# LM317 vs LM350 vs LM338

#### Willheee

Joined Aug 21, 2011
5
Hello, I'm a first time poster, occasional lurker. Done some maths on the LM317,350 and 338 to drive LEDs, was wondering how close to the ball park I am or if I'm wayyyyyy out.

I have project which requires the powering of 150 Red and 50 Blue LEDs (Luxeon Star style, 1w). I was intending on making 2 panels each of around 100 LEDs (dependant on whether less makes the maths easier and the whole project less of a balls-ache).

Red Vf = 2.0-2.6v If = 350mA (actually around 0.8w)
Blue Vf = 3.2-3.8v If = 350mA (actually around 1.2w)

I've been looking at these 3 regulators for constant current sources for the LEDs and have come to the point where I'm wondering how many I can power from one regulator using the circuit (figure 4, Page 2) of the following PDF: http://www.onsemi.com/pub_link/Collateral/AND8109-D.PDF

I'm figuring I can get around 7 on one string using the LM317 with a 37V input and 4 strings (because the LM317 can only do 1.5A) so 28 LEDs total (blue) OR 12 red (for the sake of safety) on each string and 4 strings of these, so 48 red. Blue and red on seperate regulators of course..

The LM350 can draw 3A which would effectively double that of the LM317 and would allow for 8 x 7 blue (56) or 8 x 12 Red (96)

The only reason I'd want to use the LM338 would be to keep the current I'm using well away from the max (5A) that it can supply so as to hopefully keep the temperature down.

Are there any other considerations i need to make? Obviously everything will be thermal pasted, heatsinked and fanned for cooling, but is there any risk of thermal runaway or these chips going into thermal shutdown? Do these figures seem viable on one regulator or should I be looking at splitting it up a bit more? and how big is the chunk of metal likely to be to heatsink 100 of these?

Love you long tiem!!

Will

Last edited:

#### SgtWookie

Joined Jul 17, 2007
22,230
You really don't want to use linear regulation for high-power LEDs, as your efficiency will be terrible. You'll expend a lot of power just heating up the regulators and transistors.

A large high-power LED array will be difficult to keep cool all by itself. You will essentially need to mount the entire array on a large heat sink with plenty of fins to get rid of the heat.

Instead of linear regulation, you really need a switching regulator solution. There are dedicated LED current controllers available. What you might spend extra up front for the parts, you'll get back in lower electric bills within a couple of months - and the savings will continue.

You need to give more information - like what are you planning on using to power this array? Voltage/current output of the supply?

#### Willheee

Joined Aug 21, 2011
5
Hello Sgt Wookie,

As far as a power supply goes, I was going to be using a 60V 10A bench supply (set to the required V and I for the circuit) until I'd figured out a more permanent solution. A possibility was to use a computer PSU (or 2) with a sufficient power output and possibly modify if need be.

At first, it wasn't so much the power consumption that was necessary to be reduced, just that the heat coming off the LEDs (and supply) was less than that of a conventional light source of the same luminosity. Having said that though, it has become clear that the cooler everything is, the longer it will last so I will start looking at more efficient means to this end.

Spotted this guy on eBay, http://www.ebay.com/itm/DC-24V-15A-...lectrical_Equipment_Tools&hash=item5640d797ba

Would that be a tad more suitable and less wasteful than going for the proposed cheap option? Also looks like there's an adjust terminal on there, is that likely to be used in a similar way to that of the LM317 etc... Never really used ballast or LED drivers out of the box, only really just got my head around voltage regulators. (out of interest, was my theory roughly correct despite not considering efficiency?)

Thank you very much again :]

Will

#### Willheee

Joined Aug 21, 2011
5
oh quick other question,

At what quantity of 1w leds on a circuit such as the one I proposed does the design become too inefficient to implement? are we talking 1-5 LEDs being the limit before heat becomes too much of an issue, or literally from the get-go.

Thanks

#### SgtWookie

Joined Jul 17, 2007
22,230
OK, that's about 182W total for the typical led Vf, or around 91 Watts per panel, not including the efficiency (or lack thereof) of the current regulators. 85% or better is a reasonable assumption for switching regulators. You'd get far less efficiency from the linear regulator scheme.

#### SgtWookie

Joined Jul 17, 2007
22,230
OK, with a 24v supply, you could have six red and two blue in a series string.
That will leave ~3.2v for "overhead", or remaining voltage to drop with a buck-type current regulator, or (optionally) a linear regulator.
That means 25 strings total. I guess you'll need to break them up as 12 strings on one panel, and 13 on the other.

