EEVBlog #869 - Counting LED Photons!

[mux]

[mux]

Saw there wasn't a topic for this yet, so might as well start it

I've got some maths behind at which point you should theoretically be able to see a LED switch on, as well as some reasons why especially bargain bin LEDs won't ever go so low.

Fundamentally, LEDs convert electrons directly into photons one to one. Obviously not just like that (the QM is a bit more complicated), but the model relationship is one to one. This means that one coulomb (6.24x10^18 electrons) will produce 6.24x10^18 photons. For a red LED, each photon will have an energy of 1.8eV because... the forward voltage is 1.8V, i.e. each electron contains 1.8eV. Without regarding coulomb losses and that kind of stuff, this is exactly how you can regard LEDs and fundamentally this is the reason why they're so efficient. Practically, electricity-to-photons efficiency of some LEDs is in excess of 70% (which is super close to ideal).

This also means you should expect 1 photon per second at 0.00015 fA, regardless of the color. 10fA, your lowest resolution, should theoretically produce 62 400 photons per second.

There are a few very big but(t)s to this, especially when dealing with traditional LED substrates  (e.g. gallium arsenide). LED dies are extremely large for the amount of electrons commutated at these currents, so just by virtue of the thin barrier, you're going to get many fA of pure leakage through quantum tunnelling.

Bargain bin LEDs also aren't (generally) monocrystalline, defect-free or even very well cut and doped dies. In fact, only fairly recently has ion implantation (a 'better' way of doping than chemical bath doping) really started to trickle down to low-cost LEDs. This means there are a lot of small areas on the chip with tiny little (fairly high-resistance) shorts in them where there are misplaced or missing doping atoms, which can contribute to pA or fA of additional leakage current. Plus grain boundaries and other crystal defects.

Even when electrons properly recombine, they might not produce a photon, but one or more phonons ('heat particles'). This contributes to the so-called quantum efficiency, the number of photons produced divided by the number of electrons passing the barrier.

Also, photons do not get produced with a definite orientation in these substrates, so you fundamentally lose a bunch of them inside the substrate (because they're produced in-plane and never leave the die). Then a bunch get absorbed or scattered in the encapsulation. The combined effect of die and encapsulation loss is generally about an order of magnitude for old type LEDs (i.e. you get 1/10th of the light out the front compared to what was produced in the bandgap). High-brightness LEDs fix this problem almost perfectly.

There are also a few charge/magnetic effects that i won't get into; some of them are not very well understood either (e.g. LED droop)

A modern substrate, purpose-made (tiny) low-current high-brightness LED die should be capable of producing measurable amounts of light at 1fA. It still has low quantum efficiency (about 10%) and some leakage, but most of the other effects should be so diminished as to not matter.

Also, pro tip: modern CMOS DSLR camera sensors are capable of detecting as little as 2-3 photons per pixel (50% over noise). They're incredibly useful for these kinds of experiments if you don't have the money or time to get a lab grade photon counter.

[doobedoobedo]

I have a feeling that the non-linearity of the graph at the top may have more to do with the sensor as it approaches it's absolute limit than with the LED itself.

[Fungus]

I might be wrong but I think the Rigol DS1054Z gained a pulse count mode in the latest firmware. I remember people complaining it was spelled "pluse" in the menu...


nb. I'm not saying it would  do the job or anything, but it would be interesting to see

(and it's interesting that the DS1054Z is gaining free features over time).

[Froese]

No statistics in the Universal Counter?  You need the average over 100s?  Set the gate time to 100s and do the "math" in your head.

[han]

The delayed respond is probably due to the cable capacitance.
Something missing in the experiment.I'm curious is the voltage below forward voltage will make any photon or current?

[daqq]

I have a feeling that the non-linearity of the graph at the top may have more to do with the sensor as it approaches it's absolute limit than with the LED itself.

Very likely, this is known as dead time - as the frequency of events increases, you get a smaller count than that which you'd expect - events start overlaping more and more and the sensor fails to recuperate quickly enough, until you enter a point at which you get no impulses at all, because there is an effectively 100% or more overlap of events - the sensor goes effectively into saturation.

