LED output vs bias current

[George Herold]

So I was shinning some Cree leds onto a photodiode and measuring the photocurrent. Here’s a plot (opamp bias current ~20pA subtracted from data.)
https://www.dropbox.com/s/atyo4uvsb09fgd7/LED-PD.BMP

Below about 1mA of LED bias current the photo current increases as the 3/2 power of the led current. Which was totally unexpected! (I expected linear.)
I’m making up a hand-waving explanation that’s got a non-radiative recombination channel. And then observing that for radiatvie recombination, it’s not only the number of charge carriers at the boundary.. but they also have to find a ‘dancing partner’ on the other side. So some kind of factor that has do with hole and electron wave function overlap. And for two dimensions that could go as the square root of the number density. (I think.)

Does that make sense? Is this known?

Thanks
George H.

 

[George Herold]

Nice plot!

The existence of the knee is well known, but I never really paid much
attention to how the light output behaved down there. "Trap sites" and
"surface states" are the solid state physicist's version of "profit
taking" and "program trading".

You mean the knee at ~1mA where output becomes more linear with drive current?

I've been trying to get a square root factor to come out mathematically, from some overlap factor. But that looks to be linear too.. So the plot would have gone as I^2. not I^3/2. I'm missing something.

George H.

 

[George Herold]

We've done our own optocouplers, an IR led and a photodiode. They were linear
from mA to low currents, microamps of LED current as I recall. We have plots

somewhere. The nonlinearity was in the LEDS.

Hmm, Well for my 'silly model'* if there's not much probability of non-radiatvie recombo, then it should be linear starting at a lower current. I've got an IR led I can try.

George H.
*of course my 'silly model' is not working at the moment. :^)

 

[George Herold]

If memory serves, triode vacuum tubes follow a 3/2 power law as well.

<http://www.john-a-harper.com/tubes201/>

There may be a physical parallel of some kind at work.

Hmm, is that the space charge 'thing'?
(I'm mostly clueless about tubes. I could run the tube tester and figure out which tube to replace, but not much more.)

To throw a monkey wrench into it, here's the IR led data.
https://www.dropbox.com/s/i1am61wga5fhpax/IR-PD.BMP

It's much closer to 2 at low currents and then bends over.

Above 10mA I start to see warming effects of the led. The intensity droops with a few second time constant. I should do a pulsed measurement, but that too much work.

George H.

 

[Phil Hobbs]

This is interesting. One would expect the photodiode and LED are just
reversed processes.
I have always assumed photodiodes to be linear at low light levels and
that it has likely been
checked with photon counting. Could the beam pattern of the LED be
changing with current
level??

Paul C.

Nah, LEDs are more complicated devices by a lot--double heterojunctions
in nasty compound semiconductors, for a start.

The best photodiodes are good for 14 decades of linearity at room
temperature, if you're willing to wait. See Eppeldauer et al.,
http://tinyurl.com/q4wwyts (behind a paywall unfortunately).

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510

hobbs at electrooptical dot net
http://electrooptical.net

 

[George Herold]

Good to know on the photodiodes. Been dinking around with LEDs as SPADs
based on
a thread here some time ago. Have a handfull of Siemens GaAs SFH431
that breakdown at
25V or so giving nice 10mv pulses. I would like to experiment with some
quenching circuits
and had planed to check with an LED. Guess I'll need to monitor with a
photodiode.

Excellent! Hey try some AND113R's. I mistakenly said before that these don't work, but I was wrong! (I get several bags of red led's in and one baglooks like another... anyway.) If you then file the top of the led down you can get some decent count rates with a green led as light source. Oh one thing about spads is that the higher you bias them above the breakdown voltage the higher the efficiency... And the pulse height is basically equal to the bias voltage minus the breakdown voltage... so if you've only got 10mV crank up the bias voltage. If that causes the pulses not to quench thenmake the quecnh resistor bigger.

RE: LED nonlinearity. So imagine that an led has two ways for the charge carriers to recombine. one is e-h recombination the other some trap state or other 'thing'. But this second gives no light. Now the number of traps is fixed. And so the number of e's that go through the trap depends only on the number of e's. But to recombine with a hole (h) the electron and hole have to find each other. So the number of e-h recombo's should go something like the density squared. (I know I got a 3/2's power.. go figure.)

Does that make any sense?

George H.

