Review: 510nm Direct Green Diodes / Build Photos (DGH-N1, DGH-N2)

[ElektroFreak]

Especially to a camera.. but like eyeballs, cameras exhibit a great deal of variation from one model to another. I might be able to clearly see a difference in hue, but someone else might not. My Nikon camera might be able to easily show the difference between a couple nm, while my Casio camera might not.

When dealing with human perception in particular, this, in short, is the problem with science that tries to mathematically quantify that which is subjective. Of course there needs to be some scale that helps mathematically define color space for the purposes of creating balanced imagery or lighting, but when I see people throwing $50 explanations around and somewhat detailed math equations etc.. pertaining to something that is subjective and very much naturally variable it seems to me that there will ALWAYS be some unknown quantity, no matter how precise the math involved. And I love math.. I like to see it used wherever possible, whenever it's necessary. So far, no one has been able to satisfactorily explain to me why such detail is necessary, OR if it's even warranted. Because of the aforementioned natural variation I don't think it is, which is why I've taken such an interest in the subject. Personally I think that the color space chart and theory is being used as a sort of de facto standard that defines human color perception, but I don't think that it's even close to 100% accurate in that usage. Rayleigh effects, color blindness, natural photoreceptor variation, scotopic vs photopic sensitivity based on ambient light levels etc. are all examples of things that do not seem to be covered by the math and charts that have been presented here, all of which are very important parts of the issue of perception.

I'm not an expert though, and I could certainly be wrong, but so far no solid clarification of this usage and how it accounts for the variations and unknowns that have been mentioned has been put forth.

 

[lazeristasUVISIR]

When dealing with human perception in particular, this, in short, is the problem with science that tries to mathematically quantify that which is subjective.

For it me this stuff is a try to standardize all eyeballs. It works quite well for many applications, for example monitors TVs cameras. I used only tabulated values

 

[ElektroFreak]

You're absolutely right about the application defining the usefulness of these tools. What you describe is exactly what these charts were designed for, and in those applications the previously mentioned variation and unknown quantities are of limited importance. On the other hand, for the purposes of saying definitively how a specific wavelength will be perceived by the entire human population or by a single person, these unknowns are very important. Without taking them into account all that can be obtained is a basic estimate, a rough guide as to how the wavelength in question will be perceived. When dealing with basic estimates and/or rough guides, precision mathematics is a wasted effort.

 

[pullbangdead]

The chart and data you're dissatisfied with is over 80 years old, and a LOT has been done since then.

Every discussion of color perception I've ever seen here has been based on the 1931 chart, which in turn was based on 1920s experiments. If you're dissatisfied with the color perception science talked about here (all oh which I've seen is based on the 1931 chart), then go read more about the math behind it, and perhaps more importantly, go look into all the science that has been done SINCE THEN. Because there has been a LOT of work done.

Rather than complain about the misapplication of that work, look into the updated work that is much more advanced. Every critique you're given in this thread has been considered by CIE, please look into it.

Is CIE work misapplied? Definitely. Is mathematical work on color perception wasted? Absolutely not, and you can't say otherwise until you've looked into any of the work done in the last 80 years. Is every human's perception the same? Absolutely not, which is why we have statistics.

I do agree though that a lot of misapplication happens here, and perhaps that is the only point you were trying to make, but it seemed to me that you were taking issue with the science itself. If that isn't the case, then I apologize.

 

[pullbangdead]

On the subject of color difference, ie how far apart colors "look", a LOT of work has been done on that too. Here's a link to CIE work, which goes way beyond the 1931 chart, up to today's standards: Color difference - Wikipedia, the free encyclopedia


For the purposes of this discussion though, the answer is the MacAdam Ellipse. On the 1931 CIE chart, a MacAdam ellipse is the region in which colors are indistiguishable from one another. Here's a sample chart, with sample MacAdam ellipses (ie areas of indistiguishable color) plotted on it. This gives an idea over the whole area of the chart.

 

[ElektroFreak]

I do agree though that a lot of misapplication happens here, and perhaps that is the only point you were trying to make, but it seemed to me that you were taking issue with the science itself. If that isn't the case, then I apologize.

Yeah I must have a really bad way with words because that was pretty much it. I really do remember saying several times that the tools are sound, the application thereof.... often not so much. I also did say that math is necessary, but (imo) misapplied "precision" calculations? Again... not so much.

Good to hear from you on the subject though, you always have valuable input.. I'm no expert, but I know enough about human variation to be skeptical of certain claims I've heard over time.

In your opinion, how precise mathematically are the boundaries of those ellipses? Would all humans have the same perception? I can tell you that I see color variation within the topmost ellipse.