#### SgtWookie

Joined Jul 17, 2007
22,230
Well, I came up with a switching buck current regulator schematic, but it uses 19 components - and you'd basically to make 25 copies of it. That's a bit much.

#### Willheee

Joined Aug 21, 2011
5
OK, with a 24v supply, you could have six red and two blue in a series string.
That will leave ~3.2v for "overhead", or remaining voltage to drop with a buck-type current regulator, or (optionally) a linear regulator.
That means 25 strings total. I guess you'll need to break them up as 12 strings on one panel, and 13 on the other.
Compared to the linear regulator method, what sort of efficiency is gained by using (and I assume you're talking about) the 360w driver on eBay?

Well, I came up with a switching buck current regulator schematic, but it uses 19 components - and you'd basically to make 25 copies of it. That's a bit much.
Would it be possible to hear what you had in mind? I'm up for learning about this theory, I mean, it's all good to know, that's why we learn. What better way than to do something 25 times over

Also, I've just quit smoking so willing to get stuck in because it'll hopefully take my mind off it...

Thanks again

#### SgtWookie

Joined Jul 17, 2007
22,230
OK, with 6 reds and 2 blues in series, you'll have a total Vdrop across them of ~20.8v.
With a 24v supply, this leaves 3.2v remaining. This is barely enough to use a single LM317 and a resistor per string; if it's even enough - it is very borderline. There would not be enough "headroom" to use the Motorola/OnSemi application note.

Anyway, your efficiency is basically (power in / power used in the load) * 100.
With a linear regulator, you already know that the 3.2v that is left over will have to be dropped by the regulator; at 350mA current, that's 3.2*.35 = 1.2 Watts that's thrown away, and 20.8 * .35 = 7.28 Watts used in the load (your LEDs); so the total power in (less minor losses in the wiring) is 1.2 Watts + 7.28 Watts = 8.48 Watts.
(7.28/8.48)*100 = 85.85% efficient.

However, in the circuit I'm simulating, 7.638W in, 7.355W out, for (7.355/7.638)*100 = 96.3% efficient. Notice that my input power is 7.638W vs the 8.48W required by using an LM317; there's over a 10% gain in efficiency using a switching "buck" regulator. Since you will have 25 strings of LEDs, that will reduce your power consumption by 30 Watts with no difference in light output.

If you wanted to use the circuit in the OnSemi/Motorola app note, you'd have to either drop one LED per string, or increase your supply voltage by at least 1 more volt.

BTW, searching for the power supply part number "S-360-24" brought up this page: http://www.eforchina.com/v/79754
Same power supply voltage and current rating, \$34.35 shipped from China.
One thing about these supplies is that no efficiency rating is specified.

#### Willheee

Joined Aug 21, 2011
5
I did notice that there was no actual efficiency rating on those products which put me on to thinking i might be able to find something better elsewhere, but if it'll do what is needed then it'll have to do.

I think I get what you're saying, basically to wire the LEDs up in strings between the outputs of the 24v supply, but in terms of the current limiting, is the 360w going to do all of that (or rather, should it be doing all of that) with no other circuitry? or when you said

remaining voltage to drop with a buck-type current regulator, or (optionally) a linear regulator.
were you refering to taking my +v line, into buck (or linear if need be) converter, to a string of LEDs, then to -v.

Or would using the afforementioned supply with it's adjust terminal and 3.9Ω resistor to get me 320mA for each string?

#### SgtWookie

Joined Jul 17, 2007
22,230
The output of the 24v supply is voltage regulated, not current regulated.

LEDs need to have their current regulated. Their Vf (forward voltage) can and does vary over temperature; it decreases as the temp of the LED increases.

So yes, I've been talking about taking the +24v supply into a buck or linear regulator, one per string, then to -v.

You could use a fixed resistor, which would be cheap, but would provide no active current regulation.

I've attached an image of the buck circuit I've mentioned.
LED1 through LED6 plus Rmakeup represent your string of 6 red and 2 blue LEDs; I didn't have a model of your exact LEDs so simply used what I had. I've adjusted the total voltage drop across the simulated LEDs to match your actual LED specifications.

I have no idea what your level of experience with electronics is. I suggest that this circuit is not for "beginner" level hobbyists.

#### Attachments

• 95 KB Views: 178