I can see this newish technology being mostly used only in classical/vintage incandescent-looking "light bulbs". But I think it has got potential to be used also in other applications thanks to the very high efficiency of LED filaments. Are there rumours about that?

F36T8 LED filament replacement - such tube filled with a cooling gas could have (thanks to its large surface) some power - 20W or even 30W?

Lightweight streetlight fixtures of bigger sizes could also make use of these filaments, even more efficient thanks to less phosphor used. Could also be less glary than LED lanterns with very bright spots.

. . .

There are ones shaped like ordinary incandescents (still with the same filament arrangement in a normal A shaped bulb) - Those are readily available, at least here

i seen here pics of HID shaped lamps with correctly positioned "arc tube". Neat, but the advantage is unclear to me as those are less efficient than the HID lamps they aim to replace anyway

With FL consider the place in the tube : There is a LOT of it lengthwise. If the entire tube contains filaments end to end, the filaments are to be driven at very low power each, or else it will be way too bright. LED manufacturers dont like to uunderdrive LEDs, they like to drive them the hardest they can so they can use less LEDs. The alternative option would be lowering the density of the LEDs in the filament or add space between filaments along the tube, but then we are back to a tube lit in spots. You can try to diffuse it, but that would have efficacy penalty

There are ones shaped like ordinary incandescents (still with the same filament arrangement in a normal A shaped bulb) - Those are readily available, at least here

Haven't seen them here yet. These must be very low power ones or the filament used is different (denser) to the usual arrangement.

i seen here pics of HID shaped lamps with correctly positioned "arc tube". Neat, but the advantage is unclear to me as those are less efficient than the HID lamps they aim to replace anyway

Would be interested to see a picture. I doubt the compact size of a HID lamp (esp. a tubular one) would be enough to dissipate the heat. Only as highly energy saving replacements to outskirts etc. where you can sacrifice the lighting level and go down like from 8000lm to 1000lm.
I was rather thinking of dedicated fixtures for bigger filament lamps that could be made lightweight thanks to no ballast needed. The driver would be a part of the lamp, of course.

With FL consider the place in the tube : There is a LOT of it lengthwise. If the entire tube contains filaments end to end, the filaments are to be driven at very low power each, or else it will be way too bright. LED manufacturers dont like to uunderdrive LEDs, they like to drive them the hardest they can so they can use less LEDs. The alternative option would be lowering the density of the LEDs in the filament or add space between filaments along the tube, but then we are back to a tube lit in spots. You can try to diffuse it, but that would have efficacy penalty

Classical LED tubes light also in spots. This isn't probably a big issue for most people. The less dense filament might enable retrofit tubes to last much longer than those classical retrofits that are infamous for their pathetically short lives.

This is the selection available from Eurolux. They are a local brand of Chinese imports, but in the CFLs they proven to be good quality since their start somewhere in the early/mid 2000s

http://www.eurolux.co.il/page_14179



The HID ones used pretty thick and long stick. I think it would still be realistic ro get 2000Lm-ish with that (at 20W power input)



The spot distribution of tube LED lights screws the optics in most applications where they are used - High glare when the lamp is exposed, missing the reflector where the lamp is in a parabolic, and so on. This is big disadvantage

The efficacy advantage comes only when it is supposed to replace some "classic" lamps, where the only larger surface available for cooling is the bulb itself, which has to pass the light as well. For purpose made LED fixtures it is not that convenient to mess up with the gas fill or so when you need to redirect the light on one side anyway. There the "classical" LED assembly, with the front side emitting light and the back side taking care of the heat, is way simpler and performs even better (the light loss of the reflective surface behind an "A" bulb are way greater than the extra losses of a "classical" LED compare to the filament style).
And the fact you do not need that compact ballast is just a plus - regardless, which ballast concept you want to use. Plus without the size restriction, you have way greater selection of these, so are able to select the way better suited type for the task...

