Does anyone else remove the translucent diffusers found on A19 style LED bulbs? The light cast is much nicer, guessing a 175 degree? angle of directly emitted light rather than the vague 'overcast sky' wider angle of light cast through the diffuser. Anyone?
Ditch the diffuser - LED bulbs?
- Thread starter PeterPolyol
- Start date
- Status
gathermewool
Site Donor 2023
Glass or plastic?
Stalk with individual LEDs or Edison-style?
Stalk with individual LEDs or Edison-style?
I'd be sure nothing can touch the LEDs, they're going to be hotter than heck...I used to melt clippies and even lead solder all the time while messing around with discrete power LEDs (I was overdriving them when the solder melted, though).
gathermewool
Site Donor 2023
The LED's produce very little heat; it's the drive circuitry that generates heat.
I accidentally dropped one of my new Philips LED bulbs and it shattered all about the base (almost all of the glass broke off, conveniently). I gently screwed it in to a socket and it worked perfectly. Unlike some of the older Cree bulbs I have, which are an array of LEDs circumferentially-mounted around a center stalk), the Philips bulbs has four Edison-style LED filament. This makes for a better omnidirectional light output.
For reference, I replaced 11 candelabra-style bulbs with Edison-style LED bulbs with clear glass. Unlike the CFL's they replaced (6 were CFL, 5 were incandescent), the LED's light output is omnidirectional, just like the incadenscents. They not only look cool, but they use 4W each and don't need a coating on the glass to block some of the output (which seems to be your concern.)
As an aside, the payback period for replacing the incandescent bulbs in our chandelier will be around a year. The outdoor light, which has three candelabra bulbs and is on for 8 hrs each night, will pay for itself in 2-3 years (replaced CFL with LED); likely sooner, since we occasionally leave the lights on longer than wanted or leave them on constantly while we're gone, so that we have plenty of light outside when coming home from a vacation.
Added benefit, especially for the outside lights - INSTANT on. Some cheaper and older LED lights have a delay, but the newer ones I just bought (Philips) are instant-on. I'm actually not sure why the previous owners had a ton of CFL's outside. When it was freezing or colder, it would take up to a minute to reach full brightness, worse if it was below zero!
I accidentally dropped one of my new Philips LED bulbs and it shattered all about the base (almost all of the glass broke off, conveniently). I gently screwed it in to a socket and it worked perfectly. Unlike some of the older Cree bulbs I have, which are an array of LEDs circumferentially-mounted around a center stalk), the Philips bulbs has four Edison-style LED filament. This makes for a better omnidirectional light output.
For reference, I replaced 11 candelabra-style bulbs with Edison-style LED bulbs with clear glass. Unlike the CFL's they replaced (6 were CFL, 5 were incandescent), the LED's light output is omnidirectional, just like the incadenscents. They not only look cool, but they use 4W each and don't need a coating on the glass to block some of the output (which seems to be your concern.)
As an aside, the payback period for replacing the incandescent bulbs in our chandelier will be around a year. The outdoor light, which has three candelabra bulbs and is on for 8 hrs each night, will pay for itself in 2-3 years (replaced CFL with LED); likely sooner, since we occasionally leave the lights on longer than wanted or leave them on constantly while we're gone, so that we have plenty of light outside when coming home from a vacation.
Added benefit, especially for the outside lights - INSTANT on. Some cheaper and older LED lights have a delay, but the newer ones I just bought (Philips) are instant-on. I'm actually not sure why the previous owners had a ton of CFL's outside. When it was freezing or colder, it would take up to a minute to reach full brightness, worse if it was below zero!
gathermewool, like these ones w/ plastic diffusers attached by sealant 
Virtus, duly noted. There's also about 55VDC explosed on some of them...definitely for the fixtures
Virtus, duly noted. There's also about 55VDC explosed on some of them...definitely for the fixtures
gathermewool
Site Donor 2023
I guess it depends on its use, then. I didn't realize those had a downward-firing LED, and not some sort of protrusion (e.g., stalk). In certain fixtures, the slightly greater number of lumens would be worth it, while others might shine an unattractive hot spot onto an enclosure/shade/whatever or cause a distracting, bright light that obnoxious.
Like I said, the new Philips bulbs I just got have LED filaments, which are fine to look at. The opaque glass coating doesn't seem to substantially change the color temperature, either. With that said, the one I broke is in an enclosure. I haven't thought to check for exposed voltage - looked to be sealed well enough upon immediate inspection.
Like I said, the new Philips bulbs I just got have LED filaments, which are fine to look at. The opaque glass coating doesn't seem to substantially change the color temperature, either. With that said, the one I broke is in an enclosure. I haven't thought to check for exposed voltage - looked to be sealed well enough upon immediate inspection.
