Interesting new lighting options

[Brad Johnson]

We're investigating illumination options for a number of campus remodels.  The last year has really been favorable for LED-based lighting options, and darned good ones.  The units that work are still a little pricey, but the savings in electricity and maintenaince far outweigh the price.  The best part are the color temp, coverage, and intensity options.  One excellent example is the Philips A19 12.5w LED lamp designed to replace the good old 60w incandescent.  It's true 2700k illumination, it doesn't mind the cold, it's far more physically robust than the CFLs we're looking to replace, power cycles don't hurt it, it's dimmable, and field evidence is pointing to an actual service life of 12,000-15,000 hours.  I was so impressed with the samples we received from Philips that I picked up a couple for the house.

Another phenomenally good product is the Cree CR6 can light retrofit lamp.  All you have to do to retrofit an existing 6" housing is to pull the trim ring, remove the socket retaining screw (or rivit), and screw in the lamp assembly.  The assembly friction-fits into the existing can.  We just installed a couple dozen and it was dirt-simple.  For all practical purposes the quality and quantity of illumination is unchanged vs the 60w incandescent lamps previously used in the cans.  Best part is they are available mainstream now, at Home Depot of all places.

We're also about to install a bunch of LED-based 2x4 troffers as part of a conference room remodel.  The old fluorescent units don't even come close.  Color temp is much improved, CRI is off the charts better, and they are dimmable down to 10% (something no fluorescent could think of doing).  Price is a little steep compared to the old stamped steel fluorescent units, but the savings in maintenance (both bulb and ballast) will offset the difference in under five years.  Factor in the ongoing elec savings and over the expected 80,000 hr service life (a mfg claim, but we've confirmed with field results from third-party testing) it gets pretty appealing.

To say I'm impressed would be an understatement.  Two or three years ago LED lamps were a high-dollar curiousity.  Now?  Mainstream, and getting better all the time.

Brad

[HankB]

There's also this one, dimmable, available in cool or soft color gamuts:  http://solutions.3m.com/wps/portal/3M/en_US/LEDAdvancedLight/Home/Products/

[Marnoot]

After getting sick of the warm-up time on the CFL bulbs we had in our recessed lights, and learning of a $20/fixture rebate through our power utility, we went ahead and replaced all 26 of our recessed lights in the house with those 6" LED retrofit kits from Home Depot (http://www.homedepot.com/p/t/202240932?catalogId=10053&langId=-1) that Brad mentioned. I couldn't be happier with them, they work great. Good color temperature, no warm-up time, and a big one for me already mentioned is that they're not damaged by the power cycles.

[zahc]

I coughed up for a Samsung LED for the recessed fixture above my kitchen sink, but only because I couldn't find any normal, non-halogen, non-expensive incandescents in the required format, and only because the fact that the Samsung sticks down out of the recessed fixture is not obvious due to the location.

Much better than a CFL and gotta love the lack of heat. However, it still makes the colors look wrong. It might be 'good enough' and it might be something that some people can 'get used to', but the colors are still wrong.

[Brad Johnson]

Much better than a CFL and gotta love the lack of heat. However, it still makes the colors look wrong. It might be 'good enough' and it might be something that some people can 'get used to', but the colors are still wrong.

We've had to be very careful about color temps.  There are a ton of "cheap" LED lamps out that have truly abysmal color temps.  I've noticed that a lot of mfgs, and Samsung is one, put out bulbs that are supposedly "warm white" but the color temp is more along the lines of 3100-3300k (incandescent is 2700k).  It's been our experience that units advertised only with a subjective color temp (i.e. Warm White, Cool White, Sunshine, etc. and no actual numerical rating) are probably best left on the shelf.  You pay a little more for the bulbs with a trustworthy color temp rating, but they're worth it.  If you have an A19 style lamp, try the Philips unit.  True 2700k.

I forgot to mention, if you have a can fixture that you need to keep for architectural or decor reasons, Philips also makes the BR30.  It's a straight PAR 30 replacement lamp that's available in a couple of color temps including 2700k.  Best of all they look like a traditional bulb, not the honking huge grey metal surround with a couple of LEDs poking out of the middle like their previous PAR 30 unit.

Brad

[zahc]

It's not color temperature that bothers me. It's the spectrum.

Color temperature is just something your eyes get used to. You can white-balance for color temperature. But if the spectrum is wrong, the colors look wrong, no matter what the color temperature.

[Perd Hapley]

I usually look for the higher color temps, because they seem brighter to me.

[Brad Johnson]

It's not color temperature that bothers me. It's the spectrum.

Color temperature is just something your eyes get used to. You can white-balance for color temperature. But if the spectrum is wrong, the colors look wrong, no matter what the color temperature.

