Constant-current LEDs?

Thanks for the circuit. That's four parts though, same as my circuit. Might come in handy if someone needs a more accurate current source than mine.

Cheers, James Arthur

Right.

If I understand your problem, a beta-limited current source wouldn't work because a constant Ib isn't handy. With a 2:1 binned transistor and a reasonable temperature range, constant base current and beta limiting can ensure maybe 3:1 or 4:1 collector current spread. That's good enough for, say, running a bandgap reference, or an led where brightness variations don't matter much.

Beta limiting is like anything else: if the numbers work, you can do it.

Jfets sometimes have a 10:1 Idss spec spread at room temp. THAT's bad!

That's a cute idea, they come in SO-89, and the typical bias current is 4-5mA!

Minimum bias current isn't spec'd, which is slightly uncomfortable, but might be okay. I don't imagine it'd ever be half IRL, but they do surprise us when they update parts, don't they?

Cheers, James Arthur

V(br)gss is 20V, a little low here--it could see >22V normally, and >27V if the LED buss is shorted. I like the topology though. FETs are nice for bootstrapping.

Cheers, James Arthur

In production?

Cheers, James Arthur

Yep. On a lark I scanned through Idss for Fairchild's JFETs last night. Surprisingly, I still managed to sleep.

There's no clear winner to the immediate contest. I like the NXP 10mA part, except for the 10mA and for it being boutique-y. The LM317L with a resistor is next. Phil's 78L05 is a temptation, but I can't pull the trigger without a minimum Ibias spec. Designing to depend on that wouldn't fly.

The original's a kludge, but it's rugged & jelly-bean.

I'll have to find out the assembly cost of adding a BOM and PnP line item versus placing two more parts.

Cheers, James Arthur

Thermal runaway, at 5 mA? Unlikely. The transistors ARE isolated, of course, or the six pins wouldn't have six different labels. We know it's not monolithic, because there's no designated 'substrate' pin.

I'd make it 58C versus 44C. The current ratio would go from 1:1 with matched temperature to 1:1.11 (so, about ten percent). According to that FLIR measurement, it's not a big problem.

Not a 'very good' mirror, but adequate. It doesn't hurt, in this case, to add emitter resistors. A fraction of a volt drop in those is generous.

Oh you mean I was supposed to read the original problem description? Sorry about that :)

Make a constant current switcher using a solenoid coil and a normally closed reed switch.

HTH

Cheers

tched temperature

ent, it's not a big problem.

It's much worse than that. Near room temperature, the mirror ratio shift wi th delta-T is

dR/dT = 2.1 mV/K / 25.6 mV = 8.2%/K.

Theta is usually around 300K/W. Without degeneration, the output device run s away when its dissipation gets to roughly

P_crit = 1/(theta dR/dT) = 35 mW,

which at 5 mA is only 7 volts. Half a volt or so of degeneration will fix it, as you say, but it's a real problem.

Cheers

Phil

I never said it was practical. :)

But hey, shunt [voltage] regulators are practical, in some circumstances. At the time, I thought it was interesting that there should be both shunt and pass type voltage and current regulators. I'd never seen this type before, so I dare claim it is original.

Tim

I like the symmetry of this one, the zeners come in at less the resistors in small quantity. Precision should be adequate for driving an indicator LED. Three parts on bom. Dunno about your headroom, a done deal at 5.5V.

Please view in a fixed-width font such as Courier.

. . CCR . | . | . ---+--- . | | . --- [680] . ZD / \ | . --- | . | |< ZD = BZX84C2V4 or equiv. . +-----| T1 . | |\ TEMPCO -1.6V/oK TYP. . | | . \| | . T2 |-----+ T1/2=BC847BPDW1T3G or equiv. .

Thermal runaway need not be destructive, except to the functionality of the circuit.

I would wild-guess that the cross-theta ( (c) (TM) ) is maybe 150 K/w, so the "dual" transistor is maybe 2:1 better than separate ones.

Ok, thanks for another pointless circuit... ;)

This one creates a more deterministic voltage divider across the zeners, pretty much guaranteeing start-up, but it's one more part, and just a little more current:

Please view in a fixed-width font such as Courier.

. . CCR . | . | . ----+---- . | | . --- [680] . ZD / \ | . --- | . | |< ZD = BZX84C2V4 or equiv. . +-------| T1 . | |\ TEMPCO -1.6V/oK TYP. . | | . +--[47k]--+ . | | T1/2=BC847BPDW1T3G or equiv. . \| | . T2 |-------+ .

Huh, that's kinda cure Fred. You've got equal currents flowing down each "leg". Do you need to double the R's to get 5mA? (2.5mA from each leg.) Or I'm missing something.

George H.

ote:

rs in small quantity. Precision should be adequate for driving an indicator LED. Three parts on bom. Dunno about your headroom, a done deal at 5.5V.

"leg".

VZ=2.4V and Vbe=0.66 for (2.4-0.66)/680=2.6mA each emitter for a 5mA total. I added a 47k across collectors for start-up, so now an impure CC with abou t 47k output impedance- which is not bad for a LED driver. Advantage is it handles a LOT of voltage, and power dissipation is divided equally between the two transistors- each conducting half the current at half the voltage- so runs pretty cool.

Does that start up reliably?

Replace one of the zeners with an LED! Or both.

There are a number of LDOs available, most with much lower bias currents than 5 mA so that can be ignored. The TI LP2951-33DRGR might do the job with a 30 volt max at about $0.25 qty 1000. What was your minimum voltage again?