This is generally a problem with similar systems - geiger tubes, PMTs and similar devices can give you the impression that there is no signal, when in fact there is an abundance of it. See: https://en.wikipedia.org/wiki/Geiger%E2%80%93M%C3%BCller_tube#Quenching_and_dead_time or similar.

[jdraughn]

I wonder if you could block some of the light so you could increase the current.  Kind of like a voltage divider but with light

[wolf32d]

I have a feeling that the non-linearity of the graph at the top may have more to do with the sensor as it approaches it's absolute limit than with the LED itself.

Very likely, this is known as dead time - as the frequency of events increases, you get a smaller count than that which you'd expect - events start overlaping more and more and the sensor fails to recuperate quickly enough, until you enter a point at which you get no impulses at all, because there is an effectively 100% or more overlap of events - the sensor goes effectively into saturation.

This is generally a problem with similar systems - geiger tubes, PMTs and similar devices can give you the impression that there is no signal, when in fact there is an abundance of it. See: https://en.wikipedia.org/wiki/Geiger%E2%80%93M%C3%BCller_tube#Quenching_and_dead_time or similar.

In general that is true. However, in the datasheet of the detector (http://www.optics.rochester.edu/workgroups/lukishova/QuantumOpticsLab/homepage/apd_spcm_aqr.PDF) they say that the device has a 50 ns dead time. They also give you the formula to calculate the correction factor: (actual_rate * quantum_efficiency) = measured_rate * corr_factor - dark_count
where correction_factor = 1/(1-measured_rate * dead_time) = 1/(1-measured_rate * 50ns)
At measured rates below 200kHz the correction is <= 1%  i.e. it is negligible.

My guess is that the increasing voltage & current enable the production of higher energy photons resulting in a shift of the centroid of the output spectrum. This spectral shift may result in a slightly lower detection efficiency (quantum efficiency) and therefore a lower measured count than expected. It would have been interesting if Dave reported also the voltage data (V across the led).

Suggested further tests to check my "spectral shift theory":
- same led + attenuator (dark glass)
- different center wavelenght led (e.g. a green 550 nm led)

[EEVblog]

For a red LED, each photon will have an energy of 1.8eV because... the forward voltage is 1.8V, i.e. each electron contains 1.8eV.

But that voltage varies depending upon the current.


Thanks for the extra technical info 

[wolf32d]

For a red LED, each photon will have an energy of 1.8eV because... the forward voltage is 1.8V, i.e. each electron contains 1.8eV.

But that voltage varies depending upon the current.


Thanks for the extra technical info 

Normally the bulk of the emitted photons have an energy which is slightly above the bandgap of the material  (which is the lower limit for photon energy). Their energy distribution spread is roughly KbT i.e. the juction temperature * boltzmann constant. The voltage at the junction is only an upper limit for the photon energy, but at voltages higher than the bandgap the sprectrum remains roughly the same (neglecting thermal effects).

The only major factor that determines the color of the output light is the bandgap ot the die material (& doping) e.g. GaAsP for red leds, SiC for some exotic blue leds ...

ps: At voltages very close to the bandgap the thermal ditribution in the  photon spectrum gets truncated abruptly and I think this is what leads to the nonlinearity you measure with the photon counter (at higher voltages there are more photons that have a shorter wavelength than 700 nm -which corresponds to 1.8 eV in energy- i.e. the max efficiency waveleghth of the counter)

[tszaboo]

And LEDs themselves can be used as light detectors. Just hook it up to a DMM and measure the voltage. They are really sensitive. Once I was doing a test, the LED in a cm thick wooden box, detecting light. It was outputting a different voltage when I was standing between the window and the box.
I guess at the lower end, this effect, like reflected light, can become significant.

[daqq]

In general that is true. However, in the datasheet of the detector (http://www.optics.rochester.edu/workgroups/lukishova/QuantumOpticsLab/homepage/apd_spcm_aqr.PDF) they say that the device has a 50 ns dead time. They also give you the formula to calculate the correction factor: (actual_rate * quantum_efficiency) = measured_rate * corr_factor - dark_count
where correction_factor = 1/(1-measured_rate * dead_time) = 1/(1-measured_rate * 50ns)
At measured rates below 200kHz the correction is <= 1%  i.e. it is negligible.

