GHRS Instrument Handbook
The previous chapter provided the information needed to estimate an exposure time to achieve a given level of signal-to-noise. We reiterate several factors having to do with the detectors that must be taken into account to achieve the best data quality. Note that it is not necessary to explicitly specify these parameters (except for FP-SPLIT) because the defaults that apply to each mode of operation will automatically invoke them. Moreover, you should not deviate from the defaults without good reason.
The Digicon detectors have faceplates with some granularity (uneven response). The diodes onto which the faceplate is imaged also have response irregularities and some of them have been turned off because of misbehavior. Both of these effects are relatively small but enough to prevent you from obtaining a spectrum with S/N much in excess of about 30. They can also produce "glitches" that can mimic spectrum features. The FP-SPLIT parameter causes the carrousel to move slightly between each of the two or four separate subexposures (depending on the value you choose). The COMB parameter suppresses diode-to-diode gain variations and allows one to work around the dead diodes. Both features should normally be used, especially since they cost little or nothing in exposure time and improve data quality.
The Digicon diodes also undersample the spectrum by about a factor of two. The parameter STEP-PATT causes electronic motions of the spectrum so as to sample the spectrum fully. It is possible to STEP-PATT at two samples per diode width, but we recommend using four samples per diode to yield optimum results, and again at no net cost. You can always rebin a quarter-stepped spectrum into a half-stepped one during your data analysis, but the process cannot make a quarter-stepped spectrum out of a half-stepped one. Deconvolution has worked best with quarter-stepped spectra (the default); see Gilliland et al. (1992). STEP-PATT also determines the way in which the background is measured (see Section on page 112).
We also remind you to break up long exposures into subexposures that are no longer than about 5 to 10 minutes each, so as to defeat the effects of geomagnetically-induced image motion. Bear in mind that a 20 minute exposure, for example, specified with FP-SPLIT=4 will result in four 5-minute exposures.
Remember to have wavelength calibration exposures precede the ACCUM to which they apply. This minimizes the time interval between these two exposures, so that wavelength drifts do not occur. Use SEQ <exp list> NON-INT as a Special Requirement to make sure the two lines as a group are not split.
Wavelength calibration exposures cannot ordinarily be scheduled during Earth occultation. The reasons have to do again with the interruptibility of GHRS exposures and the fact that the sequence of exposures that actually executes bears only a casual resemblance to what RPS2 shows you due to SAA passages. In a few cases we have executed WAVE exposures during occultation, but the manual effort needed is substantial, meaning that the science requirements have to be unusual and demanding.
There are calibration exposures made called SPYBALs that can often substitute for a WAVE exposure. A SPYBAL is an exposure of the wavelength calibration lamp, but it is made at a fixed setting for each grating as a means of centering the spectrum on the diodes in the direction perpendicular to dispersion (SPectrum Y BALance). Your WAVE exposure, on the other hand, would be made at the specific wavelength you're observing at. Clearly the latter is superior, but in many cases the SPYBAL exposure contains enough information to correct the wavelength zero point of your spectrum very nearly as well as the WAVE exposure would have.
SPYBALs are executed every time a new grating is used and about once every other orbit thereafter (the details of the thereafter depend on how you write your exposure lines and are impossible to generalize about). Sometimes an observer wishes to ensure that the grating carrousel is not moved at all until they have done all their exposures at a particular setting. That can be achieved with SPYBAL=NO. However, the defaults almost always provide satisfactory results.
A problem can arise when a science program specifies a large number of separate GHRS exposures. This problem is caused by the relatively small amount of memory available on HST in which to store GHRS commands. It is usually possible to break up such a program into several visits so that the separate exposures are not all together, but occasionally the science goals cannot allow that and some other compromise must be made. Roughly speaking, about 40 total spectra can be scheduled in a single visit (a WSCAN with n set-points counts as n exposures and an FP-SPLIT counts as 2 or 4). Once that number is exceeded the remaining observations must be scheduled in a new visit, and that means a new target acquisition will be needed, with the resultant overhead time. Using a large number of iterations in an ACCUM does not cause a memory problem, but what does is lots of different instrument settings.
- Specify the aperture as "2.0" (LSA) or "0.25" (SSA). The object will automatically be moved to the correct aperture even if the acquisition was into the other aperture. If a SSA spectroscopic observation follows an LSA spectroscopic observation, we recommend an ACQ/PEAKUP in the SSA with SEARCH-SIZE=5 before beginning an ACCUM.
