Hacker News Reader

Comment by CuriousSkeptic | Hacker News Reader

Warm white refers to some particular average. But how do you find led’s that don’t hide a spike of blue light in that average? Most places where I run into led-lights don’t advertise the spectral distribution.

View on HN

This is a really fun and easy experiment you might like:

https://www.cs.cmu.edu/~zhuxj/astro/html/spectrometer.html

Oh that's awesome! I have a colleague writing a semester on science projects for kids, this is definitely going to fit in with that!

See for yourself:

https://www.ebay.com/sch/i.html?_odkw=Diffraction+&_from=R40...

I was skeptical, but great link. I might order one to play with. I didn't even realize that was a thing.

My understanding is that LEDs have a very specific, peaky distribution. Heating a wire results in a wide distribution of color temps, but solid state lighting is a really narrow band of color.

I have a photo panel that needs two separate color LEDs to produce a balanced feeling color temp.

As I wrote (https://news.ycombinator.com/item?id=19919819) LEDs usually have a single color: blue. However the resulting spectrum generated by the phosphor can have almost any shape, some high-quality LEDs even have a pretty perfect black body spectrum without a blue line. Of course they are more expensive and also less efficient.

> some high-quality LEDs

Where can I get these?

In terms of commercially available bulbs, the 'Philips Master LED ExpertColor' (all in spotlight 'GU10' etc. format) have an extremely high quality halogen-like spectrum.

Search for "high CRI LEDs". This should turn up LED module vendors which usually have the data sheets which should include a LED spectrum. The warmer the LED temperature is, the smaller the blue peak. For example here: https://www.led1.de/shop/files/Datenblaetter/Osram-OS/Oslon/... or here https://lumstatic.com/9p/Vt/uRNvyjbnQnfzzgO6rg.pdf.

Soraa.

I really hope, someone makes a cheap spectrometer as a smartphone accessory. That would be the best way to quickly assess the properties of light sources. Beyond the health impact, one wants to know the light spectrum when e.g. selecting lamps for fish tanks or plant lighting.

As a very coarse but easily available poor-mans spectrometer, I have been taking pictures of a greycard with my camera in raw-mode. Looking at the histogram balanced for a fixed color temperature, e.g. sunlight, for the r,g,b-channels gives very different results depending on different light sources. While you don't see a true spectrum, just the strength of the blue vs. the green and red channel usually is very revealing.

It's cheap and works without the smartphone

https://www.ebay.com/sch/i.html?_odkw=Diffraction+&_from=R40...

Yes, that is a good start - I would like to have curve with the relative strengths of the spectrum. But maybe it is enough to carefully calibrate the smartphone camera on the diffraction sprectra to generate a course graph.

Also, if no diffraction grating is at hand, a random audio CD usually is up to the task :)

Any compact disk works as a spectrometer.

https://www.cs.cmu.edu/~zhuxj/astro/html/spectrometer.html

I've never seen this. Thanks!

That only shows differences in white balance and tone. It can't give you any information about the CRI since you only have 3 channels of colour sensors in your camera. If you use a colour swatch palette that would be different.

Yes, of course. It doesn't give you a "spectrum" at all. Only a very rough guesstimate on how the colors are balanced. But good enough to see, whether there is a large blue peak (blue line much stronger than red and green) or not. Just using your camera and a rgb histogram.

But using a color palette is a great idea. With some calibration work, this could give a rough spectral curve with a single picture. I need to investigate this :). One would require to get a palette, where the colors are generated not by rgb pigment mixing but by absorbing/reflecting distinct wavelengths. Of course, the MacBeth color targets would fufil that condition, now how to get a wavelength calibration from this? Have to read up on things...

Another option would be to use colour correction gels like from Rosco. They publish transmission curves for their gels: https://emea.rosco.com/en/products/filters/r3220-double-blue... (click on the little i). You could take pictures of the light source through a set of different gels and reverse to an approximation of the spectrum of the light source.

Interestingly they are also now offering gels to correct the colour casts of LED lighting: https://emea.rosco.com/en/product/opti-flecs

There are samples swatches of their gels, which can be had by either asking a theatrical lighting supply company or on eBay.

CRI is a common metric for LED lighting. I'm not intimately familiar with the standard, but would expect a large spectral aberration to negatively impact the CRI.

When I was building up some high-CRI architectural fixtures, finding fully characterized LEDs wasn't much of an issue. At the consumer level it can be tough to get good details beyond color temperature.

I am designing my own lighting and I found the same thing. There is a huge discrepancy between what you can get at the source from manufacturers and what ends up in "consumer" hands. Most of what is sold in the stores is poorly-specified crap, built to be as cheap as possible.

Just guessing from what I know about LEDs and efficiency, I suspect they all just hide a spike of the same blue light in that average.

Reply on news.ycombinator.com