We have analyzed archival EXOSAT and HEAO 1-A2 observations of Cyg X-1. We find, in agreement with Ebisawa, that the X-ray spectrum of Cyg X-1 exhibits a "high-energy excess" above 8 keV, similar to that seen in Ginga spectra of active galactic nuclei. We show for the first time, using a likelihood ratio test, that the data are significantly better fit by a Compton reflection model than by a partial covering model. Even with the energy resolution of the EXOSAT GSPC we find that the derived iron line energy and shape are very dependent on the continuum model used. Using the best-fit Compton reflection model we find that the Fe line is not required to be either redshifted or broadened, although the data cannot distinguish between the skewed line profile expected from an illuminated accretion disk and a Gaussian line shape. The extended, 2-60 keV, energy response of the HEAO 1-A2 data can constrain the intrinsic continuum form. We find that the best-fit temperature for a Sunyaev-Titarchuk Comptonization model is significantly larger than that found in previous analysis based primarily on data in the 30-150 keV band. This suggests that the 5-50 keV continuum is not well fit by the same spectral form applicable at high energies and that a multiple temperature thermal Compton model is necessary to fit the broad-band data. The relative intensity of the Compton reflection component, combined with the known parameters of the Cyg X-1 system, is consistent with several models of the system, namely: 1. A point source origin of the X-ray emission situated above a flat disk with twice cosmic abundances, or 2. A spherical, optically thick source at the center of flat disk in which only Fe is over abundant, or 3. A optically thick source at the center of a flared disk with twice solar abundance. The measured Fe edge in the reflection spectrum is predominantly due to iron more ionized that Fe xv, inconsistent with that expected from a Shakura-Sunyaev disk. However the relatively low equivalent width of the Fe line, 44+/-28 eV is considerably smaller than that expected from a disk also responsible for the high-energy continuum. Only the line predicted from (3) gives a line within approximately 3-sigma of that observed, but a similar effect seen in other X-ray binary systems suggests that the lack of direct connection between the line intensity and reflection spectrum normalization may be a more widespread problem in models of X-ray illuminated accretion disks.
