03-17-2012 L-Prize Bulb partial teardown (UPDATE May 06, 2012- Emissions Testing, Melatonin, SAD (Seasonal Affective Disorder)):

With time, technology improves, and it was about time to do a long awaited overdue update.

In 2007 CREE rocked the status quo with the LR6 downlight, it was years ahead of its time. It produced 650 lumens, consumed only 10.5 Watts, and measured out at 62 lumens per Watt, with a Color Rendering Index of 90%, at 2700K, dimmable to 20%, and had a 50,000 hour lifetime. Internally it employed Red LED assist, was probably the first mainstream lighting market to use this approach for increasing the CRI of the light source.

Shortly thereafter, CREE released the LR6-DR1000, producing no less than 1,000 lumens @ 80 lumens per watt, consuming just 12.5 Watts, a CRI of 90%, a color temperature of 2700K, and dimmable to 20%

A couple of years ago CREE introduced the LR24-HE, producing 100 lm/W, then later came the CR24 40L HE @ 110 lm/W (4,000 lumens). But neither of these higher efficiency models produced the golden incandescent look of their earlier product, the 2700K color temperature.

Lets turn the page and see what the winds of change have brought.

After the massive European company (European- Royal Philips) acquired the US home-grown LumiLEDs from HP/Agilent, and took full control, acquiring many companies in a buying frenzy by the behemoth European Royal Philips company. Some years later, the US Government offered up the L-Prize competition in an effort to accelerate lighting technologies.

Unlike the many cheap Asian bulbs on the market, Royal Philips Electronics gainfully employed much of the technology they purchased, employing their engineers to actually design an LED based light in the A19 form factor that would last, and entered it into the astute L-Prize competition. The carrot held before the horse was a 10,000,000.00 dollar (US) prize for the first viable A19 that met the requirements and testing of the L-Prize program.

The L-Prize competition sets a minimum standard to which a bulb must comply, and many of the bulbs parameters are tested and monitored over time. Surprisingly, they included many different but good quality, CFLs in the same test fixtures for a basis/reference, and the majority of the Fluorescent CFLs were dead before the minimum time/testing was over with. This is not surprising, CFLs rarely meet their claims or Warranties (that is another story especially using the Warranty, if you can find/contact them). None of the L-Prize bulbs failed.

More on the L-Prize testing and results can be found here (look carefully and you can see the burnt out CFLs on their page):

Source: http://www.lightingprize.org/philips-winner_photo-gallery.stm

The packaging lists this astute bulb at 940 Lumens @ 10W, which gives this bulb 94lm/W, including a CRI of 92% at the same time, raising the bar for everyone. Right now the price is a bit steep, ~49.97ea, I'm sure the European Royal Philips company is very proud of their accomplishment-and to boot, it is Assembled in the USA, as per the L-Prize requirement.

Unlike the older Philips bulb on Home Depot's shelves (grey body), the L-Prize bulb is not Made in China, but it is assembled in the USA. This was required by the 10 Million dollar L-Prize competition.

Bulb Markings:

The first glimpse of the Royal Philips LumiLEDs Luxeon Rebel LEDs are shown here, with the Lemon Meringue Pie colored remote phosphor removed:

Notice above, there are 6 LEDs in each section, 3 with the Red die that emit Red (much like CREE's LR6 approach), and three silver looking Rebels, two of them emit blue, and one emits another blue (it's appearance is different), maybe a combination of 470nm blue and 450nm blue (it looks like they may all be 450nm blue (aka Royal Blue, from the spectral plots below, many YAG phosphors perform better with 450nm blue). It is interesting to make note that Royal Philips licensed a number of CREE patents some time ago.

Here are the two bulbs, one sporting "Made in China", the other "Assemble in USA"

One may realize by now, with one section of six LEDs, and 3 sections, there are 18 3 Watt LEDs (54 Watts), yet the light only consumes 10 Watts. When you under-drive an LED especially a Power LED (excluding most OSRAM Dragons), it is common knowledge that you can up to double the LED's efficiency (lm/W). With the LEDs having a capacity of 54 Watts, they are being under driven 5.4:1, roughly just 1/2 of a Watt, which definitely makes them more efficient. Likely they are under driven further, as the power electronics in the base of the bulb also consumes power not accounted for here.

