-- Steven W. Brown (swbrown@nist.gov), B. Carol Johnson (cjohnson@email.nist.gov), and Howard W. Yoon, Optical Technology Division, National Institute of Standards and Technology, Gaithersburg, MD 20899
The Earth Observing System (EOS) is an 18-year international, multi-instrument, multi-satellite program in global remote sensing of the Earth. The overall goal of the program is to advance the scientific understanding of the entire Earth system and its changes on the global scale. To accomplish this goal, EOS instrumentation will measure the Earth's radiation flux at wavelengths ranging from the UV to the IR, producing global, long-time-series, remote sensing data sets. To correctly interpret the scientific information produced by a variety of instrumentation on different satellite platforms over the 18-year life of the program, it is critical to be able to discriminate between changes in the sensor performance used for the global change study and actual changes in the Earth's radiance. Consequently, accurate calibrations of sensors used to measure radiance properties of the Earth are central to the success of the mission. A variety of calibration methods have been developed, including ground-based calibrations prior to launch, ground-truth measurements of satellite-based sensors after launch, periodic lunar and deep-space observations, and incorporation of on-board calibration sources.
As part of the overall sensor calibration scheme, several portable radiometers have been developed or are under development at the National Institute of Standards and Technology (NIST) to provide ground-based radiometric traceability to NIST of EOS instrumentation radiance responsivity scales [1,2]. In practice, these radiometers will travel to National Aeronautics and Space Administration (NASA) instrument calibration and validation facilities and measure optical sources used to calibrate the sensor responsivities. The result is a validation of the radiance scale of the ground calibration sources. Sources are typically lamp-based integrating sphere sources for the visible and shortwave infrared wavelength regions, and large area blackbodies for the thermal infrared.
A total of three radiometers, the Visible Transfer Radiometer (VXR), the Short-Wave Infrared Transfer Radiometer (SWIXR), and the Thermal Infrared Transfer Radiometer (TXR), are currently planned to cover the wavelength range from 0.4 µm to 10µm. Radiance scales will be verified in the visible using the VXR, a six-channel filter radiometer with channel wavelengths ranging from 411 nm to 870 nm. The VXR is a modified version of the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) Transfer Radiometer [3]. The TXR, a two-filter radiometer with channels at 5 µm and 10 µm, will measure cryogenic blackbody working standards either in air or in a vacuum environment[4]. In this article, we will focus on design and characterization details of the SWIXR radiometer under development for radiance scale validations of integrating sphere sources in the 0.8 µm to 2.5 µm wavelength range. In Section II, the system layout and design will be described. In Section III, details of the system characterization and calibration plans will be presented.
A radiometer is in general composed of three basic components: collection optics, a spectral filter, and a detector. The SWIXR instrument is equipped with all-reflective input optics, a double-grating monochromator for spectral selectivity, and a liquid-nitrogen-cooled indium antimonide (InSb) detector. Each of these components will be described briefly.
The input optics for the SWIXR, shown in Fig. 1, consist of an entrance aperture (A); an optional insert to hold a filter wheel and a chopper (B); baffles with defined apertures (C and D); two gold-coated mirrors, the first a 300 mm focal length spherical mirror (E) and the second a 100 mm focal length off-axis parabolic mirror (F); and an exit aperture (G). The 6 mm by 6 mm entrance aperture is imaged to a 2.2 mm square spot at the entrance slits of the monochromator located 6 mm behind the exit aperture (G). The aperture stop is located at aperture D while the entrance slits of the monochromator act as the field stop. The f/# of the radiation imaged onto the monochromator entrance slits is approximately f/3.95, closely matching the f/# of the monochromator for maximum throughput and spectral resolution. The system has a full angle field-of-view of 5.2o, resulting in a 9.6 cm by 9.6 cm square entrance window at the exit port of the integrating sphere source if the SWIXR is located 1 m from the sphere. The input optic housing is 36.7 cm long and 16.7 cm wide. The overall height is 6.1 cm. The optional chopper/filter wheel assembly extends an additional 6.8 cm above the top of the housing.
[Fig. 1. Schematic diagram of the input optics module for
the SWIXR transfer radiometer. A is the entrance aperture (6
mm x 6 mm); B is the optional filter wheel/chopper insert; C is
a baffle with a 15.75 mm diameter aperture; D is a baffle with
a 21.7 mm diameter aperture; E is a 300 mm focal
length, gold-coated spherical mirror; F is a 100 mm focal length,
off-axis parabolic mirror; G is the exit aperture.]
