A nanophotonic comeback for incandescent bulbs? | Hacker News Reader

A nanophotonic comeback for incandescent bulbs?

news.mit.edu

62 pointsskybrian42 comments

This development was discussed on HN about 3 days ago. On that thread I commented that a merit of the incandescent is its continuous output spectrum. The newer lamp technologies produce discontinuous spectral output, CFL having prominent green, and LED blue emissions.

The difference is reflected in the CRI (Color Rendering Index) where 100 represents natural daylight. Incandescent lamps are generally very close to 100, but most fluorescent and LED sources are much lower, 90 or less. CRI is crucially important in the visual arts (museum/gallery lighting, photography, etc.) and other industries (color matching).

There are fluorescent and recently LED lamps that have high CRI ratings. These are relatively expensive items different from the "bulbs" sold at the hardware store. For the future, "remote phosphor" LED lamps look promising, though likely there will be a market for improved incandescents as described in the article.

Looks like advancements in LED, and now incandescent, technology could produce lamps that provide the best of both worlds: high CRI along with significantly extended durability.

Note that due to how CRIs are calculated, you can have an arbitrarily bad color rendition even with any CRI <100 with discontinuous spectra: if your object reflects mostly on the bands where the spectrum is weak, it will appear black (or you could have a daylight e.g. red object appear e.g. blue or any other color). That's why incandescents are really the standard for color rendition.

High CRI white LEDs have been a thing for a while though. While not completely continuous the high CRI GE "reveal" lights are pretty good [1].

[1] http://www.cnet.com/news/shining-a-light-on-high-cri-led-bul...

I bought and returned a GE Reveal bulb. High CRI or not, it rendered colors terribly. Slate green turned purple. Bluish paint did something else nasty -- I forget exactly. It didn't look at all like incandescents, daylight, or other LEDs.

We returned it and tried a high CRI Cree lamp. It looked normal.

Interesting, we've got a mix, The recent ones we put in are all Cree, but I have some of the Reveal ones in my workshop (with the Hue bulbs I got). If you could point me at an image of something with the kind of green you're talking about I'd love to try it under the Reveals I've got.

A literal piece of green slate.

The Reveal in question was IIRC a 6" can downlight -- the kind that replaces the entire can trim.

Note that by definition incandescent bulbs have a CRI of 100.

The CRI is calculated by looking at the colors reflected by a particular set of 14 paint chips under the lamp being analyzed, and comparing it to either illuminant D65 (representing daylight) or to an incandescent bulb, depending on the lamp’s color temperature. The color from the paint chips undergo an outdated type of chromatic adaptation transform, and are then compared to the colors under the reference illuminant, using colorimetric distance in the outdated 1964 CIEUVW color space. To find the final CRI take the arithmetic mean of all the distances, scale it, and subtract from 100. Most of the choices in this process are somewhat arbitrary.

CRI numbers are an okay first approximation to judge how much a particular lamp will muck with color relationships, but when you get to the precise details it gets pretty arbitrary, and the spec is heavily gamed by lighting vendors, because it’s one of the main values they report. Take CRI numbers with a big pinch of salt, and try looking directly at a chart of emission spectrum if you can, or ideally get a bulb and try looking at various objects under it for better understanding.

I've studied lighting tech on and off since getting into planted aquariums in 2006 and remember that some researchers at Philips also lamented the issues with CRI as a metric of light quality and presented their own system which I remember as being a color wheel with a lot of (possibly 100+) marked points. For each point, they'd present a vector as the difference from "actual" (as represented by the appearance in midday sun on a clear day) for a light source, so you could easily see how much a particular light skewed particular colors with fairly fine granularity. I neglected to save the source and haven't been able to find it in subsequent searchings though.

I don't know what you consider high, but FYI Cree has some relatively new bulbs that you can buy for under $10 at Home Depot with a CRI of 93 [1]. I've found Cree bulbs to be excellent in general. Tested a bunch of brands, and my house is full of these now.

[1]: http://www.homedepot.com/p/Cree-TW-Series-60W-Equivalent-Sof...

