RGB luxeon + T5 fluorescent ceiling lamp

[Wim Hertog]

This summer I decided I needed a new light for my bedroom. It needed to be something unique, something nobody else had

I found LEDs were finally ready to be used as general area lighting. Of coarse, I wasn't satisfied with the color rendition of normal white LEDs, so why not RGB?

Last year I did extensive research on the color rendition of white light mixed with RGB LEDs. I concluded that despite the very low CRI value of such mixed white light, the color rendition isn't that bad at all. In fact I was pleasantly surprised.

The color distortion you get when mixing RGB LEDs makes all color more saturated and vibrant which I find a benefit. White LEDs and most fluorescent lighting makes some colors look too pale which is in my opinion much worse than too saturated.

Another benefit of the RGB approach is selectable color temperatures and even rainbow fades etc.

Part 1: selecting a good luminaire.

I decided to use an aluminium ceiling lamp designed for 1 x 54W T5 HO lamp.I only needed the "enclosure" so the ballast was removed. The lamp had to be made out of aluminium for optimum heat transfer and spreading.

Part 2: The LEDs

As I had a cheap source for very high binned luxeons at that time, I decided to use 24 luxeon 1 LEDs: 8 red, 8 green and 8 royal blue. Why not lux III's? Lux I LEDs have an isolated base and are way more efficient than the higher power parts. They were quite a bit cheaper to...

All the LEDs were mounted on the 2 profiled using heat conductive epoxy and the alu profiles were mounted on both sides of the luminaire.

Part 3: Electronics

Last summer I e-mailed back and forth with the inventor and creator of the Chromoflex RGB controller. It's a cheap and very flexible controller with 3 PWM output channels and 4 input channels to select 8 programs. These programs can be made on your PC or laptop and are than downloaded in the controller.

As the controller outputs a PWM signal with constant voltage I had to design a current source for the LEDs. A simple current source with 6 LDO regulators does the job perfectly.

Part 4: Mooorrre light!

24 luxeons should be enough to light my room to reasonable levels, but when I needed more light, an extra fluorescent lamp would be nice. Besides, the lamp is designed for 1 54W T5 lamp, so why not use it...

I bought a dimmable (0-10V, 3-100%) ballast from Philips Lighting and a very nice 54W T5 color code 950 (5200K, CRI of 92).

Part 5: Results

ceiling view



Controller



power supply and current source



LEDs with plexi diffuser (90% translucent)



T5 fluorescent at 100% power



Wall mounted mode selector




I programmed 6 different color temperatures (visually compared to a 2700K incandescent, 3200K fl, 4000K fl, 5200K fl, 6500K fl and 8000K fl) and 2 rainbow fades. The color rendition of both the 5200K T5 and LEDs is excellent. The power consumption of the LEDs is about 15W at 5200K (more or less the same for other CCT's) and the efficiency is 35lm/W. This equates to a little less than 500lm, diffuser losses included...not bad

White balance for all shots below was set to "Sun"

5200K T5 950



2700K LED



3200K LED



4000K LED



5200K LED



6500K LED



8000K LED



I'm extremely happy with this light and after 6 months of "testing" (+4hrs/day) there's absolutely no color shift...great!

 

[DNova]

That is freaking awesome! I love it. What was the total cost? What would you do differently if you were doing it again?

 

[winny]

Good job man! :rock:

 

[2xTrinity]

Discerning color temp from a photograph is very difficult to do -- simply because the camera can't compensate for different lighting conditoins the way that the eyes can. As a result, the 5000K images seem to look the best -- that is because that is closest to the sunlight setting on the camera, so that color is set to be "white"

Either way though, the LED color rendering doesn't look to bad at all. It looks slightly weaker in green than the fluorescent, but other than that, much better than I would have expected. Hopefully a phosphor-based LED lamp will come out with more like a 4000k color temp instead of the 6500K plus we've seen so far -- that would be a massive improvement, and may even result in higher efficiency as the eye is more sensitive to the yellow/greens than the deep blue that is thrown by the standard LEDs.

I found LEDs were finally ready to be used as general area lighting. Of coarse, I wasn't satisfied with the color rendition of normal white LEDs, so why not RGB?