Originally Posted By: gathermewool
The LED's produce very little heat; it's the drive circuitry that generates heat.
...
As somebody who has designed a number of LED drivers, I have to disagree with you here.
If you're at max brightness and you're losing more heat in your driver than your LEDs, you're doing something really wrong!
Should be talking about 90% or so efficiency in a well designed circuit at high output.
As I mentioned above, when I tested my circuits the LEDs could get hot enough to melt plastic on probes and even lead solder if I really pushed them.
If you want to argue that not that much of the power in the LEDs ends up as light...yeah, I think you're quite right there.
The LED's produce very little heat; it's the drive circuitry that generates heat.
...
As somebody who has designed a number of LED drivers, I have to disagree with you here.
If you're at max brightness and you're losing more heat in your driver than your LEDs, you're doing something really wrong!
Should be talking about 90% or so efficiency in a well designed circuit at high output.
As I mentioned above, when I tested my circuits the LEDs could get hot enough to melt plastic on probes and even lead solder if I really pushed them.
If you want to argue that not that much of the power in the LEDs ends up as light...yeah, I think you're quite right there.
gathermewool
Site Donor 2023
Hoe close are the circuits you built to what’s in consumer lights? I’m telling you, you can put your fingers on these LEDs and they barely get warm. It’s the driver’s heat sink that can be too hot to touch for long.
Virtus is right on that- LEDs, especially multi-watt, multi-chip 'beads' and COBs, at their rated voltage release a lot of heat and need to be mounted to a heatsink of some type so they don't burn out. The newer model LED chips, especially the ones clad in plastic composites, might be underdriven to control heat but they still need an 1/8-1/16" piece of aluminum to sink the heat into.
The so called 'filament' style LEDs are a strip of metal (originally glass) with a row of about 20?25? tiny LEDs mounted on them- with such a small amount of mass to back it, they'd HAVE to be very low power individual emitters.
Fun fact about filament style LEDs; they were originally a cool idea to increase the efficiency of the LED array, and that's why the emitters were originally mounted on glass rods. They called it chip-on-glass (as opposed to chip on board) and it was supposed to allow light emitted from the back of the chip as well as the front, supposedly saving wasted photon emission from heating the substrate instead.
But, in mass production, the Chinese LED manufacturers decided to mount them on strips of metal instead, and then coat the whole stick with phosphor resin. They are indeed nice to replace lower output candelabra bulbs where the filament is traditionally visible with the even distribution you like
Have put a few of them through the ringer too
Used them in an old school worklight, but dropping the light would always break the little stem inside that the LEDs mount to. Note the bent filaments mounted on metal, not glass. (still doesnt work bent though)
Here are poor representing pics of standard fixtures with short, diffuserless bulbs:
A double bulb unit, with one 3000K and one 5000K w/ frosted glass removed. Normally the bulb would be sticking out of the bottom.
Again, shorty bulb not sticking out of the bottom. What hits the frosted glass offers nice moderate omnidirectional scatter while the direct beam below gives a nice flood light (pics are sad and don't do justice)
The so called 'filament' style LEDs are a strip of metal (originally glass) with a row of about 20?25? tiny LEDs mounted on them- with such a small amount of mass to back it, they'd HAVE to be very low power individual emitters.
Fun fact about filament style LEDs; they were originally a cool idea to increase the efficiency of the LED array, and that's why the emitters were originally mounted on glass rods. They called it chip-on-glass (as opposed to chip on board) and it was supposed to allow light emitted from the back of the chip as well as the front, supposedly saving wasted photon emission from heating the substrate instead.
But, in mass production, the Chinese LED manufacturers decided to mount them on strips of metal instead, and then coat the whole stick with phosphor resin. They are indeed nice to replace lower output candelabra bulbs where the filament is traditionally visible with the even distribution you like
Have put a few of them through the ringer too
Here are poor representing pics of standard fixtures with short, diffuserless bulbs:
A double bulb unit, with one 3000K and one 5000K w/ frosted glass removed. Normally the bulb would be sticking out of the bottom.
Again, shorty bulb not sticking out of the bottom. What hits the frosted glass offers nice moderate omnidirectional scatter while the direct beam below gives a nice flood light (pics are sad and don't do justice)
Originally Posted By: gathermewool
Hoe close are the circuits you built to what’s in consumer lights? I’m telling you, you can put your fingers on these LEDs and they barely get warm. It’s the driver’s heat sink that can be too hot to touch for long.
