Preprint: Paper to be published in Proc., CIE Expert Symposium on LED Light Sources, June 2004, Tokyo

MEASUREMENT OF TOTAL RADIANT FLUX OF UV LEDS

ZONG, Y., MILLER, C. C., LYKKE, K. R., OHNO, Y. Optical Technology Division National Institute of Standards and Technology, Gaithersburg, MD 20899, USA

ABSTRACT 2.1 The source-based method We have developed a source-based method Figure 1 depicts this source-based method. and a detector-based method for total It employs a spectroradiometer and a radiant flux measurement for deep-blue and spectral irradiance standard FEL lamp as an UV LEDs, using a 2.5 m integrating sphere. external calibration source. The same Several UV LEDs with peak wavelengths of principles as the Absolute Integrating 375 nm and 390 nm were measured using Sphere method [2] for the luminous flux both the source-based method and the measure-ment at NIST is used but applied detector-based method with a relative spectrally. Since the flux from an LED is expanded uncertainty (k=2) of ~6 % and ~5 relatively low and the sphere throughput is %, respectively. The results of the two low (due in part to the large size), we need methods agreed within 2 %. an instrument with a very high sensitivity. We used an array spectroradiometer Keywords: integrating sphere; LED; total employing a back-thinned CCD array, which radiant flux. gives sufficient signal-to-noise ratio for the LEDs measured. The total sphere system 1. INTRODUCTION (spectroradiometer and the integrating There is an increasing need for accurate sphere) is calibrated against the spectral measurement of the total radiant flux (W) radiant flux of the beam introduced from the and efficiency of LEDs in the deep-blue to external spectral irradiance standard FEL UV region. NIST has already established a lamp, which was calibrated at the NIST calibration facility for the total luminous flux Facility for Automated Spectro-radiometric (lm) of LEDs using our 2.5 m integrating Calibrations, in the direction of its optical sphere [1]. The current absolute integrating axis at a distance of 0.5 m. sphere calibration method cannot be used The external beam and the LED emission for radiant flux measurement in the UV have very different spatial profiles and region or even the deep blue region illuminate different parts of the sphere; thus, because the photometer signal is very low a uniform sphere responsivity is critical to and the uncertainty is too high. Total radiant reduce the error from the spatial flux measurements, in general, can be dissimilarity. In order to achieve a uniform realized by radiometric measurements with sphere responsivity, the cosine-response of a goniophotometer or by measurements of the fiber probe of the spectroradiometer on total spectral radiant flux (W/nm) with a the sphere wall is extremely important. We spectro-goniophotometer. However, these use the same diffuser, a surface-ground facilities are yet to be established at NIST. opal glass, as was used for total luminous To accommodate the urgent need of flux calibrations [2]. The surface-ground industry for calibration of UV and deep-blue 2.5 m LEDs, we have established two calibration Integrating methods using our 2.5 m integrating sphere Sphere Optical Fiber facility. Baffle

Spectroradiometer 2. MEASUREMENT METHODS LED “Hot” Two independent methods have been Precision spot Aperture developed for the measurement of the total Spectral Irradiance radiant flux of deep-blue and UV LEDs. One Standard is a source-based method using a spectral FEL Lamp irradiance standard lamp. The other is a d = 0.5 m detector-based method using a spectral irradiance responsivity reference detector. Figure 1. Source-based method

