|
Spectral Color Measurements |
 Figure 1. NIST spectroradiometer for LED color measurement. |
Measurements using spectrometers are subject to various sources of error such as wavelength scale shifts, stray light, bandwidth, scanning interval, detector nonlinearity, and imperfection of input optics. The uncertainties of color measurements depend not only on the type of instrument used but also how it is set up with the test artifact and how measurements are performed. Some of the errors can be corrected but others are difficult to correct and should be minimized. NIST recently developed new practical methods for correction of bandpass and stray light of monochromator to improve uncertainty in color measurements significantly. |
Effect of bandpass in measurement of light sourcesThe treatment of the scanning interval and the bandwidth of spectrometers is critical in measurement of light sources containing emission lines and narrowband peaks. There are two mechanisms that cause error related to bandpass. One is broadening of measured spectra, which occurs regardless of scanning interval. The other is due to the mismatch between bandwidth and scanning interval, which causes error in measurement of emission lines and narrowband peaks.Figure 2 shows the effect of broadening of spectra on color measurement caused by a bandpass of 5 nm and 10 nm (FWHM) of a modeled spectroradiometer measuring several different light sources at 5 nm and 10 nm intervals. Therefore, these results are for the condition where the bandwidth and scanning interval are perfectly matched. Errors are shown as the distance in (u',v') chromaticity diagram. While the errors for a Planckian source (very smooth spectral distribution) are negligible, errors can be significant for narrow-band sources. The errors increase nearly proportional to the square of the bandwidth. The level of errors with a 5 nm bandwidth is acceptable for most practical applications. Errors at a 10 nm or larger bandwidth for these sources are not acceptable for many applications. If bandwidth and scanning interval are not matched, serious errors will occur for fluorescent lamps, which have strong emission lines. Data in further details are available in references [1,2]. |
 Figure 2. Errors in (u',v') due to a 5 nm and 10 nm triangular bandpass of a spectroradiometer at 5 nm scanning intervals. |
|||
Correction of bandpass errorSeveral methods are available to correct for bandpass errors for measured chromaticity. For object color, ASTM E308 [3] provides tables of weighting factors (only for object colors) that correct bandpass errors for data interval of 10 nm and 20 nm and under the condition that the bandwidth and scanning interval are matched. For other intervals and for light sources as well as object color, the method by Stearns and Stearns [4] can be used under the same matching condition required. A more flexible method has been developed at NIST that does not require that bandpass and scanning interval are matched and can be applied to any shape of bandpass function [5].Effect of stray light in monochromatorDue to imperfection of a monochromator such as scatter from the grating, mirror, and reflections from detector array, etc., a spectrometer, at a given wavelength, responds to radiation of wavelengths other than from the bandpass. Such unwanted radiation within a monochromator is called (spectral) stray light. Such stray light is prominent in single-grating instruments, in particular, array spectrometers. The effect of stray light in color measurement becomes significant when measuring spectral distributions having high levels in some region and very low levels at other regions. Such an example is shown in Figure 3, comparing spectral distributions of a red LED measured with a diode-array instrument and with a double-grating spectroradiometer. In this case, corresponding chromaticity error is 0.0026 in Δu'v.Figure 4 shows results of simulation for stray light errors with several different sources measured with a modeled spectroradiometer having a stray light level of ~10-4 and a 5 nm triangular bandpass. Stray light errors are not significant for white light sources, which has spectral power in all wavelengths. Stray light error becomes significant for sources like LEDs that do not have any emission in some region in the visible region. Data in further details are available in ref. [2]. |
 Figure 3. Spectral distribution of a red LED measured with an array spectroradiometer and a double monochromator.  Figure 4. Error in chromaticity (u,v) caused by stray light of a spectrometer at ~10-4 level. |
Correction of stray light errorA practical method has recently been developed at NIST. A summary is described below for an example of application to an array spectroradiometer. First, a spectroradiometer under test measures monochromatic radiation at about 20 nm intervals (e.g., using tuneable lasers) spanning the whole spectral range of the spectroradiometer. A function of measured relative signals is called the Line Spread Function (LSF). Then, the LSF is normalized to the total signals in the in-band region (e.g., all signals higher than 1 % of the peak), and the signals in the in-band region are removed. The resulting function is called the Stray-light signal Distribution Function (SDF). Figure 5 illustrates the LSF and SDF.Then, the SDFs (measured at 10 nm to 50 nm intervals of laser emission lines) are interpolated for the measurement intervals of the instrument. With the number of pixels n of the instrument, this interpolation produces a n×n matrix, called SDF matrix, D. Each column of D is the SDF functions at a given excitation wavelength. Each row of D forms the spectral stray light response function for each array pixel. With Ymeas a column vector of measured signals, the stray-light corrected signals Ycorr (a column vector) is simply given by Ycorr = [I + D]-1 Ymeas where I is an identity matrix. The inverted matrix C = [I+D]-1 is called the stray-light correction matrix, C. Once this matrix is obtained, a stray light correction is achieved by a simple multiplication of matrix C to the measured signals. The correction with this method is applied simply to measured signals of a spectrometer, therefore in case of a spectroradiometer; corrections need to be applied for measurement of a calibration source (typically a tungsten halogen lamp) and measurement of a test source. Some results of correction using this method for measurements of LEDs using an array spectroradiometer are shown in Figure 6. This case demonstrates that stray light errors are reduced by one order of magnitude. Note that this method uses the spectrometer itself to obtain correction matrix. Therefore, stray light from radiation outside the spectral range of the instrument is not corrected. To ensure this method works, incoming radiation outside the spectral range of the instrument should be filtered out. |
 Figure 5. Illustration of Line Spread Function (LSF) and Stray Light Signal Distribution Function (SDF).  Figure 6. An example of results of stray light correction for LEDs measured with an array spectroradiometer. |
|
|
For further details of this stray light correction method, see reference [6]. Various other issues on spectrometers for accurate measurement of light source color and object color, uncertainty evaluation, etc., are discussed in reference [2]. References
|
||
|
For technical information or questions, contact: |
||
|
Yuqin Zong Phone: (301)-975-2332 Fax: (301)-840-8551 Email: yzong@nist.gov |
Yoshi Ohno Phone: (301)-975-2321 Fax: (301)-840-8551 Email: ohno@nist.gov |
|
Return to Optical Sensor Group
OTD Home Page |
Technical Inquiries |
Site Comments
Online: May 2007
,
Proc. 25th Session of the CIE, San Diego, 1(D2), 44-47 (2003).