Some time ago, back in about 2006, when 70 lm/W was the best for production White LEDs running at 350mA, I ran some tests to demonstrate the increased efficiency in lumen per watt, when the power LEDs are underdriven, and for the most part the scaling holds true for the current volume production 160 lm/W White LEDs. In this example the device produced 74 lm/W @ 350 mA, but when underdriven at 20mA it produced 108 lm/W. So under driving increased the efficiency on this old LED by 30%.

The above is from another page on this site: CREE 2006 lm/W over current

If one were to not have an "infinite heatsink" the gain would be even more for the old CREE above, and as well for the new Rebel below.

An example using a new generation, the Royal Philips LumiLED Rebel a year later in 2007, when underdriven it increases it's efficiency from 88 lm/W to 108 lm/W, an 18% increase in efficiency:

More details of the newer Rebel here:New Rebel LED lm/W over current

Here is an image of the bulb lit, but with the remote phosphor section still removed, showing the Royal Philips LumiLEDs Luxeon Rebels at work:

Here I have stopped down the camera's iris to better show the 3W LED die barely breaking a sweat @ 1/2 Watt each:

This is the ceramic based "PCB", a close-up of the top side:

Here is the backside of the ceramic board:

Here is a side view of the ceramic board, yes, the flex came mangled like that, but it was folded up like an accordion when it was installed (only one section was mangled like this):

A view from the top, showing the grey thermal pad that sits between the ceramic board and the metal bulb housing. Yes, that is residue that was left on the board at the factory where they assemble the pieces in the USA:

Here I installed 4 of the Royal Philips 75W LED bulbs (1100 lm @ 75W, 64lm/W). Then I installed one in the back right, it is the 10W Royal Philips L-Prize bulb:

Below is the top of the bulb with the cover and all the ceramic LED boards removed, you can see the boards are wired in series, for 9 Blue LEDs in series and 9 Red LEDs in series:

Here is the top of the top board, showing the series hook-up for the 3 sets of 3 Blue (9 total) Royal Philips LumiLEDs Luxeon Rebel LEDs:

Next is the top of the top board, showing the series hook-up for the 3 sets of 3 Red (9 total) Royal Philips LumiLEDs Luxeon Rebel LED:

Lets move on to the insides, where the input electricity is converted into a form more proper for driving LEDs:

This is at the input of the board, showing they even utilized a fusible component (the reddish-orange device), rated for 1 Amp and 250 Volts:

For the bulk storage, Royal Philips even chose a RX30 series 22uF 250V Rubycon -25 to +130 degree C rated Electrolytic capacitor (2,000 or 4,000 hour lifetime) This capacitor is probably around 50C and based on the 2,000 hour life at 130 degrees C, using proper de-rating, it should last for 128,000 hours. However, it is also not running near it's full ripple current rating, so the lifetime may be even longer. This definitely is not the cheap junk capacior you find failing in CFL bulbs, computer motherboards, or computer power supplies. It was good to see their very high quality choice.:


This is part of the input stage, showing the STMicroelectronics internal Zener protected N-Ch MOSFETs (there are one pair), STD4NK60Z-1, 600 V, 1.76 Ohm, 4 A SuperMESH(TM) Power MOSFET in IPAK :

Backside of the board, showing the isolated topology design for Cypress CY8CLEDAC03L (03 has enhanced dimmer support, reduced BOM, enhanced safety features, higher efficiency- look on Google, Cypress has removed most references to this part, and see AN70005) (L is optimized for 80 to 175VAC, and the pulse chopper MOSFET in light chopping mode is 1/4 the normal pulse width) PSoC controller, as is found in the 12W and 17W low-efficiency LED bulbs, as well as a few 1000Vrrm 1A Switching Rectifiers from Gulf Semiconductor (labeled GUM1E), and a variety of zeners, diodes, and other MOSFETs (if one has the time to look them all up):
Click here for a link to CY8CLEDAC03, the part an Ap Notes were removed from Cypress parts search engine, even though the Ap Note was just updated late last year. The part has also been removed from Future Electronics website.
As of November 11, 2011 CY8CLEDAC03L and H, was well as 02 and 01 were transferred to iWatt. World Wide Contact: Sal Sestito, VP of NA / EMEA Sales, ssestito@iwatt.com or sales@iwatt.com.