 

[George Herold]

On 2013-10-04 11:52:00 -0700, Phil Hobbs said:

Good to know on the photodiodes. Been dinking around with LEDs as SPADs
based on
a thread here some time ago. Have a handfull of Siemens GaAs SFH431
that breakdown at
25V or so giving nice 10mv pulses. I would like to experiment with some
quenching circuits
and had planed to check with an LED. Guess I'll need to monitor with a
photodiode.



Paul C.

Hey are those SFH431's still being made? (I can't find 'em.)

George H.

 

[Bill Sloman]

There's a really good paper I found on the interweb

"Avalanche photodiodes and quenching circuits for single-photon detection"
Applied Optics / Vol. 35, No 12 / 20 April 1996.

Next time, quote the author's names. Sergio Cova and his team at the Milan Polytechnic are magic, and their work has been recommended here before.

http://physics.ucsd.edu/~tmurphy/apollo/doc/Cova_1.pdf

<snip>

 

[Jasen Betts]

So I was shinning some Cree leds onto a photodiode and measuring the photocurrent. Here’s a plot (opamp bias current ~20pA subtracted from data.)
https://www.dropbox.com/s/atyo4uvsb09fgd7/LED-PD.BMP

Below about 1mA of LED bias current the photo current increases as the 3/2 power of the led current. Which was totally unexpected! (I expected linear.)
I’m making up a hand-waving explanation that’s got a non-radiative recombination channel. And then observing that for radiatvie recombination, it’s not only the number of charge carriers at the boundary.. but they also have to find a ‘dancing partner’ on the other side. So some kind of factor that has do with hole and electron wave function overlap. And for two dimensions that could go as the square root of the number density. (I think.)

Does that make sense? Is this known?

Do another experiment, block or divert 3/4 of the light from the led,
does start of the linear section move.

 

[George Herold]

On 2013-10-04 12:52:21 -0700, George Herold said:


Make sense? I don't know. You might try running the same LED as
photodiode, some of them work rather well.
Use a real photodiode and an LED-as-photodiode to monitor the same
light source. Would you still get the same
non-linearity??


Paul C.

Hey you read my mind. That is the intent of the LED. (to show that a the count rate of spad is linear with light intensity.) The spad has a few issues... it's not as good at low intensity, and second it poops out at highercount rates due to the built in dead time. I've got data.. I'll have to plot it up.

On the 3/2 power law front. I got a message from someone on another forum that the 3/2 power law is known and the explanation is that the non-radiative trap recombo is *not* linear in the charge density. If one assumes a random (Gaussian) trap distribution then the number of recombo's through the traps goes as the 3/2 power of the density... Which then leaves a 3/2 powerfor the increase in the number of photons. (Though I'm hoping to find a paper.)

George H.

 

[George Herold]

Do another experiment, block or divert 3/4 of the light from the led,
does start of the linear section move.

Oh, you want me to test the photodiode linearity?

George H.

 

[George Herold]

On 2013-10-07 06:21:59 -0700, George Herold said:



Cool. So, in photon counting mode do you expect a 3/2 law?? BTW I would
be happy to send you a few of the IR LEDs. My email is with
earthlink.net

Yeah.. well 3/2 till something near 1mA at which point it becomes linear.
(You can do a decent job measuring PD DC current with the uA input on a decent DMM.)

Here's a plot of count rate vs photodiode current (where I used the same LED current for both measurments.)

https://www.dropbox.com/s/vrwf2mwp51qrd14/SP-PD.BMP

Almost linear. (slope should be 1.0 'in theory') There are several thingsthat 'screw up' the spad counts. At low count rates there is after pulsing which gives you more counts. And then at high count rates you need to take the dead time of the spad into account. I've 'semi'* corrected the datafor the dead time effects.

Thanks for the IR led offer, but if I can't buy ~100's of them I'm not really interested. But drop me an email and I'll put some AND113R's in an envelope. (gherold-at-teachspin.com)(That's what gave the above data... In what might be called an extreme case of 'trud' polishing I filed the plastic lens of the 'spad' down to just above the active element and got lot's morecounts with green photons.)

George H.

*so to first order the dead time correction says that the 'real' time I wasmeasuring is the (gate time)-(dead time)*(number counts). But then there's a second order effect that a second pulse during the dead time strecthes the dead time by a factor between 1 and 2... I ignored the second order correction.