Optoelectronics hold a range of possibilities for the hardware experimenter — indeed who among us hasn’t added LEDs aplenty to our work? What many of us may be unaware of though is that an LED is also a photodiode, and can even be persuaded to generate usable quantities of power. [Voltative] takes a look at this phenomenon with a series of experiments.
Lighting up an LED from a set of other LEDs is pretty cool, as is powering a calculator, or even the calculator powering itself from its on-board LED. But what caught our eye was using two LEDs as a data link, with both of them acting as transmitter and receiver (something on searching we find we’ve seen before). The possibilities there become interesting indeed.
Given that we are now surrounded by LEDs, from OLED screens to LED lighting, we can’t help wondering what the photodiode performance of some other types of part might be. Would the large area of a lighting LED give a better result for example, or would the phosphorescent coating of a white LED make it useless. We feel there’s more scope for experimentation here.
“Would the large area of a lighting LED give a better result for example, or would the phosphorescent coating of a white LED make it useless?”
That’s an interesting question. I suspect that the coating would at least somewhat hinder the LED’s power generating capabilities. And I think it would seriously reduce an LED’s capabilities as a data receiver, given that phosphorescent materials tend to have a high degree of persistence. I have several LED bulbs here at home which take on the order of 5 seconds or longer to stop emitting light that I can see.
much of this delay is from capacitors in the LED driver. i tested some while phosphor-based LEDs at one point, and their response is pretty fast
Bingo. I tested the spectral decay of white LEDs while building a LED equivalent of an airgap high-speed flash.
The blue (raw LED output) component decays as fast as the input waveform: 100 nanoseconds. The yellow phosphor decays with about a 1 microsecond time constant. The red component is just a smidge slower, at 2 microseconds.
…All prompted by seeing my projectile images having a distinct yellow-red ‘anti-trail’, where the sharp projectile image had the correct color, but ahead of it in the image was a yellow-red blurred border. Thought to be chromatic aberration in my lens, but it turned out to really be the LED light source.
It should hinder it considering that the wavelength to excite it must be the same or shorter. White LEDs have a blue chip and since the coating works on a similar principle as it emits longer wavelengths than the excitation hardly anything would arrive.
There are monochromatic COBs though that should work fine.
When I saw you had written this up, it reminded me of page 66 of this: https://www.worldradiohistory.com/Archive-Poptronics/70s/1974/Poptronics-1974-03.pdf
Good ol’ Forrest Mims.
Ahh Forrest…that was the first electronics project I did that worked
Yeah, I thought the same.
For a college science dropout, climate change denier, anti-abortion creationist, Mims has left a remarkable, paradoxical legacy.
Built that one back in the day. (A decade later found it in the junk and replaced the led’s with modern high efficiency ones, and then it actually worked quite well.)
Forrest Mims saved my customers a _lot_ of money with that one weird trick.
Went on to use blue leds as PV generators to directly switch power fets in batteryless solar tracker – the high voltage vs silicon photodiodes makes in viable.
Another product made a high current passive optocoupler with leds glued together switching a powerfet.
And used in the 10’s of millions as a serial data input (and output) via the front panel indicator leds.
Also used a single led in ultra cheap high volume waterproof data loggers as the serial input/output device.
LDRs and selenium cells are lesser known light-sensitive parts, too! 😃
Lots of semiconductors have photovoltaic properties. When I first got into the business I watched a colleague nearly go nuts debugging the weirdest intermittent issue in a high gain analog circuit that used some spiffy new op amps.
The weird thing was, it was exactly the opposite of the usual pattern. It was bulletproof in the field but was intermittent in the nice quiet lab. He eventually tracked it down to insufficiently opaque packaging in the prototype chips.
Out in the field, sealed inside a dark housing, no problem. But on a bare board in the lab enough light leaked in that when you had them in the exact right spot under the bright fluorescent lamps of his bench, the 120hz flashes could kick the circuit into oscillation. A bit of old-school floppy disk write inhibit tab saved the day.
yep a more recent example of this happened to raspberry pi’s being photographed, the flash freaked out one of the chips causing them to reboot
I also nearly went nuts decades ago trying to fix hum in an audio amp. As soon as I went near it, the hum stopped. Turned out to be an OC71 (only oldies will remember those germanium transistors) in a not-quite-opaque black glass encapsulation picking up flicker from the room light.
I recall that older OC71, Circa 1965, had a clear gel and if you scraped the black paint off the glass it worked great as an opto sensor. Later they replaced the clear gel with an opaque white paste and that no longer worked. Funny they sold photo transistors as well with clear gel but i recall they were more expensive.
I built a simple synthesizer on a breadboard and used a yellow LED as a voltage reference diode. You can guess what happened next.
