High-output, low-duty cycle LED strobe circuit

b

Thanks, Jon that's a very helpful analysis. I'll have to scope it out in the next couple of days. And I'll consider a higher voltage rail (that just means soldering that 7805 on the MCU board, where I'd left it out before... oh, yeah, and probably decoupling caps if I'm snapping off pulses like that...)

If 500-700 mA gives me the brightness I want, fine. I'm not looking to laser etch anything. I just want LEDs that aren't pitfully dim. Your discussion gives me hope that I can get there.

I actually don't have a current-limiting resistor on the collector leg. When occasionally a software glitch leaves the LED on (I switch it off quick), it's bright. Once it even started going a bit green, but I shut it down just in time. (That color change is unmistakable.) And it gets hot after just a few seconds.

As for the LED analysis, I'm only slowly figuring out how to model LEDs in my head; I'll go through your discussion above in detail when I have time.

Does anyone here have a favorite resource or book about circuit design with LEDs? I've found a few good websites, but many of them are mostly app notes for somebody's IC. I've had a great time with TLC5940s, but they're overkill if all I want to do is flash.

Thanks again.

--Mark

Firstly, I'd put a trickle through the LED at all times (maybe half a milliamp) so the storage capacitance doesn't have to be reloaded each flash. Then, with a blocking diode, connect to a flyback switched inductor. The current in the inductor builds, when it reaches Ipeak you turn off the input, and the current continues to flow through the only other connection, the LED with its blocking diode now forward biased.

Taps on the inductor will allow impedance matching to both the charging supply and the LED. It's just about the same kind of circuit as an old auto ignition.

That's how the old flashlamp strobes worked; of course, the dynamic range of a discharge lamp allows lower duty cycles, at really large peak currents.

On a sunny day (Mon, 4 May 2009 13:18:22 -0700 (PDT)) it happened mj wrote in :

Yes, i2c is cool, I was one of the early adaptors... Made the i2c external too, the i2c is cool as you can drive strings of logic remotely from say a par port of a PC, while the task switching of the OS has no effect, because basically i2c is static, and a delay makes no difference. I had 50 units is a building with clock and sound all controlled by external i2c. Designed and sold video equipment with i2c interface to PC running Linux too... and I use it in my house for control. For very long cables you need to draw up a driver :-)

I have noticed the modern motherbourds still have a par port.. but no RS232... So it is still very useful as interface to embedded too.

No problem. I'm a hobbyist, though. So be warned. I'll probably catch some flak from someone on the above. But I'll learn from it, if I do. So that's fine.

It's not hard and it can be a great help in hand-computing a few important things that will help you in a realistic design. One thing is to drive different currents through the LED (assuming you have a desk power supply that lets you set current limits and monitor voltages, as I do) and list a few interesting data points on a sheet of paper. From there, you can develop a very simple and workable guess about the ON voltage and the intrinsic resistance (which isn't realistic across all possible circuit design needs, but is very simple for planning purposes and will cover your needs here.) I'll talk about this model more, below. Also, it will tell you a lot about what your 2N2222s are really doing for you.

Well, I imagined you already had a 5V rail. A higher rail would be more than that and the 7805 probably isn't right.

Well, I don't think it will. I'm guessing that you are already seeing a little more than 150mA. Tripling that will definitely help, but as our eyes are logarithmic against that change, so I'm betting you will imagine more like about 30-40% brighter. Not 3X. You need to really crank hard on the LED or else use a longer period of time for your pulse -- but I think you already mentioned that causes its own problems in your application.

I'm betting you need a higher rail. Question will be, will the LED last long or stay as bright with this kind of abuse? There are a lot of failure mechanisms, but you are probably looking at delivering a couple of milli-Joules in a very short time. Optical output degrades with higher pulsing -- for example, HP's book on LEDs suggests that

30% loss of brightness occurs when driving at 5X the rating for 1000 hours, but where less than 3% loss occurs driving at 1X for that long. That's direct gap LEDs. The indirect gap LEDs are worse. There is moisture creep into the package and if that vaporizes from some pulse, it's not so good. Plus thermal fatigue due to different expansions of lead frame, plastic, junction coating, die, bond wires, and so on. Then there is annealing of metals, which is strongly temperature dependent. And at higher temps, the plastic, which normally has a fairly stable expansion coefficient, can see it's coefficient amplified crossing over some unknown T. Chemical degradation... well, the list goes on. In any case, if you really do manage to whack them with 1.5A per pulse, you may see some significant reduction in overall lifetime. Just keep it in mind, assuming the thing doesn't just blow up, of course!

Well, time will tell.

There are other considerations in terms of what humans perceive. A lot of it is about contrast. I don't know exactly what your setup looks like, or how it is expected to be used, but there often are a lot of things you can do to improve the contrast and thus improve visibility instead of relying only on hammering your LEDs with huge currents. You might want to investigate those aspects before planning on using an electronic jack-hammer on your poor LEDs.

I gathered that. I didn't include them in the schematic, either.

I bet. ;)

A very simplified LED model is:

V(I) = R*I + Von

You just need to supply Von and R. That's why I wrote 3.3V and 2 ohms, as one of my guesses. More realistic models will use the y-axis cross point for the 'saturation' current (it's a theoretical offset, usually not directly measured but instead taken using a straight line slope from some set of measured points back to the y-axis current) and an Rs value, a funky N factor (emission coefficient, I think I recall) which shows up in the power of an expression using e, and a few other values. But the above equation is often good enough for "government work." It just says, you "need Von voltage at least to get anything out and R*I more than that for any chosen I."

