Has anybody ever done this experiment?

[Andrea B.]

Dark room, no ambient light.

Stock camera.

3 Visible LED torches:  Red, Green, Yellow.

Reflective subject like PTFE, or perhaps just white paper.

 

Shine yellow LED on PTFE and make a visible photograph.

Shine both red & green on PTFE and make a visible photograph.

Look at the raw histograms in Raw Digger.

******

 

Note that whatever red & green LEDs you are using might not produce exactly the same yellow as from your yellow LED. But that's OK. I'm just curious about how a mixed yellow signal is recorded as compared to a pure yellow signal. 

 

The experiment can be made with other colors of course. 

 

((I'm trying to figure out where is my own little set of color LEDs.))

 

 

 

[lonesome_dave]

True monochromatic yellow is hard. I think most of the 'yellow' LED lights you find will be red + green LEDs. The only true yellow sources at my house are HeNe gas lasers at 594nm and the new yellow laser pointers at 593.5nm. Many people see these as yellow-orange or amber though.

 

Yellow LEDs exist but I've not seen them available in a torch.

 

There might be a blue LED/yellow phosphor combo light that produces yellow but it would be fairly wideband.

 

One source could be a wideband light filtered by a narrow bandpass filter at around 580nm.

 

You really need to verify the spectrum with a spectrometer for those usually 'mixed spectrum' colors like yellow and purple.

[lonesome_dave]

On 11/13/2023 at 3:09 PM, Andrea B. said:

I'm just curious about how a mixed yellow signal is recorded as compared to a pure yellow signal. 

 

I've always wondered about this regarding human eyesight. We have the same problem as the cameras in that we only really 'see' yellow as input on the red & green sensors. Could you replicate any shade of 'pure yellow' in human eyesight using only monochromatic sources of red and green, say at 650nm and 540nm? I don't know know but I suspect you could.

 

Incidentally, there are many yellow LED lights advertised and many references to yellow phosphor lights. Most of the time these are 'white' LEDs using a phosphor coating over a blue/UV LED that fluoresce in the green and red. When the LED is off the phosphor coating appears yellow in color but shines white (sort of) when on. If it is a multi-LED array the phrase corncob or 'cob' light is often used because the yellow elements on a cylinder look like a corncob when it is off. Those 'white' LEDs that use a blue LED as the exciter will usually pass some of the blue light through to give the RGB version of white. Those that use a near-UV LED as exciter use a mixture of red, green and blue emitting phosphors to produce white.

 

Still hoping someone here finds a true yellow LED mini-torch.

[Stefano]

Basically this is about color theory. You want to know if you can find a metamer pair with red+green and yellow, with almost monochromatic sources (LEDs), or truly monochromatic ones (lasers or monochromators), and compare how a human eye and a camera sensor see that.

 

The color perceived by human eyes as a function of wavelength has been scientifically described. The CIE 1931 color space is often used for that. We perceive colors in a non-linear way, you can't add two colors by adding the RGB components, you add the XYZ components. To go from RGB to XYZ or vice versa, you use a non-linear gamma function, combined with a matrix multiplication (which is linear).

 

To accurately display some spectral colors using primary colors, you would need a negative amount of one of the primaries (maybe two of them in some cases, but I'm not sure). That's why screens cannot display some spectral colors, and that's why if you add a gray background you can reproduce colors better, you make up for the negative amount of primaries (I may post some images tomorrow).

 

Andy helped me learn this. Maybe he can help you.

 

As for the camera, it tries to mimick the response of a human eye, but probably there isn't a perfect match. I noticed that in LED traffic lights (the ones we have here, but probably this applies in other countries too), the green light is a shade of green on the blue side, maybe around 520 nm. I see it as mostly green, but my phone's camera sees it as cyan, much bluer than I see it. Also, some cameras see violet (below ~430 nm) as blue.