My guess is that the increasing voltage & current enable the production of higher energy photons resulting in a shift of the centroid of the output spectrum. This spectral shift may result in a slightly lower detection efficiency (quantum efficiency) and therefore a lower measured count than expected. It would have been interesting if Dave reported also the voltage data (V across the led).

You are right, I did not take into account the actual numbers, was thinking in lower speed things 

It would be an interesting measurement to see the rest of the plot, say, until 1mA? At what current through the LED does the detector go into saturation?

[mux]

And... the color also actually changes Although a bunch of the slack in the voltage-current curve is simply internal resistance.

Its really cool to see how the spectrum changes with current, too. I'm currently building a USB spectrometer, if I finish that project before you get a similar product sent I'll make sure I send you one so you can try looking at that too. Highly recommended follow-up.

[Synthetase]

Regarding measurement technique saturation: had similar issues doing cell counts back in the day. Quick and easy cell counting is done using absorption of light. Too many cells in a tube will interfere with each other and give you lower count values than actually exist, resulting in a plateau very similar to what Dave had in his graph. We used to dilute the cell samples out so as not to overload the system. Not sure how to do that with light, though

[wolf32d]

Regarding measurement technique saturation: had similar issues doing cell counts back in the day. Quick and easy cell counting is done using absorption of light. Too many cells in a tube will interfere with each other and give you lower count values than actually exist, resulting in a plateau very similar to what Dave had in his graph. We used to dilute the cell samples out so as not to overload the system. Not sure how to do that with light, though

If a light source is too strong for a photon counter you can use an attenuator, then you multiply the photon count by the inverse of the attenuation factor   
If the ligh intenisty is way too strong you may use another type of photodiode/photodetector

[silent]

A multi-channel analyzer (MCA) would be useful here - it would provide peak height distribution. I'm curious what's the difference in distributions between various LEDs.

[Kleinstein]

The counter somewhere saturates. However the photon counter is much faster than 15.000 counts per second. It could be the scope that is limiting the counts by adding a dead time. So maybe the real counter would be the better instrument. Even if you have to do the statistics by hand (e.g. use a 10 s gate time).


LEDs actually start to emit light well before the voltage reaches the band gap equivalent voltage. Even at reasonable currents (e.g. 1 mA) it can still be below. Besides quantum mechanics you also need to look at the thermodynamics - there are funny effects. A very good LED (especially IR) might actually get cold when operated at low intensity, as some of the energy can come from thermal source. So there is a thermodynamics reason that quantum efficiency has to go down at low intensity / voltage.

Not every recombination of an electron and hole produces a photon, as there are competing non radiative mechanisms. Though some LEDs can be very good at that point. The more difficult part is getting the light out, before it is reabsorbed. Also quite some light might get lost coupling to the sensor.


For better LEDs, I remember that a high efficiency green one was visible at 100 nA -  nearly useful at 10 µA.

[Smokey]

I might be wrong but I think the Rigol DS1054Z gained a pulse count mode in the latest firmware. I remember people complaining it was spelled "pluse" in the menu...


nb. I'm not saying it would  do the job or anything, but it would be interesting to see

(and it's interesting that the DS1054Z is gaining free features over time).

This is true.  I was excited about this new feature since it would make it much easier to count the resolution of unknown rotary encoders... but... It only counts to 256... or maybe it was 512.... either way, it's not high enough.  bummer.

[android]

Dave, perhaps you could borrow one of Fran Blanche's ancient Monsanto MV1 red LEDs and see how it compares to the modern offerings? See it at about 5:18 in her vid...
https://youtu.be/yusrrOPM43o

[NiHaoMike]

Also test some deep green and blue LEDs. In my experience, they start to visibly glow at much lower currents than red and regular green LEDs.

[alsetalokin4017]

I might be wrong but I think the Rigol DS1054Z gained a pulse count mode in the latest firmware. I remember people complaining it was spelled "pluse" in the menu...


nb. I'm not saying it would  do the job or anything, but it would be interesting to see

(and it's interesting that the DS1054Z is gaining free features over time).

