Testing frequency response of the human eye

[Stonius (Markus)]

This popped up in the horsehead thread, but I thought I'd make a separate thread here.


The discussion was around how people's eyes have different sensitivities at different frequencies. Some people are more sensitive in the area where red starts heading towards the infra-red which is where a lot of emission nebula emit their light. TBH, I'm still a little foggy on what causes this variability in terms of physiology but it got me thinking - would it be possible to create an experiment to measure such frequency dependent sensitivity?


Presumably, all one would need is a standardised broadband light source and a range of filters, perhaps red, Ha, S2, maybe Hb too.


If the light source were baffled such that the only light emitted was through the filter, one could set it a standard distance away from a (standardised) white card in a dark room and by turning the light on and off one could determine detection at those frequencies.


The biggest issue I see is the procurement of a standardised broadband light source as a light source will have its own emission curve.


An alternative would be to buy some near infra-red LEDs such as these ones, available in wavelengths from 670 to 1650nm. That may be simpler to control. What are your thoughts?


Markus

[Stonius (Markus)]

Or there's even this - a continuously variable LED that encompasses wavelengths from 650 to 1050 nm


https://au.rs-online.com/web/p/ir-leds/2186867

[Stonius (Markus)]

Okay, at this point I'm just talking to myself, but here's the main issue as I see it.


Say you can generate standardised amounts of near IR light using the above mentioned LED's, the source will likely *not be flat across all frequencies.


To measure that curve requires something that *is sensitive to IR light, like say, an astronomical camera. Only problem is that also has a sensitivity curve.


So the question really becomes; "can you compensate for the emission and detection curves of the source and the sensor with enough accuracy to generate a reasonably accurate plot of the human eye's response?"


I don't know if I'll actually get around to experimenting with this. The datasheet also says this thing is pretty bright and not to look directly at it. I also know that prolonged exposure to IR light causes cataracts, so something to be wary of.


Okay, I'll stop talking to myself now :-D


Markus

[JA]

Originally Posted by Stonius

This popped up in the horsehead thread, but I thought I'd make a separate thread here.


The discussion was around how people's eyes have different sensitivities at different frequencies. Some people are more sensitive in the area where red starts heading towards the infra-red which is where a lot of emission nebula emit their light. TBH, I'm still a little foggy on what causes this variability in terms of physiology but it got me thinking - would it be possible to create an experiment to measure such frequency dependent sensitivity?


Presumably, all one would need is a standardised broadband light source and a range of filters, perhaps red, Ha, S2, maybe Hb too.


If the light source were baffled such that the only light emitted was through the filter, one could set it a standard distance away from a (standardised) white card in a dark room and by turning the light on and off one could determine detection at those frequencies.


The biggest issue I see is the procurement of a standardised broadband light source as a light source will have its own emission curve.


An alternative would be to buy some near infra-red LEDs such as these ones, available in wavelengths from 670 to 1650nm. That may be simpler to control. What are your thoughts?


Markus

Hi Markus,

I've pondered similar testing involving gauging camera frequency response, BUT in the case of the human eye its output response can really only be gauged in two ways:

1. an individual's perception of the brightness at a given frequency or
2. some kind of reading of the signal strength in the optic nerve / brain

The 2nd option is not really that do-able for the home experimenter, so it leaves some sort of test involving a person telling you about the perceived brightness. Possibly something like .... "Oh that is twice as bright as the previous one you showed me" etc....

There is a similar problem with understanding the response of the human ear and forms the basis of such things as the Fletcher-Munson curves/ equal loudness countours in Audio and with various frequency weightings and Loudness contours can help accomodate response non-linearity. The reason I went in to the response of the ear/hearing testing is that it does give an insight in to a possible test method for the visual frequency response that you are interested in measuring.....

Something of a start would be to measure the lowest intensity at which a person, in a controlled dark environment (analagous to a person in a "silent" sound booth) could just perceive a colour out of the darkness ("nothing" so to speak). I'm almost certain that frequencies around Green will win the perception test, in the sense that they'll be perceived at lower intensities than reds and Blues. It's one of the bases of the two Green pixels in the RGB Bayer Matrix Array for one Red and One Blue: RGGB in effect.

Running that test of course requires a calibrated light source: one that you know the intensity and the frequency of for a given set of input conditions such as voltage/current etc to the calibrated light source (unless you use broadband natural daylight/sunlight and various colour and IR filters?) .... Then just adjust the source intensity until the person can no longer perceive the light source (or switch between a light/no-light channel to remove user bias/error) and note the intensity and frequency and move on to the next frequency.

