ETC 18.0 and pysynphot
Version 18.0 of the ETC uses pysynphot as its synthetic photometry engine, rather than STSDAS.SYNPHOT. This change can sometimes lead to differences in simulated spectroscopic observations, and very occasionally to small differences (a few percent) in integrated count rates, signal-to-noise ratios, or exposure times for imaging mode simulations, as described below. Pysynphot has been tested extensively against previous SYNPHOT results and in all cases where the results differ the pysynphot results have been determined to be more accurate.
Pysynphot implements a number of improvements compared to SYNPHOT, including the use of double-precision arithmetic and improved resampling and interpolation methods (in wavelength space) for passband and spectral data. These changes result in convolved spectra that are more accurately interpolated in areas of high-frequency features, such as narrow emission and absorption lines or steep gradients on the edges of very narrow instrument passbands. This can lead to differences from previous SYNPHOT results for spectral mode simulations, possibly changing the computed count rate in individual wavelength bins near such features by up to 30%, but in most situations, even these cases produce negligible differences in the integrated count rate of a source used in an imaging mode simulation.
On very rare occasions, such as when computing simulations involving a complicated source spectrum observed through a very narrow band imaging filter, the pysynphot results can differ from SYNPHOT by 1-2%. As stated above, all such differences that have been found during pysynphot testing have consistently shown the pysynphot results to be more accurate.
Another difference in behavior that users may notice is in the handling of simulations where either the instrumental throughput or the source spectral data are incomplete within some wavelength range necessary for the calculations. The previous SYNPHOT version of the ETC gave users no indication that anything was wrong in these instances, whereas the new pysynphot version will at least producing a warning regarding the validity of the results, or return an error message with no result at all.
Further detail about the testing process is described in the commissioning report, TSR 2009-01, available at http://stsdas.stsci.edu/tsr/2009_01.