Foward Voltage?

[GarageBoy]

What exactly is foward voltage? My L bin will light up with only 3 volts, but not that bright. I thought LED brightness was based on current

[evan9162]

The forward voltage (Vf) is the voltage across the terminals of the LED. The rated Vf (which is what binning information is based on), is the voltage across the LED when the rated current is flowing through the LED.

Increasing the current will result in a higher Vf, and decreasing the current will result in a lower Vf. Vf and current are tied together along a curve that the LED operates at. This is the I/V curve.

There are two ways of looking at it, depending on how you think of powering the LED.

First, let's look at things in terms of powering an LED properly, with a constant current power source. In this case, the Vf depends on the current through the LED. So, The I/V curve looks like this: (this is a TV1K Luxeon III)



When controlling the current, the Vf changes only a little across a wide current range. In fact, when powering the LED with a constant current source, Vf isn't of too much concern, which is a good thing (I'll explain this a bit later)


Now, lets look at driving the LED in terms of supplying a constant voltage - which is what we're used to for driving things like motors, light bulbs, etc...
This is the same data, only I've flipped the graph, so now the current varies depending on the voltage across the LED - as if you used a voltage supply to power the LED.



Now, if driving an LED with a constant voltage supply, things get quite a bit more sensitive. Changing the voltage by less than 1V results in a 10X change in current. It would be rather difficult to ensure that an LED is being properly driven by just supplying a constant voltage to it.

That's not the only challenge. Every single LED has a different Vf. Binning ranges of Luxeon LEDs are pretty big. For instance, the K-bin range is 3.51V-3.75V. For Luxeon III LEDs, this measurement is taken at a current of 700mA. So, lets say we want to drive this particular Luxeon at 700mA, using a fixed voltage supply. We would have to set our power supply to about 3.68V. That might be fine for that one particular Luxeon, but if you wanted to set up a bunch of different Luxeons to be driven at 700mA, you would have to tune each voltage source to match that particular Luxeon's characteristics. You'd also have to make sure that your power supply could supply that exact voltage without drifting due to temperature, load, etc. A very difficult method.

But there's more

The Vf varies, not only on an individual basis, but with temperature. As the LED heats up, the Vf decreases, and increases as the LED temperature decreases. The change is only a vew millivolts per degree Centigrade, but, as you can see in the 2nd graph, very small changes in voltage result in large changes in current.

In the most extreme case, the LED heats up a bit, the Vf drops a bit; which, when combined with a fixed voltage source, results in more current going through the LED. Thus, the LED gets hotter. So the Vf drops more, drawing more current, etc, etc. This is called "thermal runaway", and if not stopped, results in the destruction of the LED, either from too much current, or too high of temperatures.

We're still not done yet. [img]/ubbthreads/images/graemlins/tongue.gif[/img]

Vf also varies with time. Luxeons experience a reduction in Vf over their operational lifetime. The vast majority of the reduction happens in the first 100 hours or so of operation. I have done experiements verifying and quantifying this change in Vf. Below, is a graph that shows the above Luxeon after its Vf shifted after 100 hours:



And, the two plots together (initial Vf curve in blue, 100 hours later in red)



So, if you had started by using a fixed voltage supply to power this Luxeon, set at 3.68V, you would no longer have 700mA of current. You would now have about 1900mA of current through the Luxeon, or over 2.5 times the initial current. If left at that current level for long, that Luxeon is well on its way to destruction!


So, to conclude:

- Vf is the voltage across the LED
- The rated Vf (what you will see stated for an LED, or what a Luxeon will be binnned as) is the voltage across the LED at the LED's rated current.
- Vf and current are tied to each other across the I/V curve. The LED operates somewhere on this curve
- Vf varies on an individual LED basis. Two Luxeons from the same Vf bin will probably have different Vfs. Vf binning information is really provided to give a hint at characteristics, not specific information for powering. That's why Vf bins are in a range instead of exact values.
- Vf varies with the temperature of the LED. Higher temperatures lower the Vf (temporarily)
- Vf changes with use time (at least in Luxeon LEDs, probably in 5mm LEDs too). As LEDs age (while powered up), the Vf drops. This is a permanent change.
- Driving LEDs with a constant voltage sources is a bad idea. Direct driving LEDs from batteries is just one step from this configuration.
- Driving LEDs with a constant current source is a good idea. Changes in Vf are of no concern when using a current source.


