How do white LEDs work?

[James Wilkinson Sword]

Yes, I've googled it, but there are many different approaches. Which one is used in domestic LED bulbs?

[Graham.]

Ultra-violet LED exciting a white phosphor, much the same principle used in florescents.

[James Wilkinson Sword]

That's what I would have assumed, but when you look at a switched off white LED, it's not white. I would have expected it to have a white coating that can be seen like on a switched off fluorescent tube.

[Graham.]

Perhaps the phosphor just surrounds the die, with a clear epoxy envelope surrounding it.

[Clive Arthur]

I thought a thousand people would leap on this, but it's a blue LED with a yellow phosphor. Maybe not always, but as near as makes no difference.

Cheers

[Martin Brown]

It is usually a high efficiency blue LED pump exciting a yellow phosphor with that mix determining the nominal colour temperature. It is quite peaky in the blue and more of a wide hump around the yellow. The phosphor usually looks yellow and sits on top of the LED.

The visible light flux out of an LED die on a high efficiency LED these days is about the same order of magnitude as the sun's photosphere.

It is a lot more obvious on the devices which use a remote phosphor.

[James Wilkinson Sword]

That's not as even as I thought. I thought they'd made it similar to the sun, which looks like this:

Maybe we should be using the controllable colour RGB ones.

[Martin Brown]

They would be even worse at approximating the solar spectrum.

RGB LED emissions have a fwhm of 50nm so would have a very peaky spectrum. It doesn't normally matter except for colour matching or where you have unusually narrow band pigments.

Materials that have very different colour depending on the white light source you use are called after the semiprecious stone Alexandrite.

The other common one is neodymium doped glass used by glassblowers to see into a gas flame against the yellow-orange glare of sodium emission.

[James Wilkinson Sword]

Actually, they seem to be sold as RGBW, so they probably ave white LEDs too. But you can add a bit of R G B as necessary to make it closer to sunlight or whatever tone you prefer.

Clearly the best thing would be a phosphor mix that makes precisely the levels you get from sunlight, or many different wavelength LEDs which can be individually adjusted by the user to give the preferred output.

[Johnny B Good]

An interesting read (note the use of the expression "loosing efficiency" on page 8 regarding Fig 7a). Also of note is the copyright date of this document which is the year of their record breaking 303Lm/W led efficacy achievement in February/March of that year when the Cree Spokesperson let slip to the trade press that such laboratory achievements typically took a further 18 to 24 months of development before making their début on store shelves for public consumption.

A simple arithmetic calculation reveals that this scheduled development has slipped by a rather conservatively estimated 13 months. :-( The best Lm/W efficacy figures I've noted recently have been around the 125 to 145 Lm/W mark. The former being a 1500Lm 12W GLS B22 2700K Warm White lamp I sampled from a Home Bargains store for the princely sum of £2.99 which turned out to have an *actual* consumption figure of 14 watts.

Since its illumination power doesn't seem as impressively bright as I was expecting (even making allowances for the eye's logarithmic response to brightness), I rather doubt the additional 2 watts is hiking the claimed 1500Lm to 1750Lms as one might expect if the additional 2 watts was simply the result of overly wide tolerances in the 'electronic ballast circuit' causing the LEDs to be overdriven to a higher than designed Lumens output.

I have a sneaking suspicion that the claims of "1500 Lumens at 12W (100W eqv)" have been based on the best of a sampling of these lamps off the production line, possibly based on the maximum positive tolerance limit of the 'nominal' power consumption to boot for good measure (+10%? =

13.2W) so might more typically be nearer the 110Lm/W mark than to the claimed 125Lm/W figure. Even the claimed efficacy is not a particularly massive improvement over the 81Lm/W efficacy figures typical of many LED GLS lamps of just over three years ago.

The 145Lm/W lamps I saw were the grossly overpriced 1600Lm LES 11W examples being offered by Asda. I might have considered buying one if they'd been more sanely priced but at something like 10 to 18 quid a pop (I didn't bother trying to pin the confusing shelf price labelling down any tighter than that - it was enough to know that it was at least 3 times pricier than I'd been prepared to pay in Home and Bargain), I wasn't in the least bit tempted.

It's not so much the electrical consumption cost savings that interest me so much as the service life endurance promise standing a much better chance of being fulfilled outside of a laboratory test environment in the more demanding conditions typical of a domestic pendant light fitting complete with fancy draught excluding shade dangling in the warmest layers of air to be found in a room basking in the warmth of a centrally heated radiator or two.

I'm happy enough with the current crop of "60W 806Lm 120v 750 Hour life rated" American tungsten GLS lamp equivalent 9W LEDs where an 806 Lm lamp can provide the required illumination level (effectively replacing a 73W

240v 1000 hour tungsten filament GLS lamp in UK housing). It's all these 15 and 18 watt 1500Lm LED GLS replacements for the 100W tungsten filament GLS lamps with their more marginal temperature tolerance that give me pause in their deployment as a GLS alternative.