Yes. The one I keep handy is HP's old (old enough that they cover a whole lot and don't expect you to run around finding other references, yet not so old that it doesn't have a lot of very useful ideas that remain useful today) "OPTOELECRONICS: FIBER-OPTICS APPLICATIONS MANUAL," 2nd edition. Get it.

Best of luck with all this.

Jon

He wasn't suggesting using a MOSFET per se, but a MOSFET *driver*.

These are designed for supplying (and sinking) short pulses of high current, in order to charge (and discharge) the gate capacitance quickly.

Yup..that was the idea.

D from BC myrealaddress(at)comic(dot)com BC, Canada Posted to usenet sci.electronics.design

Firstly, I'd put a trickle through the LED at all times (maybe half a milliamp) so the storage capacitance doesn't have to be reloaded each flash. Then, with a blocking diode, connect to a flyback switched inductor. The current in the inductor builds, when it reaches Ipeak you turn off the input, and the current continues to flow through the only other connection, the LED with its blocking diode now forward biased.

Taps on the inductor will allow impedance matching to both the charging supply and the LED. It's just about the same kind of circuit as an old auto ignition.

That's how the old flashlamp strobes worked; of course, the dynamic range of a discharge lamp allows lower duty cycles, at really large peak currents.

I think you can get flash tube and a trigger transformer at radio shack for a couple of dollars.

Bob

OK, then - mid-terms. ;-)

Cheers! Rich

Bank Switching is Evil.

Hope This Helps! Rich

Kodak max single use camera for parts.

Bob

I once drove myself nuts trying to find an I2C "standard" - the best I can figure is, you make up your own!

Cheers! Rich

You're probably thinking of SPI. I2C is well defined.

Heads I win, tails you lose.

Jon

I suspect, but don't know, that there is a benefit to the dispersion angle/lens of the LED that the OP may prefer. A flash tube is great, brighter than all heck, but it may be hard to find one of the right size and with the right reflector/baffles for it. Or maybe all this is fine. I don't know. I like flash lamps a lot -- particularly their ability to be triggered very reliably simultaneously. But I wonder about this application. Plus getting that 300V charged back up may be a minor bother for a row of these and at up to 50Hz.

Jon

Just a quick, off-cuff thought. With an existing 5V rail, and my expectation that a 1.5A pulse through the LED will require something on the order of 7-8V total, we're talking about 2-3V across the inductor while the LED is lit up. (In fact, let's assume 3V to start, dropping to 2V.) I'm assuming for a moment that the Ron is about 3 ohms or so with a Vfwd of 3.3V. That suggests a dI=(3V-2V)/3 ohms or (1/3)A. The dt is 200us. So the dI/dt, or V/L, is a little more than

1500. With a mean V of about 2.5V, this is on the order of 1.5mH. Once the 200us has expired, it would be desirable to ramp up the lost (1/3)A of dI. But over the much longer period of something on the order of as long as 50ms (the OP mentioned 20Hz.) This suggests a V across the L of about 10mV, to gradually get it back up there. (The same as the 2.5V average times the .4% duty cycle.) The other option is to just dissipate the energy, in between, I suppose. Or use an even higher peak and just let it go flat yeilding an average of 1.5A.

Could you expand a little on this? I have some very vague things in mind, but I'm falling short and could use a few pointers to consider.

Jon

Well every mainstream CPU architecture in existence does it.

Oh sure, they all hide it, performing the bank switching automatically and pretending that you have "random access" memory, but underneath it all you still have cache lines, RAS/CAS and even swap.

for

I used to work on instant cameras at Kodak (before the Polaroid lawsuit that shut down Kodak's entire instant camera business.) We had something called a "quench flash". The shutter would trigger the flash via an SCR, illuminating the scene. In the camera was a photodiode that charged a capacitor. When enough reflected light had come back from the scene (that is, when the sense cap had charged to a certain point), a comparator would trigger another charged cap to apply a negative-going spike to the SCR, shutting off the flash immediately and maintaining the remaining charge on the cap. So you got just enough light for good exposure, but charge recovery time was greatly improved because the flash cap only partially discharged.

The original design wasn't mine, but I understood how it worked at the time. Alas not well enough that I could repeat the idea. Maybe some day. So a big fat cap and an array of SCRs could probably do it, but not by me.

Well, I spent two months trying to talk to a I2C color sensor, before finally giving up on it! Found one major bug in the PIC I2C libraries, and the sensor kept giving nonsense data back...

Charlie

They do that with IGBTs nowadays. In fact, I have some parts to a Canon S1 IS laying right here. The unusually small 220uF 330V electrolytic is wired to a circuit board with some fairly small parts, including a small ferrite transformer, something SOT-223, an axial diode shaped like a six amp axial rectifier but a good four times smaller, and on the back side, an array of ceramic capacitors, a diode, SOT-8 and I'm not sure what else (where'd I leave that screwdriver...). The SOT-8 is the IGBT, rated for something like

400V, 8A average, 150A peak. It drops 1V at 8A, I don't think I'd even like to put that much on the thing.

Tim