The monochromator is an ISA, Inc.[ Identification of commercial equipment does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the equipment identified is necessarily the best available for the purpose.] 0.18 m, f/3.9 double grating instrument (Fig.2). It has a stray light rejection of approximately 10-7, wavelength reproducibility of 0.1 nm, and an uncorrected wavelength accuracy of 0.5 nm. The incident light slightly underfills the monochromator's optical elements to minimize scattered light. Protected silver coatings were used on all mirrors for maximum throughput. The instrument comes with two sets of interchangeable gratings with groove densities of 600 g/mm, one set blazed at 1.5 µm and a second set blazed at 2.0 µm. The nominal bandpass of the instrument with these gratings is 7.2 nm/mm. The entrance, center, and exit slits are externally controllable, enabling user-selectable bandpasses throughout the entire wavelength range. The monochromator is 45.7 cm long, 41.9 cm wide, and 21 cm tall.
[Fig. 2. Schematic diagram of the ISA, Inc. 0.18 m
double grating monochromator: (M) Mirrors, (G) Gratings,
(ES) Entrance Slits, (CS) Center Slits, (XS) Exit Slits.]
The output from the monochromator is imaged onto the detector using an adjustable, silver-coated, elliptical mirror (not shown). The nominal spot size of the imaged radiation at the detector surface is 0.5 mm by 0.5 mm. A 2.5 mm diameter InSb photodiode mounted in an evacuated dewar and cooled to 77K is used to detect the transmitted radiation. A 2.0 mm diameter aperture is located less than 1 mm in front of the detector to reduce detector instabilities arising from light hitting the edge of the detector. A silicon diode temperature sensor located on the cold finger measures the diode temperature. The dewar is equipped with a sapphire entrance window. A cold filter with a transmittance of less than 10-3 over the spectral range from 3.0µm to 5.5 µm is placed behind the sapphire window to reduce stray light and thermal infrared background radiation incident on the detector [5]. Order sorting filters further reduce stray light in the system. The chopped output from the detector is sent to a three-stage trans-impedance amplifier and then to a lock-in amplifier. The signal measured by the lock-in amplifier is recorded with a computer.
The radiometer is approximately 1 m long, 0.5 m wide and weighs approximately 25 kg. The instrument is mounted on an optical breadboard and placed on a small adjustable table for orientation. A removable visible diode laser mounts to the front of the radiometer to aid in the proper alignment of the instrument during sphere measurements.
The measured signal S(l) from the radiometer is proportional to the area of the entrance pupil, A; the projected solid angle, W; the radiance of the source being measured, L(l); the transmittance of the spectral filter (and any other optical components), T(l); the spectral responsivity of the detector, R(l); and a detector amplifier gain factor, G:
(1)
To calibrate the radiometer for absolute responsivity, we relate the measured signal S (V) to the radiance of a calibration source L (W/m2sr µm):
(2)
where the integration is over the spectral bandpass of the monochromator. R' includes the area, solid angle, system transmittance, detector responsivity and gain factor from Eq. 1. The complete calibration of the instrument consists of measuring the absolute spectral radiance responsivity at each wavelength (or known spectral bandpass) over the entire spectral range.
In general, radiance scales can be maintained with either calibrated detectors or stable sources of known radiance. In this case, the radiance scale will be transferred from working standard lamps to a lamp-based integrating sphere source at the NIST Facility for Automated Spectroradiometric Calibrations (FASCAL) [6]. The integrating sphere source radiance will be measured every 50 nm, and also at wavelengths corresponding to the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and the Moderate Resolution Imaging Spectroradiometer (MODIS) band center wavelengths [7,8]. This calibrated integrating sphere source will subsequently be used to calibrate the SWIXR for spectral radiance responsivity.
The integrating sphere source will be chosen to have a radiance level and spectral distribution similar to sphere sources used for EOS instrument calibrations. For example, in Fig. 3 we show the spectral radiance of an Optronic Laboratories, Inc. OL 450 integrating sphere along with the maximum and low radiance levels of the ASTER sphere source [9] at wavelengths corresponding to ASTER bands 4 through 9. A similar integrating sphere source will be used to calibrate the SWIXR for absolute spectral responsivity.
[Fig. 3. OL 450 sphere radiance (open circles) and
maximum (solid circles) and low-level (solid squares) radiance levels
of the ASTER sphere source.]
