Please note that the date mentioned below that the database code was last updated is not the date the data itself was last updated.
Why Use a Driver For My White LED?
Ordinary diodes conduct current in one direction only and just sit in the dark. Light emitting diodes (LEDs) conduct current in one direction only and emit light while doing so. An LED is a diode that emits light in the process of its normal operation. Great.
An LED can be driven in a voltage controlled manner, but when LEDs get hot their operating voltage (Vf, or forward voltage) lowers, which can lead to a situation known as thermal runaway, which is most likely to affect arrays of parallel LEDs. The lower its forward voltage (Vf) the more current an LED will draw from a power source, so the more heat the LED will produce. The hotter an LED is the lower its Vf will get, which means it'll draw more current, and so on in a vicious cycle until either:
Thermal runaway is particularly bad for parallel arrays of LEDs because after the first LED blows the total current will now be shared across fewer LEDs, meaning the next LED to blow is likely to blow more quickly than the first, the next even quicker, and so on at an ever-faster rate (a "runaway") until all the LEDs have blown.
LEDs are most safely driven in a current controlled fashion. This is simplest done using a resistor for each LED or series string of LEDs, with the value of the resistor calculated based on the highest voltage the LED will be exposed to. In a torch this would be with fresh batteries. Don't be put off by the simplicity and low cost; a resistor really may be the best solution for driving your LED. As well as being simple and low cost, they're also often the smallest solution. If the input voltage is reasonably steady and not many volts need to be dropped then a resistor will likely do the job very well. See ledcalc or similar sites for an online resistor calculator.
More expensive is to use a current regulating driver. The best current regulating drivers have a fixed current output that doesn't vary as the LED changes temperature or when Vf or the input voltage varies. In other words, a constant current driver. With the current at a fixed level thermal runaway won't occur.
Misunderstanding the purpose of a driver
I've found the requirement for an LED driver is one that is misunderstood even by those who should know better. I'll give two examples.
This is absolute rubbish!
The circuit they are talking about is designed to run the LED by giving it very brief pulses of electricity 100 times a second, taking advantage of an LED's ability to turn on and off very quickly (unlike incandescent light bulbs which comparatively fade out very slowly when switched off). The pulses are ten times the normal continuous current the LED is specified for, but only last one tenth the amount of time, so 9/10 of the time the LED is off. It's a clever circuit, using a 555 timer IC.
Unfortunately the efficacy* of an LED reduces the more current the LED draws, largely due to heat reducing its performance. Conversely, the lower the current the higher the efficacy. The LED can only handle so much power, whether it's pulsed or continuous, so to get maximum brightness you should aim for whatever duty cycle will give the lowest actual current at any point in time. Of course, that'll be when the current is continuous, not pulsed. Therefore to get maximum brightness an LED should be run constantly, and at the maximum specified current.
The one exception to the way an LED's efficiency changes that I know of is the case of power LEDs (which can cope with hundreds or even thousands of milliamps) if they are run at very small currents (less than a few mA), at which point the efficacy starts to drop significantly, possibly thanks to a small leakage current. Obviously this is not a situation you need to worry about if you want to make light, especially if you want to get maximum brightness.
* Efficacy is expressed in lumens per watt (lm/W).
The second example is in an article titled How to select the right white LED driver. Tony Armstrong (product marketing manager for power products at Linear Technology) writes:
This is also rubbish and really quite ridiculous.
The white point of a light source refers to how closely it resembles an ideal white light source. Basically, it's the light source's colour. A white LED's white point is actually very stable - far more stable than manufacturing differences between two LEDs of the same type, and far more stable than the fluctuations our eyes are able to accommodate. As its input power is varied (within specification), a white LED's colour doesn't change very much, if at all. The LED can get dimmer and brighter, of course, but the colour of its light hasn't changed significantly. This is one of the wonderful things about white LEDs.
Compare that with the behaviour of an incandescent light bulb when run on varying input power. The most familiar example is a torch bulb. When run on fresh batteries its colour might be yellow-white, but when run on flat batteries its output is not only much dimmer, but its colour also changes considerably to brown. This is independent of it getting dimmer. The light from the bulb's filament is produced by it getting really hot. When the filament is particularly hot - with fresh batteries - the more visible light it will be emitting. If the bulb is not as hot - with flat batteries - most of the light will be emitted at infrared wavelengths invisible to us, so the frequencies that are being produced have changed, not just the intensity of those frequencies. This colour change means its white point has changed a lot. This is one of the horrible things about incandescent bulbs.
There are a few circumstances when a white LED's white point might possibly change significantly. The better the quality of the LED, the less likely it will exhibit any of the these effects to any significant degree.
Mr Armstrong's claim is also ridiculous because most current controlling regulators are not constant current regulators (their output current changes as the input voltage changes) and yet the white LEDs being driven with them do not exhibit white point shift of any significance.
In summary, an incandescent torch bulb could certainly do with a constant current driver for its white point to not change, but a reasonable quality white LED's white point shouldn't change under normal operation. More importantly, it's not the main reason that LEDs like drivers at all. That's simply to stop them burning out.
As for the rest of the article, it was pretty much useless for its purpose of helping someone figure out which white LED driver to choose. (I note the author is a marketing person, not a technical person.) I hope these pages with their lists of particular drivers of particular types will be of much more help.