This is true.  I was excited about this new feature since it would make it much easier to count the resolution of unknown rotary encoders... but... It only counts to 256... or maybe it was 512.... either way, it's not high enough.  bummer.

Ah... no, not quite. The "Pluses" count is limited by the resolution of the screen data. 597 "Pluses" seems to be the maximum I can get it to display.

But with some clever scoposcopy and encoder knob manipulation you can indeed use the Pulse counter feature for unknown encoders. You may have to do it in thirds or fourths of a complete rotation and add the subtotals...


Yes, the Rigol is gaining free features, and free bugs too! (But what do you expect, it's only 400 dollars...... innit?)   

[gardner]

I have always taken this to be the last word on measuring low current light output of LEDs.  I didn't watch it again but IIRC Mike was getting measurable output in the single nA range.

[Someone]

Here is some linearity data up above the photon counting range, starting from 1uA. The phosphors are not the cause of the non-linearity in this white LED but the efficacy keeps increasing with drive current until the self heating becomes evident around 10mA.

[Someone]

For better LEDs, I remember that a high efficiency green one was visible at 100 nA -  nearly useful at 10 µA.

If you sit in a dark room for 30-40 minutes and fully dark adapt your eyes, you'll see it at much lower drive currents (having gained several orders of magnitude gain in your eyes).

[westfw]

BTW, there was some fascinating data on the sensitivity of the human eye that I vaguely recall from a course I took in "vision" some years ago.
Um...  https://en.wikipedia.org/wiki/Absolute_threshold

[Fungus]

Yes, the Rigol is gaining free features, and free bugs too! (But what do you expect, it's only 400 dollars...... innit?)   

I wonder what would happen if somebody like Rigol made a scope without firmware (or just very basic firmware)

They could just publish the hardware specs and a dev. kit and let the programmers of the world go for it.

[Fungus]

Ah... no, not quite. The "Pluses" count is limited by the resolution of the screen data. 597 "Pluses" seems to be the maximum I can get it to display.

Looks like it's working from the screen data.

The DS1000Z seems to do that sort of thing a lot (eg. serial decoders).

I think there must be a limitation in the hardware where the main CPU doesn't have direct access to the captured data. The captured data is probably scaled/zoomed into a buffer using hardware and the main CPU has to work work with that. There's no direct path from CPU to captured data.

[Fungus]

Not every recombination of an electron and hole produces a photon, as there are competing non radiative mechanisms. Though some LEDs can be very good at that point. The more difficult part is getting the light out, before it is reabsorbed.

Good point.

How do they get the light out?

Is it just a case of making the PN junction as thin as possible?  Is that what determines LED efficiency?

[mux]

To a large extent - yes. Modern LEDs have extremely thin metallization on top of the bandgap area (or none at all, and just the bulk material - see illustration below), with the total active region only being a micron or so across. At that point the substrate is very transparent towards its surface. There are other cool things that are being done, like patterning the sapphire substrate to more efficiently reflect back stray photons, adding microlenses or micro-waveguides...



All this has led to ridiculous external quantum efficiencies nowadays. I've made a calculator a while ago that predicted a 75% LER (lighting efficiency, i.e. amount of light power out vs power in, meaning that a 10W LED at 75% LER only outpuyts 2.5W of heat). Including the phosphor losses, that would mean the underlying blue LED would need to essentially have nearly 100% quantum efficiency (about 91-93% calculated). Sure enough, when measuring over an integrating sphere this is actually true. If we'd only need lots of monochromatic blue light in our world, we would pretty much have the perfect light source already.

[ion]

If you want to gather data in the microamps, try using a blue LED.  The detector seems to be much less sensitive to those wavelenghts so it shouldn't saturate as quickly.

[Kleinstein]

Getting the light out there are several tricks, like having several layers with graded refractive index, patterns and multi layer structure in such a way that the lowest band-gap is in the junction region only. So outside the junction region the material is largely transparent and lower refractive index. As a downside a higher voltage is needed - one can actually see that difference: modern high efficiency leds usually need a higher voltage (at least for the same current).