Another shot of the other side:

Below is a close-up of the topside showing the LED driver side:

>>>---> For these next items, the bulbs were suspended in open air. <---<<<

How does the bulb stack up spectrally? There is a standard test called the Color Rendering Index (CRI), where 100% is the best score. You may have see things like a box of crayons under several different light sources, and you see how well each light source renders the crayon colors. This is done for the CRI test, but with a chip chart that has a calibrated color on each chip, and for the CRI there are just 8 color chips used to calculate CRI (aka Ra). The much improved SCRI (aka ECRI) standard uses 14 color chips currently used do calculate Ri (Special Color Rendering Index), however NIST is working on a new improved version of the test with new and additional colors. CRI and SCRI ignores a number of areas and NIST is attempting to get the industry to CQS, a whole new set of color chips, 15 all told. This is needed, as we have increased access to producing different portions of the light spectrum, and the standard chip chart (8 chip CRI) is becoming very lacking for real world accuracy. Lighting manufacturers can circumvent the CRI and get a high score, by playing "games" with the spectral output. When I get a chance I will add some additional measurments that better show color "faithfulness".

Jim Brodrick at the Department of Energy Office of Energy Efficiency as well as Yoshi Ohno, Wendy Davis, Cameron Miller, Yuqin Zong, Maria Nadal, and others (forgive the omission of others involved) at the US National Institue of Standards and Technology have put many hours into creating and testing under the CQS system, representing the unique SSL requirements/needs, and CQS is far superior to the CRI, SCRI/ECRI standards, and would catch a lot of the "cheating" the lighting industry does with spectrum manipulation, and it much more accurately represents how the human eye sees things.
Go to Color Rendering Index explanation and shortcommings at NIST

Back to the subject at hand, testing the bulbs performance...

At time zero, the L-Prize bulb under test produced a CRI of 90.95%, and as it warmed up for 16 minutes the CRI increased to 92.60%. After 27 minutes the CRI stabilized at 92.88%.
For comparision, the older Royal Philips bulb measured in at 80.21% CRI.
A standard Sylvannia 60W frosted incandecent bulb measured in at 99.49% CRI.
The frosted GE 60W Reveal bulb measured 82.26% CRI.

Color temperature measures the "warmth" of a bulb, where an incandescent is typically 2700K, which would be the color temperature of a black body heated to that temperature in degrees Kelvin.

At time zero, the L-Prize bulb under test produced color temperature of 2795K, and as it warmed up for 16 minutes it increased to 2727K. After 27 minutes the color temperature remained put at 2727K.
For comparision, the older Royal Philips 60W LED bulb measured in at 2623K.
A standard Sylvannia 60W frosted incandecent bulb measured in at 2765K.
The frosted GE 60W Reveal bulb measured 2598K.

Next we have the light output measured in Lumens, which is a measure the total amount of light emitting from a bulb.

At time zero, the L-Prize bulb under test produced 1123 lumens, and as it warmed up for 16 minutes the output dropped to 1089 lumens. After 27 minutes the CRI stabilized at 1072 lumens.
For comparision, the older Royal Philips 60W LED bulb settled in at 952.8 lumens.
A standard Sylvannia 60W frosted incandecent bulb measured 972.7 lumens.
The frosted GE 60W Reveal bulb measured 519 lumens.

However, numbers, especially on spectrum, or even the CRI do not give you the big picture. Normally, for light, plots are based on optical power, and not the human eye response, and I will keep that standard convention.

Here I've plotted the old Royal Philips 60W LED bulb (grey body) spectrum, against the new L-Prize "60W equivalent" bulb (white body):

Here I have plotted the 60W L-Prize bulb against the 60W Sylvannia frosted incandescent bulb, and the 60W GE Reveal incandescent bulb.

Here I have plotted all the mentioned light sources together:

Here is what an extended chip chart for SCRE aka ECRI(Extended CRI) looks like (your monitor/LCD will not render these faithfully):

In this graph below, I have plotted each bulb's measurement for each chip chart color, and for example grouped all five measurements under Color Rendering Index R1, and in the background put that chips color behind that group. I have repeated this for the rest of the color chips. Unfortunately I can't do the new much more accurate CQS color tests.

I was asked about CFLs in comparision to the L-Prize bulb.

