see Plot2.jpg), the LED current is constant with a LED input voltage range of 3.8V-12V.
You can set the maximum output current easily by changing R1. The maximum input current is simply V1max/R1. In this example it is 3.3V/2Ohms=1.65Amps.
An obvious improvement would be to arrange some way to get the op-amp to operate off a single supply rail so you don't need a negative supply. Some improvements may be possible by using MOSFET instead of the Fairchild TIP112 NPN Darlington I threw in.
As I mentioned previously in this post, this is simulated with the free LTSpice/SwitcherCADIII from www.linear.com (http://www.linear.com). Should you wish to run the simulation yourself, in the attached Sim01.zip file you will find (1) TIP112.LIB, the Fairchild subcircuit definition for the TIP112, and (2) Lumiled_01a.asc, the LTSpice/SWCADIII schematic file.
01-24-2008, 06:24 PM
Drone, I'm already using a red led because flys and bees are not sensitive to red light and I don't want to mess up their flight.
The issue of strobing is seperate to that of PWM, I want power control for the high speed camera use but I also want to be able to produce short pulses of light for some other experiments. I'm sorry my specification has not been clear but I am also not too clear on what I want at this stage.
If I go linear I will just use a voltage regulator to adjust the current so don't waste your time with spice.
Thanks for the info Joerg, I've built DC-DC converters that work up to 1Mhz but I think I must rule out switching regulators unless I know they can be turned on and off rapidly.
01-24-2008, 06:26 PM
Out of interest has anyone done any multiple exposure strobing? You do several flashes in a single exposure, the best results are when the object moves out of it's own path so there is no overlap but I was thinking what if you had three strobes in three colours, could these be separated? Could you use this method to produce a 3X speed up on a typical camera. Probably not.
01-24-2008, 06:44 PM
The circuit I posted is designed to do exactly what you have described. You select the current you want with a 0-3.3V input...
...Pulse the input if you like - No Problem...
...and the selected LED current will be constant even if your LED supply voltage (like a battery) varies. Obviously make sure the op-amp supplies are regulated.
I simulated the circuit out of my own curiosity as well.
01-24-2008, 07:01 PM
Yes I know what the circuit does but I'd sooner use a three pin device and a resistor and didn't want you to spend too much time on it for my benefit.
The camera by the way isn't really high end anymore, its a Phantom V7 www.yorkinstrument.com/images/phantom/7.bmp (http://www.yorkinstrument.com/images/phantom/7.bmp) it looks kind of retro but it will do 4000fps at 800X400 or there about and is easy to trigger etc. We have a couple of older 500fps NAC cameras but I have not used them, not fast enough and I am told they are a pain to work with in terms of triggering etc, I suspect that things may have moved on.
Now I'm not sure where things have moved to but you can bet it will be expensive. I'm a fan of the cordin type cameras with a spinning mirror and an array of sensors, you only get a few frames but they go up to millions of frames per second, it sort of makes me wonder about buying some cheap digital cameras and making a home brew version for more sensible FPS just for fun. Heck you could even spin the cameras.
Post Edited (Graham Stabler) : 1/24/2008 11:06:12 AM GMT
01-24-2008, 11:32 PM
Don't want to "poison" your post Graham, but I did a little more with the simulation posted earlier (other readers of this thread later-on may be interested).
1. Changed the op-amp supply to single-rail.
2. Changed the op-amp supply to Vcc=5V (propeller protoboard friendly).
3. Changed the NPN Darlington from TIP112 to the much cheaper TIP110.
Result: The circuit works as before for the most-part. However changing the op-amp to single rail broke the nice ILED=V1/R1 relationship, most-likely due to using the particular op-amp in single-rail mode (offsets etrc.), pick a better op-amp. But you only need to reduce R1 by approximately 1/2 Ohm to get back to 1.6A at 3.3V control input, and all else seems fine. A better op-amp selection that is single-rail friendly may help - dunno, I've got a bunch of LT2006A's from suplus so I just drop them in for quicky sims.
Things I should have said before:
4. This circuit is inherently linear, hence relatively inefficient. It is only 35% efficient at Vled=12V (car application for example) using the Lumiled part in the simulation. For example the LED dissapates around 7W, but the total draw on the 12V Vled at 1.6A is close to 20W.
