20

Filament LED bulbs use very narrow strips containing the LED, which look similar to filaments of an old incandescent bulb:

They look like a normal bulb with large "old-style" coal filaments, and are of about the same size and shape as the common standart E27 socket bulb.

filament bulb switched off, on wood - 2

There are two basic variants of these bulbs:

  • One "plain" as described, looking impressively similar to a classic bulb
  • The other has a white coating on the glass, about 15mm wide next to the metal socket.

How are these filaments build? I could imagine the it's "pretty normal" LEDs, but lots of them, directly bonded to something? (There are references to COB)

How does the heat management work? Ii have seen speculations about filling the bulb with either helium or "some patented organic gas" - could that be enough for cooling?

How does the driver - which seems to be very small - work?

In the variant with white coating - hiding some components, I assume - is already not that much space, and in the plain variant, there seems to be almost no space for something nontrivial.

Below you see some bulbs of the white coating variant, switched on:

three filament bulbs switched on
(Images with the permission of this supplier at Alibaba)


Note how the LED filament bulbs are not so much similar to the tungsten filament bulbs recently being phased out - they are more resembling the long outdated carbon filament bulbs:

Real retro - carbon fiber bulb
bd® Antike Edison 220V-240V 40W Retro Vintage Industry Style Deko Glühbirne (kl)

Volker Siegel
  • 1,072
  • 1
  • 12
  • 25
  • They look nice! – Andy aka Apr 16 '14 at 21:08
  • 1
    @AndyakaYes, I've not seen them live, but they look certainly impressive! (Confirmed by people who have them live) – Volker Siegel Apr 16 '14 at 21:23
  • @GRTech (forgot to notify you) Hmm... did not find a relation to LED bulbs when skipping through - in which part should I look? – Volker Siegel Apr 17 '14 at 02:20
  • 2
    Whatever is the driver, I warry for the power factor of this. Then about cooling efficiency and various noises. Look here https://www.google.com/patents/US20110298375 – GR Tech Apr 17 '14 at 05:42
  • @GRTech Based on the "keine Elektronik und somit langlebiger", it's probably a resistor dropper. Anyway, at 3W, who cares about PF? – Spehro Pefhany Apr 17 '14 at 12:33
  • @SpehroPefhany To me, that sentence sounded like "we left a little too much out, and ask marketing department to handle it"... But what is a resistor droppen? – Volker Siegel Apr 17 '14 at 12:37
  • @VolkerSiegel "No electronics" so just a resistor in series with the LEDs, rather than a SMPS like in a 13W bulb. My guess. – Spehro Pefhany Apr 17 '14 at 12:44
  • That would be a diode rectifier, a resistor, and a long string of small led dies bonded directly on some stick? Possibly some of these resistor plus LED strings in paralle, of course. – Volker Siegel Apr 17 '14 at 12:51
  • I remembrt somebody complaining about flicker, with certain models - and we'd expect this variant to flicker at 50/60 Hz, even in a perfect way, right? So, gladly there are other models... – Volker Siegel Apr 17 '14 at 12:55
  • @VolkerSiegel Maybe not even a rectifier, just back-to-back strings of LEDs. The position of shadows would then flicker at 25Hz/30Hz. Might not be good to read by (only 3W anyway). – Spehro Pefhany Apr 17 '14 at 13:02
  • @SpehroPefhany Ah! Got it, brilliant :) ...ok, after genuinely being impressed by this perfect minimalism in implementation, I propose we ignore this variant - seems to be unuseable mostly, and I know there are models that actually try to solve the interesting problems :) – Volker Siegel Apr 17 '14 at 13:07
  • @SpehroPefhany your view to use just one of this 3W device in earth it is a strict laboratory approach. But what will be happens if billion souch lamps used around the globe? To my perspective, the promissory LED lighting technology should be accompanied with a significant reduction of power supply losses. So please announce to other member of the community that this will be our aim: LESS POWER – GR Tech Apr 17 '14 at 14:23
  • @Andyaka Yes, and looks like we may get pictures soon! I asked a Chinese wholesale dealer, who has good detailed pictures of the bulbs in the catalogue; He liked the idea, I explained the license stuff, - and he may post some of them to as answer here, would be cool. – Volker Siegel Apr 18 '14 at 18:49
  • @SpehroPefhany If there are two antiparallel strings of LEDs, the flicker rate would be twice the line frequency, not half. So, 100 or 120 Hz, which isn't too bad. – Anindo Ghosh May 13 '14 at 05:46

