High altitude balloon platforms are capable of flying diffraction limited telescopes with numerous advantages over orbital observatories such as the Hubble Space Telescope. A requisite for such long-term missions is an attitude determination system that can operate diurnally with sub-arcsecond accuracy to provide continuous attitude knowledge. The common choice for such an instrument is a star tracker; however, even at altitudes above 30 kilometers, atmospheric scattering of daylight produces enough ambient light to prevent star trackers from operating in the full visible spectrum. DayStar, designed and constructed at the University of Colorado at Boulder, is a prototype star tracker that combines a high quality CMOS camera and custom filtered optics to provide an attitude solution during the day that is accurate to better than 1.0 arcseconds RMS. The system design is qualitatively straightforward; the blue end of the spectrum is filtered out, eliminating most of the scattered daylight and leaving many stars in the red portion of the spectrum visible. A camera with sufficient red performance can then capture the residual light. However, quantifying this system requires that the ambient background, starlight and camera performance all be characterized as a function of wavelength, which has proven to be nontrivial. In this paper the system design process for DayStar is discussed, focusing on the modeling required to quantify its daytime performance. Such a model demonstrates that, despite daytime ambient light conditions in the stratosphere, a star tracker can still operate with accuracy comparable to the diffraction limit of high performance telescopes. By using a star tracker such as DayStar, a high altitude balloon observatory would be able to match the image quality of Hubble for a fraction of the price. NomenclatureBC = bolometric correction c = speed of light DC = dark current F = total flux h = Planck's constant I = wavelength dependent flux k = Boltzmann constant λ = wavelength m b = bolometric magnitude m v = visual magnitude n pix = number of pixels covered by a star Ω = solid angle QE = quantum efficiency ρ = internal reflectivity R background = background light flux R star = star light flux RN = read noise SNR = signal to noise ratio T = temperature t exp = exposure time