1 Preprint: Paper to be published in Proc., CIE Expert Symposium on LED Light Sources, June 2004, Tokyo opal glass is a good diffuser in the UV and of incident angle on wavelength was visible regions. The diffusion property of the observed. We conducted series of spectral opal glass has very little dependence on spatial mapping measurements of the wavelength, significantly reducing the sphere by scanning a beam source [2] and characterization work of the sphere with using the spectroradiometer, and found that respect to various wavelengths. The the spatial uniformity of the integrating problem with the opal glass is the low sphere system has little dependence on the transmission in the UV (it drops quickly at wavelength (< 0.5 %). The UV regions have < 400 nm). similar spatial uniformity as the visible region. The reason for this wavelength We tested several other UV diffusers at independence of the sphere uniformity is 380 nm: a surface-ground quartz plate, a because the diffusing property of the opal surface-ground quartz dome, and a teflon glass is wavelength-independent and also dome diffuser. These diffusers had un- because both the external source and the acceptable cosine response or the diffusion LED are illuminating the regions of the property is strongly wavelength dependent. sphere wall that have theoretically equal The spectral flux of the external reference responsivity. beam, Φext(λ), is determined by The coating of our sphere is based on barium sulfate. We observed considerable Φext λ ⋅= EA ext λ)()( (1) fluorescence from the coating when where Eext(λ) is the spectral irradiance of the measuring UV LEDs (or even deep-blue FEL lamp at the aperture plane, and A is the LEDs), which caused significant errors. The area of the aperture. The total radiant flux of stray light of the CCD-array spectro- the LED, ΦLED, is obtained by radiometer is another problem. Therefore, we developed an algorithm that effectively iLED λ)( corrects the fluorescence of the sphere and ΦLED = kcorr ext λΦ )( dλ (2) λ λ ∫ iext λ)( stray light of the spectroradiometer at the same time by characterizing the measure- where iLED(λ) is the spectroradiometer signal ment system at many wavelengths using a for the test LED, iext(λ) is the spectro- tunable laser. The details of this work are to radiometer signal for the external beam from be published elsewhere. the FEL lamp, and kcorr is the overall correction factor that is determined by We also tested the spectroradiometer for wavelength error (± 0.2 nm) and for non- corr E angle ⋅⋅= kkkk spatial (3) linearity (1.5 %). These factors are used in the analysis of the measurement where kE is irradiance nonuniformity uncertainty. correction factor at the precision aperture, which is the ratio of the average irradiance 2.2 The detector-based method within the aperture to the irradiance at the center of the aperture; kangle is the correction Figure 2 shows a schematic of the detector- factor of incident angle dependence, which based measurement system. This method is the ratio of the sphere response at the employs a radiometric detector (silicon angle of incidence of the external beam to photodiode) to measure the output of the that at normal incidence; kspatial is the spatial sphere. In this method, the total spectral nonuniformity correction factor of the sphere radiant flux responsivity of the total responsivity due to the geometrical diff- integrating sphere system (A/W) is erences between the external beam and the calibrated using monochromatic radiation at LED emission illluminating inside the many wavelengths produced by a tunable sphere. laser directed through a fiber into a small integrating sphere (50 mm diameter) outside All corrections necessary for this absolute the sphere. The small integrating sphere sphere method are made spectrally for the produces near-Lambertian radiation. We spatial nonuniformity of sphere response used a portable version of the NIST Spectral and incident angle dependence of the Irradiance and Radiance Calibration using sphere response [However, the spectral Uniform Sources (SIRCUS) [3] that covered differences of the correction factors were wavelengths from 360 nm to 480 nm. The insignificant]. No considerable dependence

2 Preprint: Paper to be published in Proc., CIE Expert Symposium on LED Light Sources, June 2004, Tokyo

irradiance, Elaser(λ), at the aperture plane is and kcorr is the overall system correction determined by using a reference detector factor, which consists of basically the same calibrated for spectral irradiance corrections as given in Eq. (3). We applied responsivity, sRD(λ), using the NIST Spectral the same correction factors used in the first Comparator Facility (SCF). method for the incident angle dependence of the sphere responsivity, kangle, and the 2.5 m spatial nonuniformity, k , because we Integrating spatial Sphere used the same diffuser as was used in the first method since the cosine correction of Baffle Radiometric Detector (Silicon Photodiode) the detector for the sphere is just as critical LED with the first method. kE is for the laser Precision NIST Portable radiation from the small external sphere, “Hot” Aperture SIRCUS spot which was determined to be negligible. 50 mm Integrating The fluorescence of the sphere is cancelled Sphere out using this detector-based method, which d = 0.5 m is an advantage of this method. Optical Fiber Spectral Irradiance 3. MEASUREMENT RESULTS BASED ON Responsivity Tunable Laser Reference Detector THE TWO METHODS We measured several UV LEDs using both methods. The LED#1, LED#2, and LED#3 Figure 2. Detector-based method produce 200 mW at a peak wavelength of The total radiant flux of the tunable laser 390 nm, and the LED#4, LED#5, LED#6, and LED#7 produce 20 mW at a peak beam, Φ laser(λ), introduced to the sphere for each wavelength is determined by wavelength of 375 nm. We first measured the test LEDs for relative spectral power i λ)( distributions, S (λ), and found that the λΦ )( A ⋅= RD, laser (4) LED laser s λ)( total emission beyond 450 nm is negligible. RD Thus, a blue-violet bandpass filter (BG-12, where iRD,laser(λ) is the output signal of the 355 nm to 459 nm) positioned directly in reference detector for laser wavelength λ, front of the fiber optic was used to block and A is the aperture area. The absolute unnecessary spectral radiation of the spectral radiant flux responsivity of the standard FEL lamp into the CCD-array spectroradiometer in order to increase the sphere system, sS(λ), is given by dynamic range; to reduce significant stray i λ)( light at the deep-blue and UV regions; and s λ)( = S, laser (5) S λΦ )( also to limit system response to sphere laser fluorescence excited by short wavelength where iS,laser(λ) is the output signal of the radiation (UV and blue). The blocked light is silicon photodiode on the sphere. The < 0.2% of the total. absolute responsivity of the sphere system Major uncertainty components (all given in for a test LED, sS,LED, is given by standard uncertainty) with the source-based s method are nonlinearity of the spectro- SLED S d)()( λλλ ∫ λ radiometer (1.5 %), random noise in the sS, LED = (6) measurement of the external beam (1.0 %), SLED d)( λλ and system stability (the integrating sphere ∫ λ and spectroradiometer) in the UV (1.5 %). where SLED(λ) is the relative spectral power Major uncertainty components for the distribution of the test LED. The total radiant detector-based method are spectral flux of the test LED, ΦLED, is obtained by irradiance responsivity of the reference detector (0.9 %), random noise of the signal i LED in measurement of external laser flux into Φ LED k corr ⋅= (7) s LEDS, the sphere (1.0 %), stability of the integrating sphere in the UV (1.0 %), and where iLED is the output signal of silicon determination of the relative spectral power photodiode on the sphere for the test LED,