3 Answers3

26

These LEDs are not a single radial die, they are made with a transparent substrate with many LED dies in series (probably 25) placed on it. The whole thing is then coated in phosphor. The light isn't completely uniform but it's good enough. led filament cutaway There's a spec sheet here: http://www.runlite.cn/en/product-detail-145.html

Marcel Dejean
  • 261
  • 3
  • 4
  • There seems to be multiple ways of constructing these, I wouldn't discount the radial option. – Passerby Sep 24 '14 at 16:59
  • @Passerby do you have at least one example of a radial construction? I cannot find any example. – FarO Sep 04 '18 at 11:41
13

I purchased one of these lamps in Italy: 3W/300lumens/9,00euros

I took some pictures with very low exposition to higlights multiple "dot leds":

Detail 1

Detail 2

Box

I can count 28 dot-leds per each strip and 4 strips. This gives 112 dot-leds.

Other numbers: 100 lumen/watt 2.68 lumen/dotled 75 lumen/strip 0,03 euro/lumen

Assuming half the cost is for the electronics, we could assume, for easier calculations, 4,00 euros per 4 filaments, hence 1 euro per 28-dots/75-lumen strip.

jumpjack
  • 850
  • 2
  • 12
  • 20
2

To understand how these devices function, it helps to understand how traditional LEDs function. An LED is a Light Emitting Diode, so basically you can think of it as a simple PN junction that would be used in a diode (although in actuality the structure of LEDs is more complicated, often a double heterojunction structure).

led device structure
(source: gsu.edu)

Based on images of a similar LED "filament" bulb on Amazon image1, image2
closeup
It seems that each "filament" is a single LED with a radial structure, that is using radial layers rather than the traditional planar layers. So the core layer (cathode for example) would be composed of a conductor/metal, then the next outer layer would be n-type material, then the next outer layer would be p-type material, and finally the outermost layer would be a transparent conductor (to let light pass through it) such as ITO (Indium-Tin-Oxide).

It looks as though the LEDs in the bulb are arranged as two lines in parallel, each with two LEDs in series.

The driver would depend on the specs of the particular LED "filaments", but a driver could be as simple as diode bridge if the LEDs have a high enough forward voltage.

Because these LED "filaments" are so long and thin, they have a large amount of surface-area per volume, so that the heat is more dissipated than in a traditional LED.

wgardiner
  • 339
  • 1
  • 2
  • 1
    Can confirm from having broken open one of these, that these LED assemblies seem to get rid of heat far better than strings of conventional planar LEDs. – Anindo Ghosh May 13 '14 at 05:49
  • 1
    Heat management by gas... Helium is indeed a good candidate, with its high thermal conductivity (18 times higher than air). But is that sufficient to keep those LEDs cool enough? Any other suggestion? –  Aug 18 '14 at 10:09
  • 2
    Implementing multilayer radial LED structure with today's materials is nearly impossible. GaN-based materials are just waaaay to tricky even for planar substrates. – BarsMonster Jan 08 '15 at 11:17
  • 3
    "it seems that each filament is a single LED..." -- strongly disagree. For a link to a company that actually manufactures these, see https://ledlam.co.uk/how-do-led-filament-work – fearless_fool Nov 25 '17 at 01:01
  • 1
    While I appreciate trying to reverse engineer the structure based on the image, the conclusions are plain wrong. I haven't seen any LED "filament" as described. They ALL are rows of very small LED chips in a row either on a glass substrate (worse heat dissipation, but transparent all around) or metallic strip (see image above). No radial structure whatsoever. – FarO Sep 04 '18 at 11:39
  • 3
    I also have had several of these and can confirm that they get rid of heat much better than the string assortment LEDs. Also can confirm about the drive. For my bulb the driver was a bridge rectifier, to see how they work I have found this ..https://www.derf.com/how-a-bridge-rectifier-works-step-by-step-tutorial/ – Chipresistor122 Nov 26 '20 at 00:47