Q: how find specific LED

[Robert Baer]

I had one red T-1 3/4 (5mm) clear LED that was rather bright at 50uA
and it failed. Have no other than this.

Mouser online catalog was not helpful, and i seriously doubt DigiKey
would be better.
I see specs like 45mcd unknown current, 5000mcd 30mA, 2000mcd 20mA,
high intensity red, etc.

I got a little desperate, bought various Cree 5mm clear red LEDs
apparently all 500mcd 20mA...but dim as all heck at 50uA.
One of their standard LEDs was twice as bright as some of their high
intensity LEDs.
And i stupidly thought that Cree had brighter LEDs.
ROHM and Everlight look better, but i need to use 250uA to drive them
to visually similar intensity WRT my "reference".

Hell, the LED in the Dollar Tree LED Flashlight and Lantern
Combination SKU: 286458 visually equaled the intensity at 25uA (white,
clear, gigantic size).

Would like red, 5mm, bright at 50uA.

Ideas?
Thanks

 

Keep looking. Buy some Christmas light on clearance?

 

[Winfield Hill]

Robert Baer wrote...

Would like red, 5mm, bright at 50uA.

The LED brightness should be proportional
to current, even down to 50uA. So you can
compare them at a standard current, like
20mA. Did you consider beam width? Your
bright LED may have had a beam-narrowing
lens. Also, LEDs with scattering material
in the plastic are easily visible from all
angles, but will appear much dimmer.


--
Thanks,
- Win

 

[George Herold]

On Tuesday, March 10, 2020 at 11:43:23 PM UTC-4, Robert Baer wrote:

I had one red T-1 3/4 (5mm) clear LED that was rather bright at 50uA
and it failed. Have no other than this.

Mouser online catalog was not helpful, and i seriously doubt DigiKey
would be better.
I see specs like 45mcd unknown current, 5000mcd 30mA, 2000mcd 20mA,
high intensity red, etc.

I got a little desperate, bought various Cree 5mm clear red LEDs
apparently all 500mcd 20mA...but dim as all heck at 50uA.
One of their standard LEDs was twice as bright as some of their high
intensity LEDs.
And i stupidly thought that Cree had brighter LEDs.
ROHM and Everlight look better, but i need to use 250uA to drive them
to visually similar intensity WRT my "reference".

Hell, the LED in the Dollar Tree LED Flashlight and Lantern
Combination SKU: 286458 visually equaled the intensity at 25uA (white,
clear, gigantic size).

Would like red, 5mm, bright at 50uA.

Ideas?
Thanks

Digikey's search engine is ~100x better than mouser's.
I'm not sure any place will give you the intensity at such low current.

Is there some reason you can't increase the current?

George H.

 

[George Herold]

On Wednesday, March 11, 2020 at 9:39:10 AM UTC-4, Winfield Hill wrote:

Robert Baer wrote...

Would like red, 5mm, bright at 50uA.

The LED brightness should be proportional
to current, even down to 50uA. So you can
compare them at a standard current, like
20mA. Did you consider beam width? Your
bright LED may have had a beam-narrowing
lens. Also, LEDs with scattering material
in the plastic are easily visible from all
angles, but will appear much dimmer.


--
Thanks,
- Win

Hi Win.. I have some data on older LED's that show the above is not
true. The LED's I looked at were linear with current above about 1 mA.
At lower currents light output drops off faster.
https://www.dropbox.com/s/atyo4uvsb09fgd7/LED-PD.BMP?dl=0

The linear part is in the upper RH corner.

George h.

 

[Winfield Hill]

George Herold wrote...

On Wednesday, March 11, 2020, Winfield Hill wrote:
Robert Baer wrote...

Would like red, 5mm, bright at 50uA.

The LED brightness should be proportional
to current, even down to 50uA. So you can
compare them at a standard current, like
20mA. Did you consider beam width? Your
bright LED may have had a beam-narrowing
lens. Also, LEDs with scattering material
in the plastic are easily visible from all
angles, but will appear much dimmer.

Hi Win.. I have some data on older LED's that show the
above is not true. The LED's I looked at were linear
with current above about 1 mA. At lower currents light
output drops off faster.
https://www.dropbox.com/s/atyo4uvsb09fgd7/LED-PD.BMP?dl=0

The linear part is in the upper RH corner.

