Monday, December 13, 2010

Fun with super-cheap LED flashlights

Somehow I've become a prolific blogger lately.  I wonder if this will last...

Today at work I had a serendipitous moment of seeing two desktop objects in an abnormal light: a small, scavenged lens cell and a super cheap (2 for $5) LED flashlight.

As part of my on-going inculcation into the mystical arts of optical engineering, one of my fellow engineers demonstrated to me the ease with which an illuminated object can be projected onto a surface, if the various distances are correct.  His demonstration was with the image of the lit window that faces our cubes, but it occurred to me today that, by tweaking distances, I could probably get a similar image from my LED flashlight, writ large on a blank whiteboard.


Tah-dah!  What's more, I found that, by moving the lens slightly along the optical path, I could bring different elements of the LEDs into focus.  But lets start by analyzing this image.

First, you can see (though not very well) that the LED lenses are fouled quite badly.  Most of the speckling on the middle three LEDs is NOT the whiteboard.  More significant, however, is the fact that, of 12 LEDs, one is off entirely, one is barely on, and one is quite dim.  This is owing to the sub-par circuit design of this flashlight (for $2.50, I don't expect much, but, hey, let's learn from it!): the LEDs are basically all in parallel with each other and with three AAA cells in series.  The result of this is that the LED with the lowest forward voltage gets most of the current, until IT burns out, then the next lowest voltage, and so on down the line.  The LED that is off either has a REALLY high forward voltage or is the first casualty of this chain of failure.

Interestingly, although it's not visible here, some of the LEDs seem to be flickering a bit- this may be a precursor to failure or simply a result of bad solder joints (which is another possible failure for the "off" unit).  Rather than pads, the PCB (if you can call it that) which these LEDs are mounted to is essentially a couple of concentric rings of copper- resulting is an immense thermal mass which needs to be heated up to get a good solder joint.  Most likely, this is being done by hand in a Chinese factory (believe it or not, hand assembly is economical in many Chinese assembly plants), which opens up many possibilities for cold joints and overheated diodes.


By sliding the lens a little further along the optical axis, I was able to bring some other features into focus.  The middle three LEDs (particularly the lower left one) have clearly visible wire bonds to the center of the die.  This isn't really bad or abnormal- I've seen similar things on very expensive LEDs as well.  However, it's a pretty nifty little visible artifact of a part of the assembly process we normally don't think a lot about.

I'll also briefly mention the blue-ish color seems to be a hallmark of VERY cheap LEDs from China- I've ordered some "white" LEDs from the super-cheap vendors on Ebay before and they look almost more blue than white.

2 comments:

  1. interesting info..

    On a similar note, it seems that a source of ultrabright red LEDs for hacking projection clocks is old optical mice.

    Maybe someone needs to try this with one of the Kopin Cyberdisplay based cheap video glasses (cough Myvu /cough) and make a mobile phone projector, that would be awesome cool.
    A similar mono higher transmission display can be found in older cameras, normally including the driver board and needing -13V, 3.3V and 5V.


    The chips draw 0.2 mA

    ReplyDelete
  2. The blue pumps an orange phosphor, as you probably know, to produce a spectrum that sort of looks like |_^ (Blue spike, orange lump).

    Given that some cheaper mfr.s seem to use organic phosphors that degrade over time*, I'd be curious about color drift as the LEDs age, either in operation or not.

    *Someone (can't find link right now) found pink LEDs (a variant of white with a red phosphor along with the orange) that only had a couple hundred hours' effective operating life at nominal rated current before going almost pure blue.

    ReplyDelete