Probably more information than you bargained for [img]/ubbthreads/images/graemlins/tongue.gif[/img]

[GarageBoy]

Thanks, I'm still trying to relate this all to brightness

[evan9162]

Brightness is pretty much completely related to current, and almost completely unrelated to Vf.

Doubling the current gets you about 1.6 times the light, up to a point - about 1.3A on a Luxeon III. Brightness is diminished by higher temperatures.

[LITEmania]

evan9162

you are truly an asset here CPF.

Thanks for your lecture.

Warren

[fusionman]

Great explanation evan, all those i-v curves remind me of my ee classes. Luckily i'm not ee though. [img]/ubbthreads/images/graemlins/tongue.gif[/img]

A little too much like witchcraft for my taste.

[NewBie]

Very well explained evan9162!

[GarageBoy]

Thanks everyone!

[evan9162]

Just to expand on brightness vs. current, here's a graph that shows brightness vs. current for the same sample that the above graphs were made from.

The lux measurement was made by using a light meter at 1 meter from the bare emitter (no optics).

[chimo]

Excellently presented, evan9162!

This chart makes you wonder if a "burn in" period of about 4 days on a test jig may be of benefit (it would beat swapping batteries for 4 days). It's also interesting the curve seems fairly linear between 620 and 1540mA.

Paul

Edit: Deleted question after I read the other thread on Vf shift over time.

[Amorphous]

Evan9162.. That's an Awesome presentation.
I’ve been experimenting with the Vf and current relationship using a HP 6827A. (Adjustable * current limiting powersupply ) My result resonates well with your data. The only variable left is keeping the X/Y coordinate or Spectrum Loci on the CIE 1931 constant over time and over temperature variation. Another word, “Tint preservation” – I want to do everything to keep my priced LED color or tint the way it is for a long long time.

[cy]

NICE writeup..

[jeffb]

Thank You for the concise info!

jeffb

[NewBie]

Here is some more data in a similar situation, snagged from my tests in my Constant Current vs. PWM thread, with numbers added:



Additional information is found here:
Constant Current vs. PWM

Temperature Effects

[GarageBoy]

How does PWM work?

[NewBie]

Basically PWM is just like switching a light all the way off and all the way on. Basically the LED is just directly connected to the battery, and there is an electronic switch, thats about it.

[GarageBoy]

And how does that dim? (something with the eyes' perception?)

[evan9162]

You basically switch the light on and off hundreds to thousands of times per second. The percentage of on time is the percentage of dimming - so the on time of each pulse can be longer, shorter, or the same as the off time.

Past a certian frequency, your eyes integrate the pulses into constant light. The brightness you see is dimmer than the brightness of a single pulse, but equal to the average brightness.

[chimo]

Very nice chart, Newbie. Interesting to see the broad delta in intensity at the mid power levels.

Paul

[NewBie]

chimo,

If you look at the numbers closely, you'll see they get porportionately broader, peaking at the 3% point, running a 3% duty cycle and 3% of 1140mA (so both at 3% power). You'll note that the current dimmed LED is 228.3% more efficient than the 3% PWM dimmed LED.

Here is a chart showing the delta:

[chimo]

PWM does seem to have a large efficiency hit at very low power levels. So I suppose that if extremely long run times at very low levels is an important factor PWM may not be the way to go. It would be interesting to see the PWM plot compared to the efficiency of a variety of adjustable switching regulators (MMs, VIP, Fatman). Thanks for the graphs and analysis,

Paul

[wquiles]

evan9162 & NewBie,

Thank you both for the great charts - very well explained [img]/ubbthreads/images/graemlins/smile.gif[/img]

Will

[VidPro]

umm, how do you MEASURE a PWM curcuit with the average meter stuff?
when i PULSE, my less expencive meters cannot show the actual specs that are occuring.
but if you switch the meter to AC, you can see why (kinda).

before you toss out PWM dimming with the bathwater, the analisis should be done of the actual run time of a known capacity of storage, instead of trying to read a meter.

if your loosing POWER with a curcuit, it is usually going somwhere observable, if there is no heat blazing off the transister of your PWM device, then where is the power magically dissaperaing to?

[n_den]

anyone know the switching life of luxeons?

