use of a low-cost silicon diode-array spectroradiometer

USE OF A LOW-COST SILICON DIODE-ARRAY SPECTRORADIOMETER TOMEASURE SOLAR UV-BTerence M. Murphy, Section of Plant Biology, University of California, Davis, CA 95616Presented at an NIST Conference on “Critical Issues in Air Ultraviolet Metrology,” May,1994 (unpublished)AbstractAn Ocean Optics S1000V spectrometer, equipped with a UV-sensitive silicon diodearray (glass cover removed) and a UV-transparent optical fiber input together witha MgO diffusive reflectance disk, was used to measure global (whole-sky) UV-Bradiation. Reproducibility of the readings of the spectrometer using a standardlamp was about 5%. The calibration function [W m -2 (indicated unit) -1 ] of thesystem, determined by using a standard quartz-tungsten-halogen lamp,demonstrated interference peaks, the amplitudes of which were ±10% of the meansensitivity of the sensor. The scattering properties of the reflectance disk providedcosine-corrected irradiance values. The temperature effects on the dark current andthe light sensitivity of the sensor were significant, but not correlated, makingtemperature control important. Readings of solar radiation below 300 nm wereovershadowed by dark-current noise and/or stray light. The instrument isinexpensive and easily portable and, with care given to calibration, can be used tomeasure solar UV-B above 300 nm.IntroductionResearch on the effects of solar UV-B radiation on biological systems requires an accurateestimation of the effective incident radiation. Because the absorbance of UV-B radiation bymany biological molecular (e.g., DNA) varies greatly through this spectral region, in orderto predict the relative biological effects of different radiation sources it is necessary either tofind a sensor with a spectral response like that of the molecule, tissue, or system in questionor to obtain irradiance spectra for the sources (1,2).Spectroradiometers that are sensitive in the UV region have been expensive, generally over$10K. Less expensive instruments that are assembled in the laboratory from amonochromator and photometer are often too clumsy for field students (3, personalexperience of the author). Recently, a new instrument has been developed by Ocean Optics,Inc. that is relative inexpensive and easily portable. This paper describes an evaluation ofthis instrument for biological research on UV effects, emphasizing certain precautions thatmust be taken when it is used for field studies.EquipmentA spectrometer (Model S1000V, Ocean Optics, Inc., Denedin, FL) was equipped with aUV-transparent fiber optic input (50 µm diameter). In this spectrometer, the light from theoptical fiber impinges on a concave front-silvered mirror with a holographically inscribeddiffraction grating. The diffracted spectrum is directed to a silicon charge-coupled diodearray. In the UV-sensitive model, a glass plate is removed from the front surface of thearray. The spectrometer was mated to a laptop computer through an analog-digitalconverted (Model CIODAS16JR, Computer Boards, Inc., Mansfield, MA). The softwareused for control of the spectrometer and data acquisition was either that provided with thespectrometer by the manufacturer or the more convenient C-SPEC (Ancal, Inc., Las Vegas,NV). The portability of the spectrometer comes from its size (12 x 13 x 4 cm 3 ) and the fact

that it draws its power from the computer. List prices (Ocean Optics, Inc.; Ancal, Inc.) as of1993 were: spectrometer, $1875; optical fiber, $148; analog-digital board, $499; C-CPEC,$580. The software requires an IBM-compatible 80286 (or later model) computer. APack-in-Tell 386SX laptop computer (Computer Warehouse, Sacramento, CA) was selectedbecause it could accept the analog-digital board directly.To use the spectrometer as a spectroradiometer, the data were calibrated to give an absolutereading of fluence (W m-2 nm-1) using a quartz-tungsten-halogen lamp with a calibratedradiance traceable to an NIST standard (Optronic Laboratories, Inc., Orlando, FL).Confirming values were determined using the output from a monochromator (Jobin Yvon)and a 250 W high-pressure Hg vapor lamp. The monochromator output was measured withan Eppley thermopile radiometer, which itself was calibrated using both a quartz-tungstenhalogenlamp with known energy output (measured electrically) and a laser, the energy ofwhich was measured by a factory-calibrated Newport Corp. Model 835 Si Optical PowerMeter. A 100 W quartz-tungsten-halogen lamp (Model L7407, Gilway Technical lamp,Woburn, MA) was used as a secondary standard.Spectral data were calibrated with respect to wavelength by noting the positions of linesfrom an unfiltered low pressure Hg vapor lamp.For measurement of global (sun plus sky) radiation, a reflectance disk was fabricated from apaste of magnesium oxide in water, which was pressed into a plastic mold, smoothed, andallowed to dry. The optical fiber was positioned normal to the disk.ResultsAs measured with the spectrometer, the median width of Hg vapor lines at one-halfmaximum height was 2.0 nm (Fig. 1). This compared to the nominal resolutin specified bythe manufacturer (1.2 nm). At the highest sensitivity setting (integration time 4096 ms), thedark current was 10-15% of the maximum reading, considerably greater than specified bythe manufacturer. A set of eight readings of the secondary standard lamp (unregulatedpower supply) taken over one week and two cycles from full battery charge to “lowbattery” indication gave a coefficient of variation of 4.9%. Over the past 12 months, thewavelength calibration has drifted by less than ±0.8 nm.

