Calling out to Weezard for LED advice

[DreadedHermie]

Hey, (duh) doesn't a power supply + current regulator = "LED driver?" Dang, I'm dense sometimes. Jes' call me "Pb."

[Weezard]

I get it. I was about to suggest you call that setup a CFL-assisted-by-LED grow, though.

Yep, have thought along those lines but am obviously working my way through some other details first. (Like building a basic, working light!) Rectangular aluminum stock with forced air cooling through the center seemed more realistic for me to attempt; my plumbing seems to cause flooding.

Well, I think you just opened up some new possibilities. For some reason I was thinking the current regulators needed to be close to the load; I'd planned to mount 'em on the 'sink near the LED's. I prolly misinterpreted the data about filter caps needing to be close to the current regulator(s).


I had planned on keeping the power supply out of the grow area, but are you telling me I can keep the current regulators adjacent to the power supply?

Zackly!
The closer the better.
Eliminates the need for a tantalum cap on the input to the CCR. That in turn eliminates the need for a diode.
Same with the output of the CCR.
You do not need a capacitor because of the nature of the load.
With no output cap to reverse bias the regulator, no "protection diode" is needed. Keep it simple.
16 ga. wire would suffice, but I'd use 14 ga. stranded and a plug/socket right at the lamp.
I used a 4 pin hard drive connector with a seperate ground return for the Red and blue supplies.
The CCR will hold whatever current you wish when using thinner wire, but why waste watts in da wire?

I could mount them in a separate case and keep the light nice and neat. This would also let me use the 338's in a TO-3 package, which seem much more popular, and look like they'd dissipate heat much better.

You got it straight away!
Impressive.
It's gonna be fun working with you.

BTW, I don't see any "proper operation assumes adequate heatsinking" type disclaimers on any of the 338 family, unlike the LM317's.

Look harder, I'm sure it's there somewhere.
Might have to dissipate 16W. or more! So a heatsink is in order. And, in this case, bigger is better.

I also can't figure out what the voltage drop across the 338's is-- in one graph it looks like they'll want 1.5-2 volts at around 4 Amps. Help??

If you allow for a 3 volt drop across the regulator and the 1.25V. drop across the sense resistor in your design, you should have enough headroom to keep the 338s happy.

Now, I'm havin' fun.
Weezard

[DreadedHermie]

Eliminates the need for a tantalum cap on the input to the CCR. That in turn eliminates the need for a diode.
Same with the output of the CCR.
You do not need a capacitor because of the nature of the load.
With no output cap to reverse bias the regulator, no "protection diode" is needed. Keep it simple.
16 ga. wire would suffice, but I'd use 14 ga. stranded and a plug/socket right at the lamp.

Okay, now I'm excited because I actually understand all that.

What I'm not sure about (among many other things) is what the 15w. 660's are actually going to be pulling current-wise. They're rated at 1500mA, so if I'm running them conservatively are they pulling a bit over an amp or so each, or are these things actually meant to be run at 750mA? (I'm cautious of power ratings, I remember audio amps that were in reality about 30 watts RMS being rated at 600 watts "intermitent peak-to-peak" or some other hogwash.):wtf:

If the reds are indeed capable of pulling an amp and a half each, my 5.5 amp power supply isn't as overkill as I'd thought.

[DreadedHermie]

Might have to dissipate 16W. or more! So a heatsink is in order. And, in this case, bigger is better.

Dang! I hate to waste 16W., especially by turning it into heat! Can I improve my design? Maybe find a 24v. supply instead?

[DreadedHermie]

And I can think so much more clearly. :thumbsup:

This is why I have that bench supply! I can use it to see what my LED's want to be fed, and buy a power supply that fills that need.

The LM138 is good for higher temps than the 338 (150 C. vs, 125 C.) so I'll use those. Found some real sporty dedicated heatsinks for the TO-3 package. My driver will look like the Batmobile, or perhaps the Nautilus (esp. if I do any plumbing! )

So, Weezard, my last question, hopefully, before I start buying components for the red circuit, is:

What watt rating to I need for R1 and R2? It seems like folks are using a 5K ohm Radio Shack pot for R2. These are only rated at a half watt. However, I see some devices on your light that look like 7W resistors. How much juice is going through the R1-R2 resistor combo?

Mouser has a 5 watt wirewound pot-style variable resistor for only $4, but only up to 2K ohm. Would you recommend recalculating R1 value and using resistors/pots rated at 5 watts, or are half-watt parts sufficient here? I know the LM138 is going to have to handle some heat, but what about R1 and R2? Are they putting the brakes on any significant current flow, or are they just reference / sense components? And what the heck is a sense resistor, anyway? (Well, that was more than one question, I guess.)

[Weezard]

And I can think so much more clearly. :thumbsup:

This is why I have that bench supply! I can use it to see what my LED's want to be fed, and buy a power supply that fills that need.

Goodonya!

