Screw-in LED lamps for power line use

[Bright+]

There are many plug-in LED lamps that aren't worth getting, because they're all 5mm cluster design that will decay quickly.

Now, there are some aluminum cross drilled and finned body LED lights.
They use a handful of powerful LEDs and produce a cleanly collimated beam to accent something.

http://www.gelighting.com/na/busine...ownloads/led/energysmartLEDPAR30sellsheet.pdf

At 28.5 lumens/W to 34 lumens/W
Cost $50 each

A CFL with PAR can still get 33.4 lumen/watt including the reflector loss.
9W 300lm PAR 20 33.4 lm/w


So, seeing that LED alternatives are getting around 30 lumens/watt while similar shaped reflector lamps using CFL is gettin around 33lm/watt in smaller one and getting close to 60lm/watt on larger ones (23W 1350lm).

Does it make sense for LEDs compete against vapor discharge lamps in general lighting?

 

[blasterman]

So, seeing that LED alternatives are getting around 30 lumens/watt

A 23watt 4100K CFL has what, 1600 or so lumens of out-put? Even assuming ballast inefficiency that's about double your spec. Even cheap Chinese 3watt white knock-off LEDs are getting 80lpw.

Does it make sense for LEDs compete against vapor discharge lamps in general lighting?

First, vapor discharge lamps don't typically get screwed into conventional line voltage sockets. They use a dedicated fixture and ballast. However, screw in LED retrofits they you are referring to are castrated in terms of out-put because they are limited by the thermal constraints of the obsolete edison form factor they are trying to fit into.

Dedicated LED fixtures aren't limited by this. So, your comparison doesn't make much sense.

 

[Bright+]

A 23watt 4100K CFL has what, 1600 or so lumens of out-put? Even assuming ballast inefficiency that's about double your spec. Even cheap Chinese 3watt white knock-off LEDs are getting 80lpw.

The spec sheet for GE Energy Smart LED lamps show they're around 30 lumen per watt.

http://www.luminairetesting.com/Files/Sample__LED_Integrating_Sphere_Report.pdf

This testing example of cheapo LED "bulb" that runs on power line is only getting about 25lm/W. Where are you getting 80lm/W (line input to actual output inclusive of ballast loss) ?

First, vapor discharge lamps don't typically get screwed into conventional line voltage sockets. They use a dedicated fixture and ballast. However, screw in LED retrofits they you are referring to are castrated in terms of out-put because they are limited by the thermal constraints of the obsolete edison form factor they are trying to fit into.

Discharge lamps are the primary replacements to incandescent lamps and they include CFLs, regular fluorescent and HIDs. CFL is a vapor discharge lamp and it certainly gets used often in a socket getting around 70 lm/W including ballast power use.

Dedicated LED fixtures aren't limited by this. So, your comparison doesn't make much sense.

Dedicated luminaires should be compared against dedicated fluorescent or HID luminaires. LEDs require a ballast just like discharge lamps.

 

[LEDninja]

LEDs have made major improvements in the last few years and is still doing so.
The 5mm cluster design is using LEDs designed in the last century.
Many LED fixed lighting bulbs are still using the 1W and 3W LEDs from the beginning of the '00s. 42lm/350mA (that is actually 1.25W).
Since the middle of the '00 decade Crees, SSC-P4s, Rebels, Golden Dragons etc. are 80lm+/350mA.

So LED bulbs are improving.
The Cree LLS LRP-38 PAR 38 Lamp LED Bulb - Minimum Efficacy = 42 lm/W.
http://www.winderlumenled.com/reces...page=flypage.tpl&product_id=75&category_id=19

The main disadvantage of LED bulbs is either running more LEDs at low amperages each increasing costs or pushing the LEDs harder reducing efficacy.
The other disadvantage of the LED bulb is the cost is directly proportional to lumens required due to the increasing # of LEDs required. A 25W or a 100W incandescent bulb can be made for the same cost. A 100W equivalent LED bulb costs 4X a 25W equivalent LED bulb. That is why you do not see 1400 lm LED bulbs offered for sale (that is equivalent to 100W incandescent or 23/27W CFL).
Another disadvantage of LED bulbs is the sheer size of heatsink required. The GE PAR30 is 3.75 inches wide (and it admits it has 1/2 the lumens of a 45W incandescent bulb). The Cree bulb I linked to is 4.75 inches wide. Won't fit your standard size lamp.

