LED reference current source

xpect

I may have such an overwhelming desire - it would come as a surprise to me and my acquaintances, but perhaps we are a more robust lot than American pantywaists - but anything I may achieve in that direction pales into insignificance when compared with your extravagant pratfalls and malicious irrelevancies.

Find a specific idiocy in the post you responded to, and identify it. I haven't had a good laugh for several days, and the only way you are going to deliver on that claim is going to be with a John Fields/Jamie level of reading incomprehension.

-- Bill Sloman, Nijmegen

It seems to me you want the lowest voltage LED that has the magic tempco to produce a constant delta-V. The lowest voltage implies the smallest resistor going from the voltage supply to the BJT emitter. It follows that the smallest resistor has the least thermal noise. Thus the IR led looks like something worthy of examination.

I'm not so sure I'd run one at 20ma unless someone has lifetime stability data on this.

For sure the low current requirement is an issue. I've done references under 10uA. The problem is there are products that have to be in "shutdown", but the bandgap is still on.

If the process was old, i.e. no SOI or even EPI, then only the peripheral pins are designed to handle injection. [Guard rings, etc.]

But I'm not accustomed to getting parts back from the fab with a window. Was their UV prom on the chip?

enough/low

Award.

You obviously think you can do better then. Go for it of you want.

?-)

amps

high end

the

could

costs way

on

to

Mmmm. Flicker and popcorn noises. Even the shot noise at those = currents, not pretty.

Now i will remember even better. Thanx.

?-)

The lowest LED voltage implies the higher LED biasing resistor, which in turn means it will have the least current variation.

It also seems to me that you could bootstrap the LED biasing to make it "stiffer" via negative feedback. Add another BJT and resistor to make another current source. Feed that current source into the LED biasing resitor. Now it would have to be a small fraction of the current that would otherwise flow in the LED biasing resistor else it would perhaps have start up issues.

The negative feedback goes as follows. For increased LED voltage, the extra current source BJT shunts some current away from the LED, which in turn reduces the current in the LED and thus the voltage.

That wasn't very good; Sloman remains in the lead.

Am 11.09.2012 20:30, schrieb John Larkin:

If it is not too rf-ish and there is not much space you might consider this one

(infineon BCR400W)

regards, Gerhard

ow

d.

John Larkin modestly - if incorrectly - underestimates his own contributions.

-- Bill Sloman, Nijmegen

"You are in a maze of twisty little IKYABWAIs, all alike"

At least the current source seems to be very stable.

LEDs do degrade in brightness over time. I wonder if their forward voltage changes in any systematic way.

It's hard to understand what that part actually does. If the schematic is literal, its TC, as a current source, will be pretty bad. There's no data on its actual current output, or on capacitances.

LED degradation is strongly associated with bad 1/f noise. (Lots of degradation mechanisms show 1/f behaviour, e.g. electromigration of interstitials and defect propagation.) If you run the LEDs gently, and check a few from the reel for bad 1/f, you should be okay for quite awhile.

Assuming the noise power predicts the aging rate, you could do an accelerated aging test and find out what Vf does with degradation. It would be interesting to know whether the increased series resistance or the enhanced nonradiative recombination dominates.

Cheers

Phil Hobbs

Interesting. That makes sense--you don't want to have to wait for the RC filter to settle when you turn the part back on.

That's one reason why I really like the LM329--it's quiet and stable, and the TO92 package makes it mostly immune to board stress. (It's better with some extra lead length, so a spacer helps.) Stress due to changes in humidity would theoretically still be an issue, since the case is epoxy, but that's more a 30-nanohertz problem (summer vs winter) than the 100-microhertz-ish things I typically care about.

Cheers

Phil Hobbs

I remember it as a project in one of the UK electronics magazines.

IIRC one problem was that the physical layout did not necessarily match the address/data lines, so the image had to be reconstructed with some trial and error for a particular chip.

enough/low

Award.

Compound semiconductors, like gaasfets, have lots of crystal defects and charge trapping sites. A gaasfet gate bias can jump by many millivolts overnight. I'd suspect that forward voltage is less sensitive to defects than gate bias. But again, this application isn't very critical. I'm charging a ramp capacitor that's feeding two comparators, then a flipflop. It's sort of a 200 picosecond 555 timer.

We sometimes put foam over crystal oscillators, to lowpass filter thermal noise. It can move the jitter corner frequency down by 10:1.

ill be

urrent

table

s for a

nough/low

2. >

ference

Of

ly

ar Award.

ant.

The temperature sensitivity of forward bias does seem to be sensitive to the defect level

I can see the current source driving a capacitor and two comparators, but I doubt if there's a direct connection to the flip-flop. Flip- flops don't like slow edges.

-- Bill Sloman, Nijmegen