The LED version of the "100W GLS tungsten filament lamps now starting to appear would seem to be a viable GLS candidate if their claimed efficacies of 145 and 150 Lm per watt are based in reality rather than best hoped for efficacy.

It's been a rather disappointing wait for Cree to begin fulfilling that (probably ill advised) promise made by their spokesperson just over three years ago when they announced their record breaking achievement in LED efficacy. Here we are, some 50 percent further along than their upper timescale to get 300Lm/W lamps to market, with lamps of only half that efficacy to show for their efforts thus far.

Still, at least *some* progress has finally materialised at long last, so I suppose we ought to be grateful to finally be free of the 2013 'Time Warp' we seem to have been living in for the past 3 or 4 years. :-) Better late than never.

At this rate of development, we'll be lucky to see the next efficacy milestone of 200Lm/W being achieved within the next three years or so. Who knows? We may see a sudden spurt from Cree whereby the 200Lm/W milestone in commercially available lamps is reached within the next 12 months. Either that or else an admission that the 303Lm/W lab results were faked just to pressurise Philips Lighting into quitting the LED lamp business. :-(

[James Wilkinson Sword]

I thought the best LEDs were equivalent to 10W out per 1W in. Eg a 10W = LED bulb should be equivalent to a 100W incandescent. A 100W incandesce= nt is 1435 lumens. So your bulb is consuming 14W to give out about 105W= . Not as good as I thought.

Companies lie, they always do. Cameras state x MP, and if you take a ph= oto at full resolution, it's shit. Hard disks don't use gigabytes, but = billions of bytes, shaving off a little.

I'm currently using this sort of thing:

Might not be the best price, I just linked to the first one I found, not= where I bought them from. The LEDs are well spaced and they don't exceed body temperature, so they= don't fail like most LED bulbs.

They really should put sensible ratings on each LED bulb they sell. It = should clearly state the actual electrical consumption (it's been mentio= ned in this thread that they lie), and equivalent output (eg "=3D60W inc= andescent". The general public don't know what a bloody lumen is.

-- =

I took a vow of faith, I don't shoot any more. You don't shoot any less either.

-- Machete, film, 2010.

[Johnny B Good]

====snip====

I thought a quick follow up with some references would be in order.

If you google for "303Lm/W LED Cree" you'll be swamped by Cree's own web page hits where they seem to be trying to re-write history. However, I did manage to find an original trade press report:

Apologies for the line wrap.

Interestingly, the 'spokesperson' turned out to be no less a personage than Cree's vice president for product strategy, Mike Watson himself who was quoted as saying,

"While some of these performance improvements already impact our current developments, commercializing products with this level of performance typically takes between 18 to 24 months from when we announce a research and development result."

Having re-read that article, it is only now, with the benefit of hindsight, that I recognise the skilful use of Marketing Weasel Speak by Cree's VP for Product Strategy.

I'm guessing the hope was that their one and only competitor, Philips Lighting, would give up trying to win the race and drop out of the game (as they subsequently did), allowing Cree to just coast along and milk this promise of improved LED lamps for as long as the (technically ignorant) consumer masses would tolerate the tiny incremental improvements being drip fed into the market place.

With Philips Lighting gone, it's no wonder we seem to have been "Stuck in 2013" for the past four years as far as LED lighting products are concerned. I suppose we should eventually see 250 LPW lamps make an appearance since Cree will have no choice but to spend that margin between their best achieved 303LPW laboratory example and the current marketing period's "Best LPW offering to date".

Before anyone else jumps in with a best guess at when we're likely to see 250LPW LED GLS lamps finally make an appearance, I'll offer mine.

I reckon Cree could get away with a 20LPW improvement per marketing season which, given the size of the worldwide market they have exclusive access to, is likely to straddle a two year interval. Assuming the current LPW is now at the 150 mark, I reckon this gives Cree another decade's worth of high living before they're finally obliged to offer a

250LPW lamp sometime towards the end of the third decade of this current millenium (assuming Trump hasn't plunged the whole world into Nuclear Armageddon during the next 4 or 8 years of his presidency[1]). [1] In which case, I can foresee Cree diversifying into tallow and wick based lighting technology, always, of course, assuming enough of their board of directors survive "The Event" to keep the operation going as a vital part of the business of rebuilding a post apocalyptic society (and thereby get their snouts back into the trough).

[Brian Gaff]

As far as I am aware most approaches revolve around the doping of the part which emits light. Normally more than one led is used each giving off a fairly narrow range of colour, and then combining them to make a perceptually white light. I have also seen some that refine this a bit with internal filtering, but essentially, as far as I am aware, nobody has made a single led element that has a sufficiently wide band output to give white on its own. Brian

[Brian Gaff]

Yes that is true, but once again its because the bandwidth of an led is quite narrow. I feel this thread is a bit of a slight troll here as I think this very topic was discussed here not very many months ago, as regards lighting that could be set to different tints. Brian

[James Wilkinson Sword]

And of course I read every thread.