The main advantage of LED bulbs is relamping costs.
Incandescent bulbs 1000 hours.
CFLs 7000-10000 hours.
LEDs 35000 to 50000 hours. To 70% initial brightness. If you don't mind a little less brightness the LED can go much longer.
The relamping costs is what caused the City of Toronto to study LED in streetlighting. Not the cost of the bulb itself but the cost of the crane truck and crew.

There is a need to differentiate between LED lumens and OTF (out the front) lumens. With the dimmer PAR LED bulbs optics are used to concentrate the light to a spot. The cheap plastic optics absorb a lot of light. So instead of 80(Cree P3 bin)-139(Cree R5 bin) LED lumens/350 mA you get 33-42 OTF lumens/W. (note 350 mA =1.25W).

blasterman builds custom LED fixtures. His high brightness versions are 2000+ lumens. That is equivalent to 150W reflector incandescent bulb (or 250 W regular bulb). With so much light blasterman does not need concentrating optics so there is zero optics losses. In this case 80 lm/W at the LED means 80lm/w into the room.

 

[blasterman]

The spec sheet for GE Energy Smart LED lamps show they're around 30 lumen per watt.

As I told you in my previous reply, LED retrofits have to make serious compromises because of their limited form factor. For the most part, most of them suck, with decent ones only recently coming to market and being fairly expensive. The primary market for crappy ones in either Chinese Ebay sites or Walmart / Home Depot which don't exactly cater to intelligent consumer discretion. The good ones know who they are.

Where are you getting 80lm/W (line input to actual output inclusive of ballast loss) ?

Here's a legitimate study from the Department of Energy showing the huge discrepency between various LED lamps, with several performing in the range of the GE Lamp you linked. Several others perform above 60 lumens per watt, which I believe are mostly the Cree fixtures. The main reason these fixtures don't perform higher is because they are using 60-80lumen per watt warm-white emitters and/or high CRI mixing, which are very inefficient. If you shoved high efficiency, low CRI 110 lumen per watt cool white emitters in the fixtures you would get a drastic increase in efficiency. This performance level is also increasing steadily, which is indicted in the DOE study.

If you bolt a bunch of common 110 lumen per watt cool white emitters to a heat sink and power it with a decent LED driver, you'll get about 100 lumen per watt. I build my own fixtures because I'm assured I'm getting the right efficiency, and it's absurdly easy to do.

Discharge lamps are the primary replacements to incandescent lamps and they include CFLs, regular fluorescent and HIDs.

Which encompasses a broad range from +150lumen per watt high pressure sodium to 35lumen per watt Feit brand CFLs (????????????)

Screw in CFLs are the only ones of the bunch that can be screwed into a standard light-bulb socket and use standard line voltage via their integrated ballast. Also, just like LED retrofit bulbs, screw in CFLs represent the lowest common denominator of that technology. The difference between a dedicated T5 fixture with good reflector and ballast and a Feit Brand CFLs is enormous.

If I followed the logic of our many LED luddites who populate this forum I would {wrongly} conclude that a T5 fluorescent fixture with high efficiency ballast, given it's the same technology as a Feit brand bulb sold at Walmart, must suck because it uses the same fluorescent technology.

Dedicated luminaires should be compared against dedicated fluorescent or HID luminaires

Yeppers. And LED retrofits need to stop being used as litmus for lighting technology.

LEDs require a ballast just like discharge lamps.

LEDs don't use ballasts but either a fixed voltage source (uncommon) or a current regulated source. The good ones are 85-90% efficiency.

 

[blasterman]

The other disadvantage of the LED bulb is the cost is directly proportional to lumens required due to the increasing # of LEDs required...A 100W equivalent LED bulb costs 4X a 25W equivalent LED bulb.

Somewhat true, but misleading. Obviously two LEDs cost twice as much as one, but low efficiency retrofits such as the ones everybody seems obsessed don't follow those economics. Cree XPGs for instance can be driven at current levels in which the heat can't be mediated in a conventional light bulb format. Doubling the out-put of the emitter costs you nothing other than current, and LED retrofits sold at Walmart don't use premium flux emitters.