With the white LEDs the phosphor layer is directly on top of the LED crystal. So the light does not need to get out of the solid to be absorbed from the phosphor. It's only the yellow light that needs to get out, but this is at a wavelength that is not absorbed very much anymore.

The observed pulse count should not saturate the detector itself - it's more like Dave's scope is saturated with to many pulses. So the pulse more might be able to detect short pulses, but maybe not more than 20000 per second.

[Someone]

The observed pulse count should not saturate the detector itself - it's more like Dave's scope is saturated with to many pulses. So the pulse more might be able to detect short pulses, but maybe not more than 20000 per second.

Well caught, how far does the memory depth reduce when using the peak detect mode on that scope? I didn't see the sample rate shown during the video. At a guess with 1Ms/s for the 1 second span I get the attached simulated pileup for that situation, which still wouldn't explain the magnitude of drop off seen in Daves data (where we might expect a few %).

*Edited with simulations including the overhead from the scopes measurement function needing to see two edges to make a count.

[Vgkid]

Another video that I enjoyed, thanks Dave.

[wolf32d]

The observed pulse count should not saturate the detector itself - it's more like Dave's scope is saturated with to many pulses. So the pulse more might be able to detect short pulses, but maybe not more than 20000 per second.

Well caught, how far does the memory depth reduce when using the peak detect mode on that scope? I didn't see the sample rate shown during the video. At a guess with 1Ms/s for the 1 second span I get the attached simulated pileup for that situation, which still wouldn't explain the magnitude of drop off seen in Daves data (where we might expect a few %).

*Edited with simulations including the overhead from the scopes measurement function needing to see two edges to make a count.

I couple of years ago I had to characterize a single photon avalanche diode for my optoelectronics thesis ... We directly read the voltage across a series 50 Ohm shunt (with the aid of a fast opamp). The current spikes had roughly a FWHM = 5 ns; we used a beefy LeCroy WaveMaster 8Zi-A scope to sample the signal! We could sample 250 ms @ 1GSa/s  , that bit of gear proved really useful

[vlad777]

Don't you have a photon counter?

[vlad777]

I have been thinking about this experiment in 2013.

http://www.electro-tech-online.com/threads/photon-by-photon.132888/

[max_torque]

For better LEDs, I remember that a high efficiency green one was visible at 100 nA -  nearly useful at 10 µA.

If you sit in a dark room for 30-40 minutes and fully dark adapt your eyes, you'll see it at much lower drive currents (having gained several orders of magnitude gain in your eyes).

If i sit in a dark room for 30-40 mins i won't see anything, because i'll be asleep........   

[German_EE]

Quantum physics and quantum mechanics are way beyond my abilities but I am still enjoying the video and the comments here. I now have new respect for that little white LED on the front of my monitor.

[mtdoc]

Cool video!

For better LEDs, I remember that a high efficiency green one was visible at 100 nA -  nearly useful at 10 µA.

I suspect that is due to the fact that the wavelength of green light is closest to the peak sensitivity of the rods in the human retina, which are more sensitive (fewer photons needed to excite) than cones.

As a point of reference, it only takes a single photon can activate a rod!  Though more than one rod needs to be activated in order for the sensation of a visual stimulus to be perceived - but still on the order of only 10 rods or so IIRC.

The power of evolution to produce a molecule (the rod photopigment rhodopsin) that can respond to one photon is impressive.

[Brumby]

The power of evolution to produce a molecule (the rod photopigment rhodopsin) that can respond to one photon is impressive.

I put the dynamic range up as a strong second.



The brain's image processing is something else.  That is simply awesome.

[Dubbie]

Thanks to all the thread contributors, especially OP. Really fascinating stuff. It's almost worth a follow up video with some whiteboard action!


Sent from my phone using Tapatalk

[f4eru]

Two remarks :
- The non linearity in the high range is probably from the sensor, or the scope for that matter, who both have a blank time directly after a pulse.
- The LED not emitting for 10 seconds at first is probably due to the capacity of the LED slowly charging when switching on...