Nice data. But I'd say your plots show that
it's largely true, especially when compared to
huge intensity changes required to affect our
human perception of brightness. He can compare
at 20mA.


--
Thanks,
- Win

 

[George Herold]

On Tuesday, March 10, 2020 at 11:43:23 PM UTC-4, Robert Baer wrote:

I had one red T-1 3/4 (5mm) clear LED that was rather bright at 50uA
and it failed. Have no other than this.

Mouser online catalog was not helpful, and i seriously doubt DigiKey
would be better.
I see specs like 45mcd unknown current, 5000mcd 30mA, 2000mcd 20mA,
high intensity red, etc.

I got a little desperate, bought various Cree 5mm clear red LEDs
apparently all 500mcd 20mA...but dim as all heck at 50uA.
One of their standard LEDs was twice as bright as some of their high
intensity LEDs.
And i stupidly thought that Cree had brighter LEDs.
ROHM and Everlight look better, but i need to use 250uA to drive them
to visually similar intensity WRT my "reference".

Hell, the LED in the Dollar Tree LED Flashlight and Lantern
Combination SKU: 286458 visually equaled the intensity at 25uA (white,
clear, gigantic size).

Would like red, 5mm, bright at 50uA.

Ideas?
Thanks

Robert, searching on led efficiency and low current I found this
nice page from Don Klipstein
http://donklipstein.com/ledlocu.html

GH

 

[George Herold]

On Wednesday, March 11, 2020 at 10:49:32 AM UTC-4, jla...@highlandsniptechnology.com wrote:

On 11 Mar 2020 06:38:53 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:

Robert Baer wrote...

Would like red, 5mm, bright at 50uA.

The LED brightness should be proportional
to current, even down to 50uA. So you can
compare them at a standard current, like
20mA. Did you consider beam width? Your
bright LED may have had a beam-narrowing
lens. Also, LEDs with scattering material
in the plastic are easily visible from all
angles, but will appear much dimmer.

I've wondered if, or when, an LED just stops making photons as the
current (and voltage) go down. Any shunt ohmic component would shut it
down as the terminal voltage gets too low.

The slope for the low current end is about 3/2 (100 times the current
gives 1000 times the light.) (note log-log plot)
But here's data from an IR led.
https://www.dropbox.com/s/i1am61wga5fhpax/IR-PD.BMP?dl=0

Which seems to poop out faster.

I'm guessing newer LED's are better.. but don't know.
(data posted to dropbox in 2013)

George H.

I have seen light from a green LED at 1 nA, and the threshold may well
have been my eyes. A good LED and a PMT and a lockin would be
interesting.

A good LED is visible in modest office light at 1 uA.

A plastic machine screw might make a decent light pipe for the front
panel of a rackmount box. Just a few mA of drive will make a blinding
amount of light from a good LED, so optical efficiency doesn't matter.



--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

 

On 11 Mar 2020 06:38:53 -0700, Winfield Hill <winfieldhill@yahoo.com>
wrote:

Robert Baer wrote...

Would like red, 5mm, bright at 50uA.

The LED brightness should be proportional
to current, even down to 50uA. So you can
compare them at a standard current, like
20mA. Did you consider beam width? Your
bright LED may have had a beam-narrowing
lens. Also, LEDs with scattering material
in the plastic are easily visible from all
angles, but will appear much dimmer.

I've wondered if, or when, an LED just stops making photons as the
current (and voltage) go down. Any shunt ohmic component would shut it
down as the terminal voltage gets too low.

I have seen light from a green LED at 1 nA, and the threshold may well
have been my eyes. A good LED and a PMT and a lockin would be
interesting.

A good LED is visible in modest office light at 1 uA.

A plastic machine screw might make a decent light pipe for the front
panel of a rackmount box. Just a few mA of drive will make a blinding
amount of light from a good LED, so optical efficiency doesn't matter.



--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

 

On Wed, 11 Mar 2020 08:42:56 -0700 (PDT), George Herold
<ggherold@gmail.com> wrote:

On Tuesday, March 10, 2020 at 11:43:23 PM UTC-4, Robert Baer wrote:
I had one red T-1 3/4 (5mm) clear LED that was rather bright at 50uA
and it failed. Have no other than this.