Here I have the #1 CFL bulb which were rated as #1 in Consumer Reports Oct 2010, p.28 and carry a 9 year warranty- my reason for purchasing them. I have had many of them fail, and it a real long time to track down the real manufacturer for the warranty, who is TCPI (aka EcoSmart/N:Visioncfl/Home Depot- and they make 40 other brands of CFL).
Purchased at Home Depot
SKU printed on bulb: 423-599
Model: EDXO-14
E149698 Self Ballasted Lamp V#42836 (purchased at Home Depot)
Misc. Number on bulb: 0949305
Made in China

60W equivalent consumes 14W, and states 900 lumens and 10,000 hour 9 year life warranty.
Color Temperature 2700K

Yes I know they aren't a "big" brand name but they beat out everyone in the Consumer Reports testing. I switched to this brand after having no luck getting GE, Sylvannia, Westinghouse, and others lucky to get to one year life. These did better, but I've had over 10 of these fail.

I will use these as an example, as they are #1 in CFL now, even though I have had 10 of them fail, but once you get a hold of TCPI, they will replace them under warranty.
It wasn't long before the instant start ability wore out, that was gone the first year- even though they were installed in very open or completely open fixtures and open table lamps. I don't turn them off and on much, but if you do, you can grossly shorten their lifetime.

Well, back to the point. Here is their full spectral plot which is very representative of CFL, notice how they have very "spikey" output, and a lot of holes with no light output in their spectrum:

The color rendering is very poor, at 72% CRI, yet they did test at 2765K color temperature and after a 30 minute warm-up, they produced 1040 Lumens. When looked at for R9 color chip (red), they actually got 0! There are other areas that also lack a lot, especially "light olive" R3, yellow R10, and blue R12.
Remember, this is a top rated award winning CFL

Some people will say the sun is all "spikey" also, however when I measured it in 2006, it only had a few dips but was somewhat continuous spectrum. I doubt it has changed much since then.
I would beg to differ, and imho, I'd be more worried about the CFL's concentrated output power of 440nm and 420nm near UV and 370nm UV and it's affect on humans and the known damaging energy contained in those wavelengths. The LED bulbs look extremely safe after seeing the CFL spectrum. However, I'm not a doctor. Metal Halide, Sodium Vapor, HID bulbs, and many other "arc" types are much worse than this fluorescent bulb.

Getting back on topic, here is the plot of the sunlight:

Here is a shot of the current drawn by the bulb, with the oscilloscope set for normal sampling on the vertical channel. There are some interesting spikes showing up, so on the next plot I will switch to peak detect on the vertical channel to see if there is anything there that we are not catching:

Here is a shot of the current drawn by the bulb, but with the oscilloscope set for peak sampling on the vertical channel, now we are seeing all of them.:

Here is another shot zoomed in on the current drawn by the bulb, with the oscilloscope set for peak sampling on the vertical channel:

Zooming in yet even further, to find the repetition rate, or time, and frequency of the pulsespeak sampling on the vertical channel:

Zooming in *much* further on the individual spike/pulse itself, and we find it is actually a ring, the vertical channel is still on peak sampling, but we sample at a much faster rate, 250 femtoseconds per horizontal sample to get a more faithful result:

I haven't had a chance to do emission scans myself, but one gentileman, Thomas (T.J.) Hermann (amateur radio call sign W5TH) forwarded his EMC results. This type of testing takes a lot of very expensive equipment, a lot of skill, and is usually only found at large companies- Thank You very much T.J.! You will see the baseline, which is with the bulb off, and then you will see the plot with the higher amplitude, which is with the L-Prize bulb on. There are two frequency ranges and two antenna orientations. The emissions are not too bad, they could have been lower with a few low cost parts added. The Pass/Fail criteria is different in different areas of the country, but these are lower overall than most CFL bulbs I've looked at. The ring (which will vary a bit in frequency from bulb to bulb), is clearly showing up in the 240 to 260 MHz range. In fact it is the peak emissions area, it is too bad they didn't spend a few more cents to knock this way down, but since they pass the required FCC test levels, no one can can fault them.

There is a fellow out that is making a lot of noise about LED bulbs and a concentrated single wavelength blue emitted by LED bulbs and human health. I'm sorry I just don't buy the hype. Blue runs from about 490 nanometers (nm) to 415 nm, and I take one look at CFL and linear fluorescent bulbs, and I see some narrow very high power spikes in that region, as well as spikes shorter (lower wavelength) in the region from 400nm to 330nm, also known as UV. If anything, if he is worried about the narrow wavelength spikes, I feel he should be trying to get CFL and Fluorescent bulbs banned from the world. I do know that 450 to 470 nm blue LEDs are used extensively in Canada for the SAD syndrome, they light up the room or wear a pair of glasses to shine blue LEDs on the eye receptors during the day to suppress melatonin production, which also helps with the sleep cycle come night time. Several studies show that smooth spectrums like what LED bulbs make, are less effective at suppressing melatonin production, and if the light is white, it is also less effective than narrow spikes in the blue region are. Also, 2700K LED lightbulbs (like these), have much less blue in their smooth continous light output spectrum.