Prior to calibration, the instrument will be characterized for linear response, wavelength accuracy and repeatibility, spectral bandpass, stray light, and size-of-source error. Many of these measurements will be made at a new large-area monochromatic Lambertian source facility currently under development at NIST [10]. While we are interested primarily in the short-wave infrared region, this tunable-laser-based facility will enable measurements over the wavelength range from 0.2 µm to 12 µm.
To provide a general assessment of expected uncertainties of measurements made with the SWIXR, we estimate errors arising from wavelength error, stray light, bandpass filters, and signal-to-noise. These sources of measurement uncertainty will be combined in quadrature with other sources of uncertainty arising from the instrument calibration to give us an estimate of the total uncertainty.
For calculations of errors arising from wavelength error and stray light, we need an estimate of the spectral dependence of the measured signal. For this estimate, we neglect optical elements in the radiometer, such as mirrors, which have little or no spectral variation over the wavelength range of interest. Only the grating diffraction efficiency and the responsivity of the InSb detector show a pronounced spectral dependence. The monochromator transmittance is proportional to the product of the two grating diffraction efficiencies. For the grating diffraction efficiencies, we take the diffraction efficiency provided by ISA, Inc. for a grating blazed at 2.0 µm, measured under Littrow conditions. We also consider only the relative responsivity of the InSb detector. In this case, we allowed the relative responsivity to vary linearly from 1 to 3 over the wavelength range from 0.8 µm to 2.5µm, in rough agreement with published values. Using the spectral power distribution of the OL450 integrating sphere, we then calculate the relative spectral distribution of the signal.
To estimate effects of wavelength error on
measured radiance, we simply take the ratio of the derivative
of the estimated signal and the signal, 
Results are shown in Fig. 4. The wavelength expanded
uncertainty of the monochromator is 0.5 nm. With
wavelength calibration using pen lamps, the expanded
uncertainty should be reduced to at least 0.3 nm or less, the
limit being given by the reproducibility of the
wavelength setting0.1 nm. Expanded uncertainties in the
radiance arising from monochromator wavelength
error should therefore be on the order of 0.45% or
less throughout the entire wavelength range.
[Fig. 4. Estimated error in measured signal as a function of
wavelength error (open circles) and stray light (closed circles) for a stray light factor
of 10-7.]
For stray light, we take the ratio of the out-of-band signal to the in-band signal, and express that ratio in per cent:
(3)
For simplicity, we have replaced the integral in Eq. 2 by a summation and assumed a constant bandpass of 10 nm. We have also assumed a slit-scattering function of 10-7 out-of-band and 1 in-band. While the summation is in principle over all wavelengths, we perform the summation over the restricted wavelength range from 800 nm to 2500 nm. This is reasonable because order-sorting filters further attenuate visible light by a factor of 103 and the cold filter attenuates longer-wavelength radiation by an additional factor of 103. Based on these calculations, stray light of 10-7 should introduce errors less than 0.02% (Fig. 4).
Order sorting filters are necessary to eliminate higher order diffracted light from passing through the monohromator. Spectral selectivity is typically achieved by using multilayer dielectric coatings. Due to thermal expansion and contraction, the relative thickness of the layers will change slightly with temperature, and the filters will exhibit a temperature dependence to their spectral transmittance. While typically 'very small', this effect has not been well characterized. Interference filters, for example, will shift to longer wavelength with increasing temperature. This wavelength shift is on the order of 0.01nm/oC to 0.2 nm/oC, depending on the particular interference filter.
Well away from the cut-on wavelength, the typical filter transmittances of the two order sorting filters in the radiometer are smoothly varying with small oscillations on the order of 5% of the transmittance with an interval of 100 nm to 200 nm. The change in transmittance as a function of wavelength shift (or changing temperature) should therefore be fairly small, on the order of 0.05%/nm. For temperature variations as large as 5o C , the maximum filter wavelength shift would be 1 nm, corresponding to a change in filter transmittance of approximately 0.05%.
