Measurement of Luminous Intensity Using Filtered Trap Detector

The work described in this paper is a part of the process of lamp and photometer calibration to obtain a new luminous-intensity scale with a lower uncertainty. The procedure and the results of realizing the photometric scale with a standardized detector at the National Institute for Standards (NIS) are described. Th is method enables photometric values to be derived from absolute spectroradiometric measurements by exact computation. A filtered trap detector has been characterized and used to measure the absolute spectral power distribution of standard lamps. From these data a photometric scale of luminous intensity has been realized. The results were compared to the cert ified calib ration values obtained by Nat ional Physical Laboratory (NPL), and showed agreement within uncertainty of ± 1.8%


Introduction
The photometric base unit of the International System of Units (SI) is the candela (cd), the unit o f lu minous intensity. Since 1979, the candela has been defined as: the luminous intensity, in a given direction, of a source that emits monochromatic radiation of frequency 540x 10 12 hertz and that has a radiant intensity in that direction of (1/683) watt per steradian [1][2][3][4] . The definit ion of the candela links together photometric and radio metric quantities. The basis of accurate photometric evaluations of light sources is provided by accurate spectral measurements under specified geo metric conditions.
The National Institute of Standards in Egypt (NIS) used to maintain the lu minous intensity by a group of specially designed lamps periodically calibrated at NPL (Nat ional Physical Laboratory-England), with uncertainty of + 0.8% However the scale can be realized by using detector-based methods. Owing to recent improvements in rad io metric measurements, units for most photometric quantities can now be realized with the highest accuracy by radiometric methods. As trap detectors have been extensively described by other authors (5)(6)(7)(8) . Their working princip le is summarized only briefly here. Irradiance standard lamps can be accurately calibrated in absolute values using the trap detector and set of interference filters. As Where E v is the illu minance, I is the luminous intensity in candelas, and d is the distance between the light source and the irrad iated surface.
The photometric unit can be derived fro m the spectral irradiance distribution by means of a v isual response function V(λ) defined and tabulated by the CIE and the constant factor K m specified in the new defin ition.

Basis of the Measurement
The spectral irrradiance of a tungsten lamp E(λ) at a distance d as a function of λ, is defined as : Where ϕ (λ) radiant flu x incident on an aperture area A. Depending on the optical radiation incident on the effective aperture area, filter radio meter generates photocurrent as Where τ (λ) is the spectral transmittance of the filters, R(λ) is the known spectral responsivity of the trap detector, and λ is the radiation wavelength in air. Since the spectrum of light source used in measurements is close to Planck radiator, the spectral power distribution of the light source in eq. (3) for over the visib le range is estimated as given in the following analytical form: . .

exp( / ).
Where A 0 , A 1 , A 2 , A 3 & B are free parameters. The spectral irradiance was modeled with 3rd degree o f polynomial to compensate for deviations of the lamp used in measurement fro m that of ideal b lackbody radiator. (9)(10) Fro m eq. (4) & eq. (3) we can obtain the fo llo wing equation for the various filters as where i indicates the filters which are used in the measurements. Equation (4) must usually be solved by an iterative method. We have to min imize the relative errors between the measured and calculated values of the photo currents I m,i & I c,i to minimu m using the least square fitting: The detailed exp lanation for the solutions of these type equations can be found in ref. [11]. The lu minous intensity of the lamp can be calculated fro m spectral irradiance and the lu minous efficacy function, V(λ) fro m equation (1) 3. Experimental Technique. The filter rad io meter is basically consists of the trap de-tector, interference filters and precision aperture. The trap detector consists of three windowless Hamamatsu 13371 type silicon photodiodes; each has 1 cm 2 act ive area. In trap detector design, the photodiodes are geometrically arranged in such a way that polarization sensivity is removed. In this design, incident beam undergoes five reflections before leaving the detector as shown in Fig. 1. Therefore, reflectance of a trap detector is nearly zero. A precision circular aperture of area 0.1 cm was attached to the front of the trap detector.