The one sided distribution of non filament LED sources is not an advantage

With light sources lighting up at reflector and then the light being reflected back, the shape of the reflector determines the wanted light distribution, and the depth of the light source in the reflector controls the glare, down to near zero with good parabolic systems (including Fluorescent, Incandescent and so on)

With LEDs the control over the forward light cannot be dont with a large reflector in most designs, so it is down to either uncontrolled flat surface distribution (cos angle from normal), exposed emitter distribution (same formula, but now with extreme glare of single LED emitters in plain view), or controlled by a lens panel (as done in most road lanterns) which controls only the distribution, but not the glare

Filament lamps are among the best form factors for LEDs. As they use the proper optics allready built into the lantern, instead of trying to defeat them

Most of the "drop ceiling" lights are deszigned with either directly a diffuse bottom or with reflectors arranged so it forms the same distribution (for the fluorescents just with lower light losses than the basic diffuser). Indeed, some feed more light to the side lobes, to yield really uniform illumination, but those are not that frequent.
And regarding the optic: Designing the refractor so it does not pose glare is just matter of doing the job right. Mainly with the indoor installation (fixture spacing less than their height above the illuminated surface) it is not so difficult to design the refractor so, the light from the single chip reaches each point on the "target" by many equally bright paths (so each LED becomes a row or array of points no more than mm apart). Mainly in the stripe format that is quite easy (LED's of one segment in one row, refractor formed of 10's of cylindrical lenses, each "micro lens" segment forming the same beam pattern as the complete fixture - that way the stripe glows across it's complete surface; yet it is quite cheap to make as a single extruded piece, just cut to fixture length). Such design is way cheaper than messing with the filaments.
Other possibility (not that optimal, but better blending with the remaining fluorescents) is to cover the row of LED's by a half tube diffuser anf then add reflectors pretending it is a fluorescent fixture. It gives the same light pattern as the fluorescent, it even looks very alike (unless you really pay attention to it, you won't even notice it is not a fluorescent like many others around), yet it still is nothing more than everything placed on top of a large flat heat sink (the ballasts use to be between the "individual tubes", so the complete fixture is about half the height of the fluorescent "original") and there is no need for any special gas fill, so no messing up with fragile materials like glass or so.

And the filaments itself do not help either - alone they are way too intense light sources as well, so need the spreading too. And for that to utilize the light from all directions is practically not possible (it is not possible with fluorescents either, but there is no other choice).
So you use higher efficacy LED's, but have more losses in the optics (multiple passes through the diffusing tube), so at the end you are not better than with the "classic LED" concept, yet have to mess up with the fragile glass tubes and gas fill.
At the end the same performance, but more fragile and thicker fixture (so more problematic when it is supposed to be surface mount).

All LED lanterns feed more into the side lobes

With the flat panel, there is no optic control at all, so the light distribution is a flat panel type

With the reflectors (so all the Fluorescent imitations with only the bottom half of the tube lighting), the amount of light controlled by the reflector is lower than with Fluorescents :

In Fluorescents, what escapes the lantern directly between the louvers is spread at the angle it went (i think on the order of 90 deg of the lamp perimeter), the rest hits the reflector on its way (so 270 deg of the lamp perimeter) and is redirected, mostly downwards or at narrow angle side lobes which are still not far from the lantern

In LED imitation, what escapes the lantern directly between the louvers is most of the light. So the reflectors play a role in hiding the lamp from view at wide angle, but the distribution still is wider as less light ever hits the reflector and gets redirected

Uniform lighting levels are much less important than they are thought to be. Our vision deals well with non uniform lighting. But light hitting straight from lamp to eye is where the problem is, and there most LED lighting systems do very poor job



Filaments are good for m factor exactly due to this - They emit light in all dorections, and allow better optics of the lantern, which are much bigger than the lamp, to deal with controlling the light, which they do way better

I dont think it is interesting to replace Fluorescents with LEDs anyway. Present day Fluorescents (T5HE, top grade ballasts and optics) are as efficient, but they provide good optical control and good spectral output as basic features, while the LEDs not

In the case of Incandescents, the LED filament lamp, with size and distribution similar to a real Incandescent, would get the same optical performance as a real incandescent in a lantern with good optics made for incandescent (parabolic spotlights and such). Now that is business



And i dont see why being made of glass and having glass fill is bad. Filament LED lamps have good optics and efficacy compared to most ofther LED products, and we want lamps that perform well, not ones that senselessly throw Lumens in wrong directions. And up untill now we were ok with gas filled glass lamps, what changed ?

