Sorting white LEDs by colour

M

MartinM57

Moderator
One of for collective...

I have a batch (>1000 :)) of white LEDs from a reputable manufacturer and reputable supplier.

Each LED in isolation is qualitatively "white" but with say 20 all powered up and compared to each other they are a different white - some are warmer (yellower - lower degrees K) some are cooler (bluer - higher degrees K). But I don't know the range - maybe 7000K to 1000K?

I need to sort them into batches that visually look the same - I assume this is batches with a small range of colour temperature e.g 7001-7500, 7501-8000, 8001-8500 etc

What (cheap) colour sensor can I plug into a PICAXE to point each LED at, one at a time, to measure the colour temperature to that sort of accuracy will allow me to allocate them into batches that visually look the same "whiteness"

..or am I on a hiding to nothing?

For those wanting more detail, the LEDs' data sheet is at
http://docs-europe.electrocomponents.com/webdocs/12d3/0900766b812d38bc.pdf
and the part number is C503D-WAN-CCbDb232 which equates to LEDs in colour bins W2 and W3 on page 5 of the data sheet which look to be from about 7000K to 10000K, hence the numbers above
 
M

MartinM57

Moderator
Common voltage via resistor per LED. If I guess correctly where you're going, there may be different currents in each LED, but that's not the cause of the difference in colour.

If I wind up/down the voltage to the LEDs, within reason, (i.e. adjust the current) it's only the intensity that varies - the underlying colour/degreesK of each LED doesn't change

It's all consistent with the data sheet - it's possible that within a batch of C503D-WAN-CCbDb232 LEDs, the colour temperature can vary between 7000K and 10000K - and I (currently ;)) believe I can see that and want to sort them into batches with minimal visual difference within each batch
..I could do it by eye but:
- it hurts my eyes to look into them ;)
- it's qualitative assessment not quantitative measurement
 
techElder

techElder

Well-known member
I'm with you on the color temp, but in another life I spent much time certifying UV lamps with radiometers. The UV sensors were somewhat sensitive to high intensity light differences.

I'm used to measuring in nanometers around 555nM. It wouldn't be difficult to set up a jig with a white light sensor and a radiometer. That would give you numbers for intensity at a certain wavelength. That might be enough for sorting.
 
Last edited:
M

MartinM57

Moderator
@eclectic
Three of us are all in agreement as to which LEDs are at the cooler/bluer end of the spectrum and which are at the the warmer(yellower) end - and which are in the middle etc

What colours we are individually actually seeing though is pretty much impossible to know (I think)

BBC link is a 404 for me :(

@all
To bring this back on track I was thinking about a ...
http://www.adafruit.com/products/1334
...with the (so called) white LED turned off and my LEDs pointing at it, one at a time (maybe some light tube to stop external lighting affecting the reading)
 
techElder

techElder

Well-known member
So, for that price try it. Just reminding you that you light emissions have bandwidth.

Another thing to consider with the Adafruit module is that it looks to expect reflected light rather than direct illumination.
 
erco

erco

Senior Member
Real color sensors may be pricey. A low tech solution is to use a prism and bin the LEDs visually by their chromatic display.

 
J

Janne

Senior Member
A first idea that comes to my mind would be to hook up a (video)camera to computer, and see if there are any easy solutions for finding a colour temperature from an image. I suspect many imaging libraries have functions for such use. It also seems possible to calculate the colour temperature, if you know the values of R, G and B. http://dsp.stackexchange.com/questions/8949/how-do-i-calculate-the-color-temperature-of-the-light-source-illuminating-an-ima
Three fotosensors with R, G and B filters should also then work, but finding those filters might not be easy. But that would be a picaxeable-solution.
Anyways, it sounds like a very fun problem to tackle =)
edit. I've bought cheap but decent prisms from ebay in the past.
 
B

BESQUEUT

Senior Member
Janne said:
A first idea that comes to my mind would be to hook up a (video)camera to computer, and see if there are any easy solutions for finding a colour temperature from an image. I suspect many imaging libraries have functions for such use. It also seems possible to calculate the colour temperature, if you know the values of R, G and B. http://dsp.stackexchange.com/questions/8949/how-do-i-calculate-the-color-temperature-of-the-light-source-illuminating-an-ima
Three fotosensors with R, G and B filters should also then work, but finding those filters might not be easy. But that would be a picaxeable-solution.
Anyways, it sounds like a very fun problem to tackle =)
edit. I've bought cheap but decent prisms from ebay in the past.
Click to expand...
Any smartphone can do the job. There are some apps on the web
cassiopeia.camera
colormeterfree
magiclightmaster
You have to isolate from ambiant light and illuminate a white paper.
 
Last edited:
techElder

techElder

Well-known member
I need to sort them into batches that visually look the same - I assume this is batches with a small range of colour temperature e.g 7001-7500, 7501-8000, 8001-8500 etc
Click to expand...
I don't understand the emphasis on "color temperature" when you want a visual sort.

Measure the intensity of the LEDs at a constant current with a photometer and sort them out with the numbers that you get.

This cheap ebay photometer measures with the "CIE Photopic. (CIE human eye response curve)". Perhaps it will do what you want when you get back to what you want.

http://www.ebay.com/itm/200-000-Digital-Light-Meter-Luminometer-Photometer-Luxmeter-Tester-Lux-FC-US-/161411064045?hash=item2594d960ed
 
rq3

rq3

Senior Member
Texasclodhopper said:
I don't understand the emphasis on "color temperature" when you want a visual sort.

Measure the intensity of the LEDs at a constant current with a photometer and sort them out with the numbers that you get.

This cheap ebay photometer measures with the "CIE Photopic. (CIE human eye response curve)". Perhaps it will do what you want when you get back to what you want.

http://www.ebay.com/itm/200-000-Digital-Light-Meter-Luminometer-Photometer-Luxmeter-Tester-Lux-FC-US-/161411064045?hash=item2594d960ed
Click to expand...
Because the human eye is very sensitive to color temperature, and not very sensitive to brightness, unless you've shifted into rods vs cones discussions. My wife recently bought a fairly expensive LED fixture for the island in our kitchen, and the 3000K color temperature drove me up the wall, at all dimming levels. I added a Lee 202 theatre gel to the fixture, shifting the temp to roughly 4000K, and I'm a happy camper. My wife can't see the difference.
 
G

Goeytex

Senior Member
Seem unusual to have that much difference in color temperature within the same batch. I have used hundreds of Cree XR-E series LED's and have not experienced a noticeable difference in color temperature given the same part number and binning. For quality 5mm LEDs I have used Avago and also have not experienced a noticeable difference in color temp over thousands of LED's. It's odd that Cree allows such loose specs on these.

Given a tight budget, I would probably sort them visually, not by looking at them directly, but by looking at the light reflected from a sheet of paper or other surface. Find one LED in the middle of the color range as a control then compare others as warmer, cooler or close enough. How you do the sorting depends upon how close you need the color temp to match within your sorted groups.
 
J

johndk

Senior Member
Take a look at TCS3200
https://www.jameco.com/webapp/wcs/stores/servlet/StoreCatalogDrillDownView?langId=-1&storeId=10001&catalogId=10001&search_type=jamecoall&freeText=TCS3200.

It's an inexpensive module that does a decent job of measuring RGB intensities. With a little mathematical tinkering, you can approximate a course spectral reading. This module comes with white leds, as it's meant for reflective color sensing. But the leds can be turned off and it works well on reading subtle color changes in incident light. I haven't found the bare sensor yet, but I'm sure it's out there.
 
You must log in or register to reply here.
Top