LED drivers

Well, the whole point of LEDs is to be somehow more efficient, and by burning maybe half your power in a linear regulator, you've screwed yourself below the level of halogens, i.e. they will be cheaper in the long run.

Tim

-- Deep Friar: a very philosophical monk. Website:

You can work it out. This is a buck converter, so the power dissipated by this is going to be (very worst case) the power supplied to the LEDs multiplied by 1 minus the efficiency. You can get the efficiency from the graphs. It will be worst case because there are going to be resistive losses in the inductor and diodes which the efficiency has taken into account.

If we assume 4.5V per LED and you have 2, this is a power of 3.5W, so if you run at 12V that's about 92% efficient, which means the LM3407 could dissipate 0.28W. The next thing is to look at theta(JA) which is quoted as

50 degC/W, for a heatsink of thermal pad to 2 sq inches of copper, therefore a temp rise of 14 deg C above ambient. If you assume worst case ambient of 40 deg C that's 54 deg C, which is well below its maximum junction temp of 125 deg C. You should be able to work out what theta(JA) is for little or no PCB heatsinking is by Googling for the thermal resistance of various PCB heatsink profiles and calculating what the theta(Jpad) is for the package (pity they didn't provide that, may be worth asking National what it is).

So AFAICS you'll need no external heatsinking other than a little PCB area to mount the thermal pad to.

Mark.

A buck converter can and usually does have a higher output current than input current. For a lossless buck, input power equals output power.

John

On a sunny day (Mon, 11 Jan 2010 10:18:48 -0600) it happened "Tim Williams" wrote in :

That is only *one* poin tof LDs, the main pint is longer life and more reliable then normal bulbs, and lower voltage too.

and by

Who cares if you run a 2 kW heater at the same time to fight global cooling, anything helps to heat the room:-)

Well, the whole point of LEDs is to be somehow more efficient, and by burning maybe half your power in a linear regulator, you've screwed yourself below the level of halogens, i.e. they will be cheaper in the long run.

Tim

I'm still confused why the driver is important OTHER than the increase in power consumption efficiency. (which doesn't seem to be that substantial for only a couple of LEDS)

Longer life for the LED?

Seems like a switching supply is overkill to me.

I realize that a linear supply isn't very sexy and is old school but certainly that wouldn't play into it?

I guess as John says the higher output current to input current is a prime factor.

But isn't that because the buck needs a higher input voltage?

Electrically, if the supply characteristics are identical, it doesn't matter.

Sure. Less power dissipation = less heat = longer life.

A switching supply can be pretty damn cheap, and it doesn't have to be overwhelmingly complicated. A simple hysteretic regulator can be constructed from a few transistors. (Depending where you are manufacturing, that method may end up more expensive due to the total parts count required).

Like I said, it's all about efficiency. LEDs are still damned expensive and if you aren't going to use a switching supply, you're better off spending your money on halogens.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

On a sunny day (Mon, 11 Jan 2010 12:42:52 -0600) it happened "Tim Williams" wrote in :

I absolutely disagree. For example my RGB LED strips are not very expensive, and have a resistor in series to limit the current. ftp://panteltje.com/pub/col_pic/RGB_LED_strip_img0906.jpg Yes thay are PWM'ed for intensity (=color_ control), but that switches the resistor :-) And there is little relationship between temperature and lifetime too.

!

You know what -- I'm forgetting about the heat in a power LED. !

Sometimes, things fly right by me.

I'm so used to standard LED's and 20ma that my thought process disregarded it.

So -- with that in mind -- I see the light now -- it's not really about the supply but the friggin LED.

If you're switching on and off, wallah, less power consumption. (and those bad boys can get damn hot).

So -- what is a typical switching frequency and why do these drivers run at such high freqs? (400k I think I saw)

After all, I think it's about 2k or so when your eyes can't detect switching.

Wouldn't the efficiency aspect be about duty cycle rather than switching freq?

Why can't I just switch my LM317T with a damn 555 or something?