 

As for yellow LED flashlights, I couldn't find one doing a quick search, but as Dave said yellow LEDs (native yellow, not phosphor-converted) can be easily found. If you have a power supply with current limitation, you can run them that way. I have 10 W LEDs in many colors, including red, yellow and green, I have a PTFE square but I don't have a stock camera that can shoot raw, and in order to control the brightness of multiple LEDs, I would need multiple power supplies, I only have one.

 

For low power LEDs, you could regulate the brightness by running them at a high voltage (like 12 V) and by putting a potentiometer in series. You could get some yellow and red+green bi-color epoxy LEDs and adjust their brightness until they look the same, or as similar as possibile.

[colinbm]

@Andrea B. I tried to do this, but my camera has a rolling shutter & the LEDs blink at a high frequency 
 

[Stefano]

If you put a capacitor in parallel to the LEDs you can reduce the flicker. Do you have access to the LEDs?

[colinbm]

9 minutes ago, Stefano said:

If you put a capacitor in parallel to the LEDs you can reduce the flicker. Do you have access to the LEDs?

Thanks Stefano
I have this RGD flashlight.
https://www.amazon.com.au/SecurityIng-Flashlight-Aluminium-Multicolor-Backpacking/dp/B08LYTN7N7?th=1

[Stefano]

If you can, you may disassemble it and put a capacitor in parallel to the wires that power the LEDs, if you have three of them (RGB), you would need three capacitors. If the capacitors are polarized, you have to be careful about polarity. I don't know why LED dimmers based on PWM (pulse-width modulation) don't have capacitors already. I don't like flickering LEDs, even though your eyes won't notice it if the frequency is high enough.

 

Anyway, if you take a photo with a long exposure (with low ISO), does the banding go away?

[colinbm]

@Stefano Thanks what size capacitor would be recommended please.
I'll try the longer exposure & low ISO too.

[lonesome_dave]

Stefano, thanks for those details about human color vision. I just found a Wiki page that describes how it works pretty well. Like everything else it is more complicated than you think.

 

I see they also have a Wiki page on 'Yellow'. Last item is a description of 'yellow snow'.

[Stefano]

7 minutes ago, colinbm said:

@Stefano Thanks what size capacitor would be recommended please.
I'll try the longer exposure & low ISO too.

I'm not sure how much capacitance is enough, it depends on the power the LEDs draw and the frequency of the flicker, probably you would need at least 1 µF, and 1 mF should be more than enough. 1 mF electrolytic capacitors won't fit inside your flashlight though.

 

I would try to take a longer exposure photo first, and if that doesn't work I would try the capacitors.

[ulf]

@Stefano Your assumption that a capacitor would solve the flickering problem is a bit dangerous for the torch.

It would be OK for an old style indicator LED design with high supply voltage and current limiting resistors. That is not how this torch is designed.

 

A PWM is used because this is based on power LEDs to make the design reasonably efficient and at the same time give a fine control of the average light emitted from each colour LED.

The PWM ratios are likely controlled by a small digital processor.

Adding capacitors by the LEDs might in worst case kill the driver circuit, or at least screw up the function.

 

Naturally my guesses might be wrong as I have no idea about the details of the electronics design, but I would personally not do that experiment if the torch was not very cheap and I had studied the circuit and understood it's function

I would at lest seek other alternative solutions first.

 

Almost always things are more complicated than you assume in the first place. 

To quote Murphy:

If anything can go wrong it will.

  If there are several things that can go wrong the one causing most damage will be the one going wrong.

    ......

 

 

[Stefano]

Thanks Ulf. I don't know how the LED driver circuit in a flashlight works, if you think it is better not to attempt to put a capacitor I second you.

[Stefano]

About colors, here's some data obtained using the 1931 CIE color matching functions modified by Judd (1951) and Vos (1978). By converting from XYZ to sRGB, we get the RGB values (between 0 and 1) as a function of wavelength.

 

Here's a plot, between 380 and 825 nm:

 

As you can see, sometimes the values go above 1, and sometimes they go below 0. For example, for green light (~520 nm) you need a negative amount of red primary.