The signal-to-noise (S/N) ratio can be expressed as the ratio between the effective flux incident on the detector and the noise equivalent power (NEP). The effective flux is proportional to the radiance of the source, the area of the entrance pupil, the projected solid angle, the bandwidth of the spectrometer, and the transmittance of the optical system:
(4)
We will estimate the S/N at 2.5 µm, a wavelength where the transmitted flux is small. For a radiance value, we take the radiance of the OL 450 sphere,
~5 W/cm3sr. In estimating system transmittance,
we include spectrometer transmittance (for a
bandwidth of 15 nm), losses due to mirror reflections, and
losses due to window and filter reflections. Taking all
these parameters into account, 
The noise equivalent power is the incident power that produces an rms output signal equal to the rms noise voltage [11]. The NEP is often expressed in terms of the detectivity D*(cm Hz1/2/W),
(5)
where Ad is the detector area and Df is the detection bandwidth. For our InSb detector, D* is at least 1012, the detector area approximately 5 mm2, and the detection bandwidth 1 Hz. The NEP is therefore approximately 0.20 pW. The minimum S/N ratio should correspondingly be on the order of 10,000 during calibrations. In measuring sphere sources, the incident flux could be an order of magnitude lower. In this case, the S/N ratio should be on the order of 1000. This estimate does not take into account other sources of noise, such as amplifier noise.
The dominant sources of uncertainty arise from the calibration of the radiometer. The uncertainty in the radiance scale transfer from FASCAL to the integrating sphere source used to calibrate the SWIXR may be as high as 1.5%, and the uncertainty in the measurement of the calibration sphere as high as 0.5%. Using these upper-bound estimates, the root sum square of the individual uncertainty components gives a maximum total expected uncertainty in measurements of sphere radiance using the SWIXR of 1.7%. While this is a rough estimate of the uncertainty budget, based on this result we anticipate having the capability to make radiance measurements of EOS calibration sphere sources in the short-wave infrared with a maximum total uncertainty of 2% or less.
A portable short-wave infrared radiometer is under development at NIST to provide ground-based radiometric traceability to NIST of EOS instrumentation radiance responsivity scales in the 0.8 µm to 2.5 µm wavelength range. This instrument is one of three portable radiometers being developed at NIST to validate radiance scales of sources used at NASA calibration facilities to measure the absolute spectral responsivities of space-based instrumentation prior to launch. This is one step in a calibration chain designed to enable accurate long-term monitoring of the Earth's radiation properties as part of NASA's Earth Science Enterprise.
[1] Butler J. J., and B. C. Johnson, 1996:
Organization and Implementation of Calibration in the Earth
Observing System (EOS) Project - Part 1. The Earth
Observer 8(1), 22 - 27.
[2] Butler J. J., and B. C. Johnson, 1996: Calibration
in the Earth Observing System (EOS) Project - Part
2: Implementation. The Earth Observer,
8(2), 26 - 31.
[3] Johnson B.C., J. B. Fowler, and C. L. Cromer,
1998: The SeaWiFS Transfer Radiometer. SeaWiFS
Postlaunch Technical Report Series, S. B. Hooker and E.
R. Firestone, Eds., Greenbelt, Maryland (submitted).
[4] Rice J. P. and B. C. Johnson, 1996: A NIST
Thermal Infrared Transfer Standard Radiometer for the
EOS Program. The Earth Observer, 8(3), 31 - 35.
[5] Model PK50 from Barr Associates, Inc.,
Westford, MA.
[6] Walker, J. H., R. D. Saunders, and A. T.
Hattenburg, 1987: Spectral Radiance Calibrations. NBS
Special Publication SP250-1, U.S. Government Printing
Office, Washington, DC.
[7] Fujisada H., 1994: Overview of ASTER
Instrument on EOS AM-1 Platform. Proc.
SPIE, 2268, 14-36.
[8] Barnes W. L. and V. V. Salomonson, 1993: MODIS:
A Global Imaging Spectroradiometer for the Earth Observing System.
Crit. Rev. Opt. Sci. Technol., CR47, 285 - 307.
[9] The source is a one meter integrating sphere
(355 mm exit port), manufactured by LabSphere,
Inc., equipped with eighteen 30 W main lamps and two
200 W satellite lamps. The inner walls are coated
with Spectraflect. The sphere is maintained by
Mitsubishi Electric Corporation, Japan.
[10] Lykke, K. R., P.-S. Shaw, L. M. Hanssen, and G.
P. Eppeldauer, 1998: Development of a
Monochromatic Uniform Source Facility for Calibration of
Radiance and Irradiance Detectors from 0.2 to 12 µm. to
be published in Metrologia.
[11] Wyatt, Clair L., 1987: Radiometric System
Design. Macmillan Publishing Co., New York, 1987, Chapter 6.