Hacker News Reader

Comment by gumby | Hacker News Reader

Many of the comments here are about displays, but what about the LEDs sold as replacement for incandescent and halogen bulbs?

View on HN

I think, white LED bulbs do use a single wavelength blue LED, but they pass it through phosphors which transform it into a similar distribution of wavelength that you would find from an incandescent bulb, or sunlight. LED's are a lot better than fluorescent lighting, which does often have very dramatic peaks and gaps.

In theory, you can look at the CRI rating, but in fact CRI only uses a small number of wavelength probes for the test, so a rating of 100 doesn't necessarily mean a smooth spectrum, it could be the manufacturer has tuned wavelength peaks in order to game the test. There are better rating systems being developed, but they're not used by manufacturers or mandated by governments yet.

I actually think it might be the lightbulbs that change colour which could be problematic. I wouldn't be suprised if some of those did generate the colour just with single wavelength blue, green and red sources. Hopefully the better ones would have individual phosphors for each r/g/b source which produces a spectrum which when combined is reasonable.

I would be particularly careful about cheap/unbranded colour-variable lightbulbs. I doubt they've been through much or any consumer testing, it seems you can buy anything straight off the boat from China through domestic web stores these days.

Yes, white LEDs are based on blue LEDs with phosphors. But I don't think they convert all the blue light, so they're still heavy on the blue output. I'd need to see a reference to believe they have less peakiness in the spectrum than fluorescent lights.

CRI may not be the best rating for spectrum smoothness, but it's the only one we have for now. It's still hard to find bulbs that even bother to rate themselves.

Check the spec sheet of the bulbs, specifically the color temperature in kelvin. Ignore the "warm white" and similar marketing terms, they are too fuzzy.

A LED with low color temperature and high CRI should have pretty much the same spectrum as incandescent. If you want really low temperatures look for some retro filament style LED lights.

https://en.wikipedia.org/wiki/Color_temperature#Categorizing...

CRI unfortunately is not a good test, it still can produce very high ratings even when the spectrum is spiky.

95+CRI is an indicator for mostly lack of spikes. Look at Yuji (yujiintl.com), they seem to be the only ones selling actual high-CRI LEDs with sane pricing (i.e., no ideological markup).

I'd like to try one of these out, but at $20 per bulb it's a a bit steep. Are you able to comment how they compare to the Feit Electric 90+ CRI bulbs sold at Costco for a few dollars a bulb?

It's tricky, though, since most LED bulbs use a single wavelength of blue LEDs which get downshifted by phosphors, even in high CRI bulbs, you can have an unnaturally large spike in the blue (often around 450nm).

Reply on news.ycombinator.com