At the end of last year, the UNSW CATS room right next to my PhD office in the Vallentine Annexe was undergoing complete renovation. As part of the renovation, they removed all of the old luminaires, most of which were destined for the scrapheap, it seemed. After some enquiries, I was quietly permitted to take a few samples for my own research purposes.
The downlight “fad” of the late-90’s and early 2000’s was one of the worst things to happen from an energy efficiency standpoint. Designed for “accent” lighting in highlighting features, the downlights were commonly misused for whole room lighting. Such low-voltage downlights mostly relied on MR16-style 50w halogen downlight globes.
Drawbacks of such downlights include:
- Only slightly more efficient than your regular incandescent globe, resulting in lots of heat and air-conditioning costs.
- Requires a transformer, either magnetic transformer or electronic transformer, which affects efficiency and can fail.
- Globes can “mysteriously” shatter and explode, which means you should opt for closed type globes if you want to be extra safe.
- Run at high temperatures which caused occasional ceiling fires in incorrectly installed units.
- Narrow angle beam necessitates high number of globes to cover a room, which increases power usage and relamping costs.
- 2000 hour nominal lifetime necessitating moderately frequent replacement.
- Colour shift as the globes age, and there is both IR and UV content in unfiltered globes which can degrade items exposed to the light.
At the time that downlights started taking off, most other luminaires had fluorescent options for better energy efficiency, but the downlights did not. Regardless, energy efficiency didn’t seem to matter as people pursued the “look” of having downlights.
Only in the past five years, or so, have LED-based replacements for the MR16 globes been widely available. These have various issues, with poor designs resulting in short lifetimes, flashing and flickering, or insufficient light output.
It seems that, at the time of pulling the lamps out, UNSW had a mixture of regular 50w halogen MR16 lamps, and several different types of “professional” LED replacements installed over several years of relamping procedures. As I was only permitted to take a few items, I decided to ferret out for unique items to take home, take apart and learn from. After all, my PhD is LED related, and learning from disassembly of commercial products is definitely valuable.
Philips EnduraLED 10W 3000K 36D/60D 10GM6DBAABBA/10GM6DBAAABA
These globes were the most powerful of the bunch, being 10w rated, meaning they will actually put out a comparable amount of light compared to the 50w globes they’re intending to replace (the luminous efficiency difference is about 5:1).
Two particular variants seem to be installed, possibly without care, with the 36 degree variant probably replacing EXN bulbs and the 60 degree variant replacing FNV bulbs. The only difference between the two beamwidths is the design of the front plastic lens. These globes were “rated” for 12VAC, claim to be dimmable, and both were Made in China.
Surprisingly, there was a low amount of metal in this globe – it feels pretty light and the rear black portion is made of plastic. Only the silver “ring” with the branding is actually made of metal.
To open the globe, you remove two rubber plugs which reveals two Philips head screws. Removing the two screws allows you to separate the rear from the metal front. This reveals exactly why it can be so light for a 10w globe – it has a FAN inside! Yikes. I don’t like fans, because they are prone to bearing failure over time, and they collect dust. The small fan is integrated such that it sucks air through the slots around the edge of the globe, across the electronics at the rear, then through to the ribs (acting as a more effective heatsink for the LEDs) and out the hole in the centre of the globe. So it seems it isn’t really easy to make a reliable 10w globe using passive cooling in the MR16 form factor, so they opted for the fan.
Removing the two screws in the heatsink allows the front plastic optic to be removed.
The front PCB contains four Luxeon (a subsidiary of Philips) Rebel LED packages, likely rated at about 3W each. The PCB is made by HTSA, dated week 23 of 2012, and marked with SD6132LP-8. The PCB isn’t a special MCPCB – it does, however, try to improve its heat dissipation by having a copper layer on the rear and relying on a large number of vias to carry heat from the top copper tracks to the button copper traces. A pink thermal pad material (not pictured) is used to carry the heat from the rear of the PCB to the cast-aluminium heatsink which is fan-cooled.
The power driving circuitry is mounted in the plastic shell, and features two potted shielded inductors (likely to avoid any noise and RFI complaints and to improve efficiency). The PCB is made by the same company, dated one week later. Four diodes in the middle seem to form a full-wave bridge, and then probably DC-to-DC converted to drive the LEDs and fan appropriately.
Testing the globe revealed it functioned perfectly, with very bright output, and a pretty much silent fan. A good result, as expected from a quality brand. Searching online seems to show it is rated for 410-445 lumens and a lifetime rating of 25,000 hours.
Philips 7W Master LED 4000K 60D 9290002381
As a result of random regular and spot relamping operations, some of the lamps were replaced with this 7w replacement, also from Philips. This one is manufactured differently, and uses an all passive cooling design with a full metal body. This unit also claims to be dimmable, with a higher colour temperature than the one above.
Yet again, the lamp is made of four LED packages covered by a plastic lens. The lens is held on by two side clips, and can be released to reveal the board, which seems to be made of ceramic and is thermally glued to the heatsink at the rear. The board design is dated 24th December 2012, the board seems to be dated Week 10, 2013, and is labelled 4up XBD_V2.0. Surprisingly, this means that the LEDs used are from Cree (a competitor) from their XB-D series. Each of the LEDs is rated for up to 3W operation, but is probably conservatively operated for thermal and lifetime reasons. To the side, there is a large sheathed electrolytic capacitor with a K-cut top, indicating quality parts.
