Humans looking at colored lights each at the same luminescence perceive yellow as the "brightest" color. Green is somewhat dimmer, and red is very much dimmer. Using a common formula for calculating mean luminance, 30% is red, 11% is blue, and 59% is green (R=645.16nm, G=526.32nm, B=444.44nm).
Theoretically, I should choose luminous intensities using ratios of Blue=90x, Red=33x, Green=17x, and I'm guessing, Yellow=15x.
Testing random LEDs from my junk box, a lux/fc meter with expired calibration, and two volunteers, Blue=152x, Red=143x, Green=23x, Yellow=15x.
The application is for indicators easily distinguished under office lighting. All will have the same package (T1,or T1.75), similar viewing angles (about 60 degrees), with the same lens (all clear/tinted, or all defused). They will be driven based on the data sheet's test current and luminosity curve.
I am assuming the all the other zillion factors like efficiency, encapsulation, geometry, orientation, etc. are specified constant (viewing angle, lens), or they are incorporated into the millicandelas (mcd) rating provided by the manufacturer.
1. Anyone care to comment on these ratios, based on real life experience?
My experimental red number calls for 3 times more luminosity than the theoretical one.
2. What mcd target values/colors would your experience suggest for LEDs viewed under office lighting?
My testing conditions can't give results in mcd, so my test data doesn't help. Once I get a suggested mcd for a color, I'll use the ratios to get the other mcd values. And of course, you can't buy exactly what the calculator says... so "rules of thumb" rule!
Update... ---> Think of a traffic light. What if drivers complained that the yellow lamp was too dim that the weren't sure when it was lit...?
I don't want my users saying they can always tell if the color1 or color2 idiot light is on, but they are never sure about color3 without getting their eye next to the device.