But this only tests for the low level response perception limit and whether it's linear. To test the response at other than the lower limit of perception, is harder and will require more input for the subject. You could try and set the Intensity at a constant brightness from your calibrated light source and then as your scan the frequencies (colours) ask then whether the red was as bright as the orange, for instance. In doing this it would be best if the person didn't have to rely on their visual memory so 2 identical sources could be used to accomodate that difficulty. You could end up possibly making your own version of Flether-Munson curves/Equal Loudness Curves for the eye as a function of light intensity.

Or not .....

Best
JA

[glend (Glen)]

Worthwhile reading on the subject.

https://en.m.wikipedia.org/wiki/Scotopic_vision

Most human observation astronomy is accomplished in low light level Scotopic vision, in which colour sensitivity is a mute point.

[Stonius (Markus)]

Originally Posted by JA

Hi Markus,

I've pondered similar testing involving gauging camera frequency response, BUT in the case of the human eye its output response can really only be gauged in two ways:

1. an individual's perception of the brightness at a given frequency or
2. some kind of reading of the signal strength in the optic nerve / brain

The 2nd option is not really that do-able for the home experimenter, so it leaves some sort of test involving a person telling you about the perceived brightness. Possibly something like .... "Oh that is twice as bright as the previous one you showed me" etc....

There is a similar problem with understanding the response of the human ear and forms the basis of such things as the Fletcher-Munson curves/ equal loudness countours in Audio and with various frequency weightings and Loudness contours can help accomodate response non-linearity. The reason I went in to the response of the ear/hearing testing is that it does give an insight in to a possible test method for the visual frequency response that you are interested in measuring.....

Something of a start would be to measure the lowest intensity at which a person, in a controlled dark environment (analagous to a person in a "silent" sound booth) could just perceive a colour out of the darkness ("nothing" so to speak). I'm almost certain that frequencies around Green will win the perception test, in the sense that they'll be perceived at lower intensities than reds and Blues. It's one of the bases of the two Green pixels in the RGB Bayer Matrix Array for one Red and One Blue: RGGB in effect.

Running that test of course requires a calibrated light source: one that you know the intensity and the frequency of for a given set of input conditions such as voltage/current etc to the calibrated light source (unless you use broadband natural daylight/sunlight and various colour and IR filters?) .... Then just adjust the source intensity until the person can no longer perceive the light source (or switch between a light/no-light channel to remove user bias/error) and note the intensity and frequency and move on to the next frequency.

But this only tests for the low level response perception limit and whether it's linear. To test the response at other than the lower limit of perception, is harder and will require more input for the subject. You could try and set the Intensity at a constant brightness from your calibrated light source and then as your scan the frequencies (colours) ask then whether the red was as bright as the orange, for instance. In doing this it would be best if the person didn't have to rely on their visual memory so 2 identical sources could be used to accomodate that difficulty. You could end up possibly making your own version of Flether-Munson curves/Equal Loudness Curves for the eye as a function of light intensity.

Or not .....

Best
JA

That's why I love this place - so many smart people in one spot.


I agree a blinking threshold detection would be best. Measuring perceptions of brightness would be tricky. and yes, the hearing test analogy is very apt. Issues with the variability of the light source and current could be avoided by using the sun, as you say.


I would imagine something like screwing an S2 filter into a 2" adapter and creating some kind of eyecup that seals completely - essentially turning your S2 filter into a monocle.


Adapt to darkness for half an hour in a dark room with no light, then bump into the furniture until you find your way outside (aim for noon to avoid atmospheric affects?). You can then stack ND filters on top of the S2 filter to test different thresholds with known attenuation.


The thing is, what you are looking at will have different reflectivity depending on composition and sun angle. I'm thinking one of those 18% grey balls they use for lighting callibration in visual effects shots might be good.


To test for threshold, someone else can block the sightline with a black card.


The biggest issue with this is variations in latitude and transparency, but I think a camera could be used to determine how much light is hitting a surface and compensate adequately for these variations.


In this case, the cost is the grey ball, a black card and a few ND filters.


Maybe that would work? Even if I never actually do it, it's fun figuring out how you *would :-)


Markus

[Stonius (Markus)]

Originally Posted by glend

Worthwhile reading on the subject.

https://en.m.wikipedia.org/wiki/Scotopic_vision

Most human observation astronomy is accomplished in low light level Scotopic vision, in which colour sensitivity is a mute point.