Last year I did extensive research on the color rendition of white light mixed with RGB LEDs. I concluded that despite the very low CRI value of such mixed white light, the color rendition isn't that bad at all. In fact I was pleasantly surprised.

The color distortion you get when mixing RGB LEDs makes all color more saturated and vibrant which I find a benefit. White LEDs and most fluorescent lighting makes some colors look too pale which is in my opinion much worse than too saturated.

This is a big drawback of the CRI scale -- it only tells how much a light source differs from the blackbody curve, not in what way it differs (making things appear more vibrant, or more dull). Certainly this lamp has infinitely better color rendering than typical high-pressure sodium lamps, which have simialr CRI value to a lot of RGB LEDs.

 

[rgbphil]

Hi,

Nice job. Can you describe your colour temp calibration process in a little more detail please? You've done a great job in setting the colour temps. The 5200K LED compares nicely with the T5 picture. I'm guessing you went a little cross eye'd fiddling with different mixes! Just out of interest, what is your favourite reading/working colour temp and your favourite general usage setting?

If doing it again, I'd consider putting amber LEDs in with the RGB and clustering the RAGB LEDs together....it does help make things a little nicer on the colour rendering issue. Only problem is amber Leds are generally dimmer.....but you don't need too much more.

Another alternative is using cool white LEDs, and mixing in a bit of green or cyan, maybe a touch of red as well. Basically you're trying to fill in the gaps in the spectrum. White LEDs have a lot of yellow and a sharp peak in the blue.

One last enhancement would be to use some holographic film diffusers, with a assymetrical beam pattern (along the long axis). This would mush up the colours nicely and are still pretty transmissive. I see some colour shadows in the pics there which I suspect is annoying in use. Have a look at www.poc.com for details.

Phil

 

[AndyTiedye]

That looks great! We need more color around here!

Adding to LED Fixed Lighting Index.

 

[2xTrinity]

Another alternative is using cool white LEDs, and mixing in a bit of green or cyan, maybe a touch of red as well. Basically you're trying to fill in the gaps in the spectrum. White LEDs have a lot of yellow and a sharp peak in the blue.

The best LED I've seen so far is the WH bin of the Cree -- those look a bit "greenish" on a white wall, but in actual practice, render color vastly beter than any other LEDs I've seen. I did a ceilng bounce with one of those at high power (~1A), and a single 5mm red LED on my keychain was enough to make the color balance almost perfect IMO. I think mixing an efficient cool-white such as the Cree or Seoul with red (most important) and possibly cyan (blue/green) would make this potentially very close to natural light.

 

[rgbphil]

For more on the idea of mixing white+colours LEDs check out the October 2006 issue of Ledsmagazine:
http://www.ledsmagazine.com/features/3/10/3
They actually recommend using warm white+green+blue LEDs....but the principle is the same.

 

[greenlight]

hope you have some donna summer handy!!

 

[Wim Hertog]

Hi,

Nice job. Can you describe your colour temp calibration process in a little more detail please? You've done a great job in setting the colour temps. The 5200K LED compares nicely with the T5 picture. I'm guessing you went a little cross eye'd fiddling with different mixes! Just out of interest, what is your favourite reading/working colour temp and your favourite general usage setting?Phil

Well, I took a laptop, the RGB lamp, an incandescent lamp and a few CFL's with know color temperature (CCT measured in lab) in one room and pointed the RGB lamp and the reference lamp to a white wall. I also put some cardboard between the 2 lamps to make sure no light from the 2 sources could mix.

Then I connected the chromoflex controller to my laptop (serial interfac) to update the RGB values in real time (8 bit/channel). Switch on the reference light and adjust the RGB values to match the reference source as close as possible.

As the temperature of the fixture never exceeds 35°C, there's no color shifts related to large temperature differences and until now I experienced no color shift related to LED aging (after +/- 1000hrs).

During daytime I normally use the 5200K setting. At night a use the 3200K setting a lot and in the morning I tend to switch it on to 6500K (wakes me up). When working I always use the 5200K setting and when I need more light I switch on the fluorescent (at 60-70% of max power).

I really like light with CCT between 5000 and 6000K, so that's what I use most of the time.

Maybe I'll use the higher CCT's more in the summer? We'll see in a few months...