[NewBie]

VidPro said:
umm, how do you MEASURE a PWM curcuit with the average meter stuff?
when i PULSE, my less expencive meters cannot show the actual specs that are occuring.
but if you switch the meter to AC, you can see why (kinda).

before you toss out PWM dimming with the bathwater, the analisis should be done of the actual run time of a known capacity of storage, instead of trying to read a meter.

if your loosing POWER with a curcuit, it is usually going somwhere observable, if there is no heat blazing off the transister of your PWM device, then where is the power magically dissaperaing to?

You asked the right question.

Here is the answer.

When PWM dimming, it is going up in the LED die as heat.

Remember, when you dim via analog (voltage or current), the Vf of the part drops, as well as the current. Not so with PWM, you are holding the current the same, and in turn, the Vf (yes the overall temperature drops, so there is a slight shift of -2 to -4mV per degree- Thats why I held the temperature constant within the slug), but the thermal time constant in the die is very short). Think P = E * I

Remember, it is a known fact that Luxeons get more efficient when you drive them less. It is even on the datasheet.

[NewBie]

VidPro, had some time, so I figured I'd explain some additional info.

Also, when you PWM, remember you are pulling full current out of the battery, in pulses.

Losses in the battery are Power= Current^2 * Resistance

So, lets use a battery with 0.1 ohms of internal resistance for example, such as the Sanyo Lithium Ion UR18650F.

When direct driving (which is what PWM is, but it is rapidly switched off and on so you don't see it), you are pulling the same amount of current out of the battery as you are delivering to the Luxeon LED.

One of the recent PWM flashlights around here has been measured at 1.6A, with some pulling 1.8A out of the cell. Lets use 1.6A for the example.

So we take Power = 1.6A^2 * 0.1 ohms, you end up with 0.256 Watts loss inside the battery.

Now, when you use the PWM to dim, you are still drawing the current at the full 1.6 Amps, but in short pulses. You still have the loss during that period, so it doesn't get any more efficient.

Thats one source of loss.

Now...

An LED gets dramatically more efficient, in lumens per watt when you under drive it with a low current.

But, with PWM, you are not reducing the current (it is being hit with the full 1.6A), so you don't get the efficiency gain, that you do with current dimming of the LED.

Unfortunately, you do get some color shift with current dimming.

But, with current dimming, the current pulled out of the battery is reduced, and thus the battery itself becomes more efficient, due to the Current^2 factor.

The best approach to the whole issue, imho, is doing both current dimming and PWM dimming, to pick up the large efficiency gains with current dimming, but not fully reducing the current, and then using PWM to dim it further, or doing a combination of the two at the same time. I've mentioned this a number of times in the past, but as it is more complex to do, it may be a while before we see this approach. This would also reduce the amount of color shift, and with some clever characterization, a person could hold the color a little tighter, playing with the trade-offs depending on whether efficiency or color tint is the more important consideration.

Now, remember the losses within the battery? You also have what are called ohmic losses within the LED die. Also the little super thin bond wires, which are much thinner than they appear, as the Luxeon dome magnifies their appearance. This is really no different than resistance, like what you find in the battery. Again here, the Current^2 factor increases the losses in the LED, so that PWMing at the full 1.6A is more lossy than current dimming.

Then you have losses due to heat itself, in the die. Remember how a Luxeon has a 13 C/W thermal resistance. Very little of that thermal resistance is in the copper slug of the Luxeon. Most of that thermal resistance is between the slug, to the ESD diode, then through the little "solder balls" to the Luxeon die, and within the LED die itself. The thermal mass of the LED die is very low, and it heats up very rapidly, much more rapidly than a person would suspect at first. That is one of the main reasons that the color doesn't shift as much when PWM dimming, since during the short period of on time, the die rapidly heats up to temp. In most semiconductors, the electrical resistance also increases with temperature. This also causes the losses in the die to increase, as compared to lowering the current for dimming.

Anyhow, thats the bigger pieces of the whole puzzle, and should give you something to think about, and why the efficiency of a current dimmed solution can be much higher than a PWM solution.

[Geogecko]

Have you ever used a Linear Technology LT1618 before? This is a nice chip to drive an LED, but I noticed it has current dimming. However, it also has a shutdown pin.

I wonder if you could just use the shutdown pin (PWM) to dim the LED, instead of using the current dimming input pin (Iadj). Would doing to allow you to drive the LED at full rated current? My question, is how fast the shutdown pin responds...