Mouser online catalog was not helpful, and i seriously doubt DigiKey
would be better.
I see specs like 45mcd unknown current, 5000mcd 30mA, 2000mcd 20mA,
high intensity red, etc.

I got a little desperate, bought various Cree 5mm clear red LEDs
apparently all 500mcd 20mA...but dim as all heck at 50uA.
One of their standard LEDs was twice as bright as some of their high
intensity LEDs.
And i stupidly thought that Cree had brighter LEDs.
ROHM and Everlight look better, but i need to use 250uA to drive them
to visually similar intensity WRT my "reference".

Hell, the LED in the Dollar Tree LED Flashlight and Lantern
Combination SKU: 286458 visually equaled the intensity at 25uA (white,
clear, gigantic size).

Would like red, 5mm, bright at 50uA.

Ideas?
Thanks

Robert, searching on led efficiency and low current I found this
nice page from Don Klipstein
http://donklipstein.com/ledlocu.html

GH

We sometimes put a flat surface-mount LED on a PC board, so a uP can
blink it, or to indicate that a power supply is up or an FPGA actually
configured. 100 uA seems about right; much more is annoying if you are
probing the board.

One recent board has an LED to warn techs that there is still 48 volts
on the board. We have one biggish depletion fet to discharge all the
caps (takes about 30 seconds) and a second depl fet driving the
warning LED at about 1 mA, which will be bright. The LED runs at
constant brightness until the very end of the discharge.



--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

 

[mpm]

On Wednesday, March 11, 2020 at 12:42:06 PM UTC-4,

We sometimes put a flat surface-mount LED on a PC board, so a uP can
blink it, or to indicate that a power supply is up or an FPGA actually
configured.

This x1000. (Big thumb's up!)

Speaking of which, here's the sort of stuff that can drive you crazy:
We recently made a small power distribution board. Nothing special, just an "in" and several "outs", all thru-hole connectorized to make things neat and tidy. This was for low-voltage DC on the order of a couple of amps per output, so some decent size connectors one could easily solder to the boards.

I asked the design engineer to place a simple LED & dropping resistor on the power rail so we could tell at a glance whether the board was powered-up, once installed in the product.

You guessed it: He picks a tiny 0402 size LED and resistor that you need a microscope to solder. WTF? There was absolutely no reason not to go old-school here. Everything else on the board is easy through-hole stuff. It's ridiculous!!

For a while, I thought he was deliberately screwing with me. But I later discovered that he honestly had not given it any thought whatsoever and just reverted to his normal practices (most of his work is on tiny, high-frequency stuff, where board space is a downright scarcity). He's also the kind of engineer who doesn't seem to understand (or at least, greatly under-estimates) the time it takes to build something, and why "manufacturing for production" is important.

Overall though, this guy is arguably borderline brilliant. He just has a few quirks! Sorry for the mini-rant.

 

[mpm]

On Tuesday, March 10, 2020 at 11:43:23 PM UTC-4, Robert Baer wrote:

I had one red T-1 3/4 (5mm) clear LED that was rather bright at 50uA
and it failed. Have no other than this.

Interesting situation.
It might be helpful if everyone reading this went into their grab-bag of T-1 3/4 (5mm) red LED's to see if they can find a particularly bright one at 50 uA.

If nothing else, it might be a welcomed change from everyone bitching about politics, or bickering over the "OT topic du-jour".

I can attest that these LED (below) are SUPER-BRIGHT... at 40 mA!!
Link: https://www.digikey.com/products/en?keywords=C503B-RCN-CW0Z0AA1-ND

I was recently tasked with converting one of our products to allow for 24-hour battery back-up operation and in the process discovered that we were wasting about 3 watts just on these LED's (and about 4.25 watts total). The design engineer had simply spec'd the wrong dropping resistor, and worse, that resistor was a thru-hole SIP pack. You start to worry about I2R heat when they're all lit up.

Luckily, it turns out that both the LED's and the SIP's can handle the current, which I guess is why nobody ever noticed it until now, or why we never got any units returned under RMA's. But they sure are BRIGHT!!

