LED reference current source

Good paper, thanks for the reference.

Cheers

Phil Hobbs

But an LED forward is not "stiff", so where's the benefit? Or is this just one of those "fun things to do"? ...Jim Thompson

We used to run 50 mA through the Cree blue SiC led's as a VME bus-activity indicator, a pair of F38s and a 33 ohm resistor or something like that. As time went on and led's got better, we started blinding our customers. We use about 1 mA now, into the Osrams.

I got dark adapted and tried to find the minimum current that a green Avago led needs to produce detectable light. It was around 800 pA.

Here are some numbers for the Osram LB T67C series, beautiful parts.

Wow, that's impressive, except that I need PNP in this case. I figured I'd use an MMBTH81, min beta 60, kinda high capacitance but tolerable in this application, since I'm charging a capacitor anyhow. I'm doing a one-shot sort of thing that's not too critical, pulse width programmable from 2 ns to 200 ps.

PNPs suck. NE97833 has the speeed and capacitance of the NESG4030, but typical beta is 30. Maybe I could live with that, since beta changes, wot, 1% per degree or something.

Great data; thanks for that one.

My orange Osram may compensate the Vbe pretty well. My board is only

The dynamic impedance is pretty low, so in the voltage-reference mode, you can wrap an op amp around it to provide its own bias. The benefit is mainly getting low noise at low impedance and low voltage without needing BFCs. You can probably do a series-shunt feedback approach to combine the V_BE drop and constant excitation current, but I haven't looked at it in any detail.

Cheers

Phil Hobbs

...

=A0 =A0 ...Jim Thompson

yeh, from Johns data it looks like they about par with Zeners in dynamic impedance

-Lasse

In my circuit, I think you can add a resistor from the PNP emitter to ground, to cancel the non-zero dynamic impedance of the LED. Too bad I'm out of room on this board. My supply is pretty stable anyhow, and I don't need 10-bit accuracy on the timing here.

Red led? I've seen this many times in Elektor.

-- Failure does not prove something is impossible, failure simply indicates you are not using the right tools... nico@nctdevpuntnl (punt=.)

--------------------------------------------------------------

It's a power-on indicator, too!

This may be due to human color perception sensitivities to wavelength shift in the red region, which is quite poor per unit wavelength shift. Discrimination between nearby wavelengths in the red region is not nearly as good as elsewhere. Did you use a spectrophotometer?

Jon

I've

;-) I used the above circuit with green (SiC) LEDs directly off a LiIon battery to keep the intensity reasonably constant. A 20% variation over the

The problem I had with Osram was that they melted in the RoHS process. KingBright had a much better package.

on? I've

LEDS are - incidentally - photodiodes, so if you can see it, it can see enough of the external illumination to compromise its value as a voltage reference. This has already been pointed out elsewhere in this thread.

-- Bill Sloman, Nijmegen

tion? I've

I've seen learning remotes that used the IR led as a receiver when learning codes from other remotes, but I wonder how much external light will affect a LED that is already on.

might wanna check if that current source gets a 50/60Hz ac on it when exposed to light

-Lasse

I've

Yes, it has. What is your contribution?

Numbers matter. Nanoamps over milliamps equals parts per million.

It's easy enough to try. I'm soldering up a breadboard now.

I've

LEDs are really crappy photodiodes, though, because they're direct-bandgap devices with only a few nanoseconds' minority carrier lifetime. High-level injection conditions (e.g. a small diode with a few milliamps of forward bias) make them a lot crappier.

It would be very difficult to get more than a few hundred nanoamps out of a LED used as a photodiode in indoor ambient conditions, even with the box taken apart. So assuming it has a few milliamps' forward bias, that's maybe 1 part in 10**4 if you're unlucky, i.e. about 2.6 uV out of a volt or so. Since nobody is proposing to use LEDs for references for

Cheers

Phil Hobbs

I've seen learning remotes that used the IR led as a receiver when learning codes from other remotes, but I wonder how much external light will affect a LED that is already on.

might wanna check if that current source gets a 50/60Hz ac on it when exposed to light

Also check under CFL light. They put out a lot of trash.