How to drive high powered LEDs as efficiently as possible

I have come across some very high powered LEDs.

My question is: for these high powered LEDs how can I create a driver circuit that will be energy effiecent or is there a driver for high powered LEDs that I'm just not finding.

The LEDs will eventually be used in a grow light for an aquaponics system so effiecency is essential.

-
I seriously doubt one LED operates at 40 V. –  Olin Lathrop Jun 28 '12 at 23:10
Maybe he meant 4.0V... –  Chris Harris Jun 28 '12 at 23:43
... or, may be, it's an assembly with multiple LEDs in series. –  Nick Alexeev Jun 29 '12 at 3:10
Do you mean 0.8A or 0.08A? –  jippie Jun 29 '12 at 7:12
From what I'm told, it operates at 40V .8A and they were very very expensive to buy –  Mike Jul 2 '12 at 1:36

For best overall LED efficiency (which is your stated aim) you want high efficiency LEDs plus high efficiency drivers. There are may high power LED modules available but they in many cases use LEDs that are less efficient than the best available. No amount of efficient driving of a very low efficiency LED will make up for its low efficiency.

You mention 0.8A x 40V (which will use multiple LED dies or separate LEDs) = 32 Watts. This input power could be obtained with about 5 to 8 of the top available LEDs and it will often be better to use the best available if top efficiency is wanted.

It will probably be most cost-effective to buy a commercially available LED driver rather than making your own, looking for good efficiency. ebay lists a vast quantity of possible candidates

You also need to specify your energy source. Is this 110 VAC mains, or a 12 Volt battery or?

The standard way of driving LEDs correctly and at maximum efficiency is straightforward (fortunately).

• Determine required LED operating current.

• From data sheets, determine maximum voltage drop that LED will have at desired current.

• Provide a switched mode constant current source that will produce the required current at at least the maximum possible voltage.

• Extra points:

• Minimise LED temperature with cooling

• Operating an LED at below its maximum current ratings will result in somewhat increased efficiency - perhaps 10% to 20% more light out per Watt in at 20% of full ratings compared to full power.

• LED lifetimes increase with decreasing temperature and with decreasing current. While more current will usually result in more heat produced, These effects are independent of each other.

To get an LED driver of essentially any desired rated current. use of an external MOSFET of suitable rating, driven by a controller that sits, will allow whatever current the MOSFET can handle. This can be very large if desired.

For very best efficiencies use "synchronous rectification" - this replaces rectifier diodes with switched MOSFETs and gives lower losses.

Efficiencies around 85% from a low voltage DC source to an LED string will be achievable.Over 90% in selected cases and with care.

Comment on the above and we can provide better information.

-

The cost of consumer commodity PSU's for PC's range from cheap and dirty $50/500W to 1200W @$300 Corsair Pro, so ranging from $100/kW to$250/kW. But LED approved drivers & PSU's tend to cost much more than this.

My suggestion is to use a good consumer 85W "universal laptop charger" (ULC)

• They range in price and this year I've purchased ones for $35 @ 85W, or$0.41/W qty=1.
• The key point here is LED drivers, supplies etc tend to cost as much or more than the LED Engines where here, I propose a cheap consumer solution.
• Power LEDs range $1~$1.5/W and higher for improved quality
• Expect these costs to shrink < $0.10 to$0.50/W in volume buys.
• I have several installed running non-stop for a few years now.
• Heatsink costs have a wide variation due to choices of large material costs with massive fins or moderate size with force air DC fans.

If you find a cheaper complete power solution for LED Engines, let me know.

• Some non-OEM chargers have all std DC voltage settings & plugs for laptops.
• I use the 4.5A @ 19V ULC's for 6x1W white or blue on aluminum clad PWB's
• Shown below with 6x3 serial string and 6x1 string on aluminum boards
• I mount these under aluminum 0.5"x1" extruded end strips uses for tiles for elimination of LED glare and additional heat-spreading and the mount under exterior hand or deck rails.

• Moisture protection is added and essential for external use.

• Used 13 strings of 6 x 1W LED MCPCB's (78 LEDs) in parallel on 19V @ 4.5A (85W)
• Result was cool LED's, cool PS, but almost like daylight, so it was dimmed to 50% with 5Ω, 25W added. Efficiency was not a concern. Cost \$1.5
• 1W per linear foot under handrail or one strip every 5~6 feet
• 400~500 Lumen per 6 x 1W strip each hidden under the handrail.
• Results exceed user's expectations, pc's below prior to modification/ installation.

To use this for powering LED's depends on your expertise in matching an array of LEDs to the V-I operating points of the ULC. If you can buy from LEDs from the same batch, they V-I curve will be matched, but Disti's refuse to do this, so it takes purchasing skill, otherwise you will have issues.

Basically, it's all about having good low thermal resistance and raising circuit ESR slightly above nominal ESR characteristics with same batch sourcing on LEDs to achieve low cost reliable operation. You need to understand open circuit voltage a rise is expected with no load by the load regulation or ESR of the charger. 2% typ. or 100mΩ for 100W PS on long cables.

Ticky Tackie Technical secrets for Powering any LED Engine,

Now let's examine @Mike's desire to use Cree's 45W Xlamp cXa2011 LED

All PN diodes have a negative temperature coefficient, k (Shockley Effect) or voltage drop due to temperature rise. The fear with designers is with CV, is "thermal runaway" so they only consider constant current drivers whereas if you have tight control of a batch and sufficient ESR with low Tja, it is stable. When ambient & junction gets hotter, the LED's internal junction voltage drops demanding more current from the voltage difference across the ESR inside.

I'll show how to overcome this effect on the XLamp and in another article show how to measure junction temperature using the LED itself as a thermometer with the V-T curve which is standard in hospitals now.

• from curves on p.5, using conservative specs.
• 40.0V TYP @ 270mA ∂V/∂I = ESR = 7Ω at 10.4W TYP. (p.5 spec)
• Thermal resistance jcn-case = 0.4 °C/W TYP.
• Junction thermal voltage, k = -35 mV/°C. TYP. (p.2 spec)
• We know that heatsink design is crucial and Rja is sum of all thermal resistance from junctions to ambient.
• Let's choose 2 °C/W as heatsink + interface starting point, so Rja = 0.4+2 = 2.4°C/W

- Note: that a decent CPU heatsink is ~ 0.2°C/W for a forced air CPU heatsink ) and an excellent one is <<0.1 °C/W.

• V forward = 40V typ, 48V max @ 270mA this is a wide range from batch to batch - So you need a good source LED engines from same batch or ask me.

• Note this a string of 12 chips in one hence 40V or 3.3V/chip

Thermal equilibrium exists where there is an intersect between the linear temperature with thermal resistance and the quadratic effects of If^2*Rs) If ambient is too high with high thermal resistance, runaway can occur. It can be prevented with low Thermal resistance and adding ESR in the cable.

-