I'll pull one out of inventory tomorrow and see if it will even light up at 50 uA and let you know. Maybe?? It's a nice LED, with even lighting - not sure why we originally selected it in this design (may have been availability, or we had some left over from some other project?)

 

[Robert Baer]

George Herold wrote:

On Tuesday, March 10, 2020 at 11:43:23 PM UTC-4, Robert Baer wrote:
I had one red T-1 3/4 (5mm) clear LED that was rather bright at 50uA
and it failed. Have no other than this.

Mouser online catalog was not helpful, and i seriously doubt DigiKey
would be better.
I see specs like 45mcd unknown current, 5000mcd 30mA, 2000mcd 20mA,
high intensity red, etc.

I got a little desperate, bought various Cree 5mm clear red LEDs
apparently all 500mcd 20mA...but dim as all heck at 50uA.
One of their standard LEDs was twice as bright as some of their high
intensity LEDs.
And i stupidly thought that Cree had brighter LEDs.
ROHM and Everlight look better, but i need to use 250uA to drive them
to visually similar intensity WRT my "reference".

Hell, the LED in the Dollar Tree LED Flashlight and Lantern
Combination SKU: 286458 visually equaled the intensity at 25uA (white,
clear, gigantic size).

Would like red, 5mm, bright at 50uA.

Ideas?
Thanks

Digikey's search engine is ~100x better than mouser's.
I'm not sure any place will give you the intensity at such low current.

Is there some reason you can't increase the current?

George H.

The idea is to have a battery monitor light that will not unduly run
down the battery.

 

[Robert Baer]

Winfield Hill wrote:

Robert Baer wrote...

Would like red, 5mm, bright at 50uA.

The LED brightness should be proportional
to current, even down to 50uA. So you can
compare them at a standard current, like
20mA. Did you consider beam width? Your
bright LED may have had a beam-narrowing
lens. Also, LEDs with scattering material
in the plastic are easily visible from all
angles, but will appear much dimmer.

The majority of the LEDs i bought were clear lens; 2 exceptions: one
dark red, the other "clear" was not specified and was slightly milky.
Angle was all over the place - 10 degrees to 60 degrees.
My comparison is what i see straight on; I jacket the LED in
reflective metal.

I came up with a crude measurement fixture using a CdS photocell that
is repeatable.

Data specs are crap:
The Everlight HLMPK101 high intensity red at 46 mcd (3mm red lens, 60
degree) gave 4.56Megs at 25uA is usable for my purposes.
The VCC VAOL-3GRE4 high intensity red at 2000 mcd (3mm clear lens, 30
degree) gave 2.30Megs at 25uA and is also usable for my purposes.
The CREE C503B-RBS-CY0Z0AA1 standard LED at 5000 mcd (5mm clear lens)
gave 0.33Megs at 25uA and 1.6Megs at 10uA is a lot closer to what i was
looking for.
The TT Electronics standard LED unspecified (5mm clear lens) was
comparable at 25uA and 10uA

So...a no-spec standard LED comparable with a fancy CREE kills the
idea of using specs, and is topped off by the 46 mcd VS 2000 mcd comparison.

Going to use the TT since the end/top is flat and produces a better
visual "spot".

 

[Robert Baer]

Winfield Hill wrote:

George Herold wrote...

On Wednesday, March 11, 2020, Winfield Hill wrote:
Robert Baer wrote...

Would like red, 5mm, bright at 50uA.

The LED brightness should be proportional
to current, even down to 50uA. So you can
compare them at a standard current, like
20mA. Did you consider beam width? Your
bright LED may have had a beam-narrowing
lens. Also, LEDs with scattering material
in the plastic are easily visible from all
angles, but will appear much dimmer.

Hi Win.. I have some data on older LED's that show the
above is not true. The LED's I looked at were linear
with current above about 1 mA. At lower currents light
output drops off faster.
https://www.dropbox.com/s/atyo4uvsb09fgd7/LED-PD.BMP?dl=0

The linear part is in the upper RH corner.

Nice data. But I'd say your plots show that
it's largely true, especially when compared to
huge intensity changes required to affect our
human perception of brightness. He can compare
at 20mA.