5. R1 dissapates around 3.8W, use a 5W resistor minimum, and beware of wire-wound resistors if you're going to PWM this circuit at high speeds.
Anyway... with this circuit you do have the options of BOTH linearly and digitaliy (PWM, whatever) to control the LED, and (arguably) reasonable constant current performance is provided over Vled changes. Again - this LED driver is constant current, given some Vled constraints (as posted earlier). I have not done a temperature sweep simulation, YMMV as a result, but methinks it should be ok given the op-amp feedback loop.
I mentioned some possible enhancements in the previous model post: Single rail op-amp (done), MOSFET - maybe but Vgsth may be an issue (one more op-amp and take the op-amp's Vcc from Vled via a regulator?), MOSFET's with low Vdson might improve efficiency.
I attach the new schmatic in .jpg and a .zip with LTSpice sims for both .op and simple .trans Vled sweep as well as the .lib for the TIP110.
No need for reply Graham, just did this post to sew-up my findings...
Best Regards, David
Phil Pilgrim (PhiPi)
01-25-2008, 12:52 AM
Has anyone ever done what you're trying to do with holography? It would involve film and a pulsed laser, so would have to be done in a darkroom, but the mechanics might be simple (i.e. rotate the flat film plate on a vertical axis as the laser is being strobed). If the pulses are short enough and bright enough, the issues arising from vibration (the usual bugaboo of holography) could be avoided. (I'm not suggesting you pioneer this approach if no one's already done it and you want results quickly!)
I like the color (excuse me: "colour" http://forums.parallax.com/images/smilies/smile.gif ) separation idea, and there's no reason it shouldn't work the way you want it to. Separating into RGB components is easy with any image-processing program. The only catch is that each image in the separation will look a little different, since your critters will reflect the three hues in different proportions.
01-25-2008, 11:00 PM
Are you thinking 3D information might also be recorded or just that it would be really cool?
The pulsed lasers from OSRAM (http://catalog.osram-os.com/catalogue/catalogue.do?favOid=000000020000389200200023&act=showBookmark) are pretty awesome, the pulse lengths are only 30ns and the hybrid nanostack gives 70W peak output and is a peach to drive as you only need a little mosfet driver as detailed in the app notes.
I don't however have any idea about coherence lengths and indeed I don't know much about holography though I have done more than my fair share of interferometry (laser doppler, wide field speckle microscope etc etc). The concept of a spinning film is definitely a good one.
Thanks for the confirmation on the multi color/colour (let's call the whole thing off ;) ) could be neat, no use for what I do as I don't have a 4000/3 fps colour camera but I think it might be neat to knock up a 90fps webcam if you can work out when to flash your LEDs. Of course it needs to be operated in the dark for best results but hey, OK for experiments.
01-26-2008, 12:40 AM
I'm finding quite a few bits and bobs but I haven't found my feet yet.
US3594060 shows a high speed holographic camera. May I recommend gb.espacenet.com (http://gb.espacenet.com) for patent downloads, no need to register and you can get full pdfs.
01-26-2008, 01:27 AM
Graham, I tried gb.espacenet.com (http://gb.espacenet.com) and can't find US3594060 nor same when searching for "high speed holographic camera". Hmm...
In addition, when I tried to download some other patent documents, the drawings were not included in the .pdf's, this is a rampant problem from every patent server I've ever used. I really would like to find a free patent download site that includes the full document, including the drawings.
01-26-2008, 01:57 AM
go to advanced search and enter it under publication number
All of the pdfs I have downloaded contain the drawings, often in their own section though not all patents have a pdf available. This is not just yet another patent site its done by the patent office and is the UK part of http://ep.espacenet.com/
Phil Pilgrim (PhiPi)
01-26-2008, 02:07 AM
I hadn't even considered the 3D aspects of holography. I was more concerned with capturing a plurality (as they say in patent lingo) of images with the simplest mechanism possible. Having read the patent you referenced, though, "simple" seems rather elusive.
Here's another thought: I don't know what kind of resolution your images require. But you could capture several small images in one focal plane by spinning a wedge prism in front of the camera lens, coaxially, as you strobe the light source. What you would end up is a circular array of images. The number of images you could accommodate this way would depend on your resolution requirements and the overall resolution of the film/sensor.
