LED regulator

Although efficiency could be higher... MOSFETs FTW.

Tim

-- Deep Friar: a very philosophical monk. Website:

There's pros and cons to all the branches here. Gel cells are expensive, you WANT it to stop discharging when V hits 5 or so, or it'll kill an expensive battery. So, the best solution will have a low V cutoff (or at least a warning). Simple limiting resistor won't do that.

SMPS chips (I'm thinking of MC34063) for DC/DC conversion can do it, if you wind your own inductor and add a current-monitor amplifier (a low-power op amp with a sense resistor) and cutoff detector (second op amp, looking at a reference to implement the gel cell protect function). It's likely that among the many pages of power supervisor/SMPS chips there's a readybuilt solution, but the most likely one you'll hear about is 'use a PIC'.

I'd switch back to three alkaline cells and use a resistor, for simplicity.

....

In my experience, L-A batteries tolerate being run flat far better than newer technologies. They don't like being LEFT that way, but...

...

Gee, that's a Grampa's axe solution.

First, gel cells are not expensive. Every 6 or 8 years, the one I have won't hold a charge; & I buy a replacement at a hamfest for a few bucks.

I do not want an alkaline battery solution; I was seeking an upgrade to the existing system prompted by the lamp failure, and the gift of the array. I was trying to SWAG if there was some reason to go beyond the obvious.

If it's not worth the work to re-wire to series and use an LED driver chip or a DC-DC inverter, then you might as well go with an LM317 set to 3.3 volts. No need for LDO at that Vout.

Ed

Err, don't I want to run them constant-current?

While it's amusing to consider series rewiring; my eyes, hands and patience are likely not up to it. But, I SWAG that would be in the region of 100 volts...

You willing to wind a coil? I think there was quite a discussion about this in .basics, last summer. Look up "joule thief" or some such.

Used a BJT with a simply-wound transformer (I've got a bunch of these laying about, now) and a resistor. We used it as a boost converter for that discussion, but I don't see why right now that the other end of the LED combo you have couldn't be instead connected to the (+) rail instead of ground and use the LED across the primary for a kind of inverter of sorts (raises above the +6V) rather than the boost position it was in, before.

Jon

This is not my favorite, but its simple. My favorite has one more transistor.

I'd post my favorite, but I do not have a web page:

Nice to see. Thanks.

A question comes to mind, though. I think I'd prefer power BJTs over power FETs because they are lots cheaper and readily available. Assuming the OP is willing to throw away power in a linear dropout regulator, why prefer a FET over a fairly simple BJT c.c. circuit that might use either one BJT and two diodes or two BJTs, one small signal and the other a power BJT? Perhaps I didn't read enough of the text but is there a good reason to choose the approach shown there, instead?

Jon

My favorite is a npn-pnp bipolar pair and 4 resistors and a reference diode.

However I have to assume the OP is uing big leds, perhaps luxeons etc So I went with the posted fet design, which actually is quite crude. Perhaps he needs a "buckpuck" or one of the 8$ premade led drivers from Digikey... I mean, a single supply opamp, a sense resistor and a few other parts would give him a nice regulated driver that has no temperature compensation issues. Oh, and power fets are cheap, but he needs to find one speced for linear operation of the gate..

Steve

more cookbook circuits:

Bob Pease:

Steve

I think I remember Pease's, from some months ago in this group. Thanks for the reminder.

Jon

What? Okay. Accepted. BJTs are just cheaper, then.

Jon

You can't have your cake and eat it too. You have 30 LEDs in parallel - you can't control the current through them unless you re-wire.

Best to run LEDs at constant current, but that requires a re-wire. Next best is to run at constant voltage, with current limiting between the source and the LED, but you can't do that without re-wiring. So that leaves you with one option: run them at constant voltage at the low end of Vf.

Right. Or, boost to say 20 volts, and use 6 strings of 5 LEDs in series, each string with a resistor to limit current. Or to ~12V with ten strings of 3 LEDs and a resistor in series with each string - or some variation. One chip plus a few external components can provide the boost. The LM2698 is one example. Trouble is that these solutions are useless if it is not worthwhile to rewire the LEDs.

Ed

Caution: The real intent of an LM334 is as a _temperature_ sensor.

Actually VREF = kT/q*ln(10) ...Jim Thompson

Sometimes, depends on the application.

The context for my response was this thread's application.

Jon

Then do not state it as general case.

I hope both of us will be able to uphold your standard of always writing all of the caveats all of the time -- every time -- even in thread context. :?

Jon

I try, i fall down sometimes; i expect to be called on it.