Spectrum conditioning is just as subjective as temperature conditioning.  Both are something you "get used to".

The spectrum for true 2700k LED lamps is similar to a tungsten-filament incandescent.  That's why I mentioned bulbs described as "warm white" rather than with a specific color temp.  Most of the "warm white" lamps (that lacked color temp spec) we tested were in the 3100-3300k temp range.  The only one in our limited test of locally-sourced lamps that metered a true 2700k was the Philips.  In a rather - actually, very - informal test of passersby, not a single one could tell the diff between it and a 60w incandescent.

I usually look for the higher color temps, because they seem brighter to me.

It's the increased contrast.  Kills colors but brings out out detail.  An analogy would be dB vs percieved volume.  A 50 Hz signal and a 5000 Hz signal might meter the exact same dB, but the perceived volume is completely different.

Brad

[zahc]

Spectrum conditioning is just as subjective as temperature conditioning.  Both are something you "get used to".

This is objectively incorrect. The issues of color temperature and spectrum quality are orthogonal. CRI is also irrelevant simply because CRI is such a bad way to measure spectral quality of newfangled light sources like CFLs and LEDs (it was created to be easy to measure, and is adequate for 'well behaved' spectra).

As an extreme example, it is possible to have a light source that is exactly 2700k, with a CRI of 100, and there can still be visible wavelengths of light that are output at ZERO (i.e. something of that color will appear black).

[Brad Johnson]

This is objectively incorrect. The issues of color temperature and spectrum quality are orthogonal. CRI is also irrelevant simply because CRI is such a bad way to measure spectral quality of newfangled light sources like CFLs and LEDs (it was created to be easy to measure, and is adequate for 'well behaved' spectra).

As an extreme example, it is possible to have a light source that is exactly 2700k, with a CRI of 100, and there can still be visible wavelengths of light that are output at ZERO (i.e. something of that color will appear black).

Sorry, no.  Perception of both spectrum and temp are purely subjective.  You like them or you don't.  You get used to them or you don't.  Likes or dislikes for the temp or spectrum are subject to a person's taste and/or perception (that's why it's called subject-ive).

Brad

[Harold Tuttle]

one of my video classrooms went from a florescent blah to HELLO LED LIGHTS ARE REALLY WHITE AND GEE THIS ROOM LOOKS CLEAN.

[HankB]

As an extreme example, it is possible to have a light source that is exactly 2700k, with a CRI of 100, and there can still be visible wavelengths of light that are output at ZERO (i.e. something of that color will appear black).

I'd like to know what the source and object are.

CRI is based on the appearance of several different standardized color test plates, so any CRI 100 light source will have to render these accurately, and they'll pretty well span the visible color space. Phosphor and incadescent systems have very broad emission spectra (no "holes" in the output) and even an R-G-B LED based system will have overlap in the three colors, so I wouldn't expect anything to look "black" with these illuminators. I suppose you could have an R-G-B laser system adjusted to provide good CRI by adjusting the power and wavelength of the three primaries - that way you'd have output of zero at other wavelengths - but it would take a really unusual object to look "black" under this illumination; the only thing I can think of off the top of my head would be a narrow band dichroic mirror that didn't match any of the lasers. Common objects would have a fairly broad reflection spectra, so they wouldn't look black at all . . .

[Brad Johnson]

I'd like to know what the source and object are.

CRI is based on the appearance of several different standardized color test plates, so any CRI 100 light source will have to render these accurately, and they'll pretty well span the visible color space. Phosphor and incadescent systems have very broad emission spectra (no "holes" in the output) and even an R-G-B LED based system will have overlap in the three colors, so I wouldn't expect anything to look "black" with these illuminators.

+1

Most of current-gen quality LED lamps have stopped trying to mix RGB diodes and gone to a phosphor based system, some native, some remote.  The most graphic example of remote phosphor tech is the Philips "yellow flower" design (the one that has the nauseatingly yellow panels in a grey metal body).  The yellow is the phosphor.  The LEDs are blue.  Actually, they are OMGCANYOUBELIEVEHOWFRIKKINBRIGHTTHESEDAMNTHINGSARE!!!!!, screaming, piercing blue.  Excited by the blue LEDs, the phospher glows at 2700k with a spectrum roughly equivalent to tungsten-filament incandescents.

Brad

[K Frame]

I thought any form of lighting OTHER than incandescent bulbs were communist plots and an attempt by the US government to turn real men into girly men...

I've been looking at LED bulbs for several applications at my house, and so far, to this point, I'm simply not sold on them...

What I'm finding is that, compared to CFLs with comparable light output, the wattage usage isn't that much different, and the prices are still high enough that any savings would come at the 8 to 10 year mark.