Cool idea, but these sound difficult to manufacture efficiently, given that you need to grow precise layers of crystal on the filament. On the other hand, you know are living in the high-tech future when even your light bulbs require a clean room!

In the age of $0.35 16MHz ARM cores (I'm sure you can get them much cheaper, that's just what I found on digikey) a bit of process complexity might not be so prohibitive. What a time to be alive!

Why doesn't everything have a supercomputer (circa 1985) built into it? At $0.35, why not!?

Because thinking of uses for computers now costs at least 1,000x more than the computers themselves.

Because it's completely pointless in most cases. The "Internet of Shit" Twitter account[1] may be satire but it highlights this unnecessary use of technology.

1: https://twitter.com/internetofshit

Because a low voltage power supply for the micro is quite expensive generally, compared to everything else. Plus you're probably adding a PCB and more BOM parts. There are some ways to make cheap non isolated supplies, but then the microcontroller can't be wired to any external low voltage I/o.

Of course, if you already have a PCB and power supply, then it makes total sense and most products that meet this criteria are microcontroller based.

http://www.antipope.org/charlie/blog-static/2012/08/how-low-...

It takes a few decades for technology to filter through to applications. Charlie Stross had a go at mapping out some of the possibilities provided by cheap low power computing devices in the blog post above a few years ago.

We're getting there. Here's a light bulb with full RGB options and a built-in bluetooth speaker: http://www.playbulb.com/en/playbulb-color.html#meet-playbulb...

I've always wondered how efficient are traditional bulbs as an actual heat source? I live in Minnesota and its -17°F outside, and would it be more energy efficient to heat the room I am currently occupying with the light bulbs of the room, keeping the rest of the house cooler than pumping a higher level of heat to the entire house?

I feel like the switch to CFLs may have increased my energy bills overall.

They're somewhere in the vicinity of 100% efficient as heat producers. Even the energy that goes into producing light just decays into heat after bouncing off a couple of walls.

Less "efficient" than using a "heat pump," and possibly more expensive than using gas (I don't know anything about energy in Minnesota). If you're using electricity to heat your home, though, you might as well use something like SETI@home and put it to use :-)

Unfortunately you can't use a heat pump (at least a normal one) below about -8C (17F), so -17F would definitely be out.

I also don't know anything about energy in Minnesota, but here in BC, gas heating is still cheaper than electric, especially if you use an efficient gas furnace (that doesn't exhaust too much heat), but not as much as it used to be. (Technically though it will always be less "efficient" than electric due to that exhaust, but that's really comparing apples to oranges.)

That's true for air heat pumps, but many people (myself included) have ground water heat pump systems; either open- or closed-loop.

My unit is fairly old (mid 1980 Tetco) and provide all of our heating and cooling. It uses the most electricity in January/February for heating and July/August for cooling, but my total electric bill is usually never more than $200/mo and that included all of the home's "normal" electricity use.

If you mean lifecycle efficiency, then it depends on the fuel used to generate the electricity. If it's coal or natural gas then pretty inefficient to use electricity for resistive heating since the heat byproduct of combustion was lost at the generating power plant.

If the energy source is wind or solar, then it's very efficient as there is no heat loss in generation.

Both scenarios involve the inherent transmit losses of electricity distribution (15%?).

Natural gas has little transmission loss, and modern condensing furnaces reclaim much of the waste heat from the combustion exhaust. You can also zone a Nat gas heating system so it only heats the areas you want at any time.

Also heating with electrical resistance will be very expensive as the cost per unit of energy for electricity is much higher than natural gas.

I sure hope so. Incandescent bulbs are much more pleasant than any fluorescents and the vast majority of LEDs on the market.

Aren't LEDs and CFLs significantly more durable than incandescent? Certainly they're more durable than tungsten-filament incandescent bulbs. Hopefully they can do better than just parity with LED/CFL power consumption otherwise it wouldn't seem worthwhile.

What do you mean durable? If you leave an incandescent and a CFL on continuously, the CFL lasts much longer. If you take both and turn them on and off over and over repeatedly, the CFL doesn’t have the same advantage. The CFL is also much more expensive to produce, and much more difficult to dispose of.

