You Pays Your Money, You Takes Your Choice
Recent research published by NIST shows that the higher the reflectance of a sphere coating, the more spatially uniform the luminance distribution is inside a sphere. “So what?” we hear you say? Let us explain.
The “perfect” integrating sphere creates a uniform luminance distribution at all points on the sphere surface. Therefore, the small proportion of light which the detector “sees” is a known percentage of the whole. For 4pi flux measurements, you calibrate the sphere by placing a near-isotropic standard lamp in the centre. An isotropic light source emits the same intensity in all directions. In comparison, no test light source will ever truly be isotropic. Most test lamps will have some variation in intensity with angle, many by design. Certainly LED luminaires are not isotropic, they generally emit in the forwards (2pi) direction only with beams of a few tens of degrees (spotlights). With a sphere coating of low reflectance, you do not achieve the same uniform luminance distribution from a directional light source as you would from a (near-isotropic) calibration lamp. Therefore, the sphere suffers from hot spots and dim patches when you put the test lamp inside (or indeed, shine it in from the 2pi port). The lower the reflectance of the sphere coating, the worse the spatial non-uniformity in the luminance distribution is. This means that you will see large differences in the measured flux (lumens) from the test lamp depending upon the direction that you point the light source and how directional (or isotropic) the beam is. Basically, low reflectance equals big errors (the effect of multiple reflections inside the sphere “amplifies” the problem of low reflectance non-uniformity).
The worse case reported by NIST was a highly directional PAR spot light (PAR stands for parabolic aluminised reflector). This had a 20° beam angle. When measured with a sphere with an 80% reflectance coating, the measured flux varied by up to +15% and -20% depending upon the direction that the spotlight was pointed inside the sphere. In comparison, when the same spot light was tested in a sphere with a reflectance of 90%, NIST saw a variation of up to ± 10%, while a sphere with a reflectance of 98% suffered from a variation in flux of about ± 2-3%.
The Labsphere lamp measurement integrating sphere that we use at Photometric Testing is coated with a form of barium sulphate called Spectraflect. This has a nominal 98% reflectance in the visible. Barium sulphate (at the necessary coating thickness) is not cheap, but it is good. On the other hand, some people resent spending any money on a proper integrating sphere. Instead they construct their own “integrator” based upon flat sheets of wood assembled into a cube or some kind of geodesic chamber. They then paint the wood with an ordinary, household white paint. Typically these paints have a reflectance of 80%. The combination of a non-spherical surface with a low reflectance paint causes these “integrators” to give very high errors.
Size is Everything
Don't just think that because you can squeeze your luminaire into a sphere you will get good measurents. Standards such as IES LM79-08 remind us that it is important to use a sphere that is big enough for the size of fitting under test. If the fitting is too big for the size of sphere, the luminous flux readings start to become sensitive to the direction that you point the luminaire and to the beam angle. The rule of thumb for 2D or 3D fittings placed in the centre of a sphere for 4pi flux measurements is for the sphere diameter to be 5 to 10 times the size of the device under test (1.5 times the length of linear fittings). For spotlights (downlighters), we can shine the light into the sphere though a 2pi (forward) flux port. In that case, the maximum size of fitting should be 1/3 the sphere diameter.
Do We Perfrom Auxiliary Correction?
You bet! The bigger the luminaire that we put into our sphere, the more the hardware starts to absorb some of the light given out by the fitting. If this is not taken into account, errors in the tens of percent are not uncommon. In other words, luminous flux will be measured very low compared to the true lumen value if we weren't to apply an auxiliary correction. We reckon that our clients would not be too happy with that. The colour temperature and colour rendering would also be wrong.
So what is auxiliary correction? An auxiliary lamp is a stable, low power lamp fitted to the wall of the sphere. It allows us to measure the relative light level in the sphere when the device under test is inside the sphere (but switched off) compared to the light level when the calibration lamp is inside the sphere (again, switched off). The difference in detector readings between the two steps mentioned allows us to calculate the degree of self absorption of the test device compared to the calibration lamp. We then apply this as a scaling factor to all subsequent measurements of that type of test device. The net effect is to perfectly correct the absorption errors that arise when putting samples inside a sphere.
Why do we use a Spectroradiometer on our Sphere?
Our Labsphere is equipped with a CCD array spectroradiometer which measures the spectral radiant flux (Watts per nm) over the wavelength range from 350-1000nm. Labsphere LightMtrX software then computes the associated photometric and colorimetric properties of the device under test (e.g. luminous flux in lumens, correlated colour temparture in Kelvin) with reference to the standard CIE observer and colour matching functions. By calculating the outputs in software based upon absolute spectroradiometric data, we avoid the large errors associated with simple filter photometers and colorimeters. Such meters attempt to match the spectral sensitivity of the human eye using a system of coloured filters. Some are able to mimic the eye response better than others, but in general, all filter photometers and colorimeters are known to suffer from potentially large errors when used to measure light sources with spectral peaks, such as LEDs, fluorescent and discharge lamps.
Standards such as IES LM79-08 and CIE 127:2007 now recommend that photometric measurements are performed using spectroradiometers rather than filtered light meters. The result is for our readings to be as accurate as possible regardless of the type of light source under test.