I'm going to be renovating my basement sometime soon (hopefully) and I very likely will put in LED recessed cans at that time, but I'm not rushing out right now to purchase a bunch of LED screw base bulbs as I'm still not seeing the benefits at this time.

[Brad Johnson]

Yeah, energy comparisons are pretty much a wash between LED and CFL given similar form factors.  The big difference is in robustness, service life, and perceived illumination quality.  It also depends a lot on location.  Locations with cold temp or constant power cycling issues are the perfect place for an LED replacement.  So are places where you'd like better light quality than CFL affords, especially if you want dimmable illumination sources.  Places that are difficult to access (read: extremely high ceilings with recessed cans) are also a good location for LED.

Brad

[zahc]

Excited by the blue LEDs, the phospher glows at 2700k with a spectrum roughly equivalent to tungsten-filament incandescents.

Key phrase: "roughly equivalent". Some people have low standards and think they are acceptable. You can still see things under LED and CFL lighting, just how you can still hear things played through a horn-loaded piezo speaker. You can also say "horn-loaded piezo speakers have a roughly flat frequency response"--which is true for very low standards of "roughly flat".

CRI is based on the appearance of several different standardized color test plates

You make my point for me. That's all CRI does. When comparing black-body spectra, that is sufficient. CRI is completely inadequate for anything else.

Don't fall victim to the "RGB fallacy". All colors cannot be expressed as combinations of RGB, or as combinations of the standard CRI color plates. When computer monitor makers advertize "2.7 million colors" they are essentially lying. Computer monitors can output 3 color peaks and nothing else. It is technically impossible for a computer monitor to display violet, for example. It cannot be done.

Unless a light source emulates a black body very closely, the color of everyday objects will look 'wrong'.

[Brad Johnson]

Key phrase: "roughly equivalent". Some people have low standards and think they are acceptable. You can still see things under LED and CFL lighting, just how you can still hear things played through a horn-loaded piezo speaker. You can also say "horn-loaded piezo speakers have a roughly flat frequency response"--which is true for very low standards of "roughly flat".

Can't say we didn't try to be nice about explaining it to you...

Image from this page - http://web.ncf.ca/jim/misc/cfl/

Center three spectra are CFL.  Far left is incandescent.  Far right side is LED.





Is that "roughly equivalent enough for you?   

(And keep in mind the linked page, and image, represent three+ year old lamp technology.  LED-based illumination has improved several orders of magnitude since.)

Brad

[zahc]

Thanks for the graphs. However, looking at plots (which may or may not be accurate anyway, especially for my lamp) does not prove anything. Plots are useful for predicting what the light will actually look like. I already know what my light looks like, because I actually use it every day. The map is not the territory.

Note that I ACTUALLY HAVE a Samsung LED light, which I bought because the Samsung was supposed to be the best on the market. I installed it in my kitchen, and it makes the color of my towels and dishes look wrong. Not "cheap flourescent" wrong, but not quite right either.

In summary:
--The plots support the conclusion that that modern LEDs have a much better spectrum than CFLs, which we all already knew.
--My Samsung LED light spectrum is better than any CFL, but still not as good as an incandescent
--Just because something has a high CRI and correct color temperature, does not mean that it will accurately render color. The spectrum still matters. CRI is only a crude proxy for color rendering accuracy
--Color temperature is almost irrelevant to color accuracy because color temperature is something you get used to (or can be white-balanced using normal white-balancing techniques). With a good spectrum, differently-colored items reflect color in proportions that are expected by the human brain.

[Brad Johnson]

Thanks for the graphs. However, looking at plots (which may or may not be accurate anyway, especially for my lamp) does not prove anything. Plots are useful for predicting what the light will actually look like. I already know what my light looks like, because I actually use it every day. The map is not the territory.

Note that I ACTUALLY HAVE a Samsung LED light, which I bought because the Samsung was supposed to be the best on the market. I installed it in my kitchen, and it makes the color of my towels and dishes look wrong. Not "cheap flourescent" wrong, but not quite right either.

In summary:
--The plots support the conclusion that that modern LEDs have a much better spectrum than CFLs, which we all already knew.
--My Samsung LED light spectrum is better than any CFL, but still not as good as an incandescent
--Just because something has a high CRI and correct color temperature, does not mean that it will accurately render color. The spectrum still matters. CRI is only a crude proxy for color rendering accuracy
--Color temperature is almost irrelevant to color accuracy because color temperature is something you get used to (or can be white-balanced using normal white-balancing techniques). With a good spectrum, differently-colored items reflect color in proportions that are expected by the human brain.

You have a bulb.  One.  And you are prepared to make a blanket statement of derision based on that single example and a couple of talking points from the internet?

You want experience?  Okay...

