LED power consumption, a revelation

[BushmanLA]

I'm studying for an Electronics test tonight. I wanted to work something out that I had been curious about for some time and the answer was what I expected but was still a little surprising.

A little bit about LED Power consumption....
First some assumptions:
We typically power our LED's from a 5V source.
We typically use a trusty 470 Ohm resistor to limit the current through that LED to about 10mA.

Consider a 5V source in series with a 470Ohm resistor and a typical LED.

Ignoring the effects of the LED, you can get a great estimate of how much current this circuit will draw with a simple V=IR or I = V/R
I = 5/470 = 10.6mA

Power is given as P = VI
5*10.6m = 53mW

Now, all is well with the world until you look a little closer and realize that of that 53mW only about 30% is being used by the LED!! The rest is being wasted as heat in the resistor! Oh the horror!

One good model for the wierdness of the diode is to replace it with with a voltage Vd and a resistance Rd and an ideal diode to show how it will only let current go one way. I'm way to lazy to type out all the calcs I did with this model so..

Another good model is just to replace the diode with its forward voltage drop of about 1.7Volts (this value varies wildly) and an ideal diode to show how it will only let current go one way.

So, Imagine a 5V source, then a 470Ohm resistor, then a 1.7V source, with the two voltages opposing each other.

The current can be given as (5-1.7)/470 = 7.02mA

The power supplied by the 5V source (this is the total power the circuit consumes)
5*7V.02mA = 35.1mW

The power lost in the 470Ohm resistor
4.3V*7.02mA = 23.2mW

The power used by the LED
1.7V*7.02mA = 11.9mW


Power of the LED / Total Power
11.9mW/35.1mW = 33.9% :(


A pretty quick way of arriving at this answer is just dividing the supply voltage by the LED voltage drop.
1.7V/5V = 34%


So what have we learned here?
Bushman types too much crap.
Also, the higher your power supply voltage, the more mW you are wasting.
Also, the higher the voltage drop of your LED, the less mW you are wasting.

If you ever have to power more than one LED at a time it is way better to put them in series with a single resistor.

For example, two LED's in parallel each with their own resistor will waste 23mW in each resistor.

However two LED's in series with a single resistor (with a value set to allow the same 7mA to flow) will only waste 11mA or so in the resistor.

You guessed it, double the effeciency, yay!


I'll shut up now.

[I_Gotts_Ta_Fly]

If you're studying, what are you doing writing this crap? :p :D

But you are right on, my man. LED's themselves are efficient, as they say, but only if set up right.

Stringing many LED's in series with 1 resistor will work for a regulated supply voltage. If you run directly from a battery, they may go from bright to not-so-bright quickly due to the battery discharge curve.

Therefore, you need an LED driver with a smaller value resistor for current sense and the driver IC (90%+ efficiency) to run the chain of LED's.

And now I'll shut up. :D

And go Ace that test, Bushman!

[BushmanLA]

The real way to do it is to not have a resistor at all, but bias it with a nice current source made with a MOSFET or something....

This class is getting to my head.

[rmteo]

This is why in more demanding, high power applications, PWM control is used.

[JimDrew]

The easiest way to make an LED efficient is by not using the resistor at all and pulsing the voltage using a microcontroller or some type of timing circuit. Using a micro you can adjust the brightness by varying the duty cycle. You can find the optimum efficiency and use that frequency for a PWM output. I made a light controller that can do anything from standard wing/tail/strobe lighting to an afterburner ring. It uses only 10 ohm resistors (to protect the micro). You just have to program the micro with info about the LED specifications. You could remove the resistors as well, but the micro would then be open to problems when LEDs become shorted.

[HFG]

You do not need a fancy uP to fo such a simple task, a 555 can be configures to output a PWM signal to control the brightness of an LED.

[I_Gotts_Ta_Fly]

The real way to do it is to not have a resistor at all, but bias it with a nice current source made with a MOSFET or something....

This class is getting to my head.
If you're thinking of using MOSFET in series with the LED, then it acts the same as a resistor, unfortunately. :(

The LED driver (PWM switchers) IS a current souce, so that is the "correct" way to drive LED's. ;)

So, did you ace the test?