That is why you do not see 1400 lm LED bulbs offered for sale (that is equivalent to 100W incandescent or 23/27W CFL).

The reason you don't see 1400lumen LED retrofits is the small format can't handle the heat.

Over at Nano-Tuners Evil has some custom Par 38's that if outfitted with higher flux neutral whites would likely exceed 1400lumens. Those I trust about 1000x more than the ones at Walmart or Home Depot.

There is a need to differentiate between LED lumens and OTF (out the front) lumens. The cheap plastic optics absorb a lot of light.

LEDs by nature are already directional, and unlike fluorescent bulbs don't incur a huge penalty via focusing, so I'm not sure where you're getting this. The lowest end Acrylic optics at worst are 85% efficient. Better ones over 90%... Standard acrylic diffusers are 92%. The reason you're only getting "33-42 OTF lumens" per watt from a LED retrofit is because the bulb sucks.

The Cree Fixtures all put out at least 60 lumens per watt, and use diffusers which are likely a tad more opaque than typical LED optics because they are doing more diffusion. The Cree fixtures also perform pretty much in line with the rated out-put of their emitters, which proves that optics in LED designs incur little penalty. If you're shooting LEDs through opaque plastic, then the penalty is much greater. This is actually the biggest problem I'm running into right now because LEDs are so specular they require massive amounts of diffusion for relaxed lighting, and micro etched, high efficiency diffusers are still new technology.

A spiral CFL in a reflector on the other hand incurs a huge OTF penalty for obvious reasons.

 

[LEDninja]

The other disadvantage of the LED bulb is the cost is directly proportional to lumens required due to the increasing # of LEDs required. A 25W or a 100W incandescent bulb can be made for the same cost. A 100W equivalent LED bulb costs 4X a 25W equivalent LED bulb.

Somewhat true, but misleading. Obviously two LEDs cost twice as much as one, but low efficiency retrofits such as the ones everybody seems obsessed don't follow those economics. Cree XPGs for instance can be driven at current levels in which the heat can't be mediated in a conventional light bulb format. Doubling the out-put of the emitter costs you nothing other than current, and LED retrofits sold at Walmart don't use premium flux emitters.

You did not read the whole paragraph.
A 25W incan equivalent at ~350 lm requires 1 XPG driven at 1A.
A 100W incan equivalent at ~1400 lm requires 4 XPG driven at 1A.
You are not going to get 35000 to 50000 hours life if you drive a single XPG at 4A. You can do that with an SST90 but an SST90 costs a lot more than an XPG so costs still go up.
-
What has Walmart to do with it? Walmart sells Lights of America. LOA would need a 500*5mm array to generate 1400 lumens.

-----

That is why you do not see 1400 lm LED bulbs offered for sale (that is equivalent to 100W incandescent or 23/27W CFL).

The reason you don't see 1400lumen LED retrofits is the small format can't handle the heat.
Over at Nano-Tuners Evil has some custom Par 38's that if outfitted with higher flux neutral whites would likely exceed 1400lumens. Those I trust about 1000x more than the ones at Walmart or Home Depot.

Agreed the A-19 globe format can't handle the heat. (For the OP's info JTR1962 once calculated max output from an A-19 globe LED bulb is 500 lumens from the amount of heat that can be dissipated.) So matching a 60W globe bulb at ~840 lumens is impossible with current technology.
The OP linked to a PAR30 and I to a PAR38. So size is not the problem for heat dissipation. Availability of the latest LEDs is, especially if you need large numbers of the warmer tints for mass production. I think we will see 1400 lumens in a production PAR38 before the end of the year.
-
Walmart sells Lights of America which to me is a near fly by night company. There are a couple of threads on how their bulbs last just weeks.
Home Depot Canada sells Philips. Philips together with GE the OP linked to are "Brand Name" companies. They rely on their "Brand Name" to sell products. Bad publicity about problems would drastically reduce sales across all divisions (see Toyota). So they are very conscious of reliability, design very conservatively to maintain the reliability. Despite using Rebel LEDs (60-100 lm/W depending on bin) my 7W bulb is only rated 155 lumens. That is only 22 lm/w; 39 lm/LED. I am not worried about the LEDs burning out.