-We have hundreds (very soon to be thousands) of various LED-based lamps in service.
-We have several years of hard test data to confirm/deny mfgs claims.
-We have hundreds of field impressions from actual users that span dizzyingly diverse backgrounds, professions, experiences, cultures, yada, yada....

In summary:
Our experience and field test results trumps yours, and by a significant margin.  In addition our experience and field test results show LED lamps can mimic incandescent illumination to a such degree that differences are effectively undiscernable to the casual observer, including intensity and color rendering.

You don't care for the LED lamp you have.  Point conceded.  Your satisfaction with it is your satisfaction.  Okay.  To each his own.  But a general castigation of LED lamps based on your single subjective interpretation is, well... it's pretty unscientific.

Brad

[zahc]

Please quote my "blanket statement of derision" and "general castigation of LED lamps" so that we are both on the same page.

I made a specific statement about my Samsung LED. YOU then began making general statements about spectra and color perception. I stand by my rebuttals to these, such as they are.

I get it...you think that LEDs are "good enough" for the "casual observer". That's great. Allow me to quote myself:

I coughed up for a Samsung LED ... Much better than a CFL and gotta love the lack of heat. However, it still makes the colors look wrong.

[AJ Dual]

Can't say we didn't try to be nice about explaining it to you...

Image from this page - http://web.ncf.ca/jim/misc/cfl/

Center three spectra are CFL.  Far left is incandescent.  Far right side is LED.





Is that "roughly equivalent enough for you?   

(And keep in mind the linked page, and image, represent three+ year old lamp technology.  LED-based illumination has improved several orders of magnitude since.)

Brad

Considering that CFL's just use the UV light from the Mercury/Argon arc to excite phosphors and try and produce a mix of "white" light, and that "white" LED's do the same, but with a high intensity blue gallium nitride die that excites a blended yellow phosphor, I'm wondering why CFL's just can't use a better phosphor mix that isn't "gappy" in the spectrum?

I've noticed that by illuminating some of my larger white LED flashlight's dies with a UV light source, the phosphors glow pretty yellow/white.

I'm assuming there's some technical limitation in the way CFL's work and the UV light that they can't use the same phosphor blends "white" LED's do... or they'd already be doing it. 

[Brad Johnson]

I'm assuming there's some technical limitation in the way CFL's work and the UV light that they can't use the same phosphor blends "white" LED's do... or they'd already be doing it. 

I'm glad someone else was wondering the same thing.  Fluorescent tubes are, at their most basic, a big UV lamp.  Visible color comes from the phosphors used to coat the tubes.  Must have something to do with the way the phospher can/can't handle the electron flow in the tub.  Or maybe they don't use enough bacon.

Brad

[K Frame]

I must be one of the few people who doesn't really find the light emitted by today's CFL's to be objectionable. Until recently I even was using one as a reading light on my bedside table and was very happy with it.

I no longer use it as I recent got one of the Philips sunrise alarm clocks (halogen bulb), and it doubles as a reading lamp.

[HankB]

OK, here's a graphic illustrating the relative emission spectra of a number of typical sources.

* The bar code scanning laser is red and essentially monochromatic - it's not an illuminator.
* The mercury vapor lamp has a number of peaks due to the emission spectra, and I think most everyone will agree that while it can be pretty bright, color rendering is terrible. You don't really see them used much other than as street lights these days.
* Noon sunlight is THE standard for accurate color rendition - we evolved to see it. The sun is a pretty good blackbody radiator, but thanks to things like atmospheric absorption and Rayleigh scattering (which makes the sky look blue) daylight has a color temperature of around 6500K - and in fact, this is what's referred to as the "D65" standard illuminant.
* The incadescent lamp with its ~2700K emission is essentially a modified black body, and once the eyes are color adapted, black bodies above a certain temperature threshold produce good color rendition. Cascade photopic visual response curves with either the D65 or incadescent emissions, and you get good color rendition.
* The white LED emission spectrum actually matches sunlight better than incadescent, except for the blue spike, which is typically around 450nm or so. This is leakage of the pump light, since current white LEDs use a blue LED to pump phosphors. This leakage isn't as significant as the graph suggests since relative visual response is 'way down by the time you get to 450nm.

There's a lot of work going on with phosphor development to optimize emission spectrum, blue absorption, and phosphor longevity, as well as dichroic mirrors to boost emission efficiency and reduce blue leakage.

[Brad Johnson]

Too bad they didn't include aging metal halide lamps in that graphic.  They can turn all kinds of funky colors.

What's really cool is how well they're able to use on-diode phosphors now.  For a while it looked like remote phosphors were the only way to get the proper incandescent mimic.  Now it's no problem even with surface phosphors.

Brad