All measurements carry a certain degree of uncertainty, not only from the test equipment but also the test conditions and system of measurement. In the case of luminous flux, factors such as room temperature, stray light, power supply, stabilisation time, sample variation and positioning all contribute to the repeatability of a measurement. Accuracy of a measurement is determined by it's calibrated equipment accuracy, repeatability and correlation to reference data.
The Extech HD450 lightmeter, the main piece of measurement equipment used on our goniophotometer, specifies an accuracy of ±(5% rdg + 10 digits) and measurement repeatability of ±3%. It's suitability for measuring white LED light sources is further specified by a spectral accuracy of f'1≤6%, which indicates the degree to which the meter replicates the photopic human eye response across wavelengths. The slight difference between the spectral response filter of the HD450's photo detector from the CIE human eye response curve may introduce spectral mismatch errors when measuring the light output of LEDs, due to their markedly different spectral distributions from that of the incandescent source the HD450 was calibrated against.
Correlated colour temperature, CRI, spectral power distribution and CIE chromaticity coordinates are measured using the UPRtek MK350 Spectrometer. The equipment specifications claim a CCT accuracy of ± 2%, CRI accuracy of ± 1.5%, and ± 0.0025 accuracy and ± 0.0005 repeatability for CIE 1931 chromaticity coordinate measurement. In low light conditions the spectrometer has some susceptibility to noise, this potential source of error is minimised by taking measurements sufficiently close to the light source so as to limit integration time and consequently detection of stray light.
The power and voltage supplied from the mains and power factor of each test sample is measured using a Cabac CCI Powermate power meter. Claimed accuracy is ± 0.2% with a voltage and current sample rate of 4800/s. As the measured wattage is used to calculate efficacy (lumens/watt), the uncertainty of this measurement affects the accuracy of the efficacy measurement. The product is given 90 minutes for thermal stabilisation before these measurements are taken.
Thermal images are taken using a Testo 875 thermal imager. The unit has a thermal sensitivity of < 80 mK and claimed accuracy of ±2 °C or ±2 % reading and reproducibility of ±1 °C or ±1 %. Use of a thermal imager leads to additional sources of error arising from the variation of surface emissivity and reflections from surrounding objects. Reflective surfaces such as polished metals may appear cooler in thermal images than they actually are.
The repeatability of this measurement is further dependent on ambient temperature, stabilisation time of the test sample and the infrared lens' angle of view of the test sample, as certain lines of sight may obscure elements of the test sample. Measurements are taken in a temperature controlled room, however it can vary between ± 2°C from 25°C room temperature. Maximum temperature is incorporated into product scoring as a relative value to compensate for this variation.
The TAOS TSL257 high-sensitivity light-to-voltage converter sensor paired with a Tektronix DPO2012b 100 MHz digital storage oscilloscope and TPP0200 200 MHz, 10x passive probe were used for measuring the variation in light output captured over very short sampling times. Limited by the rise time of the photosensor, the light output signal cannot reflect frequencies above 2 kHz, however this has a negligible impact on the accuracy of measuring 100 Hz flicker.
Noise in the signal via the power-supply ripple and stray light are other potential sources of error. The photosensor specifies a high power-supply rejection ratio (35 dB at 1 kHz), limiting the effect of power supply ripple to ±0.04 V and therefore error in percent flicker to ±1% for the average maximum signal level of 4 V. Stray light is controlled by conducting the flicker measurement within the goniophotometer dark room.
In addition to the possible equipment error introduced by the HD450 photometer, the accuracy of the luminous flux measurement is further determined by the repeatability of the measurement system. Repeatability and measurement uncertainty depends on conditions such as positioning of the test sample, light output stabilisation time, number of lux measurements sampled, stray light and product consistency.
Angular positioning of the goniometer arm is automated using a stepper motor, microcontroller and Windows application. The vertical height of the test sample is positioned using a level rule to ensure the centre of light emission (in the case of a globe) or emitting surface (in the case of a downlight) is aligned with the goniophotometer's axes of rotation. Random error arising from incorrect height positioning was found to have a negligible effect on measured lumens, a ±10 mm error resulting in no more than ± 3 lm difference to the calculated result. The significance of the error introduced through height positioning is small when calculating luminous flux relative to average measured quantities of 500 lm and thus ± 0.6% relative error. Test samples with dimmer overall output have the potential to yield a greater relative error; ± 10% in the case of our dimmest, 30 lm globe. Measured output angle is more sensitive to variation in height positioning, especially in the case of globes, a ± 10 mm error leading to ± 4° error.
As with other semiconductor devices, LED performance is highly dependent on the p-n junction temperature, affecting both output light intensity and spectral distribution. The change in a test light's output over time was examined in the FAQ article LED and CFL light warm up times. To ensure a stable light output while running the goniophotometer, products were switched on for 90 minutes before testing, giving time for the p-n junction temperature to settle.
The accuracy of the test is limited to 90° azimuthal sampling increments, above the maximum 22.5° step-size specified by ASTM, CIE and IES guidelines. Lux measurements about the polar axis are sampled at 1.8° increments, satisfying the maximum 10° specified and increasing the detail of the luminous intensity distribution plot. Polar positions range from 18° to 342°. Luminous flux calculations assume no globes emit a significant amount of light output outside the sampled solid angle. The large sampling increments taken about the azimuthal angle assumes a high degree of radial symmetry in the light output of all test sources, averaging over just 4 data points per horizontal plane. The validity of this assumption varies based on the symmetry of the test sample's light output.
Stray light is controlled by enclosing the goniophotometer measurement system inside a 1m x 2m x 2m dark room. The limited dark room space determined the dimensional constraints of the goniophotometer, limiting the arm length to 550 mm. In most cases the measuring distance from the test source was sufficient to satisfy the IES specification of at least 5 times the largest light emitting surface dimension of the SSL product so as to approximate testing of a point source. In the case of large downlights and globes, this criteria was not met, potentially compromising the accuracy of the measured luminous flux of these products.
Each product is evaluated based on a single test sample. In order for the evaluation to be representative of the product in general it is assumed that all units perform consistently for any given model. The validity of this assumption depends on the quality and performance standards maintained by the luminaire manufacturer. The production processes of LEDs create inherent variation in the light output, colour temperature and voltage of each chip. LEDs are then 'binned' based on these performance characteristics and it is up to the luminaire manufacturer to specify the acceptable performance range of the light sources used in their products. As only one sample is used for each product, our testing does not account for potential variation across units.
The overall accuracy of the LEDBenchmarks goniophotometer system is best determined by how well it's measurements agree with that of an accredited photometric laboratory. To date, only one LED product unit subjected to accredited laboratory testing has subsequently been tested by LEDBenchmarks. A randomly selected and sourced sample of Brightgreen's D900 55° Curve Downlight was tested by the Queensland University of Technology's photometric laboratory, and then forwarded to LEDBenchmarks for a comparison of the results. The accredited photometric lab measured 940 lm, compared to LEDBenchmarks' 909 lm. The relative error in luminous flux measurement for this unit was therefore 3.4%.
Back to FAQ index