 

[Lynx_Arc]

The biggest obstacle is money IMO, if it takes longer than 5 years to recoup the cost of a similar output LED vs CFL light then that is a deal breaker because even though the LEDs may last 5 years in use the electronics may not with power outages and brownouts in storms here. You can buy a CFL for $5 and an LED costs $30 that means it has to save you $25 in electricity in 5 years and unless it is a LOT more efficient and run a lot during that time I don't see it competing well. even at 10cents a kilowatt hour if you replace 50watts of CFL with 25watts of LEDs for a savings of 5 cents a kilowatt hour divide that into 25.00 difference for 500kilowatt hours and divide that 500,000watts by 50watts giving you 10k hours at 8 hours a day makes for 1250 days usage or close to 4 years. here electricity is about 6 cents/kw hour so you are only saving 60% as much so it would take 6 years. This is IF... and only IF the LED is twice as efficient and both last 6 years. If the LED light goes out the game is lost, if the CFL goes out you only lose one year off the time the LED has to catch up. I would say until LED lights get considerably cheaper they are not a good investment for standard households yet.

 

[Dave_H]

Despite using Rebel LEDs (60-100 lm/W depending on bin) my 7W bulb is only rated 155 lumens. That is only 22 lm/w; 39 lm/LED. I am not worried about the LEDs burning out.

With such low output to price ratio I'm a bit puzzled as to what is the
retail market; who's buying for what purpose, and volumes. My theory:

(1) People who don't understand what they are buying; some dazzled by
the novelty. Both sides lose if the buyer returns the product, or keeps
it but develops a negative impression.

(2) Niche requirements such as frequent on-off cycles, low heat
generation, and/or high cost/inconvenience of change-out.

(3) Early-adopters who aren't concerned if money is initially lost (or
at best break-even). Hidden saving is the overall energy reduction.
Plus there may be some bragging rights, not necessarily undeserved,
such as "Hey, I'm lighting with LEDs, how about you, still using those
oil lamps...?"

The Sylvania 2W A19 bulb I've mentioned will be a good product if
it lasts the specified 15k hours, albeit limited to low output cases. It's
about 1/3 of the Philips bulb output at 1/3 the cost. I'm surprised how
far 50 lumens goes in my hallway, with a good wall fixture and lightly-
frosted shade.

Dave

 

[Bright+]

The main advantage of LED bulbs is relamping costs.
Incandescent bulbs 1000 hours.
CFLs 7000-10000 hours.
LEDs 35000 to 50000 hours. To 70% initial brightness. If you don't mind a little less brightness the LED can go much longer.
The relamping costs is what caused the City of Toronto to study LED in streetlighting. Not the cost of the bulb itself but the cost of the crane truck and crew.

I suppose 70% cut off is a good reference if it is competing against CFLs and metal halides, but I think its a lost cause against normal straight fluorescent. The now dominant T8 fluorescent is around 97% at 20,000 hours and around 95% at failure or 50,000 hours. T5 that's gaining share doesn't last quite as long, but maintains 95% at 35,000 hours.

http://www.lighting.philips.com/us_en/browseliterature/download/p-5752.pdf

I think hours to 70% brightness should be used for metal halide and CFL replacement, but for indoor dedicated luminaires, cut off should be 95% of original lumens. The fluorescent systems today can get around 90-95 lumens from power soruce to lamp output, less fixture loss (reflector, diffuser etc). Even using LEDs, diffusers are still needed for general lighting.

There is a need to differentiate between LED lumens and OTF (out the front) lumens. With the dimmer PAR LED bulbs optics are used to concentrate the light to a spot. The cheap plastic optics absorb a lot of light. So instead of 80(Cree P3 bin)-139(Cree R5 bin) LED lumens/350 mA you get 33-42 OTF lumens/W. (note 350 mA =1.25W).

I think the binning is based on junction temperature of 25C, which will never happen in real life.

blasterman builds custom LED fixtures. His high brightness versions are 2000+ lumens. That is equivalent to 150W reflector incandescent bulb (or 250 W regular bulb). With so much light blasterman does not need concentrating optics so there is zero optics losses. In this case 80 lm/W at the LED means 80lm/w into the room.

Interesting. What is the temperature he is rating his output on?