As I can’t get to disassemble this lamp without prying off the LED substrate, as the screws were underneath, I left it there. The thermal adhesive has set hard, and just won’t let go, and I didn’t want to destroy the globe.
The light was nice, bright and stable with a very believable 7w output. It never seemed to get too hot either, which is good. The unit seems to be rated for 40,000 hours. Interestingly, if you wanted to buy it, it seems that they’re being sold at AU$15.95 at Bunnings.
Lighting Science Group Corporation RL-16D2-E0540 8W
This is not a brand I’ve heard of before, but it seems to be a very botique brand that lighting specialists might choose to use in their lamp replacements. That being said, they do have a datasheet for the lamp, claiming a 12VAC/9VDC operation for the lamp, 3050K colour temperature, 58 typical CRI, 48 minimum CRI, 38 degree beam angle and over 50,000 hours lifetime. It does say “is not compatible with all fixtures having integral electronic transformers” and has a very handy table stating compatibility (excerpted below). The lamp is backed with a one year warranty.
I’ll explore ballasts in the next part of this salvage, but it’s good to see that there is good documentation for this globe.
The globe is mostly made of aluminium fins for heatsink, with the end cap made of plastic, secured to the heatsink by two Philips head screws.
From the front, there is a reflective multi-faceted reflector optic which is made of plastic with a reflective coating on it. This is held in place by an aluminium ring which is held by the front fins.
The 8w globe is made of, what appears to be, a single emitter on a star base, secured with three screws. Thermally, it seems well managed with thermal paste on the rear, however the star PCB was a little unusual. The rear has a thick copper plate, but no visible vias, and the main material of the board doesn’t seem to be metal.
The material inside seems to be a brown “bakelite” like material, but I’m not entirely sure. The LED looks interesting as well, as there was no obvious soldering, and the package doesn’t seem familiar to me. The PCB is marked with 0601 A307600028 BBB.
Removal of the two top screws doesn’t actually allow the top pins to pull free, at least, not without some destruction. The top chamber is filled with a grey putty-like potting substance which was pliable – this seems to be for improving the acoustic performance by damping any inductor vibrations and to improve thermal dissipation by acting as thermal mass and conductivity material.
It took a lot of scraping to reveal the components – there is one large shielded potted inductor, a smaller open inductor and a transistor and resistor on this side.
The majority of the components are on the other, and are densely packed together, with several controller ICs, and diodes. No large capacitors seem to be part of this design.
Fawoo 5W (?)
Finally, the last type of globe (all found in the same room) is the Fawoo branded globe. This one doesn’t have a power rating on the globe itself, but I’d say it’s likely a 5W model. This unit was based on a nice “curved” heatsink design with a plastic base, plus a water-drop shaped front lens.
This unit seems to have a “magnifying” lens which spreads the light from the smaller phosphor area across a wider beam area. The front plastic lens is removable, by pulling on the wire clip which holds it in place. A white-grey plastic donut is used to cover the area around the LED to improve the “reflection” of light out of the unit.
Thermally, the LED was well managed, with just the right amount of thermal paste applied and a metal core circuit board, but only two screws. The two screws featured plastic washers, however, the washers are split.
The LED seems to make use of remote phosphor technology, and several individual LED dies (see indents), which prevents “hot spots” and glare.
Some of these have failed around the campus, resulting in a dead light or some “flashing” which can be quite annoying.
Experiment: Voltage vs Power Consumption
One of the issues with LEDs is the need to regulate the current flow through the LED as they are more of a constant-current style device. As a result, the power consumption should really stay “steady” across a range of voltages, to keep the brightness constant despite voltage changes. For dimmable LEDs, depending on whether it’s a waveform based or voltage based dimming, the current consumption is expected to correlate with voltage.
As a result, I used the Agilent U1241B to measure current to the globe, as I varied the supply voltage (by stepping upwards) to each of these globes. The product of voltage and current was used to produce the graph of power. Do note that the globes seem to have some hysteresis, so when reducing voltage, you can reduce it slightly further and still have usable light than the graph would imply.
Each of these globes seem to have a different story. The two Philips LED globes seem to share similar regulation tactics – namely, that they start producing useful light at about 5v, peak in power at about 9V, and maintain a similar power level to the rating right up to overvoltage. As a result, voltage based dimming is possible with the globe, but seems to result in some audible noise from the globe itself and maxes out by 9v.
The Fawoo lamp seems to show a system which is compatible with voltage based dimming, with peak brightness just before 12v, and useful light above 8.5v. It seems that this means that only part of the dimming range is usable with the Fawoo, but the regulation through overvoltage is excellent.
The LSGC lamp seems to have failed in regulation, with useful light being produced at 7.5v, and the rated power rating being reached by 10v. Providing a higher voltage seems to cause linearly increasing power, which might shorten the lifetime of the LED, especially when considering most ballasts provide 11.5v to 12v.
Downlights are terrible things when it comes to energy efficiency, and a renovation of a lecture theatre provided an opportunity to salvage some gear. Most of it was destined for the skip, which was surprising, given the number of LED replacement energy-saving lamps with useful remaining life. As I was allowed to salvage a limited number of items, I decided to sample one of each sort for my own investigations to understand their design elements and performance. Of course, it seems that the performance of the above units are proven based on their many-years of operation in a commercial setting, and their designs seem appropriate for the needs.