Seldom will only one LED fit a particular application, so even if bare minimum cost is the only requirement, designers will still have to ask themselves some questions to focus on likely candidates.
One of those questions has to be, which cost needs to be minimised: bill-of-materials (BOM) cost; ex-factory cost; life-cycle cost; or carbon footprint?
It turns out that the question cannot be answered until you know who the end customer is, as the needs of commercial lighting are different from the desires of consumers. Commercial operations are far more likely to look beyond initial price.
Paul Scheidt, product marketing manager at Cree, points out that commercial customers realise that an LED product can last twenty times longer than a halogen bulb.
“There are lot of scenarios when after the first two conventional lamps are replaced, an LED would have been cheaper,” he says. “Wal-Mart is looking at LED parking-lot lighting. It reckons the cost difference between old lights and LEDs is $500/light, while re-lamping is $1,000 if it is scheduled, and $2,000 if it is not scheduled. If it is working for Wal-Mart, it should be an indication that the money savings are there.”
Philip’s lighting application group manger Charles Schrama agrees: “It’s lumen maintenance: how often do you need to replace the light bulb. Traditional lights are replaced every 2,000 to 3,000 hours. LEDs last much longer and reduce the cost of ownership.”
The second factor is efficiency. “With professional lights – office and streetlights – there is a strong tendency for very efficient lighting to reduce the power bill,” says Schrama.
On the other hand, consumers tend to buy on price. “The price of LED bulbs is very high for home use, although manufacturers are trying to reduce initial cost,” says Schrama. “Mass adoption will be when the price gets down to the point where compact fluorescent lamps started to be adopted: $12.99, or even better: $7.99.
“The price will have to come down a little bit more to reach the tipping point,” says Jason Posselt, vice-president of marketing at Californian LED array maker Bridgelux.
So how do you minimise the cost of LEDs in a product?
“In general, minimising the LED count is a good way to reduce BOM, usually by putting more current through the LEDs,” says Schrama. “Some LEDs cannot be driven hard because their construction can’t dissipate a lot of power. The best LED to reduce BOM is the LED that can be driven as hard as possible.”
“For us, it is better for us to use a large die and drive them very hard to reduce number of LEDs,” says Scheidt.
“There is a trade-off here as the harder you drive LEDs, the less efficient they get. People are willing to pay for a certain level of energy saving, but no one knows what that is yet. Do they want an 80 lm/W bulb or a 60 lm/W bulb? You need more LEDs or a higher bin for more efficiency. Are people willing to pay for it?” he asked.
And lower efficiency means more heat and therefore, all things being equal, higher temperature and lower reliability.
“We have a lot of data to show that reliability is not affected by temperature as much as people imagine,” says Scheidt.
Cree recently released an application note which compares two routes to identical-looking approximately 940lm ceiling lamps (100W bulb equivalent), one with 12 LEDs running at 350mA, the other with five LEDs at 1A.
They share similar metalwork and optics. Power, PCB temperature and calculated life were 14.5W, 52°C (Tj=57°) and 900,000 hours for 12 LEDs, and 74°C (Tj=89°) and 300,000 hours for five LEDs – with greater than 60,000 hours extrapolated measured life for both.
|Bridgelux, Cree and Philips come at LED lighting from different directions
Philips, always a lighting company, bought into HP LED technology to form Lumileds, makers of Luxeon lighting LEDs. Cree is a silicon carbide device maker which diversified into lighting LEDs, and then bought lighting firms.
A lower LED-count is not always the way to reduce ex-factory cost at the lighting plant. Particularly if high-reliability is lower on the priority list.
“Some bulbs are small, like an MR16 or GU10, so you have to engineer the correct solution to fit,” says Scheidt. “An A19 light bulb can contain a big board, and in Asia it is fairly common to see them stuffed full of backlighting LEDs.
“Backlighting LEDs are focused on cost-efficiency and last for between 5,000 and 10,000 hours to 50% intensity. Lighting LEDs last 50,000 hours to 70% intensity.” he says.
And if there is room for more LEDs, smart manufacturers can use cheaper, tinted LEDs, which have fallen outside the desirable ‘black body’ characteristic in production test, and intelligently mix them inside the lamp to produce an un-tinted product. Changing the number of LEDs makes only a small difference to hardware costs.
“Usually, heatsinks are made of aluminium and price depends on mass. “If the LEDs are more efficient, the heatsink price is not that much different,” says Schrama.
He adds: “Some manufacturers try to go to sheet steel to reduce cost. This has very different thermal properties.”
As well as different technologies in the die, there are different LED packages, none of which is the ‘right answer’ for all applications.
Retrofit and new-build
There is the retrofit industry, producing LED-based light bulbs that fit onto sockets designed for incandescent bulbs – much like compact fluorescent lamps are designed to do – and there are the luminaire manufacturers that are making complete light fittings that will only ever be powered by LEDs. In the latter case, the luminaire industry was until recently mostly low-tech: metal forming and plastic moulding, with snap-together or screwdriver assembly, and soldered, push-in or screwed termination.
For those that remain low-tech, or have small production runs, there are LEDs in packages that can be screwed down, like the original Luxeon Star, or an increasing number of multi-die rectangular ceramic tiles, frequently called ‘arrays’, with holes for screws and pads for solder connections.