 

Usually the values are clipped between 0 and 1. below a plot of the clipped values:

 

This is how the clipped RGB values look like as colors. They look nice, but there's some distorsion and clipping. This shows how a screen cannot accurately reproduce all spectral colors.

 

To compensate for the negative amount of primaries, we can add gray and reduce the brightness of the colors. By displaying colors with less saturation (mixed with gray), we can display them more accurately. This image is very close to how projecting a rainbow on illuminated paper (not in the dark) would look like:

 

Using the plots with the non-clipped RGB values (first plot), I found the cyan and yellow points (when you have the same amount of blue and green, or green and red).

Cyan = 491.15 nm

Yellow = 570.38 nm

 

I hope I didn't make mistakes. This can be confusing, I haven't done this for a while.

[colinbm]

https://scontent.fbne6-1.fna.fbcdn.net/v/t39.30808-6/400565371_10159918125381645_5048706546619380853_n.jpg?stp=dst-jpg_s600x600&_nc_cat=102&ccb=1-7&_nc_sid=5f2048&_nc_ohc=zqN1EkwPX_wAX__92YC&_nc_ht=scontent.fbne6-1.fna&oh=00_AfBAhGdcK8XXvUyrwYQVuiuSr3eQ_JU4u8tp12QXUplKEA&oe=655AB11C

[Stefano]

59 minutes ago, colinbm said:

https://scontent.fbne6-1.fna.fbcdn.net/v/t39.30808-6/400565371_10159918125381645_5048706546619380853_n.jpg?stp=dst-jpg_s600x600&_nc_cat=102&ccb=1-7&_nc_sid=5f2048&_nc_ohc=zqN1EkwPX_wAX__92YC&_nc_ht=scontent.fbne6-1.fna&oh=00_AfBAhGdcK8XXvUyrwYQVuiuSr3eQ_JU4u8tp12QXUplKEA&oe=655AB11C

Here the brain is performing a white balance. The background outside is slightly blue, the lines inside the "yellow" circle are white. If you try to white balance the image on the slightly blue background, the yellow circle will become slightly yellow.

[DKoch]

I recall in school we used to play with these Metameric matches..exactly your example: a red and green to match a monochromatic yellow spectral line. As Stefano illustrates with the human eye, there's so much variation in sensor spectral response! In the photochemical days, we used Minolta  "color meters" that basically had  R,G,B filtered photodiodes in them.They seem to have done a reasonably good job of matching the spectral response of the meter to Kodak color films...at least within limits!

 Strangely, it seemed to always work even with non black body sources! The most common example is matching our HMI ( nominally 5500K) to existing fluorescent lights, most commonly" cool white". Here we had a semi broadband spectrum from the phosphors with spectral lines from Mercury on top; the 546nm green being the most prevalent. 

  Our color meter would recommend a green filter for our lighting to match , but of course this is a broadband dye based green filter that is attempting to mimic a clean spectral line.  Weirdly it generally worked quite well... 

  If you really studied it, you could easily match whites by this method, but the reproduction of other colours, particularly skin tone, did not match exactly...

this has become an issue with contemporary LED lighting as well: the colour rendering is quite different from different manufacturers 

[lonesome_dave]

Relating back to beginning of this thread, I put together a set of LED flashlights for educational use that use native LEDs, not color filters. I labelled each one with the peak wavelength emitted.

I'm hoping the folks in Shenzhen will notice a missing yellow LED flashlight and start making one at around 580nm to make my set more complete. Another one at around 490nm (cyan) would be extra credit.

[Stefano]

Very cool flashlight set. I think the LED driver in those flashlights would work fine if you swapped the LED inside with another one with the same current rating and similar voltage drop (similar IV curve).

 

For example, a yellow LED might require a similar forward voltage (maybe slightly higher) compared to a red LED. A 490 nm LED would work almost the same as a green or blue LED. If that's the case, you can buy those LEDs (easily available) and make custom flashlights. David once said he made a 385 nm torch this way.