- If wavelength accuracy is needed that exceeds the default (see Section on page 62), then specify WAVE as the target with an aperture of SC2. Specify a WAVE before the ACCUM to which it applies.
- Specify the grating to be used, either first-order or echelle. If you wish to force an echelle observation to be done in an order other than the default, you may do so by specifying the grating as, for example, ECH-B24, where 24 was the order chosen.
- STEP-PATT may be chosen as a number from 1 to 15, and specific pattern numbers go with specific spectrograph configurations. We recommend using the default that pertains to the setup you have chosen (STEP-PATT=DEF). The details of how the substepping is performed and the background measured are given in Section on page 112.
- FP-SPLIT=STD is recommended if you hope to achieve a signal-to-noise level of about 30 or better. The default for FP-SPLIT is NO, which will not yield a spectrum with the best signal-to-noise. However, restoring your observations to their full spectroscopic resolution depends on having enough signal-to-noise in each individual subexposure to achieve a satisfactory cross-correlation. In other words, you should not rely on a priori knowledge of carrousel position to shift the individual FP-SPLIT exposures back to a common wavelength zero point. For this reason you should probably not use FP-SPLIT at all for signal-to-noise levels below about 30.
- COMB=FOUR is the default value and is recommended for the best results.
- Once you have chosen the STEP-PATT, note the minimum exposure time in the last column of Table 8.5 on page 114; this is the time it takes to go through one complete cycle of the pattern, and your total exposure time should be an integral multiple of this number. If you are using FP-SPLIT, however, multiply the cycle time by the FP-SPLIT value (2 or 4) to get the minimum exposure time. For example, STEP-PATT=5 is frequently used, and its minimum exposure time is 27.2 seconds (these values are based on using COMB=FOUR, which is recommended). If FP-SPLIT=STD is used, the minimum time is 4 times 27.2, or 108.8 seconds. If you wished to spend about 20 minutes on an object, the exposure time to choose would be 1196.8 seconds, which is 11 cycles. If you specified a time of 20 minutes, the actual value used would be 1196.8 anyway, even if RPS2 makes it appear as if the full 20 minutes is used (this aspect of RPS2 may change by the time you prepare a Cycle 6 Phase II proposal).
- DOPPLER=DEF is recommended. This activates compensation for the velocity shifts of astronomical spectra over the course of an orbit but turns it off for internal exposures.
- STEP-TIME may be specified as a number from 0.2 to 12.75 seconds, in increments of 0.05 seconds. STEP-TIME specifies the length of the individual subspectra that are accumulated to form the final spectrum, and there is no good reason to not use the default of 0.2 sec.
- Specify Wavelength as the central wavelength (preferred) or wavelength range for the exposure, in Ångstroms.
- The CENSOR parameter may also be specified. The default is NO, which is appropriate in almost all cases. If CENSOR=YES is used, individual subspectra are examined onboard the spacecraft and are discarded if multiple counts have occurred within a 8 ms interval. This allows for the lowering of background noise in cases where the object being observed is very faint, i.e., less than about 0.1 counts per second per diode. Rejected exposures are repeated by the GHRS, leading to a longer total elapsed time for the observation, but only by about 2%. See Section on page 116 for more information on CENSOR.
- A special commanding option called FLYLIM can also be used to reject noise in cases where the object is substantially fainter than the background. See Section on page 118.
- If you have any doubts about the manner in which your program will be executed, we recommend that you remove the ambiguity by explicitly indicating all the Optional Parameters and Special Requirements.
Use of WSCAN can result in a spectrum covering a broader total bandpass than is possible with a single exposure. All the parameters listed above for an ACCUM exposure are available in WSCAN mode. The most important parameter to specify is WAVE-STEP, which is the spacing (in Ångstroms) between each subexposure. If WAVE-STEP=DEF is specified, the central wavelengths of the separate exposures will be equally spaced so as to cover the range of wavelengths that you specify, with at least 20% overlap from one subspectrum to the next.
You may also explicitly give a WAVE-STEP value. If lmin is the central wavelength of the shortest-wavelength exposure, and lmax is the central wavelength of the longest-wavelength exposure, then choose these values in concert with WAVE-STEP so as to yield an integral number of WAVE-STEPs between lmin and lmax.
OSCAN mode makes it possible to scan across echelle orders at a fixed value of ml, where m is the order number and l is the wavelength. It is rare that adjacent orders both have features of astrophysical interest and so this mode is primarily used for calibrations and not for science observations. If you do use this mode, all the parameters of an ACCUM observation are available.