This is a pretty reasonable list of complaints about CFL lamps. http://www.richsoil.com/CFL-fluorescent-light-bulbs.jsp

(Many of the other sites around the web complaining about CFLs are full of wacky exaggerations.)

CFLs and some LED lamps are very heat sensitive. Some are engineered to only operate in an upright position.

If you have fixtures other than table lamps, like wall sconces, CFLs usually suck.

Supposedly they last forever, but they don't seem to deliver on the supposed 10+ year lifespans -- I have no bulb older then 3 years in my house.

Out of 12 bulbs, I've had three of the CREE LED bulbs from Home Depot fail in a "disco strobe" mode in less than six months :(

I actually gave up and bought a bunch of incandescent bulbs at a clearance sale.

Even accounting for wasted energy, financially I'll be in the black for 5-7 years based on my experience with CFL/LED fail.

FYI I had that happen to one of my Cree bulbs. I didn't want to deal with returning a single bulb to HD, so I emailed Cree, and they just sent me a new bulb, no questions asked.

I don't know if my experience is atypical, but nearly every light I my house is a CFL. They have survived 4 moves in 7 years, with only one going out.

LEDs are great, but retrofitting them into the old bulb form-factor is the cause of most problems. The power supplies get too hot, the LEDs get too hot, and everything is generally cramped together with little room for cooling.

My LED bulbs work well (Philips), but have one flaw: they do not redden when you dim them. I would like the color temperature to be much lower than 2700K when they are dim- to more approximate an oil lamp.

Anyway, perhaps this work will lead to efficient incandescent bulbs that provide nicer light when dimmed.

Does this address the short lifespans of incandescent bulbs? Since the filament portion is essentially the same, wouldn't these new bulbs suffer from this same issue, and wouldn't that make them less efficient overall due to replacement costs and waste?

Interesting technology, and it'll be fun to see how they market the technology since people just started understanding that incandescent = lots of energy = money. I wonder if LED technology will be able to keep up.

They've developed a photonic crystal [1] which reflects IR while passing visible light. That should be useful. Using it to to try to reflect IR back to the filament of an incandescent bulb is probably not going to be cost-effective. Photonic crystals have detailed structure like an IC, with waveguides and cavities, and are made by similar methods. That's a lot to make just to get an IR mirror for a light bulb, since you'd need a large surface area of the material. Probably more expensive than fabbing LEDs.

This sounds like an interesting development overhyped by the PR staff. Again. See [2].

[1] http://ab-initio.mit.edu/photons/ [2] https://news.ycombinator.com/item?id=10916894

Nah, I bet this can be done with a 1D photonic crystal where they can get away with simply using alternating layers of uniform thin film thickness. I forget exactly but I'd think all that's needed is alternating dielectric constant materials with a periodicity ~700nm which isn't that crazy. More expensive than glass sure, but it's not like it'll need special etching to get microlenses.

FWIW, slide 21 from this presentation on the site you linked has the picture I have in mind since at that size scale the glass is effectively flat.

http://ab-initio.mit.edu/photons/tutorial/L1-bloch.pdf

Feel free to correct if this is oversimple, I am vaguely familiar with this stuff but it was by no means a speciality.

That sounds like an ordinary bandpass interference filter.[1] Those have film layers and only pass a specific wavelength of light. Unlike ordinary filters, the angle at which the light hits the filter affects the passband, because it's the spacing of the layers that makes it work. The receive side of LIDAR systems use those. But those absorb the light they reject; they don't reflect it, much.

[1] http://www.edmundoptics.com/optics/optical-filters/bandpass-...

Gotcha, that makes sense. I suppose most of it will come from a narrow solid angle so if it were perfectly spherical you'd be okay but that would only work for the first reflection unless it was all reflection and no emission...

It reminds me of the halogen infrared (HIR) bulbs used for automotive lighting. I'm not clear on what advantage a theoretical mass-produced photonic crystal bulb would offer for this application that HIR does not. Any thoughts?

It will be interesting; but I'd pay a little more to have that nice yellow incandescent light in my house without the whopping electricity bill to boot.

Reply on news.ycombinator.com