Design Article
Teardown: LED light shrinks size, cost with non-isolated driver
Margery Conner, www.designingwithleds.com
2/22/2013 4:33 AM EST
Page 2
For comparison purposes, Figure 4 shows the driver electronics for a LED light teardown from about a year ago. Not only is the packaging quite different, but there are a lot more electronics. For example, there are three electrolytic capacitors in the older design and a very large transformer.
Figure 4
Figure 5 shows the two different generations of drivers side-by-side: The Insignia driver has just two relatively small e-caps. This begs the question: What LED driver IC does the bulb use, and how does it allow such a tiny driver?
Figure 5
Figure 6 answers this question: The “SUL B” on the tiny IC is the marking for the Texas Instruments LM3445. There is no transformer, indicating that the LED driver design is non-isolated.
Figure 6
There is certainly precedent for non-isolated light bulbs: The incandescent bulb design itself is non-isolated. If you broke the glass in an incandescent bulb while it was plugged in, you would have direct access to the AC line power. Clearly, non-isolated designs can be made fully compliant with UL specifications. (While a non-isolated AC-DC LED driver design can be both safe and meet UL specifications, keep in mind that developing and testing a non-isolated offline LED driver in the lab requires stringent lab safety procedures.)
To get an idea for the actual LED driver circuit design, we can turn to the TI application note AN 2061 LM3445 A19 Edison Retrofit Evaluation Board (PDF) for a non-isolated buck converter. The schematic is shown in Figure 7.
In order to have the LED drive electronics all fit into the base, the Insignia's designers broke the single rectangular board up into the two circular-shaped pc boards (shown back in Figure 3) which fit into the base of the bulb, separate and mechanically isolated from the LEDs. The only wires that come into the bulb are the low-power DC wires that vary between approximately 8-9V. More details and photos on the dimming control performance are here.
In summary
- Placing the Cree LEDs on the three heat sinks that were part of the surface of the bulb. This quickly gets rid of the heat and keeps it separate from the power electronics.
- Forcing the bulb to meet the same, familiar envelope of the incandescent light, which in turn forced the power control electronics to fit into the tiny base.
- Going with a non-isolated buck topology to lower the parts count. Fewer parts = lower cost, smaller size, and greater reliability.
The Insignia bulb is currently one of the best values for an LED bulb on the market.
About the author
Margery Conner has covered the LED industry since its beginnings. She has a BSEE from the University of California, Irvine, and produces www.designingwithLEDs.com. You can read in detail about the Insignia (Part 1, Part 2, Part 3) and other lighting tear-downs there.
Related links
For comparison purposes, Figure 4 shows the driver electronics for a LED light teardown from about a year ago. Not only is the packaging quite different, but there are a lot more electronics. For example, there are three electrolytic capacitors in the older design and a very large transformer.
Figure 4
Figure 5 shows the two different generations of drivers side-by-side: The Insignia driver has just two relatively small e-caps. This begs the question: What LED driver IC does the bulb use, and how does it allow such a tiny driver?
Figure 5
Figure 6 answers this question: The “SUL B” on the tiny IC is the marking for the Texas Instruments LM3445. There is no transformer, indicating that the LED driver design is non-isolated.
Figure 6
There is certainly precedent for non-isolated light bulbs: The incandescent bulb design itself is non-isolated. If you broke the glass in an incandescent bulb while it was plugged in, you would have direct access to the AC line power. Clearly, non-isolated designs can be made fully compliant with UL specifications. (While a non-isolated AC-DC LED driver design can be both safe and meet UL specifications, keep in mind that developing and testing a non-isolated offline LED driver in the lab requires stringent lab safety procedures.)
To get an idea for the actual LED driver circuit design, we can turn to the TI application note AN 2061 LM3445 A19 Edison Retrofit Evaluation Board (PDF) for a non-isolated buck converter. The schematic is shown in Figure 7.
Figure 7
Click on image to enlarge
Click on image to enlarge
In order to have the LED drive electronics all fit into the base, the Insignia's designers broke the single rectangular board up into the two circular-shaped pc boards (shown back in Figure 3) which fit into the base of the bulb, separate and mechanically isolated from the LEDs. The only wires that come into the bulb are the low-power DC wires that vary between approximately 8-9V. More details and photos on the dimming control performance are here.
