Constant-current LEDs?

Couldn't you just run the LEDs off a constant voltage that's pretty close to the Vf, and then switch them with a high-side MOSFET?

I've run LEDs off constant voltage regulators before and it seems to work okay. Ain't no law says you can't.

LND250s have Idss around 1.6 mA, much more repeatable than jfets.

They make nice capacitor bleeders, in series with an LED. The LED stays bright through most of the cap discharge, then winks out.

I sometimes use DN2530 for higher power things. Good for 3 watts on a copper pour.

Well, if you have a few volts extra headroom, you could run it from the ground pin of a 78L05 with its output floating. Still SO8, but only one chip. ;) They even come in BGA!

Cheers

Phil Hobbs

But that's not an idealogically pure current mirror; it takes more parts.

People sell "digital transistors" with integrated base resistors. Somebody should sell transistors with integrated emitter resistors.

Still, with a quad pack resistor for emitter degeneration, it could be reasonably competitive.

Cheers

Phil Hobbs

The other thing is a BF862 with a BAT54 in series with the source (gate connected to cathode). That comes in pretty reliably at about 3-5 mA. Using BAT54Cs, that's only 1.5 parts per channel.

Cheers

Phil Hobbs

You sort of want an LM3909, only backwards. ;)

Cheers

Phil Hobbs

That's cool.

I did something like that recently, but I also used the regulator output.

When I want to light up an LED, I look around for some current somewhere that I can hijack, instead of adding a resistor.

You can also float the adj pin of a 3T reg and use it as a higher current limiter, sort of a self-resetting fuse with advantages over a polyfuse.

Of course there are all sorts of creative ways to get input overvoltage--iirc George had a good one a year or so back. ;)

Cheers

Phil Hobbs

the 862 is only a 20V part

I don't remember an over voltage problem, perhaps you're thinking of my method for defeating the thermal/current shut down of the LM395.. which involved putting the "right" size coil/inductor on the output. Maybe you can blow up a '317 in the same way?

George H.

I tried all sorts of ways to damage an LM1117 in this configuration, and none of them worked.

Hand-select a JFET for 5 mA with source and drain shorted?

A ~10 ohm, 8 inch diameter air coil can take out an lm395... It can't shut off the current.... (Big diode in reverse across the output fixed it.)

George H.

For a 5mA source, ignore the 'thermal connections'; we don't need 1% matching, it's driving unmatched LEDs. Matched pairs aren't expensive (this one is under 0.1 USD in quantity) but it's a different story for quads or higher. So, each of two matched pairs gets an emitter degeneration resistor. 60 ohms or more should do it. The current regulator, probably cheapest to go with LM317 and a (1.25V/.005A) = 250 ohm program resistor. BOM is now: one IC, two dual transistors, three resistors. You could use a quad 120 ohm resistor pack instead (two in series for the LM317).

That data sheet carefully avoids mentioning thermal coupling. It's not even clear if the transistors are electrically isolated from one another. There is no transistor-transistor max voltage spec. There is no single-transistor theta or max power. Thermal runaway could wreck the matching specs.

Here's a thermal image of a similar NEC part:

One transistor is hot, one is cool. It wouldn't make a very good pure current mirror, little better than two separate transistors.

So, use big emitter resistors.

The LM317 and LM395 are the same chip with different metallization.

IIRC you hit the thermal limit, the chip tried to turn off, and the inductive kick killed it by overvoltage. Like I said, creative. ;)

I might well have been bitten the same way--one time I had a power stage oscillating so hard that the decoupling bead melted its way right through a Pomona clip.

Cheers

Phil Hobbs

I can only take half the credit. The lm395 drives a 50 ohm cell heater circuit with banana plugs for the heater. Like the heater, the big coil also has banana jacks, mix in physics students, and it was only a matter of time. :^)

George H.

Grins :-). It's not a bad idea really, but where would I get the constant base drive? John's circuit is for a fixed rail.

Cheers, James Arthur