3 Preprint: Paper to be published in Proc., CIE Expert Symposium on LED Light Sources, June 2004, Tokyo distribution of test LED (1.2 %). Including the 4. CONCLUSION repro-ducibility of the test LEDs (1.2 %), the A source-based and a detector-based overall relative expanded uncertainty (k=2) method for total radiant flux measurements is 5.8 % for the source-based method, and of UV LEDs were developed with overall 5.2 % for the detector-based method. relative expanded uncertainties (k=2) of Figure 3 shows the results of the measure- 5.8 % and 5.2 %, respectively. The ments, given as the ratios of total radiant comparison shows that the two methods flux using the source-based method and that worked well within the estimated uncertainty. using the detector-based method. The For the source-based method, the agreement between the two methods were uncertainty can be further reduced by using within 2 %, which is well within the overall a diffuser with higher UV transmittance, relative expanded uncertainty of both correcting the nonlinearity of the methods (The error bars show the expanded spectroradiometer, and improving the uncertainty of the source method, 5.8 %). stability of spectroradimeter in the UV Errors from common correction factors, such region. For the detector-based method, the as the incident angle dependence factor, uncertainty can be reduced by calibrating kangle and the spatial nonuniformity factor, the reference detector with our SIRCUS kspatial cancel and won’t affect the results. facility, increasing the power of the tunable However, they are not dominant laser, and use of a diffuser having more components of uncertainty. Thus, the spectrally flat transmittance. Further work for comparison results verify that both methods such improvements is planned at NIST in worked well, and either method can be used order to achieve an overall relative for the measurement of UV LEDs. expanded uncertainty (k=2) of 2 %. The source-based method is simpler to REFERENCES implement, not requiring tunable lasers except to characterize the sphere [1] C. C. Miller and Y. Ohno, Luminous Flux fluorescence. It also has an advantage that Calibration of LEDs at NIST, Proc. 2nd CIE the total spectral radiant flux (spectral Expert Symposium on LED Measurement, distribution) of the LED is obtained. May 11-12, 2001, Gaithersburg, Maryland, However, stray light of the USA, 45-48 (2001). spectroradiometer and fluorescence of sphere coating can cause large [2] Y. Ohno and Y. Zong, Detector-Based measurement errors, and should be Integrating Sphere Photometry, corrected. The detector-based method does Proceedings, 24th Session of the CIE Vol. 1, not have these problems. However, it Part 1, 155 -160 (1999). requires a tunable laser covering the [3] S. W. Brown, G. P. Eppeldauer, and K. wavelength range of the LED emission. R. Lykke, NIST facility for spectral irradiance Also, the relative spectral power distribution and radiance response calibrations with a of the test LED must be measured uniform source, Metrologia 37, 579 (2000). separately. ACKNOWLEDGEMENTS The authors thank Steve Brown for the 1.10 Travelling SIRCUS, Thomas Larason for U(k =2), detector-based 1.08 calibration of two silicon photodiode 1.06 detectors, and Charles Gibson for 1.04 calibration of two FEL lamps. The authors 1.02 also thank John Scarangello and Ling Zhou 1.00 of LumiLEDs Lighting for helpful discussions 0.98 of UV LEDs. 0.96

0.94 detector-based value detector-based Authors: 0.92 Error bars: U(k =2), source-based Ratio of soure-based value to to value soure-based of Ratio 0.90 Yuqin Zong [email protected] LED#1 LED#2 LED#3 LED#4 LED#5 LED#6 LED#7 C. Cameron Miller, [email protected] Test LED Keith R. Lykke, and [email protected] Yoshi Ohno [email protected] Figure 3. Ratios of results using the two methods 4 Preprint: Paper to be published in Proc., CIE Expert Symposium on LED Light Sources, June 2004, Tokyo

National Institute of Standards and Technology 100 Bureau Drive, Stop 8442 Gaithersburg, MD 20899-8442, U.S.A. Phone: 301-975-2332 Fax: 301-840-8551

5