Fig. 1. Line spectrum from an unfiltered low pressure Hg vapor lamp,placed ca 30 cm from the fiber optic input. Integration time was 2048 ms;10 scans were averaged. Inset: logarithmic plot, showing the symmetry ofemission lines and the limits imposed by dark current noise and/or straylight.The sensor of the spectrometer was sensitive to temperature. By placing the spectrometer atdifferent temperatures in incubators in the laboratory, I found that the dark current was thesame at -7 o and 12 o C, but was higher at 25 o C and much higher at 30 o C. In contrast, thereading from a standard lamp (less the dark current) was fairly constant from 12 o to 30 o C,but was substantially lower at -7 o C (Fig. 2). Changes in dark current could be compensatedby using the values given by the “dead pixels” (pixels 1-15 and 1048-1100 on the diodearray), which also varied with temperature, but this procedure did not work for the lightresponse. In practice, it was best to insulate the spectrometer and minimize the differencesin temperature between the times that the instrument was calibrated and used.

Fig. 2. Effect of temperature on spectroradiometer readings. Open squares,average of a set of 52 light pixels. Closed squares, average of dead pixels 1-15. Closed triangles, average of dead pixels 1048-1100. Each reading wasmade after the sensor housing had equilibrated to a chamber of a measuredtemperature; the optical fiber and the quartz-tungsten-halogen lamp remainedat a constant room termperature, and the position of the fiber relative to thelamp did not change when the sensor housing was moved from one chamberto another. Integration time was 4096 ms; 10 scans were averaged. In thedark, the readings of the dead pixels were precisely the same as in the light,and the readings of the light pixels were simlar to those of the dead pixels.Used directly, the 50-µm diameter optical fiber had an input sensitivity as a function ofangle on incidence that was approximately triangular, dropping to zero for angles ofincidence greater than 7.5 o from the fiber axis (Fig. 3). The maximum sensitivity of theinstrument was determined by aiming the sensor directly at a standard lamp (Fig. 4). On anenergy basis, the sensitivity at 300 nm was 0.52 times that at 450 nm; on a photon fluencebasis, the sensitivity ratio was 0.52(450/300) = 0.78. The calibration function showed aperiodic structure that was due to thin-layer interference by the SiO 2that formed the diodearray (M. Morris, Ocean Optics, Inc., personal communication).

Fig. 3. Sensitivity of the optical fiber-spectroradiometer system as afunction of the angular deviation of the lamp direction from the fiber axis.Fig. 4. Calibration function, showing the absolute sensitivity of thespectroradiometer system with the fiber pointed directly at a point source.The specified irradiances of a standard quartz-tungsten-halogen lamp weredivided by the spectrometer readings at the corresponding wavelengths (•).To approximate a continuous calibration function, a modified black-body

spectrum (6) with color temperature chosen to match the spectrum to thespecified irradiances of the standard lamp was divided by the spectrometerreadings (|). Finally, a function (B) was derived with a trend taken from asecond-order regression formula for the standardized irradiances (A) and aninterference component with amplitude and periodicity chosen to match thecontinuous function. The formula of the calibration function shown is:B =- 28.12 - 0.099λ + (1.001x10 -4 )λ 2 + (1.617 - 0.0028λ) sin (2π[6500/λ]- 0.5)To measure global solar radiation from sun and sky, the fiber was directed downward to ahorizontal MgO diffusive reflectance disk. In this mode, the response of thespectroradiometer was proportional to the cosine of the angle of incidence of the lightnormal to the disk (Fig. 5). The sensitivity of this system was calibrated by directing theoutput of the secondary standard lamp onto the disk. The reflectance of the disk was 0.70-0.95% and was considered to be a smooth function from 290-475 nm (Fig. 6 inset--thesharp peaks from 300 to 350 nm are though to be random variation from the low intensityof the reflected radiation of the standard lamp measured in this mode.)Fig. 5. Sensitivity of the diffusive reflection disk-optical fiberspectroradiometersystem as a function of the lamp direction from a linenormal to the disk (and along the fiber axis).

