Blue LED at 3.3V

That pretty much demands a voltage boost of some sort. Depending on how many lights you have, how much power you're willing to waste, how much design time you want to spend and how expensive you want the final product to be, your choices sort of boil down to a switcher with inductors and diodes and all that, or a current pump.

Most of us would solve this problem by looking for a suitable IC. _Some_ of us would do it with two transistors, an inductor, and a cap, then brag about only needing one $.001 resistor instead of three.

--
Tim Wescott
Control system and signal processing consulting

And some of us would brag about doing it right ;-) ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |

I have some nice Osram blues that are OK at 1 mA and 2.65 volts, bright at 10 mA, 3.1 volts. So you could just get by with a resistor or current limiter from 3.3. You could use one of my famous beta limiter circuits.

I sometimes make my "3.3" volt supplies actually 3.5 or 3.6. Most other parts don't mind.

Or use a booster.

John

Probably in a lot of applications where people choose a blue LED there is a higher voltage (eg. 5V) supply present as well.

Here is one way to do with about 2-3 cents worth of parts (3 tiny SMT jellybean parts, no inductors) if you have a microcontroller doing the driving:

+3.3V

| | | .--|--. | | | Cs | V | Rs | - | || ___ | | | eg. BAV99 -||--|___|--|- + | || | | | Port pin | V | | - | '--|--' | | V LED (Blue or White only) - | | === GND

AC on the port pin => ON, either level of DC => OFF

Or you could search on, say, LTC's website and find a $5 chip which will be designed for the purpose (blue LEDs are electrically the same as white LEDs in most cases, so all those white LED drivers will typically work equally well with blue LEDs).

One resistor:

ftp://jjlarkin.lmi.net/LED_boost.JPG

John

On a sunny day (Tue, 25 May 2010 11:00:42 -0700) it happened John Larkin wrote in :

Yup, my blue one drops 2.66 V Extremely bright at 3.5 mA. ftp://panteltje.com/pub/low_current_LEDs_img_1964.jpg in that picture it is at 10% PWM with 180 Ohm in series fro ma 3.3V PIC output.

Google "Joule Thief". This is the way it's usually done in small LED flashlights - the sorts powered by one or two AA or button cells.

--
Dave Platt                                    AE6EO
Friends of Jade Warrior home page:  http://www.radagast.org/jade-warrior

tim@servo:~$ ftp jjlarkin.lmi.net Connected to jjlarkin.lmi.net.

421 Service not available, remote server has closed connection ftp>

--
Tim Wescott
Control system and signal processing consulting

How is it at cold, though?

--
Tim Wescott
Control system and signal processing consulting

AFAIK a white LED is just a blue LED chip with some fluorescent material in the package that makes enough "whatever else" to make it look white.

They look exceedingly blue to me -- I don't know if that's because they are, or because I'm color deficient in green and don't see them the same as other people do.

--
Tim Wescott
Control system and signal processing consulting

Usually a blue chip with yellow phosphor. There maybe some that are made differently.

They make them with different color temperatures these days, especially those designed for illumination.

ow

FTP link does not work...

On a sunny day (Tue, 25 May 2010 11:58:36 -0700 (PDT)) it happened rich wrote in :

Works OK here.

A quick look at a blue LED data sheet shows a 3.3V nominal forward voltage at 25C, with a 20% increase at -20C and a 40% increase at -40C.

"Fading blue"?

--
Tim Wescott
Control system and signal processing consulting

I can just doubleclick on the link in Agent, and Firefox opens the image.

It's like Spehro's idea, but with a schmitt-trigger oscillator.

John

Not here - with Firefox or Opera

That link works here (and it is bright!)

Since LEDs get more efficient when they're cold, there is a thevenin drive impedance that results in nearly constant brightness over temperature.

John

1: Check out the latest and greatest offerings by Cree and Nichia. I somewhat remember that many of these have forward voltage drop typically 3.2V at 20 mA.

And consider their forward voltage drop at 4-5 mA, or even less.

Further, consider that these typically have ratio of photometric output to current peaking at 1.6 to 4 mA, and at 4 mA this ratio is usually about

1.3 times that at 20 mA. (If characterized at 20-30 mA and with maximum continuous current anywhere around 30 mA.)

One more thing - good latest-and-greatest Cree and Nichia ones have about twice the efficiency of most of the others. (Although their good ones may be mostly through-hole ones.) (Don't forget about 4-lead 7.5-7.62 mm square ones having .2 inch lead spacing, characterized at 30 mA and having maximum current at least 35 mA. The good ones of those do wonders at 4-5 mA and have decently wide viewing angle, and Cree ones are available from Digi-Key.)