These firms have no choice but to use high-tech manufacturing, even if they are using screw-in arrays, because they have to incorporate power supplies and meet the thermal tight constraints of semiconductors.
There are also ‘light engines’ – arrays of separate surface-mounted LEDs on thermal substrates made by high-tech third parties for old-tech luminaire makers.
And there are luminaire firms in countries where low wages make labour-intensive assembly cost-effective.
Last, there are established luminaire makers that have skilled-up, and new luminaire makers that use surface-mount parts themselves.
Arrays vs single die packages
“The LEDs run from a quarter-watt single die to 50W arrays, and applications run from nightlights to streetlights, from a few lumens to several thousand lumens. The corner cases are obvious, but there is a lot of middle ground,” says Posselt.
While Lumileds, Cree, Nichia and Osram are power LED companies that grew from signal LED companies, Bridgelux was set up from scratch to make only ceramic tile array LEDs for lighting, aimed specifically at manufacturers wanting low-tech assembly. Several other firms, including now Cree, make array products.
“Our premise on the lighting array is we offer a solution to lighting makers. If you throw them a handful of LEDs, they are not going to know what to do with them,” says Posselt.
“Industry today buys light sources that are pretty consistent, like 60W bulbs. We target applications that roughly match them – one source that delivers 800lm to replace one 60W halogen, for example.
“The customer picks an array, replaces the ballast in his luminaire with a power supply, designs the optics, and leverages his channel to market.”
As such, the firm is more likely to be talking light metrics than electrical metrics.
“When we set up, everyone was talking about watts, but this is somewhat irrelevant to selecting how many to use, so we focused on lumens, and we make sure we give enough lumens so one array is usually all they need,” says Posselt.
“The idea is to reduce cost at the system level – one array for 1,000lm, not 10x100lm emitters. One array eliminates non-value-add stuff. You don’t have to mount 10 emitters on a PCB and you don’t need 10 lenses or 10 lead frames.”
To further ease the use of arrays, connector companies including Molex, BJB and Tyco have developed screw-mount holders which simply fit over an un-mounted array.
These not only apply mounting pressure to the array, but also make contact with its solder pads and offer push-in wire receptacles for fast assembly.
Bridgelux and Molex actually teamed up with to produce Helieon – a bayonet-fit system that give LED luminaries light-bulb-like plug-in convenience, allowing consumers to swap LEDs of different beam angles and colour temperature.
In a similar vein, Philips has Fortimo Twistable and GE has Journee.
|Not a lot of people know that
Have you ever wondered why power LEDs are tested at 350mA, and where the de facto standard 20mm hex board came from?
So arrays are always better?
“Where arrays start to loose is lower and lower lumens. It is hard to compete against one emitter,” says Posselt. “You probably wouldn’t propose an array for nightlights, and you might for a streetlight, but maybe you would go for emitters in a streetlight, depending on the optics.”
And overall? “We have seen we can compete against the ‘bag of emitters’, and in many cases drive lower cost,” says Posselt. “We compete well at 800, 1,200, 2,000, 3,000 and 5,000 lumens.”
For applications that need beams, the exit diameter of the optics – which is the minimum diameter of the product – is proportional to the width of the light-source and loosely inversely proportional to the beam angle required.
So for flashlights the only technically-feasible answer is a single die, or perhaps a 2×2 die array. At the other end of the scale, with retail accent lighting for example, an array with a 10mm emissive diameter is likely to produce a more compact product that 10 single emitters, each with its own small lens.
In between these two examples there is plenty of room for arrays to compete with single emitters.
Posselt has advice that applies to costing all LED lighting designs, no matter what kind of LED is used.
“The luminaire manufacturer has to decide what they want to achieve: do they need to meet a specification like Energy Star or Part L; do they need a certain number of lumens,” he says. “Then they need to work back through the trade-offs.”
These include optical losses if the light needs a diffuser or physical protection, thermal losses against heatsink size, and driver losses.
He points out that there is a tendency for luminaire designers to go easy on these parameters on the first pass and come up with an LED requirement that is prohibitively expensive.
“When they trade off losses, we may find we get asked for a 135lm/W warm white, says Posselt, “and we have to say maybe they should spend more on the optic, a larger heatsink, or active cooling, to make the led to more cost effective.”
The specification may need some attention as well, in colour temperature for example. “You take a bit of an efficiency hit when you go from 3,000K to 2,700K. Maybe 90lm/W hot with 80CRI to 80lm/W,” he says.
The bottom line
Consumers ten d to like low cost and enough efficiency, while professionals think long-term and value high reliability for low maintenance, and high efficiency for lower energy cost. Selling price is best with a few LEDs driven hard, whereas more LEDs doing the same job at lower current will last longer, waste less, and run cooler.
Arrays are best suited to traditional luminaire maker working methods, while surface-mount single emitters and surface-mount 2×2 arrays match well with high through-put electronic industry production techniques.
Which results in cheapest product will depend on labour cost, production equipment, expertise, the number to be made, and many other parameters – there is seldom a right answer.
And if the customer can take a slight hit in specification, they might save money in the long term.
“The name of the game is, you can always throw more LEDs at the problem to get any intensity, any colour rendering index, and any colour temperature; it just depends on how big and expensive the product can be,” says Cree’s Paul Scheidt.