That's actually a really good point. Which would seem to conflict with the idea that some people are less sensitive to certain frequencies - which was posited as the reason why some experienced observers can't see the horsehead, even in large telescopes.


But now you mention it, the horsehead is a H beta nebula, which emits in at 486nm, which is in the *green part of the visual spectrum, yet all the RGB pictures show a red background nebulosity, which I'm struggling to make sense of.



But then wikipedia has this to say; "Infrared, as implied by its name, is generally considered to begin with wavelengths longer than visible by the human eye. However there is no hard wavelength limit to what is visible, as the eye's sensitivity decreases rapidly but smoothly, for wavelengths exceeding about 700 nm. Therefore wavelengths just longer than that can be seen if they are sufficiently bright, though they may still be classified as infrared according to usual definitions. Light from a near-IR laser may thus appear dim red and can present a hazard since it may actually be quite bright. And even IR at wavelengths up to 1,050 nm from pulsed lasers can be seen by humans under certain conditions."


So if it's bright enough, you can even see longer light wavelengths, but is this a case of the Mesopic (colour vision) or Scotopic (luminance vision)?


Markus

[multiweb (Marc)]

I have a feeling the majority of people not sensitive to red lights given the number they run through in my area.

[sharkbite]

found this by working the google on the internet machine...

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5566267/

this is more around the correlation between sensitivity to bright light and headaches....

but this method could be used to test the low end as well...

[By.Jove (Jove)]

I’m sure an experiment can be devised analogous to a hearing test and I can recall doing something similar with an optician many years ago regarding my own vision anomalies.

What I would suggest is this:

1. Take a broad-spectrum source, and set up a slit and spectroscope so a person looking into an eyepiece sees a nice big patch of 1 colour. You need a spectroscope that has some sort of calibrated dial so you can determine what colour they are being shown; and you want to do this remotely so that cannot tell when you changed the colour - or not. These devices exist but not cheap.

2. You will also want a means to dim the light source too; the slit width/length would probably do.

3. Put a shutter in the path, electrically operated that you can operate remotely, so you can turn the light on, at will, randomly.

4. Give the test person a button to press each time they see a coloured patch in the eyepiece.

Now, the way this works is that each time the person sees a colour patch, they press the button. This has to correlate with the moment you opened the shutter. You start with orange, move into red, and continue through the spectrum…

If they don’t press the button when you opened the shutter, or pressed the button with the shutter closed that’s a false response.

If they’re scoring above say 95% correlation, obviously they are seeing what you show them. As you change the frequency (colour) this correlation will change if they aren’t seeing the patch of colour.

The question is, at what level are they not seeing anything reliably ? 80% ? 75% ? 50% ? This is standard statistics and confidence-interval stuff for the mathematically inclined.

[Merlin66 (Ken)]

Markus,
the Horsehead is a H alpha emission object.
There are also (in all H alpha nebulae) resonance emissions in H beta ( close to 50% of the H alpha, can be affected by dust etc).
As the eye is more sensitive to the H beta that's why it "seems" to be better seen in H beta rather than H alpha.


I have in my library, what is said to be the definitive work on colour vision:


Researches in colour vision and the trichromatic theory, by Sir William de W. Abney. Longmans, Green and Co, 1913


You may find a copy in archive.org. Here you go: https://archive.org/details/researchesincolo00abnerich

[LewisM]

No idea, but I do know that :
a. my night vision is particularly acute to the point I can see foxes, rabbits etc in pitch darkness that others cannot. My dark adaption also is quite short, even when ruined by someone with a red torch (see next )

b. any red light at night is FAR too bright for me. Even when I flew commercial aviation, I couldn't use a red torch as it blinded me, unless I put MANY MANY layers on the torch.


Consequence of (a) is i find using large aperture instruments visually nauseating as I can see WAY WAY too much - too many stars gives me trypophobia (a condition I don't suffer in regular life). I used to use a Mewlon 300 quite often and also Matt's Massive Mother of All Telescopes, and whilst the views were amazing, they were overwhelming. I have on occasion literally had to sit own for a period after using them. Might explain why I have NVER had aperture fever even for imaging. Content with my Wee Tak



Consequence of (b) is I don't even use a red torch at night or if necessary, a VERY filtered one - the complimentary red key ring lights Alex Massey gives with orders - a simple LED - are significantly too bright for me - I have darkened mine a LOT with many layers of clear red lacquer.