Introduction

The details of these procedures are not yet finalised. In particular there are good reasons for using a rather bright star, but peculiarities of the software which can complicate this. These are noted where relevant.

A. Observing procedure for CGS4

1. Choose and point to an A star, with 6.5 < V < 7 near the zenith. (At this spectral type the K magnitude is about the same as the V mag in the catalogue.)
2. In a manual config, set , 1st order, 2x1 sampling (say: its irrelevant), NDSTARE. The slit width can be 1 or 2 depending on seeing: say 2 if over 0."6 from Z_rms. (We have not explored the effects of slitwidth on the results very much, but they do not appear to be very dramatic.)
3. Set the Fast Guider mode to autofocus.
4. Now move the crosshead (NOT the telescope) to put the four images produced by the Autofocus lens array in roughly the right place on the CCD for autofocus. The shift is roughly (-9, +3) arcsec from the normal fast guide position.
5. Start running autofocus. (This also does fast guiding.) Peak up on a suitable CGS4 row using "Peakup" and 1s x 5 or a similar combination. When finished use Manpeak to set the exposures: 1s x 13 should be OK.

   NB: Temporary remarks: Because of Manpeak peculiarities it must be a prime number; the intent is to get an exposure which is an average over ~30s, to smooth out short-term seeing effects, using a star which is bright enough that the sky background is relatively unimportant, but which is not so bright as to saturate the array.

   If Manpeak will not allow the selected combination it may be necessary to select a fainter star and use a few, longer, exposures. For a K = 10 star, using 3 ten-sec on chip exposures has worked well, but the sky background on such long exposures reduces the dynamic range available for determining the centre of the focus curve (see below, C. Data Analysis.)

6. On the Bottom End Controller screen set Fine Focus to (CGS4!) "default" - 0.8mm.
7. Switch guiding off and on again to re-start the autofocus cycle: the shorter integrations at the start of the cycle will speed up focus correction for the new setting.
8. After the 32s exposure (or when happy with the autofocus convergence) record the "signal" number on the movie screen along with the "fine focus" setting.
9. Increase "Fine Focus" by 0.2mm and repeat operations 7 and 8.
10. Continue up to "default" + 0.8mm (for nine points).

• This whole process takes 15 or 20 minutes.
• The important points are those well away from focus where the slope is steep (see data reduction) and contains most information on the side-to-side position of the whole curve. These points are also least affected by seeing fluctuations, as the seeing adds roughly quadratically with a big fat image.

Back to "Description of Telescope: Autofocus."

B. Observing procedure for UFTI and IRCAM/TUFTI

This is similar to that for CGS4 except that images are examined using GAIA. From experience so far the best-behaved image quality measures seems to be the central intensity of the fitted Gaussian or the FWHM. The latter should behave the best when the images are well out of focus, and should be unaffected by cirrus.

1. Choose and point to an A star, with 6.5 < V < 7 near the zenith. (At this spectral type the K magnitude is about the same as the V mag in the catalogue.)
2. Point the telescope so as to place an image roughly centrally in the field of the imager (the default 512sq in the case of UFTI).
3. Set the Fast Guider mode to autofocus.
4. Now move the crosshead (NOT the telescope) to put the four images produced by the Autofocus lens array in roughly the right place on the CCD for autofocus. The shift is roughly (-9, +3) arcsec from the normal fast guide position.
5. Load and run the EXEC called AUTOFOCUS. This will insert the Br gamma filter and set the exposure time to 10s x 3 coadds, then BREAK.
6. Start running autofocus. (This also does fast guiding.)
7. On the Bottom End Controller screen set Fine Focus to "Default" - 0.8 mm.
8. Switch guiding off and on again to re-start the autofocus cycle: the shorter integrations at the start of the cycle will speed up focus correction for the new setting.
9. After the 32s exposure (or when happy with the autofocus convergence) hit "RESUME EXEC" on IRTCON.
10. When the eposure is complete the image will appear on the GAIA display on KOKUA. Enter the "View" menu and select "Pick Object". Click on the small "z" until the scale in the widget window is 3x (i.e. make the window a lot bigger than the star image).
11. On the main GAIA display park the cross-shaped cursor on the star image and click the left mouse button. GAIA will do a 2-D Gaussian fit to the image, and the "Pick Object" panel will display "peak intensity above background" "FWHM X:Y:" Record the peak and the average of the FWsHM, along with the "Fine Focus" setting.
12. Incease "Fine Focus" by 0.2 and repeat steps 8,9 10 and 11. operations.
13. Continue up to "default" + 0.8 mm (for nine points).

Back to "Description of Telescope: Autofocus."

C. Data analysis

1. Hand reduction:

1. Plot the signal numbers (CGS4) or the FWHM (or Central Intensity) (for UFTI and/or IRCAM/TUFTI) (all hereinafter "signal") vs the Fine Focus setting at which they were secured.
2. Draw a smooth curve through the data points (smoothing between any ragged data points).
3. Select three or four "signal" levels at which the data are reasonably well behaved (i.e. not too near the peak, where seeing has the biggest effect on signal level) and determine the Fine Focus values at that level at the LHS and RHS of the plotted curve; average these numbers to find the centre. The different centre values from a given curve should agree to 0.01 or 0.03 depending on how ragged the measures were.
4. The mean centre value is the new estimate for the Autoguider Fine Focus setting for that instrument.

NB: Quite similar overall results are obtained by just interpolating linearly between the adjacent data points to get the LHS and RHS values.

2. Using a spreadsheet (as of 2000 Mar 1 not yet demonstrated).

1. Enter signal or FWHM numbers opposite their Fine Focus settings.
2. Plot the measures vs their Fine Focus setting.
3. Fit a parabola to the plot and determine its axis of symmetry: this is the answer.

NB: If there is a lot of background (e.g. if long on-chip exposures were used) a Gaussian or Lorentzian may need to be fitted instead of a parabola.