The LM138 is good for higher temps than the 338 (150 C. vs, 125 C.) so I'll use those. Found some real sporty dedicated heatsinks for the TO-3 package. My driver will look like the Batmobile, or perhaps the Nautilus (esp. if I do any plumbing! )

So, Weezard, my last question, hopefully, before I start buying components for the red circuit, is:

What watt rating to I need for R1 and R2? It seems like folks are using a 5K ohm Radio Shack pot for R2. These are only rated at a half watt. However, I see some devices on your light that look like 7W resistors. How much juice is going through the R1-R2 resistor combo?

Mouser has a 5 watt wirewound pot-style variable resistor for only $4, but only up to 2K ohm. Would you recommend recalculating R1 value and using resistors/pots rated at 5 watts, or are half-watt parts sufficient here? I know the LM138 is going to have to handle some heat, but what about R1 and R2? Are they putting the brakes on any significant current flow, or are they just reference / sense components? And what the heck is a sense resistor, anyway? (Well, that was more than one question, I guess.)

A sense resisror is so called because it hooks to the sense input of a regulator.

Here's how the LM317 CCR circuit works;

The lm317 is designed to be a variable voltage regulator
and it "wants to see" 1.25Vdc between the output pin and the ADJ/Sense pin. before it will begin regulating anything.
We take advantage of those facts for our current regulator.
We tie a restor to the output pin and draw the current to our load through that resistor.
If we run a wire from the load side of the resistor to the adjust pin, that pin will sense the voltage drop across the resistor.
So, ohm's law.
If we want a 1 ampere current limit, we use a resistor value that will drop 1.25V at 1 amp.
Once the adjust pin senses 1.25v.it regulates voltage to keep that voltage drop, and thus the current, constant.
1 A. X 1.25 V. = 1.25 W.
While a 2 watt resistor would suffice, a 5 or 10 watt resistor will run cooler and thus regulate tighter.
I had some 10 Watt, 1 ohm power resistors laying around so I thermal glued them to the heatsink.
Overkill? Perhaps,. But overkill is a good thing in this case .
Gives me about 1.2A. which is exactly what I wanted.
Seems to be the sweet spot in photons per watt.

What is this R2 you speak of?
The current limit configuration only requires one resistor.
[attachment=o213372]
And, I think this all applies to the 138 series as well.

It's very simple.

[DreadedHermie]

Here's where I'm getting that R2 stuff:[attachment=o213389]

[attachment=o213390]


Here's the TO-3 heatsinks:[attachment=o213391]

Looks like R2 is like a trimmer to make the voltage variable(?). Of course, once it's dialed in you can measure the resistance the pot's providing and replace the pot with a fixed resistor, unless you want to be able to change how much your blues are putting out, for example. Somehow I think PWM might be more efficient at doing this, but I'm not that far along yet.

[DreadedHermie]

[attachment=o213392]

What about this bad boy? (The 12v. version.) Looks like all the goodies are built-in. A little pricey, and maybe marginal headroon for 4 660's. Would you still need CCR's on top of this?

[Weezard]

Here's where I'm getting that R2 stuff:[attachment=o213389]

[attachment=o213390]

Oh, O.K., those configurations are for regulating voltage.

We want to regulate current.

Lose R2 completely, we don't need it.
And R1 should be about 1/2 an ohm. for 2 amps or 1/4 ohm for 4 amps.
If you can't find the exact value you want, you can always wire larger values in parallel.
2. 1 ohm resistors in parallel = 1/2 ohm, etc.

The schematic in my last post is complete.
One regulator, one resistor, no "salad". :jointsmile:
It's all you really need for a non-inductive, non-capacitive load.

Here's the TO-3 heatsinks:[attachment=o213391]

Looks like R2 is like a trimmer to make the voltage variable(?). Of course, once it's dialed in you can measure the resistance the pot's providing and replace the pot with a fixed resistor, unless you want to be able to change how much your blues are putting out, for example. Somehow I think PWM might be more efficient at doing this, but I'm not that far along yet.

Quite right! PWM is a better choice, but, in a pinch, a high current low resistance trimmer with a limit resistor will work for varying current.

You might be over-thinking the CCR part.

Disclaimer:
This works a treat with the LMx17 series and should work just as well with the LMx38. But, I have yet to try it with the x38s.

See easier that you thunk, ya?

Weezard

[DreadedHermie]

You might be over-thinking the CCR part.

Yes, I am positively devoted to overthinking (everything) now, not after I've smoked $150+ of LED's!

The schematic in my last post is complete.
One regulator, one resistor, no "salad".
It's all you really need for a non-inductive, non-capacitive load.

Very elegant in its simplicity. Without your guidance I would be making this way harder than it needs to be. Thanks again!


BTW, read somewhere you're on nursing duty this week. Please give my best to your patient, and I hope you're better at nursing than I am at EE.

I found a CRAZY thick aluminum cake pan last night, 13' x 9", except it's also non-stick. What's your take on grinding the teflon off so I can epoxy to it?