The HIRES spectrograph at Keck is a grating cross-dispersed echelle spectrograph. Albeit there is a lot of useful information on the HIRES webpage, here are a few more tips that may help you plan your observing run.
If you are observing wavelengths above 6000 Å, one important concern is second order contamination. The HIRES spectrograph uses two gratings to disperse light, so that you can see fine details in the spectra. The main grating, the "echelle grating", is a diffraction grating optimised for high spectral orders. Those high orders overlap, so you cannot distinguish them (they would fall at the same observing angle, i.e., at the same place in the detector). A second grating, which is perpendicular to the echelle grating, is used to disperse the echeller orders, so that they can fall at different regions on the detector. This second grating is the "cross disperser". Notice that a prism could be used instead as a cross disperser. For example, the MIKE spectrograph at the Magellan telescope uses a prism to disperse the echelle orders.
When you are preparing your HIRES configuration, you must be careful about potential second order contamination (caused by the second grating -the cross disperser). "Second order" wavelengths below half of your maximum wavelength will contaminate your "first order" spectra. This doesn't happen with the MIKE spectrograph, because the cross disperser is a prism. You must worry about second order contamination with HIRES only if you are observing for lambda_1 (first order) > 6000 A. The reason is that for lambda_1 = 6000 A, the second order wavelength is half that value, lambda_2 = 3000 A (the ultraviolet atmospheric cutoff). So, Earth's atmosphere would be your "filter", blocking the second order light below 3000 A.
In practical terms, you must define first your maximum wavelength and then use a filter to cut the second order contamination of wavelengths half of that maximum value. For example, if you would like to observe up to lambda_1 = 9000 A, then you must use a filter to block lambda_2 ≤ 4500 A. There are several filters available with HIRES, so that you can use the more convenient. For example, if you want to observe the OI triplet at 777 nm, then you may want to set up a maximum lambda_1 = 778m, corresponding to lambda_2 = 389 nm, and so you may choose the kv389 filter. Notice, however, that the filters are not perfectly sharp, so that there would be some small second order contamination around the OI triplet. This contamination is minor, so you may be OK with it. All depends on the level of extra "noise" that you're willing to accept. For the majority of observers aiming at low-moderate S/N (let's say, S/N ≤ 300), this should be OK. If you want to exclude any minor contamination to the OI triplet, then you must use a redder filter. In the example above, you would need a filter with lambda > 389nm. The problem is that if you use a filter that is too "red", then you may also block interesting regions of the first order light. For example, if you use a filter to block lambda < 400 nm, then you will miss the H and K lines.
My favourite configuration with HIRES/red is to use the cross-disperser angle xdisp = +0.24, echelle angle = 0.0, and filter kv389. In this configuration the useful spectral coverage would be from about 390 - 780 nm, so you can get data from the H and K lines in the UV to the OI triplet at 777nm. The HIRES mosaic detector would actually cover a few more orders in the red, but they would be contaminated by second order light. So, my reddest useful order in this configuration would be the order of the OI triplet.
Regarding the spectral resolution, you can achieve R ~ 90 000 - 95 000 with HIRES. This is based on my own measurements taken in 2004 - 2006. The highest resolution is achieved with the 0.4" slit, corresponding to about 2 pixels, so you MUST use 1x1 binning (i.e., no binning). For the highest resolution I recommend the E4 slit, that is long enough for sky subtraction. If you're using a slit ≥ 0,8 arc sec, then you can bin by two (2x2 binning) without loosing any spectral resolution.