Suggest you look at the "comparison" of a 45 mcd VS 5000 mcd LED.
Everlight 45 mcd: 10uA reads 18.5Meg, 25uA reads 4.56Meg, 50uA reads
1.73Meg.
ROHM 5000 mcd: 10uA reads 7.6Meg, 25uA reads 1.86Meg, 50uA reads 0.41Meg.
Note the 45 mcd at 50uA reads near what the 5000 mcd at 25uA.

 

[Robert Baer]

George Herold wrote:

On Wednesday, March 11, 2020 at 10:49:32 AM UTC-4, jla...@highlandsniptechnology.com wrote:
On 11 Mar 2020 06:38:53 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:

Robert Baer wrote...

Would like red, 5mm, bright at 50uA.

The LED brightness should be proportional
to current, even down to 50uA. So you can
compare them at a standard current, like
20mA. Did you consider beam width? Your
bright LED may have had a beam-narrowing
lens. Also, LEDs with scattering material
in the plastic are easily visible from all
angles, but will appear much dimmer.

I've wondered if, or when, an LED just stops making photons as the
current (and voltage) go down. Any shunt ohmic component would shut it
down as the terminal voltage gets too low.
The slope for the low current end is about 3/2 (100 times the current
gives 1000 times the light.) (note log-log plot)
But here's data from an IR led.
https://www.dropbox.com/s/i1am61wga5fhpax/IR-PD.BMP?dl=0

Which seems to poop out faster.

I'm guessing newer LED's are better.. but don't know.
(data posted to dropbox in 2013)

George H.



I have seen light from a green LED at 1 nA, and the threshold may well
have been my eyes. A good LED and a PMT and a lockin would be
interesting.

A good LED is visible in modest office light at 1 uA.

A plastic machine screw might make a decent light pipe for the front
panel of a rackmount box. Just a few mA of drive will make a blinding
amount of light from a good LED, so optical efficiency doesn't matter.



--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

What did you use ("photocurrent")? Motofotoplier?
The best i had for sensitivity was that ugly CdS fotoresistor.

 

On Wed, 11 Mar 2020 19:33:04 -0700 (PDT), mpm <mpmillard@aol.com>
wrote:

On Wednesday, March 11, 2020 at 12:42:06 PM UTC-4,

We sometimes put a flat surface-mount LED on a PC board, so a uP can
blink it, or to indicate that a power supply is up or an FPGA actually
configured.

This x1000. (Big thumb's up!)

Speaking of which, here's the sort of stuff that can drive you crazy:
We recently made a small power distribution board. Nothing special, just an "in" and several "outs", all thru-hole connectorized to make things neat and tidy. This was for low-voltage DC on the order of a couple of amps per output, so some decent size connectors one could easily solder to the boards.

I asked the design engineer to place a simple LED & dropping resistor on the power rail so we could tell at a glance whether the board was powered-up, once installed in the product.

You guessed it: He picks a tiny 0402 size LED and resistor that you need a microscope to solder. WTF? There was absolutely no reason not to go old-school here. Everything else on the board is easy through-hole stuff. It's ridiculous!!

For a while, I thought he was deliberately screwing with me. But I later discovered that he honestly had not given it any thought whatsoever and just reverted to his normal practices (most of his work is on tiny, high-frequency stuff, where board space is a downright scarcity). He's also the kind of engineer who doesn't seem to understand (or at least, greatly under-estimates) the time it takes to build something, and why "manufacturing for production" is important.

Overall though, this guy is arguably borderline brilliant. He just has a few quirks! Sorry for the mini-rant.

We have at least three design reviews for everything.

PDR preliminary design review before the real design

CDR critical review when the schematic is done

PCB review of the PC layout when it's ready to Gerber.

so we tend to catch things like that.




--

John Larkin Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet.
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

 

[Martin Brown]

On 12/03/2020 02:17, Robert Baer wrote:

George Herold wrote:
On Tuesday, March 10, 2020 at 11:43:23 PM UTC-4, Robert Baer wrote:
I had one red T-1 3/4 (5mm) clear LED that was rather bright at 50uA
and it failed. Have no other than this.

    Mouser online catalog was not helpful, and i seriously doubt DigiKey
would be better.
    I see specs like 45mcd unknown current, 5000mcd 30mA, 2000mcd 20mA,
high intensity red, etc.