All LED lanterns feed more into the side lobes

With the flat panel, there is no optic control at all, so the light distribution is a flat panel type

With the reflectors (so all the Fluorescent imitations with only the bottom half of the tube lighting), the amount of light controlled by the reflector is lower than with Fluorescents :

In Fluorescents, what escapes the lantern directly between the louvers is spread at the angle it went (i think on the order of 90 deg of the lamp perimeter), the rest hits the reflector on its way (so 270 deg of the lamp perimeter) and is redirected, mostly downwards or at narrow angle side lobes which are still not far from the lantern

In LED imitation, what escapes the lantern directly between the louvers is most of the light. So the reflectors play a role in hiding the lamp from view at wide angle, but the distribution still is wider as less light ever hits the reflector and gets redirected

Uniform lighting levels are much less important than they are thought to be. Our vision deals well with non uniform lighting. But light hitting straight from lamp to eye is where the problem is, and there most LED lighting systems do very poor job



Filaments are good for m factor exactly due to this - They emit light in all dorections, and allow better optics of the lantern, which are much bigger than the lamp, to deal with controlling the light, which they do way better

I dont think it is interesting to replace Fluorescents with LEDs anyway. Present day Fluorescents (T5HE, top grade ballasts and optics) are as efficient, but they provide good optical control and good spectral output as basic features, while the LEDs not

In the case of Incandescents, the LED filament lamp, with size and distribution similar to a real Incandescent, would get the same optical performance as a real incandescent in a lantern with good optics made for incandescent (parabolic spotlights and such). Now that is business



And i dont see why being made of glass and having glass fill is bad. Filament LED lamps have good optics and efficacy compared to most ofther LED products, and we want lamps that perform well, not ones that senselessly throw Lumens in wrong directions. And up untill now we were ok with gas filled glass lamps, what changed ?

Maybe people think that gas filled lamps have bad optics.

@WattMiser:
The thing is:
By a refractor you may efficiently shape only the light going about in the same direction as the target. The light going to the opposite direction you can not control by that. So the refractor has to be placed between the light source and the target.
By a reflector you may efficiently shape only the light going about the opposite direction than the target, so the reflector has to be behind the lamp from the target. So it can not shape the light going straight from the lamp to the target.
 With a lamp emitting the light only into half-space you may easily arrange the system either so all the light gets shaped by the refractor, so make the design so all the light gets reflected by the reflector. In either case you may shape all the light emitted by the light source, so have most freedom to shape it to get any light coverage pattern you need.

With a light source emitting all directions you have a problem: For part of it you would need a refractor, for other part a reflector. But you can not have both at once, one block or scatters light of the other one.
So you may have either beam shaping refractor and let the reflected light to just be scattered (so form a nondirectional beam) then the reflected light won't be much shaped (so will form a kind of "base" beam), the refractor then has to be engineered so it just complements that base beam.
Or use all the shaping elements in the reflector and use no refractor at all. Then the direct light then forms the "base" and the reflected then has to supplement it.
In both cases it means more complex beam shaping and that means higher losses (light, which is obstructed or scattered completely away). And these light losses are losses, so worsen the efficiency. Not saying it is not usable or so, with light sources emitting light into all direction you just can not go better.
It just means from 100lm emitted from a "classic" LED you may easily (so with no scattering and other principal losses) redirect all of the light into virtually any beam shape you want, so may get all the 100lm into the useful beam.
With a 100lm filament you will be able to redirect "just" about 80lm into the useful beam (bright spots with uneven illumination I count as losses too). So to get the 100lm into the beam, you need 120lm light source.
Well, there is one thing not completely fair: There is no light generation method sending light only into one half space or so. Neither in the "classic" LED's. What is there instead is a reflective layer behind the junction, redirecting the back light forward. Of course, that means some light losses (with modern LED design about 20% or so). So the "100lm" LED I have mentionned as the example means, the junction itself emits 120lm (count after phosphor conversion), the 20lm then get lost straight inside of the LED.
On the other hand the "filament" design uses no reflector inside of the LED, so the "120lm" LED in the example means the LED does emit the 120lm, but the optic afterwards looses the 20lm.
So the net result is practically the same. Why is this important: The LED rating counts all the light going out of the LED, so with essentially the same LED, the filament appears 20% better in the catalog, although the useful system efficacy is at the end exactly the same. You just should remember the "120lm/W" filament is not better, but exactly the same than the "100lm/W" classic, when you intend to use optic really well designed specific (for each light source type) optics.

