CAUTION: Please consider the following to understand limitations of the StarCAT ETC test spectra.

(1) The photometric error (per resol) was heavily filtered to produce a smooth and relatively featureless curve ("smoothed 1σ photometric error").

(2) The flux density tracing was smoothed by a triangular profile of width comparable to the local spectral resolution (which can vary with wavelength if different resolution datasets were spliced to form the final spectrum).

(3) The photometric error (per resol) was similarly filtered and divided into the smoothed flux to yield a monochromatic tracing of S/N.

(4) The spectrum then was resampled onto a wavelength grid with uniform spacing of 0.2 Å over the range 1150-3150 Å. For each bin of the output spectrum, the flux density value is the maximum of the greater of the actual flux, if S/N > 3, or the smoothed 1σ photometric error, otherwise, taken over all the original points falling within the wavelength limits of the output bin. Thus, the resampled spectrum traces a flux envelope that always lies on, or above, the statistically significant ("3σ") fluxes, even though the resampled spectrum is highly undersampled in wavelength compared to the original. In the resampling procedure, if the original spectrum had a gap, or all of the fluxes within the limits of the output bin were flagged as bad, then the output flux density would be set to zero.

(5) Finally, the resampled spectrum was filtered to retain only the wavelengths with non-zero fluxes. Thus, gaps of varying size can occur within the resampled spectrum. The resulting dataset, however, is compact and suitable for rapid uploading to an ETC session. IT IS NOT SUITABLE FOR SCIENTIFIC ANALYSIS.


Mind the gaps! They should be obvious in an ETC run, but could affect the global count rate (underestimating it) if there are broad gaps within the free spectral range of the simulated mode. Also, any target feature falling within, or partially in, a gap will not be calculated properly, as far as the local count rate is concerned. Again, such cases should be recognizable by similar gaps in the ETC output.

A second consideration is that since the compacted spectrum always lies above the heavily smoothed 1σ photometric error, in situations where the intrinsic fluxes are closer to zero, and the spectrum is dominated by a few isolated emission features, global count rates could be overestimated, and thus always should be treated as conservative values. (Most such cases would not trigger a global count rate limit in the first place, however.)

At the same time, a local count rate calculation always should be correct (except in gaps) because the resampled fluxes are peak values from the original spectrum over each interval corresponding to the resampled bin. At wavelengths of sharp absorption lines, clipped by the maximization procedure, the calculated local count rate normally will refer to the surrounding continuum. Note, in addition, that the TOTAL COUNTS in an emission feature will be overestimated if the true profile is narrower than about 0.4 Å.

Finally, the user always should check the StarCAT preview spectrum of an object before conducting an ETC run: there are examples for which all, or part, of the original STIS exposure was essentially blank (object too faint or a failed ACQ), and consequently the processed ETC spectrum will trace solely the photometric error, which is highly unlikely to properly render the object's true energy distribution. Such cases are easily recognized at the preview level, however.