Fig. 6. Calibration function for use with the MgO diffusive reflection disk,calculated as the product of the calibration functin for direct radiation (B,Fig. 4) with the reciprocal of reflectance (C, inset and formula below). Thereflectance was measured by placing the sensor the standard distance fromthe MgO disk (8 cm) while illuminating the disk with the secondarystandard lamp at a measured distance (0.25 m) and angle (9.2 o ). Data forreciprocal of reflectance were obtained by the following calculation,Reciprocal of reflectance = [S(λ)/J(λ)][cos(9.2 o )π(0.5) 2 /π(0.25) 2 ],where S(λ) was the reading of the standard lamp measured by pointing thefiber directly at the source from 0.5 m and J(λ) was the reading of thereflected radiation. The right-hand expression corrects for the distance andangle of incidence of the standard lamp radiation on the disk. The functionC represents a third-order regression fit to the reciprocal of the data forreflectance and follows the equation,C = -938 + 8.97λ - 0.024λ 2 + (2.05x10 -5 )λ 3Fig. 7 shows the spectrum of global solar UV (and blue) radiation in April in Davis, CA; thestructure of the spectrum in the UV range is similar to that produced by more elaborateinstruments. The upper inset shows how the measured irradiance compared to thatpredicted by a semi-empirical analytical function (4,5). The lower inset shows a logarithmicploy, demonstrating the limits imposed below 300 nm by dark current noise and/or straylight. A comparison of spectra taken at Lake Tahoe, CA and Davis, CA (Fig. 8) shows aratio, greater than predicted by the program of Björn and Murphy (4) on the basis ofaltitude alone, that may reflect the presence of ozone or other pollutants in the CentralValley.

Fig. 7. Global solar UV and blue spectral intensities, measured at Davis on26 April 1994, 1200 PST; data corrected for spectral sensitivity andreflectance of MgO disk (Fig. 6). Upper inset: UV-B portion of thespectrum, compared to values predicted by the DAYLIGHT program ofBjörn and Murphy (4), using appropriate parameters and three values, 3,4,and 5, for “aerosol.” Note that an aerosol value below 4 overestimates theamount of UV-B reaching the Earth’s surface. Visibility at that time wasover 80 km in all directions. Lower inset: logarithmic plot.

ConclusionsFig. 8. A comparison of representative solar spectra in Davis, Ca (15 July1993, 1215 PST, 10 m altitude) and Lake Tahoe (25 July 1993, 1145 PSTam, 1920 m altitude). Six Davis spectra and four Tahoe spectra werecompared by t test: the difference at 340 nm was significant at the 5% levelof confidence. Inset: ratio of the Tahoe spectral values to the Davis spectralvalues. Although visibility was excellent in both locations when the spectrawere taken, the higher ratio in the UV-B band suggests that the difference isdue to an accumulation of pollutants, including ozone, in the more populatedCentral Valley around Davis.The S1000V spectrometer provides the basis for an inexpensive, easy-to-operate, fieldportablespectroradiometer. Although subject to uncertainties generated by the limitedspectral resolution and stray-light rejection characteristic of a single-grating monochromatorand by temperature effects on both light and dark currents, the sensitivity[signal:(noise+stray light)] of the instrument above 300 nm for measurements of globalUV-B radiation reflected from a MgO diffusive reflectance disk, and the disk gives a goodcosine response. In using this instrument, the large uncertainty at wavelengths less than 300nm must be acknowledged.Is an instrument of this type useful for studies of the biological effects of UV-B? Theanswer depends on the biological weighting function of the phenomenon being studied.This instrument cannot measure solar radiation at those wavelength most affected by ozone;however, a substantial proportion of solar damage to DNA and photosynthesis occurs fromwavelengths above 300 nm. Also, the spectroradiometer has some advantages relative to abroad-band meter, since its measurements can easily be adapted to different light sourcesand to living systems with different biological weighting functions. The low cost, simplicity,and portability of this instrument make it inevitable that it or similar instruments by othermanufacturers will be used increasingly by physiologists and ecologists.

The complexity of the calibration curve (Fig. 4) emphasizes the importance of having astandard lamp readily available. The high cost of absolute spectral standards for the UVwavelengths presents a danger of data appearing in the literature from instruments that havenot bee calibrated or have been calibrated too infrequently. There is a need for lessexpensive standard lamps of reasonable, if lower, accuracy (5%).Acknowledgement--I am grateful to Jana Steiger, Laboratory of Chemical Biodynamics,University of California, Berkeley, for providing the standard lamp.References1. R.D. Rundel, “Action spectra and estimation of biologically effective UV radiation,”Physiol. Plant. 58:360-366, 1983.2. Y. Furusawa, K. Suzuki, and M. Sasaki, “Biological and physical dosimeters formonitoring solar UV-B light,” J. Radiat. Res. 31:189-206, 1990.3. W.F. Kaufmann and K.M. Hartmann, “Low cost digital spectroradiometer,”Photochem. Photobiol. 49:769-774, 1989.4. L.O. Björn and T.M. Murphy, “Computer calculation of solar ultraviolet radiation atground level,” Physiolgie Vegetale 23:555-561, 1985.5. A.E.S. Green, K.R. Cross, and L.A. Smith, “Improved analytic characterization ofultraviolet skylight,” Photochem. Photobiol. 31:59-65, 1980.6. J.C. DeVos, “A new determination of the emissivity of tungsten ribbon,” Physica20:690-714, 1954.