 

Ulf maybe knows better about LED drivers (I'm saying this in a friendly way), and could help you.

---

 

Anyway, I have 10 W LEDs in many wavelengths, including 490 nm. I roughly measured the peak wavelength with a homemade spectrometer and it was about 500 nm, I don't know how accurate that measurement was. It doesn't look cyan, it is greener, if I had to classify it as either blue or green I would say it's green. It looks cyan on camera. Other people in online reviews have said that, so don't expect it to look cyan like (0, 255, 255). Probably 490 nm does look cyan, but in that region of the spectrum the color is very sensitive to the wavelength, meaning that 500 nm is already too green and probably 480 nm is too blue.

[lonesome_dave]

Yes I think cyan might be a difficult color to perfect with an LED. The most accurate cyan light I think I've seen is from the 488nm laser diode (also the Argon gas laser). But those are narrowband. The wider spectrum from an LED at that wavelength would include quite a bit of blue and green so I'm not sure how pure it would look.

 

I'm sure you're right about swapping out the LEDs but I'm afraid soldering on one of those little flashlights is beyond my skill level.

[Stefano]

There's a lab in my university with a monochromator. Currently it is not being used so I briefly used it to look at colors. It is probably an old one, with a manual crank, and you need to input your own light source. I used a zoomable LED flashlight I brought from home, in the zoomed-in configuration to collimate light. The monochromator clearly uses gratings. I cranked it to infrared, and at around 900 nm I started seeing blue again (second order diffraction).

 

I don't know how accurate it is, given it is old, but being a mechanical instrument I suppose it still works well.

 

The calculations I did in a previous post suggested pure yellow is located around 570 nm. Using the monochromator, I identified pure yellow at 575 nm. This is one of the regions of the visible spectrum where hue is very sensitive to wavelength. 570 nm looked noticeably greener and 580 nm started looking amber.

 

There's no color that looked cyan to me like (0, 255, 255) does on a screen. Blue turns to green looking at most turquoise. It's possible that at higher brightness it looks different. Anyway, 490 nm looked about in between blue and green. My "490" nm LED looked similar to 500-510 nm (from memory, I didn't have it to compare), and I think 488 nm (a laser wavelength that is often described as cyan, which I set on the monochromator out of curiosity) looked bluer.

 

Pure green was in the 540s, 520 nm is a bit bluer. Also, blue at 450 nm looks solid, deep blue, but turns clearly violet at 420 nm. I didn't expect color to be so sensitive there.

 

I understand that all considerations above are subjective. Other people may see colors differently. I know I am not colorblind (never had problems with colors, and I did an Ishihara test a few years ago in the studio of an eye doctor), but we are not made equal.

 

Monochromator set at 575 nm. To me, this was pure yellow, but my phone picked it up as a greener color. That's more or less how 570 nm looked like to me, yellow on the green side.

[Andrea B.]

cool beans, Stefano! Thank you. 

 

Stefano wrote:  There's no color that looked cyan to me like (0, 255, 255) does on a screen.

Now that is quite interesting !!!

 

I've read a lot about how we all see color. For the non-colorblind there are indeed some variations in color perception. But generally it seems that non-colorblind color perception is similar across humans. It is not easy to determine that though because we all have our own cultural overlay on color perception. Some cultures don't have names for some colors. And some cultures "lump" a color range under one name. 

 

Please, don't anybody take what I just wrote as any kind of scientific "truth". It is simply a general comment on a very fascinating topic.

 

I sometimes make up color patches or displays to test my own discernment of color. Here is one example. The wavelength to RGB color converter which I use is found here: LINK.

I will reproduce the author's warning: 

A frequent way of referring to colour on computer screens is by using the RGB system. In this model, each colour is given a value for each red, green and blues components ranging from 0 to 255, giving a total value of 16.7 million possible colours. However, due to the very complex way in which the eye perceives colours, we can see colours which are outside of the gamut of the RGB scheme - there is no unique mapping that definitively converts a wavelength to a colour, and as such the above tool should been seen as more of an approximation than a rigorous resource.