[James Wilkinson Sword]

With a few junctions and a few phosphors, I can't see why we can't make a real sun spectrum. All we have is cool white and warm white.

[NY]

We've just got a couple of Philips Hue bulbs which can be set to a variety of different colours (as well as being dimmable), presumably by varying the proportions of different coloured LEDs. It was my wife's idea and they are a bit of a gimmick. I tend to use the one in my study mainly on fairly neutral white (ranging between warm white and daylight) but my wife likes to set hers to lurid purple or red as background lighting when she reads on her Kindle (which produces its own light, so the colour of the room lighting doesn't affect the colour of the text). We got them partly to test the technology, controlling by mobile phone app or Alexa voice recognition; if we were to get any more we'd go for much cheaper fixed-colour ones.

What is interesting is that even the whitest light, which appears slightly blue to the eye, shows on a digital camera as being a little warmer than full sunlight (about 5000K). A "daylight" CFL alongside it (OK, I was curious so I did a comparison!) looks slightly warmer by eye but the camera sees it as being cooler than the LED.

This goes to show that the eye is a very poor indicator of colour temperature, and that it has a lot of auto-adjustment built into the eye/brain mechanism.

Of course, if the digital camera is set to auto-white-balance rather than a preset "tungsten", "fluorescent", "sunlight", "shade", it adjusts too.

Some time I'll have to try some test photos using the same Hue bulb on various colours and brightnesses, and see what the colour rendition of real subjects is like in various conditions.

When I've done it in the past, using sunlight, shade, tungsten bulb, warm white fluorescent tube, daylight CFL and daylight LED GU10, the camera's auto-white has made them all look fairly similar in terms of overall colour cast, but red objects tend to be a bit darker and less vibrant with CFL and LED.

I dread to think what the spectrum of some of these bulbs is like, but I bet there are a lot of holes in the spectrum compared with a black-body radiator like a tungsten bulb or the sun (ignoring very small gaps in sunlight due to absorbtion lines of the atmosphere). I dimmed a tungsten bulb from full brightness to barely lit, using a conventional thyristor dimmer, and the camera's auto-white and auto-exposure made all the test photos look pretty well identical, which is how it *ought* to be with LEDs and CFLs.

I presume LEDs vary their brightness by varying the duty cycle of a square wave (and maybe even varying its frequency too). I've always wondered why this doesn't causing any banding or beating when those lights are used as studio lights in a TV studio.

[Dave Plowman (News)]

Colour temperature to the eye is subjective. The real problem with many of this sort of light source is they ain't continuous or smooth over the visible light spectrum. Which can make colours - like paint - appear a different colour (or shade) than in daylight, or halogen.

This may not matter much in a domestic living setting, but certainly can in a workshop, etc. Or even a kitchen.

[NY]

Or if you are viewing certain precious stones which change colour dramatically under sunlight or tungsten or one of the many CFL or LED lights.

The other thing you have to be careful of is pulsed lights with very short persistence. I first learned this from my grandpa who had a lathe for making model engineering models. He had his garage workshop illuminated with fluorescent strip lights but he had a tungsten bulb that he could shine on the work. He showed me how important that light was by running the lathe a certain speeds which were an exact multiple of the mains frequency. Under the fluorescents along, the work appeared to be stationary (and therefore safe to touch); under the tungsten light you could see enough blur to make it obvious that it was spinning and therefore dangerous to touch. And he was catering for that moment of inattention; normally when you have your brain devoted to the task, it's blindingly obvious that if you can hear the motor, the chuck is spinning.

I gather that in situations where pulsed light (eg fluorescent or LED) is used as the only light in engineering works, they have circuitry which throws in occasional "extra random heartbeats" into the mains-fed lights, which is enough to give some blur or jitter on the work in the lathe to make it clear that it is spinning, even though dead-regular mains at 50 Hz would freeze it stroboscopically. I heard of this when someone was filming a video in an engineering works and got all sorts of flicker even though the camera was set to a flicker-free 50 Hz refresh. He had to get H&S to sanction temporarily disabling this safety feature during filming because it was noticeable even though most of the light came from the filming lights.

My wife has an LED desktop lamp with LEDs that are supposed to give better colour matching more like daylight or tungsten (ie with fewer peaks and troughs). But it does have an annoying side-effect. I don't know what frequency they pulse the LEDs at, but occasionally if you move your eyes rapidly from one thing to another you can see a trail of sharp images, especially if it's dimmed so there's probably more space and less mark in each cycle of the lights. It's like you see with some car tail lights. or with the "red/green man" signs on the pole of pedestrian lights - the large bright lights that you see from the opposite side of the crossing are fine, but the little telltale light on the pole beside the button has bad flicker that is visible out of the corner of your eye.

[James Wilkinson Sword]

Why do you hate CREE so much? They appear to be the ones at the forefront of LED technology.