Forums

Any linear solution will waste you huge ammounts of power.

For example, let's say that you connected 3 strings in series to achieve a load voltage of around 22V. When the current on all the LEDs in series/parallel is 1A, you would be wasting almost 6W on the elements between your source and the LEDs, be them resistors and/or linear regulators.

So, for this kind of high power application, your best bet is to go to switching solutions.

For example, take a look at this component:

http://www.ti.com/product/tps40211#description

Or you can take a look at the complete line of solutions for LED driving:

http://focus.ti.com/paramsearch/docs/parametricsearch.tsp?family=analog&familyId=480&uiTemplateId=NODE_STRY_PGE_T

Switching regulators take too long to get through approvals.

Anyway, the leds are always flashing, and the flash duty is 0.3 and the overall power is just  26W.

6W is dissipated in the linear regulator.

The  battery on the van is re-charged by the alternator so theres no problem.

Our  flash sequence is 44ms ON, 10ms OFF, 44ms ON, 10ms OFF, 44ms ON, 298ms OFF ,

..THEN REPEAT.

..Switch mode solutions that flash on/off this quickly need lots of components to stop them  overshooting when flashing back on so they are not cost effective.

It is still not clear for me your configuration. You write 26W, so I assume approx. 7.5V x 3.2A (all "strings" in parallel?, although numbers don't close).

But you could put 3 strings in series, and making two groups of 3 strings, int that case it would be 22.5V x 1.06A total current, and the same power.

Providing 1.06A would be well in the realm of the old faithfull LM317 configured as current source, which I think could cope easily with the current load changes.

Can you confirm or describe your LEDs configuration? It is crucial for determining Linear regulator requirements.

its six strings of three-leds-in-series

How much current for each string?

Do you connect three strings in series, then make a parallel of two strings?

All leds have equal current

There are six strings

these six are in parallel

each string has three leds in series

Total current is 3.2A....so this divides between the six strings

In that case, instead of putting all the strings in parallel, you can put 3 strings in series, with a resistor (group A). Another 3 strings in series, with another resistor (group B). Put group A and group B in parallel, their current consumption would be around 1A. Feed both groups with an LM317.

How will PWM function be handled? Is the intent to PWM a linear regulator via an On/Off (or Enable) pin? If so, then you need to be sure that the regulator has some specification for on/off rate, or times, even if it's just 'typical'. It's not so much that regulators 'go unstable' with rapid on-off cycles, it's more that the internal on/off circuitry is not designed to be PWM'ed and they hang up.

I believe that the suggestion to put the LEDs into a ~tighter~ array has merit, and power efficiency is something we believe has value, but there may be some physical placement limitations that we are not aware of.

Another choice might be a discrete current sink circuit ... more pieces, more design work, but easily scalable.

The LM350 is turned on/off by simply using a bjt to ground the adj pin.

There is no need for any of that "resistors in series so that LED currents are balanced".

I assure you that LED Vf's are much tighter than we are lead to believe...........LEDs can  be paralleled!

....without any current equalisation circuitry, and from a single current source, try it and see, they are well toleranced as long as they are all the same type......eg, all XPEWHY types