In summary
- Placing the Cree LEDs on the three heat sinks that were part of the surface of the bulb. This quickly gets rid of the heat and keeps it separate from the power electronics.
- Forcing the bulb to meet the same, familiar envelope of the incandescent light, which in turn forced the power control electronics to fit into the tiny base.
- Going with a non-isolated buck topology to lower the parts count. Fewer parts = lower cost, smaller size, and greater reliability.
The Insignia bulb is currently one of the best values for an LED bulb on the market.
About the author
Margery Conner has covered the LED industry since its beginnings. She has a BSEE from the University of California, Irvine, and produces www.designingwithLEDs.com. You can read in detail about the Insignia (Part 1, Part 2, Part 3) and other lighting tear-downs there.
Related links
. LED performance boosted by wireless bonding
. White LEDs printed on paper
. OneLED - Waterproof LEDs advance design of outdoor displays
. LEDs tailor automotive lighting, improve safety
. Power and controlling high brightness LEDs
. LED drivers provide control, circuit-protection design options
. Controlling multiple LED strings with C2000 MCUs
. Understanding LM-80 to evaluate LEDs
. Efficient method for interfacing TRIAC dimmers and LEDs
. Compatibility per watt, meaningful metric for dimmable LEDs
|
|
|
|
|
Navigate to related information
Bert22306
2/23/2013 6:19 PM EST
Wow. It takes a lot more eletronics than I would have expected, to drive these LEDs. One obvious problem is that they don't use 120 AC or 170 DC, and another must be that you need a current source to obtain predictable light output from individual LEDs.
Also, a lot of filtering going on there, with chokes and caps all over the place. Hopefully, this will prevent any significant amount of RF racket from being emitted.
Sign in to Reply
Daryl Anderson
2/26/2013 1:41 AM EST
I went to the best buy site. From what I can tell the bulbs are 13w not 8w. What am I missing?
Sign in to Reply
dleske
2/26/2013 8:32 AM EST
It takes a lot LESS electronics to drive a LED with CC than it does for a CFL (compact fluorescent lamp) to excite a gas tube with HV/HF waveforms!
Sign in to Reply
OleAlexander
2/25/2013 3:32 AM EST
I can't help always returning to the irony of LED bulbs;
Q why are they more efficient?
A because they generate much more light and much less heat.
Q what is that huge ugly metal bit?
A A heat sink. because it neets to get rid of so much heat.
??
Sign in to Reply
Bert22306
2/25/2013 3:46 AM EST
Consider the difference in operating temperature of a PN junction compared with, say, that of a white-hot thread of tungsten.
Sign in to Reply
dleske
2/26/2013 8:53 AM EST
Filament bulb: 2% of energy emitted as light, 88% as infrared, 10% as conducted heat. Filament at 2600°C, glass at ~100°C (regular bulb) to ~300°C (halogen). Bulb operates by being hot, so heat doesn't worry it. (It can melt lamp shade though.)
LED: ~20% of energy emitted as light, ~80% conducted heat. LED is very small, doesn't like being much over 100°C, so needs careful design to get heat away.
LED lighting can be cooler, safer and more efficient. (Some poor designs are not!)
Sign in to Reply
Jack.L
3/3/2013 9:09 PM EST
Keep in mind that radiated power is a function of the temperature of what is radiated. Heat sinks are relatively cool so they do not radiate much heat and much rely on convection (and some radiation). A bulb radiates most of its energy into the environment.
To dleske below, actually we are up to 30-40% in some cases for modern LEDs in terms of percentage of power that is light, though high CRI, low color temp bulbs are generally lower today.
As the LED efficiency goes up, power supply requirements go down linearly, but heat sink requirements drop quicker as less and less power is generated as heat.
Sign in to Reply
Tamza2
3/11/2013 2:43 PM EDT
because the heat (even if a lot less) has to be essentially conducted away rather than lost by radiation.
Sign in to Reply
Pinestream
2/25/2013 11:42 AM EST
“Dimming is an important bulb characteristic for the US market, and is an EnergyStar requirement."
This is not true!!! ENERGY STAR Lamps Draft 3 Version 1.0 Specification does NOT require dimming.
