Yeah the parallax propeller 1 (p8x32a) had a forum post on this around 2007? Where they pulse the LED and then measure the voltage drop across the led by making the signal line high impedance. This was super useful for turning a led into a photo detector or even a light switch (continually pulsing and measuring the reflected light).
The number of photons in a normally lit lab should be well over a billion per second, so it's definitely not registering individual ones. 50 per second would be what a SPAD should read in pitch black. Interesting all the same
Is there a way you can demonstrate that it is does really have the capability to detect single photons? If you place a LED in a well lit room it will produce a current of 10uA or so. -- This would require a huge number of successful photon/LED interactions per second What may be happening here is the barrage of photons is slowly charging up the junction voltage until it reaches a breakdown; you can even see an RC charge curve before every breakdown event on your scope. One test that could be performed is place the LED in a dark room with a very low rate photon source ~dozens per second, and see if it's possible to detect if the source is there or not with your LED
That’s typically avalanche behavior of diode, it’s just one pixel, it’s not a sipm arrays, in this condition it’s long overwhelmed, so it’s the dynamic range its weaknesses is in not sensitive. Around 450nm, all silicon based ones quantum efficiency is around 40%. And yes it generates dark count due to thermal act, that’s why people invent sipm to achieve higher sin to noi ratios. I understand why he is jumping into the project, as one can do gamma spectrometry by a bpw 34( kind of doing it), that is 1 cents of cost compared with sipm chips, quite tempting. Again apd is a apd, if it’s working in Geiger model, it can detecte 450nm single photon quite well, actually bpw34 can resolve Photo numbers by pulse height, again the weakness is noises and dynamic range not quantum efficiency. As apd has to be quenched after each pulse it’s in dead time during discharge.
Hi Prof. This is crazy I've worked extensively with laser diodes but to use a humble LED for this is cool! It's just a transducer so can work forwards and backwards...I'm off to try my LEDs out. Thanks!
Me too. My goal is to build a dirt cheap but reliable and reasonably sensitive radiation detector from stuff that you could find in scrapped equipment. This looks very promising ❤
@@yeyuan6273 I meant the blinking light from flourecent tubes, probably HF drivers. Flesh of a hand is really suboptimal at blocking light, especially in the longer wavelenght of visible spectrum.
thats 50hz or 120 for old energizer, about 200 to 500 khz for morden switching one, he is meauring 20 to 40 counts per 50us, doesnt add up. and its typical guss. response regarding spike. and i dont think flourent tube use pwm to dim it@@Paxmax
@@yeyuan6273 Good points! Florescent drivers has gone thru the entire frequency spectrum really, depending on age the "HF drivers" started at 50kHz and has crept upwards the newer it is.
Going to go through my LED stash to try this. Red LED, that would be GaAlAsP. Wonder if GaInN green blue or UV LEDs would work? In theory the more narrow the emission bandwith the better the detector would be at similar wavelength.❤
so how woulkd you take a picture with 100 x 100 LEDs like this? Would you measure intensity on each of them and transform it into brighter/darker pixels? Also, with LEDs with different wavelengths, could you detect colour? Even in dark room?
You should look into sipms. But I don’t see the implementation in images as now. You can imagine that each and every pixels character has to be almost identical with each other to work. Or one pixel discharging will pull all the other pixels voltage to ground............. and that is why sipm is so fking expensive.
in principle yes. but impractical. you will need a large bias voltage and a register to collect the counts from each "pixel". you will need miniLEDs, or have a huge optical lens to project the image onto the LEDs.
It will have a photodiode transconductance region then you will hit the avalanche region. For LEDs this will very a lot. You would need to severely limit the peak current, but a high ohms resistor and a tiny capacitor, or even the intrinsic capatance of the LED itself should work. Follow the pulse with a JFET amplifier and sense resistor and go from there. ❤
Unfortunately the AND114 LEDs seem to be out of stock everywhere, especially in Europe. Is there another type of LED which is known to work reliably ? I tested some of the LEDs i have, very few of them work, and even if i test some LEDs from the same type only a few of them work. Some types don't work at all (e.g. yellow ones) or maybe they have a very high breakdown voltage, which i did not reach in my test. I'm thinking about building a random number generator based on this principle.
LTL-4223 red LEDs work. They are very cheap, and today are sold everywhere for a symbolic price, like no more that 20 cents something for each. I bought 10 of them recently ( 1-2 days delivery for extra $8).
@@andrewvolunteer2185Do you know what the breakdown voltage for the LTL 4223 LEDs is? I am trying to replicate the experiment in the video with a few of those diodes and was having trouble recreating the situation. I think I might not have a high enough load voltage.
You will want to use ones with a 10nm or less peak emission bandwidth. The yellow ones have a very large bandwidth so the gain region would be highly diffused.
Yeah the parallax propeller 1 (p8x32a) had a forum post on this around 2007? Where they pulse the LED and then measure the voltage drop across the led by making the signal line high impedance. This was super useful for turning a led into a photo detector or even a light switch (continually pulsing and measuring the reflected light).
Thank you for your amazing video. It was clearly presented and I liked the way you stepped through the enhancements to the circuit. Bravo !!!
The number of photons in a normally lit lab should be well over a billion per second, so it's definitely not registering individual ones. 50 per second would be what a SPAD should read in pitch black. Interesting all the same
Using a LED as a SPAD is simple, cheap (instead of paying hundreds or thousands) but not efficient, probably < 1%.
