Constant-current LEDs?

The FET's nicer--more accurate. But, accuracy doesn't matter here. Just an indicator...

Cheers, James

spike

oks

imit

rces, and feed the fourth with 5mA to set up the common base voltage. Getti ng the constant 5mA is left as an exercise for the reader, but a PNP, a zen er and a couple of resistors would do it, or a 10mA CCD and a 130R parallel resistor.

with the emitter (which could be a resistor array), and a 1.23V voltage ref erence to drive all the bases would work too. Two duals, a quad resistor pa ck and one voltage reference is four parts.

Unfortunately the LED indicators are powered by three 0-30V separate rails and the cathode has to float, hence the desire for a two-terminal CCS.

The LM317L looks like a decent choice. But I've got to weigh non-electroni c considerations like adding to the BOM, real estate, etc.

Cheers, James Arthur

Yeah I know... I was thinking some one would put it all in a three terminal pac for me.... Hey I'm a consumer, it's something I want... :^) George H.

On-Semi NSI45020

Diodes Inc. AL5809

And Similar parts from TI, Supertex etc...

A quick search at Digikey shows many with widely available current ranges..

Steve

I don't believe that is what Tim is referring to as a "shunt" device. The current regulating diode you are talking about is a two terminal device that goes in series, no? I guess my reply was not clear I was responding to that part of his post.

--

Rick

Consider several mosfets, or a multiple-section switch, switching to ground. Each drives one LED through a resistor, from the erratic V+ supply.

Some intelligent machine, a uP maybe, sees V+ and varies the switch duty cycle. The V+ sense mechanism could be the obvious ADC, but could be simpler.

This is an "inductorless switcher", simple and inefficient.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Thanks. I'd included the 10mA version of the OnSemi part in my original post. The AL5809 is slick, but 15mA--I'd really like 5mA. I might go for

10mA, but 15mA is really too much heat for my sealed box.

I'd looked under PMIC/Current regulators. Hadn't thought to look under 'LED drivers,' thanks. Still none @ 5mA.

NXP has an integrated version of my circuit(!)--NCR401U. 6-pin package, adjustable, with a 10mA (min) internal current-setting resistor. But it can't handle 22V x 10mA at 70*C, with the heat-sinking available. It's only derated to handle 150mW @ 70*C, not 220mW.

That's the problem with 10mA--it doubles the dissipation.

Thanks for the ideas, James Arthur

Awww crackers! I switched datasheets while flipping from browser to PDFs. This last comment applies to the OnSemi NSI50010, NOT the NXP part. Sorry.

Cheers, James

So, the current source you have, runs into a multiple-output current mirror (four matched transistors, one for input and three for output). You then turn the LEDs off with a switch in parallel. It takes 20 mA all the time, though.

Discrete current mirrors usually don't work very well. Real monolithic quads are expensive. Most multiple transistors are separate chips, with terrible thermal connections.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

That's what Bill suggested, but I can't do it. Each LED indicates power status for a different power source--there's no common.

A diagram would probably help.

V1 V2 V3 (0 V~> LEDs --- --- --- . | | | LED_BUS . '----+----+-------------->.>--/\/\/--. | . | External LED .-' . V~> ^ ZD1 . --- | 5.1V | === ===

The on-board LEDs drive the LED bus, which drives an external indicator if and only if one of the power sources is present.

This is a wired-OR bus shared potentially among many boards, so it's not okay to drive the LED bus unless power is actually present. Simple cc sinks referenced to the LED bus would power the bus if one of the inputs failed, e.g.:

| | | |/ |/ |/ Vr --|----|----| |>. |>. |>. | | | R1 R2 R3 | | | '----+----+----> | .-' ^ | ===

That's not allowed. More complicated circuits fix that but I'm trying to save parts, not add more!

Cheers, James Arthur

There's utterly no need for matched transistors--that's way overkill. A volt across an emitter resistor fixes Vbe variations more than well enough to run an LED!

Cheers, James

I'm surprised JL has not posted his "beta limited" driver, just a transistor with binned beta BCX-something IIRC. I think it ended up with a 2:1 led current variation.

--

John Devereux

??? Maybe someone can come up with a modified planar layout that would guarantee a more behaved Idss spread ("by design")?

Nah, just one.

In the left-hand case, Vbe regulates Io, so it's pretty sloppy, but not too bad for a first try. On the right, a diffamp compares to a bandgap reference for very stable regulation. (Of course, you can also use the TL431 directly, as it's already an "NPN" device.)

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

I also have an old circuit for that:

The huge ratio on the transformer actually suggests an unshielded (SMT or THT) inductor plus a turn or two on the PCB. Alternately, a current transformer, perhaps?

I don't think efficiency was ever much better than running it at DC without positive feedback, though.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

Each one could be wired as a monostable, with the time set by a resistor from +V rather than from VREF.

Some buck controllers do this, including quite advanced ones like LTC3810. It's ideal because, for fixed delta-I, the high side on-time must be inversely proportional to the voltage difference.

You could use a multi-RS-F/F like CD4043, where the reset comes from the Q output, delayed by such an RC. The set inputs are all triggered from a common clock (let's throw in an LMC555 for cookbook cred), which are gated by AND gates for what lights are actually active.

Meh, still sucks. There's probably an LED driver chip that does it, and includes SPI or whatever.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

TL431s are 11 cents in quantities over 100 at Mouser.

Sorry, 15 cents.