@Ash:
What appears to be flat panel may well be accurately engineered optic (cylindrical Fressnell lens type - so flat panel, but accurately reshaping the light), if you look closer. Of course, "may be" does not equal "always is", it depends if the maker cares or not. What I describe is assuming the maker does care, although that does not represent majority of the cheepeese market offerings.
Plus that remark is not limited to just the classic LED's, but applies for the filament type as well (if someone does not want to bother with accurate optics, the light won't be that optimal, so for the given minimum level you will need more lumens, so more power).

What is wrong with gas filled lamps? They have to be made of something, which seals the gas in and at the same time is transparent. Because the gas has to be lightweight, the only known material to suit that is glass. But the glass is rather fragile material, mainly when talking about larger things, so complicating the logistic.
So why to bother with gas filling, when the same performance of an appropriately designed fixture you may get from the "classic" concept using tough metal and plastic parts (try to bend glass sheet compare to plastic sheet).

And why "the filaments were so good and now they aren't...":
Here we are speaking about two completely different applications, so no wonder concept ideally working for one is not that well suited for other.
If the application is "an incandescent retrofit", the gas filled filament is indeed the only one really 100% compatible with the incandescent - beam pattern, look,... There really I do not know any better concept, when the universal compatibility with existing "incandescent standard" is the goal, there the gas fill and glass envelope is common, so not much complication for the user, mainly when no other method is able to achieve that goal.
But the title here says "Except retro light bulbs", so I assume special fixtures designed from scratch on purpose for the given light source, without any limitation to any compatibility with any existing technology or so. Just fulfill the final task as best as possible. And there the gas fill is just a complication, when the same overall performance is achievable without that fragile glass or so.

The thing is, with real incandescents, fluorescents, nor HID's you had no choice than design the optic around 360degrees omnidirectional light source, otherwise big part of the light will be just lost. So the fixtures we know are looking how they are, because they are just (more or less) optimized for these all direction light sources. But with LED's we have the choice.

That is one of the big reasons, why I think the "LED fluorescent retrofits" are a complete nonsense. It does not bring any significant efficacy improvements (unlike the incandescent retrofits), but it is expensive like hell and brings many performance issues (mainly reliability, but as well poor optical matching). Of course, for a fluorescent retrofit the filaments will most likely work better than the classic components, but what is the point, when at the same cost as the retrofit tubes you get the complete, way better performing purpose made LED fixtures.

Wouldn't it be cheaper to make them out of plastic and metal, as well as eliminating gas filling? I read somewhere on candlepowerfoums that the raw materials used to make the above costs even less than something as simple as an incandescent GLS.

The gas filling is to eliminate the metal heatsinking so the gas filling cools the LEDs instead making the lamp looks exactly like an incandescent lamp and allow 360deg light distrubution. As a result, the outer heat more, as the heat from the LEDs, convected from the plastic to the outer so it needs to be made from glass. I think that LED filament lamp, is better solution than the traditional LED lamps with heatsinking.

The reason why glass is needed is not the thermal properties of the glass vs plastic (many plastics are way better than glass in this aspect), but practically only the fact it has to hold the gas inside. Because of the small atoms needed to have efficient cooling gas, you need some material without gaps. And there all plastics are practically like paper for the air - the gas passes through. For the LED cooling itself the best gas would be Hydrogen the second best Helium. But Helium is so small (beside it's cost), even glass can not hold it, it leaks through.
And as Dor wrote, the small and light atom gas fill is there essential to transfer the heat from the filaments to the bulb surface. Without that you will have to occupy big part of the bulb surface by some heat sink, so reduce the beam angle and at the same time have worse cooling.

But if you do not need the light to go all around (unlike with the retrofit bulbs, that is the case for the complete light fixtures), then indeed using the back side as a heat sink and then use just plastic lenses on the front side is easier and less fragile. That is the main background about the difference of "filaments" suitability for one vs other use.
Even for many specific uses inside a "incandescent" fixtures a LED emitting only to one side may perform better (when properly aimed) compare to the "filament" bulb, but that is specific for those cases and the lamp has to be properly adjusted for that. And for that is essential, the one installing it knows exactly what he is doing and is able to check the light pattern and asses the suitability of such replacement, as he is in fact altering the optical properties. So no direct replacement at all (with incandescent you just screw it in and it works), the filament style makes really a direct replacement with well predictable performance, no different from the incandescent.