Here's an interesting presentation on differences between "LED spec" vs LED performance measurement in lighting industry. Page 21

http://www.cormusa.org/uploads/CORM...M80_and_Future_Standards_CORM_2009_Y_Ohno.pdf

He said LEDs don't need a ballast, but that is not true. Fluorescent lamps are constant current devices too and the ballast is the "driver" (driver or ballast is a matter of semantics). Once the lamps are operating, the ballast/driver provides a constant current supply needed to the device, and in this process, some energy is lost reducing system efficacy.

 

[jtr1962]

He said LEDs don't need a ballast, but that is not true. Fluorescent lamps are constant current devices too and the ballast is the "driver" (driver or ballast is a matter of semantics). Once the lamps are operating, the ballast/driver provides a constant current supply needed to the device, and in this process, some energy is lost reducing system efficacy.

If dimming or absolute constant current isn't important than the "ballast" can be as simple as shown below:

OK, the current might vary 10-15% if line voltage fluctuates from 110 VAC to 130 VAC but your eye will hardly notice it. The circuit as shown will drive 24 small LEDs at ~20 mA. To drive power LEDs you just scale up some of the components. Efficiency is well into the 90s because the only losses worth mentioning are in the diode bridge. Of course, it won't work with a triac lamp dimmer but who cares? Instead, just have one or two or three banks of LEDs which can be switched on or off if multiple light levels is so important. The point is that a "ballast" like this is dirt cheap and easily made by anyone who can solder together a couple of LEDs. There really isn't much stopping anyone who is reasonably good at soldering from building their own fixtures instead of relying on the mostly crappy stuff being foisted on the general public these days.

 

[Bright+]

If dimming or absolute constant current isn't important than the "ballast" can be as simple as shown below:

OK, the current might vary 10-15% if line voltage fluctuates from 110 VAC to 130 VAC but your eye will hardly notice it. The circuit as shown will drive 24 small LEDs at ~20 mA. To drive power LEDs you just scale up some of the components. Efficiency is well into the 90s because the only losses worth mentioning are in the diode bridge. Of course, it won't work with a triac lamp dimmer but who cares? Instead, just have one or two or three banks of LEDs which can be switched on or off if multiple light levels is so important. The point is that a "ballast" like this is dirt cheap and easily made by anyone who can solder together a couple of LEDs. There really isn't much stopping anyone who is reasonably good at soldering from building their own fixtures instead of relying on the mostly crappy stuff being foisted on the general public these days.

In North America, many longer fluorescent lamps requiring a higher starting voltage than the line voltage uses an inductor, but I believe series capacitor limited ballasts have been put into use in 230v countries which allows direct starting of fluorescent lamps without the starting voltage boost.

Also, in the R-C series circuit shown, how much of the 50 or 60 Hz ripple gets through it? LED lights, like the Christmas tree decoration type flickers badly. If it flickers, then it ranks a few levels lower than modern CFLs.

 

[Dave_H]

Also, in the R-C series circuit shown, how much of the 50 or 60 Hz ripple gets through it? LED lights, like the Christmas tree decoration type flickers badly. If it flickers, then it ranks a few levels lower than modern CFLs.

Christmas lights use half-wave rectification without filtering (ones
I have seen or heard about) which causes flicker. Ones with
full-wave rectification without filtering have flicker at 120Hz
which is not visible. The above circuit is full-wave rectification PLUS
filtering, so any small amount of ripple should be undetectable.
I'll leave it to jtr to confirm.

As for the ripple, it could be measured or calculated. I recall
doing the latter decades ago, solving for the sine wave and
exponential discharge into the load...ends up as a sort of sawtooth.

Dave

 

[Lynx_Arc]

In North America, many longer fluorescent lamps requiring a higher starting voltage than the line voltage uses an inductor, but I believe series capacitor limited ballasts have been put into use in 230v countries which allows direct starting of fluorescent lamps without the starting voltage boost.

Also, in the R-C series circuit shown, how much of the 50 or 60 Hz ripple gets through it? LED lights, like the Christmas tree decoration type flickers badly. If it flickers, then it ranks a few levels lower than modern CFLs.

that is a full wave rectifier, should be no flicker at all if the filter capacitor used is large enough as AC changes from - to + two of the diodes allow voltage to pass one way and from + to - the other two allow it to pass the other way with the filter cap acting like a battery driving the LEDs between switching phases.