If doing it again, I'd consider putting amber LEDs in with the RGB and clustering the RAGB LEDs together....it does help make things a little nicer on the colour rendering issue. Only problem is amber Leds are generally dimmer.....but you don't need too much more.

Another alternative is using cool white LEDs, and mixing in a bit of green or cyan, maybe a touch of red as well. Basically you're trying to fill in the gaps in the spectrum. White LEDs have a lot of yellow and a sharp peak in the blue.

One last enhancement would be to use some holographic film diffusers, with a assymetrical beam pattern (along the long axis). This would mush up the colours nicely and are still pretty transmissive. I see some colour shadows in the pics there which I suspect is annoying in use. Have a look at www.poc.com for details.

Phil

yeah, using RGBA clusters was tempting, but too difficult. You'd need a 4 channel driver+there's always more than 1 way of reaching your desired CCT....which would make the calibration of the LED array a nightmare....

Is there a big difference between your RGB arrays vs RGBA arrays regarding color rendering?

Thanks for the link!

 

[Wim Hertog]

That is freaking awesome! I love it. What was the total cost? What would you do differently if you were doing it again?

Cost breakdown (in Euro):

- Fluo fixture: 50
- LEDs: 110
- power supply 24V 4A: 35 (ebay)
- Dimmable ballast: 20 (ebay)
- controller: 35
- materials+odds and ends: 50

Total: +/- 300 euro

If I I was doing this again I would probably use SSC RGB LEDs instead of luxeons (less color fringing and mixing problems), hide all the wires a littlle bit better and mount the diffusers in a different way...but I'm quite happy with it as it is now

 

[jtice]

WOW oo:
That is a REALLY nice setup you have there !!!!!
Looking at it from below looks very professional

Really nice how you can the different color temps.

~John

 

[tenfour]

Wow - what a cool project! Great description and pictures, too. I think you should go for the SSp4 version now =)

 

[2xTrinity]

For more on the idea of mixing white+colours LEDs check out the October 2006 issue of Ledsmagazine:
http://www.ledsmagazine.com/features/3/10/3
They actually recommend using warm white+green+blue LEDs....but the principle is the same.

Odd. With the warm white LED, you're generating blue, then converting almost of it to amber/green/red. This tends to be less efficient than cool white due to phosphor inefficiency. To then add back in the blue that you removed by using warm white seems like it woudl make things worse. Warm white LEDs also still seem to be deficient in red, they just have stronger amber/green.

Cool white would be more efficient as that gives off a lot of the blue (rather than using a separate blue emitter), and dependong on the tint of LED you get, a lot of green. The only real deficiency is in the red, and filling in red spectrum with extremely efficient red LEDs seems like it woudl be better off than putting more power into a less-efficient warm white LED. Takign things a step further with the Cool LED setup, it's possible to add cyan LEDs which are also very efficient, this will boost the apparent amount of "green" and fill in a "gap" in the spectral output.

 

[rgbphil]

Thanks Wim for the calibration description. Interesting you prefer warmer/yellower lights for non-work use at night....same here. Must be something to do with caveman instincts.

I've definately found RAGB LED clusters nicer than RGB, although the RGB makes colours 'jump out' at you as you've no doubt found, sometimes it's a little too much. The RAGB colour rendering feel nicer. Haven't experimented with Cool White/G/R or Warm White/G/B yet, just noted the literature on it. I wonder if anyone here has done White LED mixing?

I've been working on building a spectrometer using an old LIDE scanner (noted in another thread) and finally bought some diffraction gratings from Surplus Shed to make it a reality. The idea is that the device will hopefully read the spectrum of input light and do automatic calculations to determine CRI etc.

In looking up that stuff, I found some interesting techniques that the astronomy people use that anyone with a camera, a diffraction grating (or prism) and a bit of time with photoshop or paint shop pro can do.

Take a picture of your RGB setup at a far enough distance so each LED is effectively a point source through your dispersing element. The colours will then spread out (astronomers do this with stars). Import the pic into your photo program and isolate the individual smears for each of the colours. Use the brightness analyser on the photo program to then get a profile across a line. This will be the spectrum of the LED.
You should then be able to convert to Excel and do some sort of mathematical trickery integrating with the published algorithm to calculate true CRI and CIE coordinates....a lot of work but possible without special equipment.

Phil