Sign in to Reply
conner
2/25/2013 11:49 AM EST
You're right - I was thinking of the L-Prize requirement. This is being fixed in the article.-- Margery Conner
Sign in to Reply
anne-francoise.pele
2/26/2013 9:39 AM EST
FYI, this has been fixed... Thanks for your interest. Anne-Francoise Pelé
Sign in to Reply
EVVJSK
2/25/2013 1:12 PM EST
It would be interesting to know an estimated cost of the electronics and if it would be possible to put all the electronics in one package (i.e. the base) that is separate from the actual LEDs and come up with something that could long-term keep the cost of replacement devices down when the LEDs finally burn out. That is, make the LED and/or Light producing part of the design replaceable, but the base resusable. It would probably require a more stable design that is not going to need to change over time so the electronics could be assured of being reusable. Lighting is one of the bigger reoccuring costs that people know are going to happen. Any way in which costs can be significantly reduced ,over the lifetime, would be a big help.
Sign in to Reply
fotoguru2
2/25/2013 6:24 PM EST
I'm guessing those power supplies may burn out due to power supply spikes before those well regulated LEDs do. Or the failure of the supply is what will burn out the LEDs. Who knows? We might need replaceable power supply units for our 100 year LEDs!
Sign in to Reply
cyclone08
3/6/2013 12:10 PM EST
This is correct - the power supply will burn out before the LEDs.
Sign in to Reply
Tamza2
3/11/2013 2:49 PM EDT
Myth of 50-year LEDS has to be nixed. Specially if additional electronics in the power circuit (such as motion-sensitive switching) spikes will shorten life of 'bulb' electronics. My CFLs last 'forever', but those with motion-sensor switching.
Sign in to Reply
DrQuine
2/26/2013 10:11 PM EST
This bulb having the same form factor as a traditional light bulb is a great benefit. I do hope that they prove as reliable as promised. While the LEDs may have a long life, most of the package is other components. I know that I've returned at least 10% of my 7 year guaranteed compact fluorescent lamps (CFLs) because of failures within 2 years. The CFLs also suffered from slow start-up (anything over "60 watt" replacement could take a minute or more to become bright despite their packaging claims to the contrary). Dimmers and remote controlled outlets, of course, were non-starters.
Sign in to Reply
Etmax
2/26/2013 11:10 PM EST
Interesting, 2 electrolytic capacitors. From memory the life span at 100degC is 2000 to 5000 hours for the good ones. I think this needs to be re-thought.
Sign in to Reply
Dave1010101
2/28/2013 9:34 PM EST
I think the caps will not be that hot but eliminating them--or making them smaller--is probably a design project somewhere.
I know that there was much work done in multi-phase power converters to power very high current (but low voltage) with fast voltage adjustment.
Perhaps more IC integration will simplify the overall architecture and make the system more reliable and easier to manufacture.
Sign in to Reply
Jack.L
3/3/2013 9:03 PM EST
I have evaluated several LED bulbs and have found the temperature in the power supply to vary between about 65C and 90C across a range of designs in a typical installation environment.
If you are using 2000 hour 85C caps then you are going to have issues in all of those bulbs depending on how tight the required tolerance is of the power supply. On the other hand, if you have 5000 hours 105C capacitors, most of the designs will see lifetimes that match the product claims.
Sign in to Reply
Tamza2
3/11/2013 2:51 PM EDT
These little (sometimes large) caps are the bane of most power electronics; will need some alternative or a lot development for renewable energy to really kick in. As an example windmills typically have a 25 year 'warranty', but the power electronics is covered under a 'maintenance' agreement.
Sign in to Reply
JeffL_2
5/28/2013 12:36 PM EDT
And you're surprised? I worked in that industry awhile, windmills are generally sited where there's a high probability of high sustained winds ie up on hills and at or near the end of long 480V 3PH feeders. To translate that's kind of like putting up a lightning antenna network! You quickly realize these little "surge protector diodes" don't help you A BIT, your protection devices need to actually dissipate significant amounts of power (gas tube protectors help a little). Ever try "recovering" a backplane that's suffered a near-direct strike? Keeping these things up SOUNDS easy until you actually have to do it, then you start to appreciate that it's MUCH harder than it looks!