I worked on automotive stuff back then, so DC/DC instead of AC/DC...doing offline conversion is going to tend to lower efficiency somewhat, but I still see offline parts claiming 90%. Edison LED filaments were not around at that time as far as I know, and I had to test my part to over 2A and only had 300mA LEDs to use in our lab...yeah, I fried a bunch of those things. Still, even at 300mA I was careful not to touch them. Putting the LEDs in parallel didn't really help as the forward voltages varied a lot and adding ballasting resistors was not going to match what the customer was going to do...this is why using series strings is very popular in LED applications and also why parallel applications usually use a number of current matched parallel drivers (expensive). The light output of a given type of LED correlates very strongly to its current while the LED voltages at that current can easily vary +/-10% even within LEDs from the same batch...combine that with a very nonlinear V/I curve for these diodes and it turns out driving them in parallel is a real PITA.
Surface area is your friend when trying to get rid of heat, so I can see a number of long filaments doing a good job of taking heat away from the LEDs while the driver is going to be a point heat source that might get a small heatsink cooking pretty good.
I believe you when you tell me that the LEDs themselves aren't hot, but if that driver is dissipating more than 20% of the total power in the circuit something is really wrong...even 80% efficiency is really not too good. The LED light output is a portion of their total power dissipation, but really not very much...it's still mostly heat.
Hoe close are the circuits you built to what’s in consumer lights? I’m telling you, you can put your fingers on these LEDs and they barely get warm. It’s the driver’s heat sink that can be too hot to touch for long.
I worked on automotive stuff back then, so DC/DC instead of AC/DC...doing offline conversion is going to tend to lower efficiency somewhat, but I still see offline parts claiming 90%. Edison LED filaments were not around at that time as far as I know, and I had to test my part to over 2A and only had 300mA LEDs to use in our lab...yeah, I fried a bunch of those things. Still, even at 300mA I was careful not to touch them. Putting the LEDs in parallel didn't really help as the forward voltages varied a lot and adding ballasting resistors was not going to match what the customer was going to do...this is why using series strings is very popular in LED applications and also why parallel applications usually use a number of current matched parallel drivers (expensive). The light output of a given type of LED correlates very strongly to its current while the LED voltages at that current can easily vary +/-10% even within LEDs from the same batch...combine that with a very nonlinear V/I curve for these diodes and it turns out driving them in parallel is a real PITA.
Surface area is your friend when trying to get rid of heat, so I can see a number of long filaments doing a good job of taking heat away from the LEDs while the driver is going to be a point heat source that might get a small heatsink cooking pretty good.
I believe you when you tell me that the LEDs themselves aren't hot, but if that driver is dissipating more than 20% of the total power in the circuit something is really wrong...even 80% efficiency is really not too good. The LED light output is a portion of their total power dissipation, but really not very much...it's still mostly heat.
gathermewool
Site Donor 2023
Thanks for the info, guys. Great information!
I've used a Belkin watt-meter to check the load drawn by samples of each bulb and they appear to be accurately advertised (within half a watt +/- error). Regarding heat dissipation vs lumen output, it appears that some are just better-designed than others. The base of the old (original) Cree LED bulbs (the ones with the radial cooling fins surrounding the base), as well as the cheap LED's we got for free during a home energy audit get hot enough to be uncomfortable to touch.
The new Philips LED's (with the resin/filament-style) look identical to the old-style incandescent bulbs they replaced. There's no perceivable base (just like the candelabra-style with the same resin/filament LED's) and nothing on the bulb is more than just warm, even after being on for an extended period of time.
I've used a Belkin watt-meter to check the load drawn by samples of each bulb and they appear to be accurately advertised (within half a watt +/- error). Regarding heat dissipation vs lumen output, it appears that some are just better-designed than others. The base of the old (original) Cree LED bulbs (the ones with the radial cooling fins surrounding the base), as well as the cheap LED's we got for free during a home energy audit get hot enough to be uncomfortable to touch.
The new Philips LED's (with the resin/filament-style) look identical to the old-style incandescent bulbs they replaced. There's no perceivable base (just like the candelabra-style with the same resin/filament LED's) and nothing on the bulb is more than just warm, even after being on for an extended period of time.
I'll have to check out the new Philips LEDs, thanks for the info! I have some glass bulb LEDs that look a lot like old style bulbs when on, but they are not Philips and I'm not sure what the LED configuration is inside. I bought some small base bulbs with the LED filaments for my daughter's study room lamp, but they oscillated badly when dimmed (despite being advertised as dimmable) and I stuffed them away somewhere and am not sure if I could easily find them now.
- Status
Similar threads
- Locked