The limiting spectral resolving power, R ~ 165,000 (1.8 km/s), is achieved with two pixel (27 μ) sampling. A standard slit width (0.4 arcsec) provides a resolving power of R ~ 90,000 (3.3 km/s).
The efficiency curves of the system measured by wide-slit observations of standard stars are shown in Figure . The efficiency includes the atmospheric transparency, the throughput of the telescope and the spectrograph, and the quantum efficiency of the detector. It does not take into account the I cell, which has a throughput of about 85%. The efficiency curves for the blue and red setups of the spectrograph cross at about 4400Å.
In the Figure the efficiency curves of the ADC and the two image rotators are shown. Very recently we have discovered a degradation of the efficiency of both (red and blue) image rotators, so be careful to employ them in your program. This effect has been implemented into the Exposure Time Calculator (after 1/28/14).
Three image slicers (IS) are available, and can be inserted int the optical path. Their transmission curves are also shown in Figure below.
As shown in Figure , and in the Appendices (Section ), typically 1700Å and 2500Å are covered by one exposure for the blue and red setups, respectively. Note that the central wavelength region (one or a few orders) is not observed because of the gap between the two CCDs. For wavelengths longer than 7200Å, the free spectral range is not covered by the CCD, hence a continuous spectrum cannot be obtained with one exposure.
The maximum slit length as a function of wavelength is given in section for each setup using blue or red cross disperser.
Single order observing mode is also available by selecting order with some narrow band filter. For this observing mode, one can use the plane mirror in stead of cross dispersing grating. The maximum slit length is 60 arcsec for this case.
Even for the ``same'' setup of the spectrograph, the spectrum format can shift as a result of changes in the inclination angle of the echelle gratings and the cross-dispersing grating. The format may also be affected by changes in the collimator mirror. The repeatability of the spectrum format through the changes of these setups is within about one pixel of CCD on the detector.
The spectrum format can shift even if the setup is fixed. The shift is primarily dependent on the temperature of the Nasmyth enclosure (temperature of the spectrograph). The measured shift along the dispersion direction is about 1.4 CCD pixels per degree. The temperature in the Nasmyth enclosure is not actively controlled, but the temperature variation is expected to be less than 0.1 degree per hour.
The continuum profile is mainly determined by the echelle blaze profile. But it is known that the profile changes during observation. The quality of the flux calibration using standard stars is limited by this problem. The change of the continuum profile is sometimes as large as 10%. The reason of this problem is under investigation, but it is likely casued by changes of the optical path within the telescope, and related to the telescope position during observations (alt-az) and/or focus length.
Since the temperature of the detector is well controlled, the spectrum, including the fringe pattern originating on the surface of the CCDs, is reproduced by different exposures. Therefore there is usually no problem obtaining calibration data for flat-fielding before and/or after the observation (see Section ).
The readout noise of the CCDs is 4.4 e. A time variation of 2-4 e has been measured in different bias exposures. This variation can be corrected by use of the data in the over-scan region (see Section for details). The dark current of the detector is lower than 1 e per hour. However, there is some emission in the Nasmyth enclosure or some leakage of light from outside which causes almost homogeneous dark level about 10 e per hour. The gain (conversion factor) for the readout is about 1.7 e/ADU and is slightly dependent on the output amplifier (see Section ).
A good linearity is confirmed in the data recorded for electron numbers less than 10,000e (~6,000ADU), significant nonlinearity appears for higher electron numbers. The nonlinearity effect is of the order of several percent at 50,000e. No clear pixel-to-pixel difference is observed in the nonlinearity effect in a given chip; however, there is a notable difference in the nonlinearity effect between two CCD chips used in the instrument. No significant time variation in the nonlinearity effect is found from our measurement up to June 2010.
See Tajitsu et al. 2010 for description of the effect. An IRAF CL-script for the correction of the efect is also available.