I also wonder if discrete images are even necessary. If there were a computational means to separate overlapping images, that might suffice. For example (and ignoring parallax effects for the moment), suppose you had a fixed array of cameras, each with a linear polarizer in front of it, and with each polarizer rotated at a slightly different angle. The strobed light source would also have a polarizer, spinning rapidly. Each camera would record the same sequence of overlapping images, but the images recorded by each one would be modulated in intensity, based on the relative polarizer angle at each strobe. It might then be possible to separate the overlapping images by correlating each pixel or group of pixels with their corresponding pixels in the other cameras, along with knowledge of the relative modulation phases. If you had fewer cameras than imags, the system would be underconstrained, mathematically, but by considering neighboring groups of pixels, that might be overcome. It may also be possible to combine this method with spinning prisms to obtain some actual separation before the math is applied.
Of course, if your subject matter screwed up the polarization upon reflecting the illumination, as would happen with a primarily diffusely-reflecting subject, all bets would be off. Yet, it may be possible to achieve the same results with a modulated bi-colored light source whose color intensities "rotate" 90-degrees out of phase while strobing. In place of polarizers on the cameras, one would use color filters — or just color cameras.
Okay, that should be enough pie-in-the-sky hand-waving for one morning! http://forums.parallax.com/images/smilies/smile.gif
01-26-2008, 02:32 AM
I've considered producing multiple images of that sort, there are lots of ways to do it (most from the 60's) when you use film you can have quite small images and still get good resolution. I very much like the idea of the wedge prism, the other alternative is to just spin the film. High speed Polaroid film would be a good start (assuming it exists), I actually picked up an old Polaroid camera to look into this but I wasn't thinking of pulsed light sources so it was going to be more complex.
I just know that some of the "old school" high speed methods could be combined with modern sensor technology and image processing to create practical and low cost cameras, it's just a matter of doing a lot of thinking. Unfortunately as you can spend big books on nice high speed cameras most researchers will stop thinking and start saving/begging.
Love the polarizer idea, I guess that's both of the ways of encoding the light sorted, light and polarization. But multiple perspectives are not ideal not least because of the space all these macro optics take up.
Have you heard of image segmentation high speed cameras? These are great, the simplest type puts a sort of grating type mask in front of the film, the part that allows light through is roughly 1/10th of the width of the part that blocks the light, if you expose the film through the mask you get an image with 1/10th the resolution in that dimension, move the grating 1/10th and flash again, then repeat 10 times. What you get is 10 interleaved images which must then be decoded by sort of reversing the process, this is easy if you digitize the image or do it with a digitized film. There are other methods that distribute the image in 2D but I can't remember enough details to explain them.
Problem for me is that I want to do macro and I can't really arrange a mask over the sensor when the pixels are 8um across roughly and it would make triggering the flash rather tricky. Perhaps with a projection lens and a proper sized film there might be some chance. For more normal stuff you could magnify, apply the mask and then de-magnify to the sensor.
I like pie!!
Phil Pilgrim (PhiPi)
01-26-2008, 03:38 AM
Is the motion you're trying to capture repetitive (e.g. the beating wing of an anchored fly)? If so, could you synchronize the strobe to the motion, advance the phase gradually, and snap pictures at a more leisurely pace?
01-26-2008, 06:02 PM
You can do tethered experiments but who knows how a fly reacts to being glued to something, would you walk normally glued to a stick? But it has been done with strobes and it is something I have considered as you can use a very high resolution camera. I have some record player pickups to use as a means of creating a trigger signal for the strobe which can then be phase shifted to create a sequence.
I've looked a little more at holography, there is certainly more potential than I ever realized:
1. You can produce images with a high depth of field, allowing you to create an image of a dynamic subject that can later be examined under the microscope looking at the full volume
2. You get 3D information as you essentially capture the whole field
3. There is potential for very high resolution
4. You can do a sort of lens less imaging which may be great for high speed work
I'm still a little iffy on the whole thing but it sure is interesting.
01-26-2008, 08:18 PM
We are getting off topic I suppose but we must remember the prop cam might be used in some of these applications.
Just found out about digital holography: