90 Watt LED Driver - Need a bit of help here :)

[Chris Smith]

Ahh the smell of peanuts from the gallery. Fresh and roasted.

Never an accomplishment, just a pretend trashing, or was that a hallow bark from the junk yard dog?

Amateurs always speak loudest about their lack of understandings and non accomplishments in life, to fit into the clique.

[LTC]

Um - guys... no need to fight here. I do not think that PWM is not the best approach for this application although it has many other very valid uses. I have read up substantially on the circuit design for this circuit and it seems to me that the best approach is to go with a boost type circuit. The main issue is cost... and what safety elements I can add in. I assume a thermal fuse and a voltage fuse will be enough? Are there cheap, resettable thermal fuses? Keep in mind that I have either 12, 24 or 48v input and I need to drive 18 K2's using an ATX power supply. Several of the approaches I have seen are very very good but require very high input voltages or have 70% or less efficiency. Is there a simple circuit that can run all 18 LEDs in series that can use a standard ATX power supply and can drive each LED at 1200 mA? Surely this is possible with a minimum number of components. I just dont quite "get it" just yet.

[rshayes]

Hello LTC,

The most efficient way to operate the diodes would probably be a series string of 18 fed with a boost type of converter, as you suggested.

The reason that I did not suggest this, and probably the reason that dyarker didn't also, is that this would probably require the design of custom magnetic parts, such as inductors and transformers, and the design of a stable feedback loop to control the switching power supply. The efficiency would probably be around 80 percent, which is a little better than the linear approaches, but not by a factor of two or three. The peak current in your 12 volt switching device would be about 45 amps. This would require a little care in design and layout.

Some of the references that dyarker supplied show this type of arrangement and give design information.

The trade off with the switching supply is longer design time to get better efficiency. If you are planning on building many units, this may be worthwhile. For one or two units, probably not. The cost may be similar with either approach. The switching approach would use fewer parts, but custom magnetic parts will be disproportionately expensive and may make the costs about equal.

[Chris Smith]

We love to argue here on the NV,.... LTC, .....I love them more when they get an education, instead of rejecting it out right.

The whole concept of learning Im sure?

I suggest you look up the major systems already in place.

Las Vegas and others have pulse systems in place for many reasons including dimming, graphics, and longevity of the LEDS, and I think the longest working display is almost a 1/4 mile long. [about a million leds?]

I know they don’t use standard daisy chains with simple resistors, trading up for the pulse methods and it long life.

Changing a 1/4 miles worth of LEDS Im sure would take a 1/4 of a decade or so to replace Im sure, so reliability is their issue as well.

Go with the simple and unreliable[P-G], or go with the industry standard?

[ecerfoglio]

The most efficient way to operate the diodes would probably be a series string of 18 fed with a boost type of converter, as you suggested.

The reason that I did not suggest this, and probably the reason that dyarker didn't also, is that this would probably require the design of custom magnetic parts, such as inductors and transformers, and the design of a stable feedback loop to control the switching power supply. The efficiency would probably be around 80 percent, which is a little better than the linear approaches, but not by a factor of two or three. The peak current in your 12 volt switching device would be about 45 amp.

You don't have to convert 115 V AC to 12 VDC and then to 45 VDC.

You may try to modiffy your AT power supply so it provides 45 VDC:

Just locate it´s output transformer, identify its secondary windings (there should be al least one for 5 V and other for 12 V, maybe others for the negative voltages), count how many turns there are in each and remove them.

Then make a new winding with the appropiate number of turns (eg: if the 5V winding had 3 turns, and the 12 V one had 7, you will need:

3 x 45 / 5 = about 27 turns for 45 V, or

7 x 45 / 12 = 26 1/4 (about 26 turns)

Of course, you will have to replace the rectifier diodes and filter capacitors with ones rated for the new voltage, and "scale down" the feedback voltage (insert a voltage divider made with a pair of resistors) so the regulating circuit "sees" 5 V instead of your new voltage.

SAFETY NOTE:

Remember that the power supply has "live" line voltages inside

Remember that there are capacitors charged at line voltage, even after you disconect it from the mains.

USE COMMON SENSE AND SAFETY PRECAUTIONS AS IF YOU WHERE WORKING INSIDE A TV SET

[Thontia]

First, it is a pleasure to meet so many people that care about LEDs
and supplying current. Felt very alone in my experiments...

I keep hearing about the K2 LEDs from some of you guys??
Has anyone actually gotten hold of a white 140 lumen version??
I have been on backorder for MONTHS....
They are now saying JUNE for delivery...

By the way, I am a PWM kinda guy....

thanks

[Chris Smith]

They are working as we speak [University of Utah] on a new technology to produce a seven fold increase in the luminosity from the white LED, using the exact same amperage as before.

They are in the midst of offering these patents to other firms so we should see some nice NEW LEDs on the horizon.

Photonics Spectra June 2005

[LTC]

First, it is a pleasure to meet so many people that care about LEDs
and supplying current. Felt very alone in my experiments...

I keep hearing about the K2 LEDs from some of you guys??
Has anyone actually gotten hold of a white 140 lumen version??
I have been on backorder for MONTHS....
They are now saying JUNE for delivery...

By the way, I am a PWM kinda guy....

thanks

7 - 8 week leadtime on the 120s.

Gotta wait until 2007 for the 140 binned ones.

[LTC]

I really do not want to convert a 12v ATX power supply into a 48v+ supply, go the PWM route or use just resistors. I have seen one circuit that is getting close to what I need. But I really want to adhere to KISS on this one. Some concepts, while great, are taking things a bit too far.

I want to keep this SIMPLE, CHEAP and EASY. Take 12v in - drive 18 of the K2's. Is it the voltage variation that is tripping this up? Come up with a circuit that does this with minimal components, a stock ATX power supply and runs at 75% or better efficiency. Can it be done? I think so. Can I do it? Sadly, no. But I know one of you guys could do this in a jiffy. Is the transistor based circuit I saw above the best that can be done? The problem with it is the input voltage has to be pretty crazy. 12v is really preferred here but 24 or 48 is possible. Any more than that and the cost is going to go up dramatically (unless anybody knows where to get a 90+W LED driver for under $40.

Feedback - please?

[Chris Smith]

If your willing to eject precaution for dollars, then rig up three diodes in series with a resistor for each feed line or spur.

You can build a super cap [gas] tank which will resists transient changes in voltage, and drive it with several P-Supplies of the computer type, and even use a crowbar design FET or [Bi-polar] regulator to keep the voltage constant, or even a large Zener to clamp over values and spikes.

You should use a soft start value lower than norm to warm up your Diodes, and then a final current voltage setting to protect the variations you mentioned to within your specs.

Several computer P-Supplies can feed the super cap gas tank, [one farad] and the crowbar regulator will dish out a constant current and voltage supply to the diodes.

And all of this can be done for under 40 bucks.

[LTC]

I am not really interested in throwing caution to the wind here.... this circuit is designed to run 24/7 with a MTBF of 100,000 hours+. When you say that I would be ejecting precaution for dollars, what exactly do you mean? Above all else, this circuit should have no safety concerns, even in fail mode, it should sacrifice itself rather than overheat, etc. Can you elaborate a bit?

[Chris Smith]

I’ve already noted all my concerns as to the life span. High power LEDs best run single resistor configurations, and in the PWM mode.

They even incorporate a PIN diode in some cases as a feed back circuit to maintain a equal brightness over their life span.

They also run a reduced power level by a few percent, if life span is one of your essential goals. Never 100% of the recommended or allowable power ratings.

[rshayes]

The lowest stress on the LEDs and the condition used to test and estimate their life is a continuous current at the minimum value which will give sufficient light output.

White LEDs are a special case. Some designs change color when the current is changed from the value at which they are tested and sorted. If the color shift is tolerable, they can also be operated at lower currents. The color shift can be avoided by operating the diode at the test current but using PWM to reduce the average current. This will result in a lifetime intermediate between operation at full current (shorter) and operation at a continuous lower current (longer).

The 12 volt supply combined with the variation in forward voltages of the diodes forces you into a configuration using nine strings of two diodes each. A higher voltage supply would be nice, but will not be as common an item as the one you want to use, and probably much more expensive.

The most efficient method of controlling the current in each string would probably be a buck type switching regulator, one for each string. This would require a switching device, diode, inductor, control chip, and associtated components for each string. When this is repeated nine times, you will probably have about a hundred parts. You may also have to coordinate them such that they don't switch at the same time and supply enough filtering to guarentee that they don't interreact with your main power supply or each other. The efficiency of this approach would probably be around 85 percent.

Linear current sources on eack string would require a transistor and a resistor for each string and about ten or fifteen parts common to all nine strings. This is simple and reliable, but the efficiency depends on the forward voltage of the LEDs. With the upper end LEDs at close to five volts, the efficiency would be over 80 percent. With low voltage LEDs, the efficiency drops to around 50 percent.

I would sugggest the linear approach because it is simple and reliable. Further, since the control is common to all strings, there is not much additional complexity in adding a simple form of PWM. This could be used to operate the LEDs at about 70 percent brightness initially. As the diodes age, the pulse width could be increased to keep the brightness constant, untill you require full drive at the end of the LED lifetime.

No matter what you do electrically, you will have to pay a great deal of attention to heatsinks. The hotter the diodes junctions get, the shorter the lifetime and the lower the light output. If you are using forced air on the heatsink, an overtemperature sensor on the heat sink would protect against fan failure.