Sign in to Reply
PCB_Tom
2/28/2013 4:41 PM EST
Yep, those electrolytics will almost certainly be the weak link that fails in these.
Sign in to Reply
l8mackey11
2/28/2013 10:25 PM EST
I thought one of the nice attributes of the LEDs was long life (like on the order of 50,000 hours) if the heat management is designed right?
Anyone hear about the Plasma Emitting Diodes. Saw an article about them awhile back but not much since then.
Sign in to Reply
Jack.L
3/3/2013 9:05 PM EST
The LEDs are good for 50,000 hours or more. However, bulbs have force a smaller than ideal heat sink and limited ability to isolate the power supply from the heat sink.
I don't think you mean plasma emitting diodes, but plasma light sources. They are still going strong somewhat, but there are few companies working on it. As LED costs come down and efficiencies go up, their market space keeps getting smaller.
Sign in to Reply
Jack.L
3/3/2013 9:25 PM EST
The two picture of an old design and a new design DO NOT represent generational changes, simply different design choices. I have a driver extracted from a LED lamp (non-isolated) that is 18-24 month old that is simpler in design than either of the examples shown. It is not the greatest design in terms of efficiency, but it works ..... other than the Rudycon ... yes Rudycon capacitor. It even passed UL, though I have a hard time believing it passed any conducted emissions tests.
Sign in to Reply
daemlo
3/20/2013 2:25 PM EDT
Does anyone see a problem using electrolytic caps in this application? Forget LM80, 50,000hrs, or 50yrs - The life expectancy on the caps will dictate when you need a new bulb.
Sign in to Reply
KGround
3/22/2013 8:55 AM EDT
A few comments ...
1) Not to be too picky, but I think it is 9 LED chips and not 6 ...
2) A large part of the power supply complexity and parts count (and much of the need for those pesky electrolytic caps) arises out of the need or desire to control power factor. To achieve an Energy Star rating the power factor must be good. This is not a requirement for UL listing as far as I know. Good power factor helps the electric company but is actually detrimental to the customer(at least in the short term) because electric meters do not measure or charge for the 'imaginary' component of power present in poor power factor equipment.
3) I do see the need, at least for the present retrofit market, to make lighting products that are backward compatible with existing fixtures, but the real future and advantage of LED lighting will be achieved with fixture and architectural designs that take maximum advantage of the desirable features of the LED source. See for example the LED 'strip lights' which are becoming widely available at reasonable cost. These are a distributed light source which is more desirable in many architectural designs, and by distributing the LEDS instead of concentrating them the heat dissipation problems go away. In my personal architectural applications I am using DC powered LEDs, which permit direct use of low voltage power source derived from solar without the inherent loss of efficiency arising from converting stored DC to AC and then back to regulated DC at the fixture. The parts count for the LED current regulator is 1 IC and no capacitors are used. But this approach requires a fully integrated design from power source through power distribution (at 12-36VDC) to LED fixture. DC designs can of course also be powered by UL listed class 2 'wall warts' allowing the choice of AC or DC power source, and providing an inexpensive and easily replaceable alternative for the short lived AC power supply.
Sign in to Reply
pyeatte
3/22/2013 9:53 PM EDT
Perhaps LED bulb costs could come down with separate low voltage lighting circuits. A step down transformer at each room would distribute low voltage for LED lighting instead of dealing with 120VAC at each bulb.
Sign in to Reply
anne-francoise.pele
4/12/2013 8:54 AM EDT
Click on the link below to check out the collection of Design Articles, Case Studies, Product How-To articles, Teardowns, etc... that have been published on Smart Energy Designline.
Check back frequently. The list will be updated as new articles arrive.
http://www.eetimes.com/design/smart-energy-design/4371743/LED-Focus
Sign in to Reply
sixtysixscrews
5/30/2013 2:25 AM EDT
Why are we stuck on screw-in replacements for Edison lamps?
Yes, there are billions of Edison sockets out there that can be filled - but the future isn't written in stone. Dump the Edison screw-in replacements and think of something that works best for PN junction light emitters - what is that? I don't know but the prize goes to whoever thinks of the best solution to the new problem, not the old one.
Sign in to Reply