Yes -- I do get pissed off at how my brain works sometimes.

--
For a lossless _anything_, input power equals output power...

JF

Naaaah! Really? Some of JL's Alzheimer's showing thru ?:-) ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

Another thing.

I do have my LED's on large heatsinks (total surface area of probably

4 - 5 sq inches per 1 watt led) (using 2 sides of a double sided PCB with a lot of plated through holes). They run very cool.

And using a linear supply.

Try charging a battery sometime.

John

High switching frequency allows smaller inductors with higher Q. On the downside, more switching means more switching losses, so you need to move things faster (switch time < 50ns), so the dV/dt is higher, which makes more RFI, but, it's easier to filter because your EMC chokes are smaller too.

I have an LED string built from 70mA LEDs, PWM'd around 5kHz:

If it were faster, I could filter current to the LEDs so it's not throwing off nasty AC fields, but the 4N25 is pretty pokey, so I can't switch it too fast before everything drunkenly slurs together.

Even when turning around extremely fast, with a low duty cycle (narrow pulses), it's difficult to see the pulsation of these LEDs. Visually, it's a practical PWM rate; electronically, it's too low for any energy storage (unless you like iron core transformers!), so you have to use resistors to limit current.

This project wasn't for lighting purposes, so I don't care that it burns 10W when the lights are on, or that the 120 total LEDs barely have the illumination power of a 60W light bulb.

Efficiency is about wasting watts. You don't gain anything by pulsing current into the LEDs. A switching regulator is pulsing *voltage* into an

*inductor*, which filters the current into the LEDs. The LEDs get a *fairly constant current*, while the supply draws as much current as it needs from the source.

Because it doesn't have an inductor to store energy inbetween switching.

Yes, they'll run cool with a sufficiently large heatsink and seperate power supply. If you were in the business of making LED lamps, you'd have both in the same place, and a lot more pressure to build it cheap as hell to remain competitive while maintaining the primary advantage of the LED, which is long life and high efficiency.

You're welcome to run LEDs from resistors or LM317's and seperate power supplies. It's an excellent novelty, and occasionally may produce enough light to read by. Just don't fool yourself into thinking it is in any way useful, efficient or cost effective.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

If you use a switchmode buck regulator circuit, of the usual kind with an inductor and a catch diode, average input current is less than average output current. For part of each switching cycle, the output current is in a closed loop that includes the load (and filter capacitor) and excludes the power supply.

- Don Klipstein ( snipped-for-privacy@misty.com)

But what if your main home heating system is more cost-effective and more efficiently using energy derived from fuel than resistive electric heat is?

In most of USA, electricity costs more per unit energy than fuel oil and natural gas. One major reason is that electric generating plants are only

40-50% efficient at converting heat energy to electric energy.

- Don Klipstein ( snipped-for-privacy@misty.com)

--
Hmmm...

You must be talking about one of those "latching batteries" where all
you have to do is put a short current pulse into it and then it'll
supply output power forever.

JF

Frequency in the hundreds of KHz (occaisionally even in the MHz ballpark now) is used to allow the inductor to be small and for output filter capacitors to be small.

A homebrew or older design can switch in the 10's of KHz, but that requires a larger inductor, likely hundreds of microhenries or so.

It is preferable to get the switching frequency past 25 KHz, so as to be inaudible to people and pets.

A switching regulator usually has an output filter capacitor.

Meanwhile, the frequency required to make a light appear steadily on is usually in the range of 52 to 66 Hz. 72 Hz is generally enough to appear steadily on, but a few people need more, rarely past 120 Hz. There is an Energy Star proposal for LED general purpose lighting products to not have significant flicker at frequencies lower than 150 Hz (mentioned a week or so ago in sci.engr.lighting).

That won't improve efficiency. The high efficiency of a switcher is because for part of each cycle, the output current does not involve the power supply, but is maintained by the inductor through the catch diode in a closed loop excluding the power supply. This gets average input current to be less than the output current.

- Don Klipstein ( snipped-for-privacy@misty.com)