Is there a way you can demonstrate that it is does really have the capability to detect single photons?
If you place a LED in a well lit room it will produce a current of 10uA or so. -- This would require a huge number of successful photon/LED interactions per second
What may be happening here is the barrage of photons is slowly charging up the junction voltage until it reaches a breakdown; you can even see an RC charge curve before every breakdown event on your scope.
One test that could be performed is place the LED in a dark room with a very low rate photon source ~dozens per second, and see if it's possible to detect if the source is there or not with your LED
That’s typically avalanche behavior of diode, it’s just one pixel, it’s not a sipm arrays, in this condition it’s long overwhelmed, so it’s the dynamic range its weaknesses is in not sensitive. Around 450nm, all silicon based ones quantum efficiency is around 40%. And yes it generates dark count due to thermal act, that’s why people invent sipm to achieve higher sin to noi ratios. I understand why he is jumping into the project, as one can do gamma spectrometry by a bpw 34( kind of doing it), that is 1 cents of cost compared with sipm chips, quite tempting. Again apd is a apd, if it’s working in Geiger model, it can detecte 450nm single photon quite well, actually bpw34 can resolve Photo numbers by pulse height, again the weakness is noises and dynamic range not quantum efficiency. As apd has to be quenched after each pulse it’s in dead time during discharge.
the LED is detecting single photons when given billions of photons. its efficiency is very very poor. but still a great demo!
@@zhenyuanyeo8386 Yes, that was total bullshit. :-)
Amazing demo, thank you very much!
Very cool video, thank you for sharing! Looks like a good candidate for some DIY quantum experiments as well.
Very interesting. You should declare the count variable as volatile if used in an interrupt routine.
Yes, I should've declared it as volatile.
Hi Prof. This is crazy I've worked extensively with laser diodes but to use a humble LED for this is cool! It's just a transducer so can work forwards and backwards...I'm off to try my LEDs out. Thanks!
Me too. My goal is to build a dirt cheap but reliable and reasonably sensitive radiation detector from stuff that you could find in scrapped equipment. This looks very promising ❤
You are looking at the HF flourescent ceiling tube drivers... Not individual photons 😂
Well if your palm can block 500 kHz switching emf……
@@yeyuan6273 I meant the blinking light from flourecent tubes, probably HF drivers. Flesh of a hand is really suboptimal at blocking light, especially in the longer wavelenght of visible spectrum.
thats 50hz or 120 for old energizer, about 200 to 500 khz for morden switching one, he is meauring 20 to 40 counts per 50us, doesnt add up. and its typical guss. response regarding spike. and i dont think flourent tube use pwm to dim it@@Paxmax
@@yeyuan6273 Good points! Florescent drivers has gone thru the entire frequency spectrum really, depending on age the "HF drivers" started at 50kHz and has crept upwards the newer it is.
you did not try on complete darkness.
Going to go through my LED stash to try this. Red LED, that would be GaAlAsP. Wonder if GaInN green blue or UV LEDs would work? In theory the more narrow the emission bandwith the better the detector would be at similar wavelength.❤
I'd like to know what specifically it is about that diode that lets it work as a spac
so how woulkd you take a picture with 100 x 100 LEDs like this? Would you measure intensity on each of them and transform it into brighter/darker pixels? Also, with LEDs with different wavelengths, could you detect colour? Even in dark room?
You should look into sipms. But I don’t see the implementation in images as now. You can imagine that each and every pixels character has to be almost identical with each other to work. Or one pixel discharging will pull all the other pixels voltage to ground............. and that is why sipm is so fking expensive.
in principle yes. but impractical. you will need a large bias voltage and a register to collect the counts from each "pixel". you will need miniLEDs, or have a huge optical lens to project the image onto the LEDs.
Is it possible to tell the wavelength of photon incident on led? If yes then how?
Can you share your paper/report?
I am interested in reading more about it
Is the 20 ish volts a reverse bias or does this LED have no polarity?
It will have a photodiode transconductance region then you will hit the avalanche region. For LEDs this will very a lot. You would need to severely limit the peak current, but a high ohms resistor and a tiny capacitor, or even the intrinsic capatance of the LED itself should work. Follow the pulse with a JFET amplifier and sense resistor and go from there. ❤
can i use this to build an optical quantum computer?
How do I find these? I'm trying to buy some, but I can't find any.
Unfortunately the AND114 LEDs seem to be out of stock everywhere, especially in Europe. Is there another type of LED which is known to work reliably ? I tested some of the LEDs i have, very few of them work, and even if i test some LEDs from the same type only a few of them work. Some types don't work at all (e.g. yellow ones) or maybe they have a very high breakdown voltage, which i did not reach in my test. I'm thinking about building a random number generator based on this principle.
Usually inferred ones, remember come back bring more information back to us.
LTL-4223 red LEDs work. They are very cheap, and today are sold everywhere for a symbolic price, like no more that 20 cents something for each. I bought 10 of them recently ( 1-2 days delivery for extra $8).
@@andrewvolunteer2185Do you know what the breakdown voltage for the LTL 4223 LEDs is? I am trying to replicate the experiment in the video with a few of those diodes and was having trouble recreating the situation. I think I might not have a high enough load voltage.
You will want to use ones with a 10nm or less peak emission bandwidth. The yellow ones have a very large bandwidth so the gain region would be highly diffused.
Could you share your Arduino code for this project please?
Can't buy such LED in Europe... ;(
Bad science
He did build a photodetector... and it detects the ceiling flourescent tubes alright 😁👍