[Dan Baldwin]

The correct way to drive an LED depends on the application. If the wasted power dissipated by the resistor is negligible as it is in 99% or so of applications, then the correct way to drive the LED is with a series resistor. The most efficient way to drive a single LED would probably be to use a constant current inductive DC-DC converter.

Simple PWM can be used to reduce the brightness of an LED, but you still have to limit the on current in some way, typically with a resistor.

Dan

[I_Gotts_Ta_Fly]

The easiest way to make an LED efficient is by not using the resistor at all and pulsing the voltage using a microcontroller or some type of timing circuit. Using a micro you can adjust the brightness by varying the duty cycle. You can find the optimum efficiency and use that frequency for a PWM output. I made a light controller that can do anything from standard wing/tail/strobe lighting to an afterburner ring. It uses only 10 ohm resistors (to protect the micro). You just have to program the micro with info about the LED specifications. You could remove the resistors as well, but the micro would then be open to problems when LEDs become shorted.

Are you considering instantaneous current during the ON time? Since you only have small inductance in the leads(may not help much), you're shorting the output to a 2V source on every ON cycle. With only 10 ohms, you're soucing about 300mA during the ON time. I wonder if the LED would last very long at that rate. :confused:

But then again, if you run the micro off of a battery that can only source 50mA(lets say), then the battery would limit the current. But this is not the right method.

Also, I'm wondering, what are you using for feedback? How does the micro adjust for the battery voltage dropping?

Eh, I guess I wouldn't really care too much about LED changing brightness. I might not even notice. :D

[cstratton]

Are you considering instantaneous current during the ON time? Since you only have small inductance in the leads(may not help much), you're shorting the output to a 2V source on every ON cycle. With only 10 ohms, you're soucing about 300mA during the ON time. I wonder if the LED would last very long at that rate. :confused:

Understand your concern, and I'm not current (!) on the details, but this is basically how it's done - full available current at brief duty cycle.

Consider what the mechanism of overcurrent damage could be - if it's basically thermal, then a sufficiently high rate PWM will be averaged out by the thermal mass even if it's tiny.

I'm trying to remember as it's been a few years, but I think with some of the cheapy LED bike taillights you can actually detect the pulsing if you wave it back and forth really fast.

You could also try it with an intentionally added series inductor...

[rmteo]

Take a look at the datasheet for just about any LED. You will typically find 2 ratings for maximum current, such as:
DC Forward Current: 30 mA
Peak Forward Current: [1] 160 mA

Note:
1. 1/10 Duty Cycle, 0.1ms Pulse Width.

[JimDrew]

Are you considering instantaneous current during the ON time? Since you only have small inductance in the leads(may not help much), you're shorting the output to a 2V source on every ON cycle. With only 10 ohms, you're soucing about 300mA during the ON time. I wonder if the LED would last very long at that rate. :confused:

But then again, if you run the micro off of a battery that can only source 50mA(lets say), then the battery would limit the current. But this is not the right method.

Also, I'm wondering, what are you using for feedback? How does the micro adjust for the battery voltage dropping?

Eh, I guess I wouldn't really care too much about LED changing brightness. I might not even notice. :D

Most micros will not provide more than 25ma of current on their output pins, and if you use a high enough frequency you can pulse a LED with current well over the maximum rated value, especially if it less than 50us per pulse.

The input voltage is typically monitored using a A/D pin on the micro. I have had designs just like this in operation since the early OTP 12C508 parts were released and those designs (with the same LEDs) are still working today.

[I_Gotts_Ta_Fly]

Most micros will not provide more than 25ma of current on their output pins...
Gotcha. That should do it then. Now, if I could just learn to start using microcontrollers, life would be simpler. :)

[BushmanLA]

If you're thinking of using MOSFET in series with the LED, then it acts the same as a resistor, unfortunately. :(

The LED driver (PWM switchers) IS a current souce, so that is the "correct" way to drive LED's. ;)

So, did you ace the test?


Yeah I realized that earlier today.
Test got postponed till tomorrow.