    I got a little desperate, bought various Cree 5mm clear red LEDs
apparently all 500mcd 20mA...but dim as all heck at 50uA.
    One of their standard LEDs was twice as bright as some of their high
intensity LEDs.
    And i stupidly thought that Cree had brighter LEDs.
    ROHM and Everlight look better, but i need to use 250uA to drive
them
to visually similar intensity WRT my "reference".


Digikey's search engine is ~100x better than mouser's.
I'm not sure any place will give you the intensity at such low current.

Is there some reason you can't increase the current?

George H.

  The idea is to have a battery monitor light that will not unduly run
down the battery.

Use a decent white or a green LED then with narrow angle beam.
Plenty of them (even from two decades back) are visible at 10uA.
(that is you can see that the die is lit up in daylight)

Putting a bleed resistor across the switch of an LED torch makes them
findable in total darkness after a power cut.

Mine with 10uA once dark adapted you can see to move in the room by the
light of 4 now quite ancient white LEDs. They have improved still more.

--
Regards,
Martin Brown

 

[Martin Brown]

On 11/03/2020 14:49, jlarkin@highlandsniptechnology.com wrote:

On 11 Mar 2020 06:38:53 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:

Robert Baer wrote...

Would like red, 5mm, bright at 50uA.

The LED brightness should be proportional
to current, even down to 50uA. So you can
compare them at a standard current, like
20mA. Did you consider beam width? Your
bright LED may have had a beam-narrowing
lens. Also, LEDs with scattering material
in the plastic are easily visible from all
angles, but will appear much dimmer.

I've wondered if, or when, an LED just stops making photons as the
current (and voltage) go down. Any shunt ohmic component would shut it
down as the terminal voltage gets too low.

It shouldn't make any photons once the terminal voltage drops below the
potential difference needed to emit the photon (plus a tiny bit).

That is about 3v for a blue photon and a shade under 2v for red. Like
the photoelectric effect with metals but in reverse.

I have seen light from a green LED at 1 nA, and the threshold may well
have been my eyes. A good LED and a PMT and a lockin would be
interesting.

A good LED is visible in modest office light at 1 uA.

Although there are plenty of not that good LEDs about. Even two decades
ago the standard white and some blue and green LED dies were visible in
daylight at about 10uA. the best have improved an order of magnitude
since. Some of those intended for high current maximum brightness seem
to have parallel resistive leakage which hammers low current efficiency.

A plastic machine screw might make a decent light pipe for the front
panel of a rackmount box. Just a few mA of drive will make a blinding
amount of light from a good LED, so optical efficiency doesn't matter.

Increasingly efficient modern LEDs with the drive current not adjusted
from their nominal 10mA are blindingly bright indoors.

--
Regards,
Martin Brown

 

[Martin Brown]

On 11/03/2020 15:28, George Herold wrote:

On Wednesday, March 11, 2020 at 10:49:32 AM UTC-4, jla...@highlandsniptechnology.com wrote:
On 11 Mar 2020 06:38:53 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:

Robert Baer wrote...

Would like red, 5mm, bright at 50uA.

The LED brightness should be proportional
to current, even down to 50uA. So you can
compare them at a standard current, like
20mA. Did you consider beam width? Your
bright LED may have had a beam-narrowing
lens. Also, LEDs with scattering material
in the plastic are easily visible from all
angles, but will appear much dimmer.

I've wondered if, or when, an LED just stops making photons as the
current (and voltage) go down. Any shunt ohmic component would shut it
down as the terminal voltage gets too low.
The slope for the low current end is about 3/2 (100 times the current
gives 1000 times the light.) (note log-log plot)
But here's data from an IR led.
https://www.dropbox.com/s/i1am61wga5fhpax/IR-PD.BMP?dl=0

Which seems to poop out faster.

I'm guessing newer LED's are better.. but don't know.
(data posted to dropbox in 2013)

I suspect the curve tends towards a slope of 2 when the drive current is
low and source impedance becomes comparable with the leakage resistance
in the device so that both I and V are falling faster and together.

Your real physical LED is an ideal perfect LED in parallel with a high
but not quite negligible leakage resistance which acts as a potential
divider once the drive current gets small.

--
Regards,
Martin Brown