 

These just-for-fun patches APPROXIMATE the differences in green which are 10 nm apart and which are 10° apart on the color wheel. I note that in the Greens, it is difficult to discern differences which are 10° apart (second set).

 

 

 

*****

 

FWIW, an observation: I don't think I can see this 120°, rgb(0,255,0) green in an actual rainbow. The rainbow green looks completely different to me when looking in real time. And any rainbow photos I've made don't seem to show this screen green either. 

Stefano's "projected rainbow" looks more like the actual rainbow green.

 

*****

 

Dave, I really like that flashlight set. That would be fun to play with.

[Andrea B.]

https://www.photonlight.com/products/photon-micro-light-ii-led-keychain-flashlight

I have a set of these little keychain lights to play with. But my original set did not have yellow. In my set the purple light is labeled 405nm. It will induce some fluorescence. The others I have are 470, 495 and 525 nm.

 

Their currently sold yellow beam claims 592 nm. 

 

I wouldn't recommend these little keychain LEDs for photography. They are bright but too small. I got mine for some experiments years ago. They are nicely made and do last a long time as claimed.

[Andrea B.]

DKoch writes:  this has become an issue with contemporary LED lighting as well: the colour rendering is quite different from different manufacturers.

 

Yes, we've gone slightly nutso trying to match existing LEDs when we needed to replace ceiling lights. (We have a lot of recessed "can lights" in this house.) Finally the SigOth bought a huge new box of them and started replacing them all. That was probably overkill, but it was slightly strange in some rooms with mixed K.        

[photoni]

9 hours ago, Andrea B. said:

cool beans, Stefano! Thank you. 

 

Stefano wrote:  There's no color that looked cyan to me like (0, 255, 255) does on a screen.

Now that is quite interesting !!!

 

I've read a lot about how we all see color. For the non-colorblind there are indeed some variations in color perception. But generally it seems that non-colorblind color perception is similar across humans. It is not easy to determine that though because we all have our own cultural overlay on color perception. Some cultures don't have names for some colors. And some cultures "lump" a color range under one name. 

 

Please, don't anybody take what I just wrote as any kind of scientific "truth". It is simply a general comment on a very fascinating topic.

 

I sometimes make up color patches or displays to test my own discernment of color. Here is one example. The wavelength to RGB color converter which I use is found here: LINK.

I will reproduce the author's warning: 

A frequent way of referring to colour on computer screens is by using the RGB system. In this model, each colour is given a value for each red, green and blues components ranging from 0 to 255, giving a total value of 16.7 million possible colours. However, due to the very complex way in which the eye perceives colours, we can see colours which are outside of the gamut of the RGB scheme - there is no unique mapping that definitively converts a wavelength to a colour, and as such the above tool should been seen as more of an approximation than a rigorous resource.

 

These just-for-fun patches APPROXIMATE the differences in green which are 10 nm apart and which are 10° apart on the color wheel. I note that in the Greens, it is difficult to discern differences which are 10° apart (second set).

 

 

 

*****

 

FWIW, an observation: I don't think I can see this 120°, rgb(0,255,0) green in an actual rainbow. The rainbow green looks completely different to me when looking in real time. And any rainbow photos I've made don't seem to show this screen green either. 

Stefano's "projected rainbow" looks more like the actual rainbow green.

 

*****

 

Dave, I really like that flashlight set. That would be fun to play with.

.

There are some problems in the two tables you inserted.
They are files without a color profile
therefore everyone will see them with the profile of their monitor (standard or calibrated) and with the gamut of their monitor.
None of us have a monitor that displays all LAB colors
(I remind you that sRGB even though it uses the same numbers is a part of AdobeRGB or P3, and these two are a part of ProPhotoRGB, which is a part of LAB)
To get an idea of what your device sees and doesn't see, check it with this LINK

.