I’ve since used the same kind of LEDs for sorting out LED bulbs that strobe. The issue with those is that even though the phosphors are now slow enough that you can’t generally see the blinking, the blue LED underneath that excites them does blink, which creates the curious effect that blue objects seem to move with a “staccato” in your peripheral vision.
Literally ANY semiconductor junction (diode, transistor, whatever) has at least some photovoltaic property.
True. I made a solar panel with a few 100 wrongly ordered glass zener diodes.
I could measure the voltage, but power was as good as non-existent.
I even harnessed sunlight into electrical current once with a buch of parallel LED’s
Hmm, nothing new.
When our first LED came out (early 1970’s) me and my friends made our first own remote controls.
We put the LED behind a small hole at the front of the Radio receiver, add pcb with relais driver transistor amplifier/trigger cascade and relais.
Grab the torch llight, switch on/ off the radio from the bed.
Great overview/demos, and cool idea: Tx/Rx.
Thanks, also, Al for the Mims link.
Allegedly they are more sensitive to shorter wavelengths than they’re designed to emit.
Also, watch out for *color*-phosphor-coated LEDs… (e.g. blue with red or green phosphor rather’n white). They do exist.
Phosphorescence cannot run in reverse. Thermodynamically forbidden. With the phosphor removed the LED is going to be sensitive to photons of the same or higher energy as it would emit.
As far as I understood, the bandgap changes with the applied voltage and current (and temperature), so the color of the LED shifts slightly by how hard you’re driving it. Does it go up or down, though? I remember I once observed a LED that was so over-driven it shifted between yellow and green, but I forget which way it went.
When you cool a yellow LED in LN2 it can turn green, so I’d imagine if you overdrive a green LED enough it might be briefly yellow.
If you over-drive it enough it will turn infrared and then black.
This makes me wonder with enough lighting could a HD/4K/8K monitor, that looks powered off, be made to act like a camera. It would probably not be a very good camera without lenses. But it is an interesting theoretical ponderance.
If you made a light-field display from diodes, it theoretically with proper electronics could be made into light-field camera. Normal 2d panels without lens would just act like a millions of light sensors. If you got rid of light spreader on panels with several leds lighting up an lcd, you could theoretically light it up pixel by pixel and make virtual aperture camera out of every led. This is only theoretically, practical considerations would probably mean you only get fuzzy image enough to know that there are some shapes in front of panel.
Resistor+LED and small photoresistor+transistor on opposite ends of a short piece of discarded cat5 outter jacket made for a cheap and good enough optoisolator once.
We aslo were using transistors as a photo diode. We simply didn’t have the dedicated photo diodes.
This is a timely reminder of an almost-forgotten technique for my OLED screen recycling project (https://hackaday.io/project/195003-ambient-oled-display-re-use) I shall shine some strong lights on the front and probe around the driver pcbs for generated voltage.
If you back feed a solarpanel
And point a camera at it
It’ll glow infrared you cant see like a giant led panel
So wouldn’t blue green and ultraviolet sensitive panels make sense
High frequency = higher energy density
Mean less sq ft of panel for same power
Yes, but you have to consider the available spectrum. Ultraviolet is only 2-5% of the light that reaches the ground, and adding blue-green gets you up to 25%-ish. This is because the atmosphere absorbs and scatters these higher energy photons more easily.
There’s more energy available per square meter if your panel is sensitive from near-IR to yellow, which is what silicon panels generally are.
I remember, years ago, when HAD wrote a ton of articles on this. One of my favorites was about using an LED matrix as a touchscreen by sensing its own light.
Rarely, flashlights have tried doing IO with their LEDs. The data rate and error rates are poor, especially if it’s meant to receive from a computer monitor. The output rate can be decent, if you have something to receive with, but on a flashlight I had that used pwm on a mosfet, the brightness was nonlinearly lower when you shortened the pulses much below a millisecond or so. I didn’t see it on a scope, but I naturally believe that it wasn’t managing to fully turn on and off in that time. A hall sensor might let you program a device using a speaker coil, a bit faster and more reliable than the LED trick without needing an exposed contact. That said, Another LED based idea is to use multiple LED pairs side by side if they’re different colors and (probably) colored plastic instead of clear. Maybe you could use RGB leds at least on one side.
Another effective improvised photodiode is a 1N914/1N4148 glass-bodied small-signal diode, which, IIRC, will produce about 0.5V under a bright light, and will switch quite rapidly. And they are quite cheaply had, costing pennies in quantity.
> What many of us may be unaware of though is that an LED is also a photodiode, and can even be persuaded to generate usable quantities of power.
I dunno about this. When I used to teach students about LEDs it was literally the second thing I taught them. Then we measured the voltage induced across an LED held near the classroom window.
Other posters have mentioned that solar cells emit light when they are back-powered. This is used to test solar panels in real life. It’s easy to do in the dark, but to test them in a brightly-lit lab requires some clever filtering.