Spectrophotometric Flux Calibration F-flux, C-count rate, S-sensitivity F(l )= C(l)/S(l) or S = C/Fstd Details: Net Count Rate, C 1. 2. 3. 4. 5. 6. 7. 8. 9. Background Flat Field, P and L Stray Light Noise and Artifacts, eg Dust and Fringing Operational Changes, eg G140L +3" ‡ -3" Wavelength Calibration Temperature Time Non-linearities, eg CTE losses Details: Standard Star Flux, Fstd 1. 2. 3. 4. 5. Teff & G from Balmer lines Finley V band normalization of Landolt LTE Koester = Hubeny LTE ‡ non-LTE WD agrees w/ Solar Analog in IR MAJOR CONTRIBUTORS TO THE UNCERTAINTY OF A FLUX DISTRIBUTION OF A SINGLE OBSERVATION - RELATIVE TO THE V BAND - IN ADDITION TO COUNTING STATISTICS OF THE OBSERVATION I. UNCERTAINTIES IN THE COUNT RATES A. REPEATABILITY IN BROAD BANDS MODE Bandpass (A) one sigma (%) G140L 350 50 0.54 0.46-.79* G230L 1000 100 0.18 0.19-.59* G230LB 1000 100 0.27 0.27-.39* G430L 2000 200 0.32 0.20-.56* G750L 2800 400 0.15 0.13-.91* * Small correction for counting statistics B. UNCERTAINTY (%) IN CTE CORRECTION TIME 100e- 1 YEAR 5 YEARS 0.6 3 1000e0.3 1.5 10000e~0 ~0 II. UNCERTAINTIES IN THE MODELS A. STELLAR TEMPERATURE uncertainties, eg formal uncertainty of +/- 1000K for G191B2B at 61200K makes <<1% flux differences. B. INDEPENDENT HUBENY, KOESTER LTE CODES now agree in the continuum to a worst case of 1% at 3microns. C. RELATIVE STIS FLUXES agree w/ models to 1% at 2000A and below. D. NICMOS SOLAR ANALOG AND G191B2B AGREE TO ~1% at 1.5microns and beyond E. COMPOSITE MODEL ERROR BUDGET is 1% to 2000A, deceasing linearly to zero at 5500A, and increasing linearly to 1% in the 1.5-3micron range.