 

[Dave_H]

I suppose 70% cut off is a good reference if it is competing against CFLs and metal halides, but I think its a lost cause against normal straight fluorescent. The now dominant T8 fluorescent is around 97% at 20,000 hours and around 95% at failure or 50,000 hours. T5 that's gaining share doesn't last quite as long, but maintains 95% at 35,000 hours.

http://www.lighting.philips.com/us_en/browseliterature/download/p-5752.pdf

I think hours to 70% brightness should be used for metal halide and CFL replacement, but for indoor dedicated luminaires, cut off should be 95% of original lumens. The fluorescent systems today can get around 90-95 lumens from power soruce to lamp output, less fixture loss (reflector, diffuser etc). Even using LEDs, diffusers are still needed for general lighting.

90-95 percent cutoff sounds great but with LEDs will result in
an unnecessary low "life". I believe the 70 percent cutoff is where
people start to perceive the drop in brightness. In critical applications
I can see a higher cutoff, but not for general household use.

Dave

 

[jtr1962]

Also, in the R-C series circuit shown, how much of the 50 or 60 Hz ripple gets through it? LED lights, like the Christmas tree decoration type flickers badly. If it flickers, then it ranks a few levels lower than modern CFLs.

C2 is the filter cap in that circuit. The 100 uF value shown means the current through the LEDs varies by about 2 mA in either direction from the mean of 20 mA ( I did a circuit simulation using CircuitMaker software ). In other words, about 10% ripple current, and no detectable flicker. In fact, you could probably reduce C2 by a factor of 5 and still not get noticeable flicker. I'm just anal about reducing ripple current.

For power LED use you might increase C1 to 10 uF or more to allow 350 mA current, and also increase C2 to 1000 or 2200 uF. R2 is only needed to discharge C1 to prevent a shock when the device is unplugged. This is only applicable to uses like nightlights. R2 can be eliminated if you're making a ballast which is hard wired to 120 VAC. R1 is generally a metal oxide resistor which both adds a little filtering against line surges, and also acts as a fuse in the event C1 fails. A PTC resistor can perform the same function. The full-wave bridge of course can be made up of discrete diodes such as 1N4005. Just use a peak reverse voltage of at least 300V for 120VAC and 600V for 240VAC ( I prefer to use 600V and 1000V, respectively, just for safety ).

 

[Bright+]

C2 is the filter cap in that circuit. The 100 uF value shown means the current through the LEDs varies by about 2 mA in either direction from the mean of 20 mA ( I did a circuit simulation using CircuitMaker software ). In other words, about 10% ripple current, and no detectable flicker. In fact, you could probably reduce C2 by a factor of 5 and still not get noticeable flicker. I'm just anal about reducing ripple current.

So, what does that come out to be in terms of flicker index (in range of 0 to 1) ? My concern is not flicker as we can see on steady objects but on moving objects or use with cameras/camcorders.

Also, how are you going to address keeping THD below 15%, and power factor above 0.95? This is something that has been addressed in fluorescent ballasts except for the integral ballast in CFLs.

 

[jtr1962]

So, what does that come out to be in terms of flicker index (in range of 0 to 1) ? My concern is not flicker as we can see on steady objects but on moving objects or use with cameras/camcorders.

If the ripple is 10% then the LED output would vary by more or less that amount ( probably a little less ). You can arbitrarily reduce the ripple and flicker however low you want by simply increase the value of C2.

Also, how are you going to address keeping THD below 15%, and power factor above 0.95? This is something that has been addressed in fluorescent ballasts except for the integral ballast in CFLs.

I'm aware of that. This is a simply DIY ballast, not a commercial ballast where such things would be important. Also, it's probably possible to add a few passive components for power factor compensation and/or THD if you want, although I haven't really looked into it.

 

[LEDite]

Bright;

I built a 115VAC 9-Watt LED light several years ago with 56 lumens per watt.

It is still operating.

With today's LED's the efficiency would certainly be improved.

I am working on an 18 Watt Fluorescent replacement now.

Hi-Power LEDs are much more reliable IMO.

LEDite