PWM pretty much is the way to go, no matter what device you use to drop the voltage, you will be throwing away power in that device. However since PWM gives an average voltage based on duty cycle you don't lose all that power. If the frequency is high enough the small inductance and capacitance of the leads and the switching time of the LED itself start to make a difference. (and it helps)

As pointed out, LED's can take a huge pulse of current if it is very short and most microcontrollers have some kind of internal current limiters anyway. Although you don't want to rely on those since you are driving things out of spec.

This thread is awesome. Nothing is more satisfying that finding the most efficient way of doing something even though it is mostly a huge waste of time. :D

[spinup]

Now, if I could just learn to start using microcontrollers, life would be simpler. :)
This is the easiest platform I've seen so far to start using microcontrollers:
http://arduino.cc/

It gives you a simple language to start with, but you actually have full use of C language and libraries, and you can dip down to embedded assembly if you want. The compiler and libraries are free and open source. At least a dozen hardware boards are available with different sizes and voltages and supporting circuitry. The standard board costs about $35 and works for development, chip programming, and deployment. All you need is a computer with a USB port.

Back on topic...I had a recent "LED revelation": Xmas lights are now a great source of cheap LEDs and drivers. Just picked up some Sylvania sets that have 15 bright white LEDs, in a 3s5p configuration, with a 3AA battery box and a driver chip that PWMs them in 8 different modes (chasing, flashing, fading, etc). $4 each. That's worth it for the LEDs alone IMO. Takes about 5 seconds to remove each LED - just pop out the bulb, bend the pins straight, and pull it out. :)

-Jeff

[Comatose]

This thread is awesome. Nothing is more satisfying that finding the most efficient way of doing something even though it is mostly a huge waste of time. :D


A really cute way of doing it, which you see a lot in inexpensive 1-AAA LED flashlights, is to use the LED (or a series of LEDs) as the diode of a simple boost regulator. This requires only a low-side transistor and a cheap inductor, plus whatever feedback device you like. The efficiency can be well above 90% and the parts cost is stupidly low.

[BushmanLA]

A really cute way of doing it, which you see a lot in inexpensive 1-AAA LED flashlights, is to use the LED (or a series of LEDs) as the diode of a simple boost regulator. This requires only a low-side transistor and a cheap inductor, plus whatever feedback device you like. The efficiency can be well above 90% and the parts cost is stupidly low.

Oh yeah, I'm feeling nerdy today.

http://www.national.com/appinfo/power/files/national_power_designer116.pdf

http://www.maxim-ic.com/appnotes.cfm/an_pk/1804

The maxim app note was pretty nice.

I think the short answer is this.
The only way to not waste power when driving an LED, is to apply a voltage across the LED such that the current is limited only by the LED's itself.

The most efficient way of regulating a voltage is with a switch mode power supply.

Thus, LED's love PWM :)

[Tomapowa]

These have to be the coolest drivers I've messed with to efficiently drive most 1 watt (350mA max) Luxeon-like LEDs...:

http://www.allegromicro.com/en/Products/Part_Numbers/6260/6260.pdf

No diode, coils, etc... just an input cap and a resistor or two. Also has a bunch of built-in safety features...

(Ok... so maybe it uses a uniquely low value current sense resistor... but these are readily available... just not in your ordinary parts bin)

[HFG]

Allegro sends out free samples if you are really cheap and dont mind waiting.

[I_Gotts_Ta_Fly]

These have to be the coolest drivers I've messed with to efficiently drive most 1 watt (350mA max) Luxeon-like LEDs...:

http://www.allegromicro.com/en/Products/Part_Numbers/6260/6260.pdf


I think that one is a linear regulator, so there is a series transistor to drop the voltage down. Again, the power is lost in that transistor.

The boost regulator by Maxim looks like a good one. 36 LED's. Not too long ago, I could only find ones for 10 LED's or less.

[Tomapowa]

Allegro sends out free samples if you are really cheap and dont mind waiting.

Wait? I got my samples in 3-4 days. While you are at it, get some samples of these nifty current sensors...
http://www.allegromicro.com/en/Products/Part_Numbers/0756/index.asp

[HFG]

I got 4 of those as my free samples, never ended up using them tho.