Here is that CFL spectrum again (shown in Red), you can see how very smooth the LED spectrum (L-Prize -in Blue, and the Old Philips LED bulb- in Orange) is in comparison, CFLs and Fluorescents pretty much all have this sort of very high output jagged narrow spike spectrum:

Also note the wavelengths above, a single narrow spike wavelength in blue is more effective, if it is at 470nm it is the most effective source for affecting melatonin. One will notice the L-Prize bulb (blue) has both a smooth spectrum, a continous spectrum, and even no spikes at 470nm.

In 2011, Mariana G. Figueiro, Basar Erdener, Asiri Jayawardena, Natalia Lesniak, Rinara Reh, Levent Sahini, Kate Sweater, Brittany Wood, Lisa Yue, at Lighting Research Center, Rensselaer Polytechnic Institute 21 Union Street, Troy, NY 12180, recenty took a batch of students and monitored their melatonin production through samples of the student's saliva. The paper was published at SID, "31.1: Invited Paper: The Impact of Self-luminous Electronic Devices on Melatonin Suppression".

Believe it or not, the Dell Trinitron (a very common CRT display monitor), due to it's spectrum would still suppress melatonin on a multicolor full brightness background. On a White background at full brightness, most of the displays would suppress melatonin. Here is the chart:

From their paper:
"4. Discussion The present results suggest that a two-hour exposure to light from a computer screen delivering less than 30 lux at the cornea will not significantly suppress nocturnal melatonin production. Lack of melatonin suppression does not guarantee, however, that the use of electronic devices before sleeping will not interfere with people's ability to fall asleep; the computer or gaming tasks themselves may be alerting or stressful stimuli that can lead to sleep disruption." (They were also able to predict within +/- 10%, how strong the suppression would be from various devices, spectrums, and light levels.)

There have been a number of studies that have shown that LED backlit monitors, if bright enough, help people to be more alert and productive, as the blue helps to suppress melatonin production.

The important part is to have them bright enough.

I still have more graphs, and more oscilloscope shots to do, but I will save that for another day as it is getting late, and we all have real jobs to get to in the morning.

Home Depot link:

Model # 423244
Store SKU # 643430
Internet/Catalog SKU # 203285540

Go to Home Depot product page

This is the consumer packaging, this one was purchased at Home Depot:

Package Information, Contact information, and Links:

Royal Philips
LED Lamps 3PM5 E335104 UL Listed
A19 Warm White 2700K
10 Watt
940 lumens
92% CRI
Model Number: 9290002097
Order Code: 10A19/L-PRIZE-PRO/2700 DIM 6/1
Outer Bar Code: 5-00-46677
SKU UPC: 0-46677
Registration ZC23-33EE55

Philips Lighting Company
200 Franklin Square Drive
Somerset, NJ 08873

Philips Customer Care (fax): (847)768-7768

Philips Customer Care
10275 West Higgins Rd.
Rosemont, IL 60018

Additional links for the 10W A19 LED-Winner of the L-Prize:

The actual government tested L-Prize bulb:

Webinar: The L Prize-Winning LED A19 Replacement—What Commercial Building Owners/Operators Can Expect in 2012

This January 18, 2012 webinar presented an update on the status of LED A19 lamp options for commercial businesses, with an overview of DOE's L Prize competition and the rigorous lab, lifetime, and field testing that went into selection of the winning lamp. During the webinar, Kelly Gordon of Pacific Northwest National Laboratory presented current A19 LED options and provided details on where LEDs are most competitive and ready. Gordon also shared insights on the L Prize competition winner, the role of partners in market development, and preliminary plans for promotions and incentives. Todd Manegold, Director of LED Lamps Marketing at Philips Lighting, outlined best applications for the winning lamp in commercial settings, Philips' plans for commercial distribution, and a financial calculator showing expected cost/payback using actual examples from field tests.



EDN Teardown by Margery Conner of the Philips sibling bulb:

A19 Philip's low cost Sibling Teardown

Main site
Main site mirror

Click here to E-Mail the Author, Doug Leeper

(or just use leeper@molalla.net if you are uncomfortable with the benign pop-up)