567

Fortunately, lumen figures are given by most lamps you can buy these days. This number gives the total amount of light emitted by a lamp as seen by human eyes. So if you want to light up a room, this figure is the most important specification. But although the numbers are probably correct for top brands like Philips or Osram, they might be grossly overstated for noname products, like it is the case for the capacity of aftermarket cell phone batteries. The reason is nobody can verify it - if you buy a pack of 100 screws and there are just 60 in it, the vendor is in trouble. But for a lamp labeled "500 lumen"? Lumen measurement is expensive. The reason is in the words of the second sentence above: "total amount of light" means you have to somehow measure all light that the lamp emits. Either you have a full sphere detector or you measure many points around the lamp and add up the measurements. The solution is to have either an integrating sphere (Ulbricht sphere) or a goniometer, scanning the sphere around the lamp. "as seen by human eyes" means that you need to have the same spectral response in the measurement as the human eye. No semiconductor sensor has that. Professional measurement tools use spectrometers for that reason. Given the eye sensitivity function, the physiologically weighted result can easily be calculated from a spectrometer readout, as can CRI, color temperature and color shift. Needless to say this gets expensive. Expect to spend 30 k€ for a professional lumen meter. But we wouldn't be engineers if we couldn't do better. As no hobbyist can afford a calibration from a certified lab, there is little point in aiming for a 0.01% accurate measurement. I think we can simplify the setup to a point where it gets real cheap, but still delivers a useful measurement. The first simplification we can make is demonstrated by the Lightspion from Visio Systems: http://www.visosystems.com/products/lightspion/ - watch the video! The key idea here is that the emission pattern of most lamps is axially symmetric, so we need to measure only around one axis, perpendicular to the lamp axis. Rotating around one axis can easily be done using a stepper motor, which can be driven by half or full steps, something that is super simple, cheap and easy to implement even by the smallest microcontroller. The motor provides the bearings for the rotation as well, all we need is a simple mounting bracket (which I'd do with 3D printed parts). I see little point in adding a second motorized axis for measuring the full spherical emission pattern. You can always rotate the lamp manually in the socket (eg. make four measurements in 45° steps) and average the four results. The next thing is the spectrometer. Obviously this is expensive, but I have to check the new MEMS micro spectrometer modules from Hamamatsu. But for the time being, I suggest using a simple photodiode instead - but one with the correct characteristics. The "lumen" is a physiologically weighted measure of the total radiant flux (in W). The weighting is added in order to account for the eye sensitivity - the lumen is meaningful only for human eyes. So for the lumen (or lux) measurement to be precise, the weighting function V(lambda) (CIE 1931 definition, see Wikipedia) needs to be accounted for. Further explaining this is application note http://www.osram-os.com/Graphics/XPic3/00039056_0.pdf/General%20appnote%20for%20Ambient%20Light%20Sensors.pdf from Osram, highlighting the importance of a correct sensitivity function for the benefit of mankind and their ambient light sensor sales. One sensor having the correct spectral sensitivity is the SFH 5711: http://www.osram-os.com/osram_os/en/products/product-catalog/infrared-emitters%2c-detectors-andsensors/silicon-photodetectors/ambient-light-sensors/sfh-5711/index.jsp In addition it has a logarithmic output, simplifying a high dynamic range readout (3-80 klx). Why is this so important? Have a look at an emission curve of a white LED, say Cree XM-L2: http://www.cree.com/~/media/Files/Cree/LED%20Components%20and%20Modules/XLamp/Data%20and%20Binning/XLampXML2.pdf All versions, the colder colors in particular, have a massive emission peak at 450 nm (this is the blue LED that pumps the yellow phosphor). Now have a look at the BPW21R photodiode, which was suggested for the recent (130109) luxmeter project: http://www.vishay.com/docs/81519/bpw21r.pdf You can see that at 450 nm, the sensitivity of the sensor differs by more than a factor of 10 (!) from the eye sensitivity. If we'd like to measure LED lamps, we'd better find a more accurate sensor. So the project suggestion is: Build a luxmeter (take 130109 as base for example) Add a stepper motor driver (real simple, voltage mode, half step only) Find a photo sensor with best possible matching to the V(lambda) eye sensitivity function. SFH5711 is a candidate, Hamamatsu has others... Write a software that does a 360° scan, either simple in a microcontroller or fancy with graphics and all on a PC Mechanical layout: Same as LightSpion (stepper motor, lamp bracket, sensor placement)