If the calculations are indeed very involved, then what is done mathematically to account for individual variation if these charts and data are to be used in an application that would benefit from that kind of thoroughness? For example, when mathematically trying to establish a precise figure for how much brighter a specific wavelength of laser will be than another to other people's eyes? I notice that none of those ellipses encompass pure wavelengths, where most lasers would reside.

By my logic, with the data I've see up until now (not including what you've posted here as I haven't had a chance to read the links yet) all that can be done is estimate with some accuracy..

 

[qumefox]

In your opinion, how precise mathematically are the boundaries of those ellipses? Would all humans have the same perception? I can tell you that I see color variation within the topmost ellipse.

This is also an example chart used to show the ranges of indistinguishable color ranges. How that actual picture looks isn't meant to be a 'test' of it. Only a representation of the data. How that image itself looks will vary from monitor to monitor.

 

[ElektroFreak]

This is also an example chart used to show the ranges of indistinguishable color ranges. How that actual picture looks isn't meant to be a 'test' of it. Only a representation of the data. How that image itself looks will vary from monitor to monitor.

OK, but that still doesn't clarify my issue, so let me try once again to state it as clearly as possible:

If a person claims (using CIE charts and data as the basis of their claim) that, just for the sake of a random example, 1W of 532nm will be 14.33x brighter to the human eye than 1W of 445nm, where does the precision come from? That .33 on the end there.. that's pretty precise, and quoting figures at that level of precision ends up, at least from my view, as an attempt to suggest that the difference will be precisely the same for each and every person on earth. IMO, based on everything I know about this, it is more than sufficient to simply estimate that 532nm will be ~14x brighter than an equally powered 445nm because of the fact that everyone sees things a little differently from one another.. and that's not even factoring rayleigh effects in at all, which changes things even further. If we do take Rayleigh effects into account we will have a more accurate estimate, but it is still lacking true precision because of natural human variation. To my eyes, 1W of 532nm is maybe 3x as bright as 1W of 445nm, which is a far cry from what is being suggested by using these tools to quantify.

At any rate, I can only state things so many times, and so many different ways before I get exhausted with the topic. That time has come.. maybe the clarification I seek will come to me in a dream sometime.

 

[rhd]

You keep bringing up this "accuracy" thing - that's not the way to look at this. Maybe this example will finally put the concept to rest.

Take a look at the data below. Four people are polled to find out how intense they think each ingredient in a recipe is. They all agree that salt is the most intense flavour, so they all rate salt as a "1", and provide their rating of each other ingredient relative to salt.

We use the data to compute the "average" response to each ingredient's relative intensity.

Now, here's my question. Do you look at that "average" response and say "Hey, how can we have 4 decimal points of precision to something this subjective?". No, of course not. Or at the very least, we shouldn't if we're logical human beings.

Suppose you made an argument similar to the one you've made for wavelength intensity, and said "we should only be stating 1 'decimal point of accuracy' in our relative intensity measurements", and then proceeded to round all your figures to the nearest 1 decimal point. What happens then?

Well, for one, you would end up with the impression that on average, people found both thyme and parsley to be equally intense, when in reality, every single person found parsley to be more intense than thyme. More importantly, you would be INCREASING THE STANDARD DEVIATION OF THE MODEL, and thus making it LESS ACCURATE. In other words it would be LESS LIKELY TO ACCURATE REFLECT EACH PERSON'S PERCEPTION.

I sincerely hope that this issue is put to rest. It is so incredibly absurd to keep harping over an issue of rounding vs not rounding. If the "decimal points of accuracy" are bothering you, then just multiply every value in the CIE tables by 1,000,000. You'll end up with no "decimal points of accuracy'" to speak of

 

[ElektroFreak]

You've already said that, so I'm with you.

But that still doesn't answer my question. HOW DO YOU KNOW, BEYOND ALL DOUBT, THAT 532nm WILL BE EXACTLY 14.33x BRIGHTER THAN THE SAME AMOUNT OF 445nm LIGHT TO THE EYES OF EACH AND EVERY PERSON ON EARTH.

You don't.

That's my point.

That point, in and of itself, removes all need for the level of precision you are working at (no matter how you want to express that precision). THAT'S what I'm harping on. There is no way to calculate what you are attempting to calculate with any real precision. The best you can do is to attempt to average the human response curves across the population, which is more than good enough to establish a standard for color space and color balance, but THIS WILL NOT WORK as a means of establishing precise values for the perceived intensity of each wavelength increment, especially if that increment is 1nm or less. THAT'S why I used rounding in an attempt (a poor attempt obviously) to demonstrate what "less precision" means. That's all. Maybe I'm not stating things clearly, or maybe people aren't reading or actually processing what I'm writing.

 

[rhd]

You've already said that, so I'm with you.