==============

If you have a higher budget for cost and for space, how about the surface-mount DigiKey-available Cree XPEBLU-L1-R250-00Y01?

That one has typical voltage drop of 3.2V at 350 mA and maybe minimum luminous output of 30.6 lumens at 30.6 mA.

At 50 mA, the typical voltage drop is down to about 2.85 volts according to the 2nd of the 3 graphs in the datasheet page marking itself as "8" in

At 50 mA, luminous output at 50 mA looks to me around 18, maybe 19% of that achieved at 50 mA, meaning likely 5.5 to 6 lumens. This is according to the second of 3 graphs in the page marked as "10" in the above datasheet. That page appears to me to indicate ratio of light output to current not decreasing or at least not by much as current decreases down to maybe 25 or 20 mA. That curve makes me think probably 2.2-2.5 lumens at 20 mA.

The light distribution pattern in the 2nd of the 2 graphs in the above datasheet, on its page marked as 11, makes me think that the light distribution pattern is close to lambertian. With 60 degrees off-axis being shown a bit bit higher than lambertian, I would guesstimate that ratio of on-axis-candela to lumens is about 3, maybe as low as 2.9 (guesstimating). So, so-far, this LED appears to me to have on-axis intensity of in or near the ballpark of 733 to 862 millicandela at 20 mA. And with

2xtheta-half (nominal viewing angle) around 125-130 degrees. And likely having typical voltage drop of 2.8 volts, maybe 2.76 at current so low.

One more thing to keep in mind - InGaN blue and green LEDs often have peak wavelength and accordingly color varying slightly with voltage. The peak wavelength usually varies slightly but noticeably inversely with current.

The above Cree blue LED has dominant wavelength (a color specification largely meaning hue) of 465 nm min, 485 nm max. This makes me think 475 nm typ. My experience with cree makes me suspect 476-477. That is a "turquoise blue", close to the color of cyan printer ink, but a bit deeper, pushing showing of a tinge of greenishness. At around 6% of characterization current, I expect the dominant wavelength to be probably at least 480 nm, maybe even 485-plus nm. This usually looks like a noticeably somewhat greenish shade of blue. At least that is much more bluish than "blue-green" and "traffic-signal- green" LEDs, with typical dominant wavelength usually 497-507 nm, often around 505 nm. If you are familiar with the color of the 486.1 nm "H-Beta" line of hydrogen, keep that one in mind. The 485-486 nm ballpark of wavelengths tends to appear to be a very slightly greenish turquoise blue *when viewed brightly* (such as looking into a 700-plus mcd LED), and gets closer to an only slightly bluish side of blue-green *when viewed dimly* (such as being in a room illuminated by this with at most a few lumens).

Another thing - at 20 mA, I would not worry about heatsinking the LED's thermal connection. If your SMT soldering can accomodate the thermal pad, then I would advise a trace 1.2 mm or .04-.05 inch wide under it, extending out to the sides, overlayed with some sort of "buterfly wings" if you are capable of achieving such layout, merely to make things a little better, as in probably capable of handling 60 maybe 100 mA. Don't forget to deploy a rectangular pad under the LED's thermal pad in your PCB layout so that you don't get solder mask in the way. And I do like solder mask on copper layout intended for heatsinking - solder mask has much higher emissivity of thermal radiation than bare metal has. And to the 3.2 mm wide (longer dimension) solder pads for the electrical connections - I would run traces of width 3 mm or .12 inch, at least in the 12.7 mm / 1/2 inch within these pads. Doing this heatsinks the LED slightly, whether or not you can do tricks with thermal design of PCB layout to the LED's thermal pad. This gets more important if you can't solder to the central thermal pad. Wide traces to the LED's electrical connections should make the LED safe to run at 30-40 mA as I guesstimate, so that has high chance of achieving (or improving) this LED's safety of being run at 20 mA.

Maybe you get enough light from this "high power" LED at 10 mA (with typical voltage drop likely in the 2.72-2.8 volt range) - at which point I see no need for thermal considerations, despite this LED differing from low power ones by having maybe moderately higher thermal resistance between the "junction" and the electrical leads (especially the cathode one in the case of many LED chip chemistries including all that I have heard of for blue).

--
 - Don Klipstein (don@misty.com)