Filter Radi ometer
Application of a bias voltage increases the linearity limit by a factor more than 5

Filter Trans mittance
Seven interference filters fro m Avian, having no minal central wavelengths at 400 n m, 450 n m, 500 n m, 550 n m, 600, and 650 n m and 700 n m, were used for the measurements. All filters have a path band of 10 n m. The spectral transmittances of the filters were measured with a Shimad zu double monochromator with 0.1 n m interval and band slit width 0.3 n m (Fig.2). Before starting transmittance measurements the wavelength calibration of monochromator was performed. The filters were checked out of the path band. It blocks radiation away fro m the band path region to better than 0.01%-

Measurement Set-Up
Two set of lamps, 1000 watt standard FEL type and Osram 200 watt Wi/ G lamps were used as light sources having high optical radiation during measurements. Lamps were operated with DC power at fixed polarity and at a stabilized lamp current. The lamp current was measured as the voltage across a calibrated reference resistor of 0.1 Oh m using DVM Flu ke 8085. The photocurrent of the photometer was measured using a calibrated d igital LMT nano-photocurrent meter I1000. Baffles were emp loyed to minimize the stray-light. The trap detector and the baffles were p laced on 4.5 meter optical bench equipped with an accurate length scale. The lamps were mounted on stages with five degrees of freedom that allow accurate positioning of the lamp filament. The optical axis of the bench was determined with an align ment laser positioned between the detector and the lamp. The detector to lamp distance was measured fro m the lamp filament plane to the plane of the aperture. Filter radio meter was used to measure the irradiance values on the aperture plane of the trap detector. e transmittance range of the filters.

Spectral Responsi vity of the Detector
The spectral responsivities of the polarizat ion independent reflection trap detector, was determined in the range 350 -800 n m (Fig. 3). Using a high-accuracy spectrophotometer, the trap detector was compared with a standard silicon detector, 71582 & serial nu mber: 492, one of the set upon which the scale of detector spectral responsivity was established at NIS (Egypt) and traceable to the cryogenic radiometer.

Linearity
Linearity of the detector was tested using a double beam aperture method [12,13] . In this technique the output signal S a , S b are measured for two equal beams A and B. The sum of these signals, S a +S b was compared with the signal when both the beams are p resent in co mbination, S ab . The linearity factor is defined as S ab / S a +S b . A wide range of incident power which corresponds to an output current range A to 10 -4 A mpere.

Spati al Uniformity
The response uniformity of the active area of light-trapping detector was investigated. Nonuniformity would produce measurement errors when the detector used in different position. The uncertainty due to response nonuniformity depends on the position and size of the beam [14][15] . The measurements were done by using a Heliu m-Neon laser. The detector was moved in x and y axis and the output radiation were scanned across the surface.

Luminous Intensity Measurements
Group of 10 lamps were used, Six FEL 1000 watt and for Wi G Osram 200 watt lamps. The irradiance of the lamp has been measured at discrete wavelengths corresponding to the filters 400,450, 500, 550, 600 and 700 n m. A fitting curve was carried out according to equation 7. The parameters obtained for the lamps fro m the minimization were considered to be the final results.
Fro m the irradiance distribution, the Illu minance of the lamp at the detector aperture can be determined at known distance d, then calculat ing the Lu minous intensity from equations 2. Five standard lamps operating at color temperature 2856K were used. The measurements were performed at a d istance of about 2 meters. The lu minous intensity resulting from the measurements of the lamps using filtered trap detector was compared to the certified values reported by NPL, as shown in Table (1). The difference between calibrated lamp at NPL & the scale based on trap detector illustrated in Figure (4).

Uncertainties
The relat ive uncertainty of the measurements is about 1.8%. The major source of error affecting the uncertainty was the spectral responsivity of the detector. This is due to lower uncertainties in spectral radio metric standards. However, it is possible to reduce the uncertainty, if more accurate spectral responsivity obtained. Also, to min imize errors the trap detector should be temperature-controlled. The diffraction fro m the aperture and fro m baffles and the interreflections has a significant effect on the measurements. Table (2) illustrate the elements of uncertainty and their contribution to the luminous intensity. Lamp no.

Relative differences
The SI unit of lu minous intensity, the Candela has been realized at National Institute for Standards (NIS) by using a filter radio meter which consists of trap detector QED200 and seven interference filters. The detector has been characterized for spectral sensitivity, non-uniformity and linearity to evaluate their stability as irradiance standard. Lu minous intensity of group of 10 la mps has been determined by filter radio metry method. These lamps which calibrated at National Physical Laboratory (NPL) with uncertainties of + 0.8 were co mpared with that measured by the trap detector. The comparison showed an agreement better than 1.8%., which lies within the uncertainty of the two scales.