But that still doesn't answer my question. HOW DO YOU KNOW, BEYOND ALL DOUBT, THAT 532nm WILL BE EXACTLY 14.33x BRIGHTER THAN THE SAME AMOUNT OF 445nm LIGHT TO THE EYES OF EACH AND EVERY PERSON ON EARTH.

You don't.

That's my point.

That point, in and of itself, removes all need for the level of precision you are working at (no matter how you want to express that precision).

YOU'RE RIGHT, YOU DON'T!

But that's NOT WHAT WE USE STATISTICS FOR!

The probability of becoming victim to a violent crime on a Sunday in two cities might be:

Darkburry: 0.002347
Lighthaven: 0.001183

Based on your argument above, when given the data on these two cities, your response would probably be:
"HOW DO YOU KNOW, BEYOND ALL DOUBT, THAT Darkburry WILL HAVE EXACTLY 1.98393914 X (0.002347 / .001183) THE AMOUNT OF VIOLENT CRIME THAT Lighthaven HAS ON EACH AND EVERY SUNDAY OF THE YEAR?"

And you would probably answer:
"You don't. That point, in and of itself, removes all need for the level of precision you are working at (no matter how you want to express that precision)."

Except that, when talking about statistics, and how statistics are meant to be used an interpreted, you're just dead wrong. Not to get all Ad Hominem on this argument, but genuine question - have you taken a stats class recently?

 

[ElektroFreak]

No, I haven't taken a stats class, been out of the university for years.

Asking me if I've taken a stats class, and calling me wrong STILL DOESN'T ANSWER THE QUESTION. You are attempting to use statistics in an effort to put a precise value on the difference in perception between wavelengths, are you not? If that is the case, then that IS obviously what YOU are attempting to use statistics for.

Clearly either no one can answer my question, or no one is willing to. Either way I have presented questions as clearly as I can, and the only thing anyone has offered in the way of clarification is to call me wrong numerous times and ask me if I've taken a statistics class.

Your answer does not justify or clarify your attempts to put a precise value on the difference in perception between two pure wavelengths. Asking me if I've taken a stats class does not answer the question. Telling me I'm illogical and wrong DOES NOT answer the question. Since you cannot possibly be 100% accurate with your declaration that (according to these charts and data) 1W of 532nm light will be exactly 14.33x brighter than 1W of 532nm light, then what is the point of calculating to that degree in the first place? It's my belief that using the CIE data in this manner is a misuse of the data.

When it's been proven that no one can answer my question, that's my cue to leave the discussion because obviously the expertise necessary to answer it cannot be found here..

 

[Arayan]

I've missed this thread, very very interesting, thanks to all :bowdown:

 

[rhd]

If you're looking for a simple, succinct and concise answer, it was best found in the portion of my previous response above that I'm including below. We include a high degree of precision in these statistics, because not doing so would:

"INCREAS[E] THE STANDARD DEVIATION OF THE MODEL, and thus making it LESS ACCURATE. In other words it would be LESS LIKELY TO ACCURATELY REFLECT EACH PERSON'S PERCEPTION."

There, that's the answer. It's sometimes hard to answer a big conceptual question in a single line, but I think that comes pretty close. The high degree of precision is there, because if it wasn't, you would have a higher standard deviation, and a less accurate model. That is what "justifies the precision" (thereby answering your question).

 

[ElektroFreak]

You're right WHEN THE DATA IS APPLIED PROPERLY, for example when used to establish a white balance and otherwise tune imagery devices. But it cannot be applied in the manner you are attempting with any real accuracy. That's all I've been saying the whole time. I think that this is a misuse or misapplication of these tools.

I can't say exactly what would or could produce a truly accurate result, so you may be using the best tool available, but I just have issues being told that, because of some math, 1W of 532nm light will be 14.33x brighter than 1W 445nm, when this is NOT even close to factual according to my eyes. What good is that estimate if it's false?

 

[rhd]

You're right WHEN THE DATA IS APPLIED PROPERLY, for example when used to establish a white balance and otherwise tune imagery devices. But it cannot be applied in the manner you are attempting with any real accuracy. That's all I've been saying the whole time. I think that this is a misuse or misapplication of these tools.

Now you're just making stuff up.
"Applied properly" ?!? huh ?!?
This is a statistical model. You don't get that?

I can't say exactly what would or could produce a truly accurate result, so you may be using the best tool available, but I just have issues being told that, because of some math, 1W of 532nm light will be 14.33x brighter than 1W 445nm, when this is NOT even close to factual according to my eyes. What good is that estimate if it's false?

Now we're finally getting to the route of the problem. You have issues being told that "because of some math", something else occurs, when that something else is contrary to your own perception of reality.

Fine. Lots of people are that way. Irrational non-scientific people, who think their perception trumps the scientific method and statistics. Honestly, I have ZERO respect for a mindset where one person places their own subjective experience above science.