Autoguiding and Acquisition at MDM
John Thorstensen, Dartmouth College
with a special guest appearance by Jules Halpern, Columbia University

2011 June; revised 2017Oct17 (epg/mdm)

0: Basic Concepts and Terminology

The MDM telescopes point pretty accurately, but not to arcsecond accuracy. They track pretty well, but exposures longer than a minute or so are likely to be trailed. It's therefore important to be able to acquire objects and guide the telescope. This document describes how to do this.

Because the telescopes are on equatorial mounts, the field does not rotate on the focal plane as the telescope tracks across the sky. This is a great simplification.

To keep the telescope from drifting off target as it tracks, the usual strategy is to use an offset guide star. This is a star in some part of the focal plane away from the science field of view, which is reflected by a "pickoff mirror", mounted on a movable guide probe), through some optics into a guide camera. An autoguiding program watches the guide star, and if it appears to drift -- indicating that the telescope is going off-target -- it sends a correcting signal to the telescope. With proper tuning, this can be very accurate, holding the telescope at a constant celestial position to within a fraction of an arcsecond. Such accuracy is necessary for good direct imaging, and is also very desirable for spectroscopy.

The figure below is a sketch of the telescope focal plane (the 2.4m and 1.3m are similar). The pickoff mirror is mounted on the guide probe, which moves on an XY stage as shown. The pickoff mirror is above the science instrument (it's in the Multiple Instrument System (MIS) guider unit), so it can't collide with the science instrument; however, it can block the science instrument field of view. This can be a useful capability (e.g. for finding a bright star when you're lost), but it's important to retract it into a safe position when you're taking data.


1: The Cheat Sheet ...

In all likelihood, when you arrive at MDM, the previous observer will have left the guider in a usable state, with the software going and everything. Assuming that's the case, here are are quick instructions to get you going.

1.1 : First night with a newly-changed instrument:

  • Slew the telescope to your target and center up to the accuracy desired. Focus the telescope at least approximately. How you do this depends on your instrument.

  • Enter the coordinates of your target in JSkyCalc, ideally using a target list. The JSkyCalc version run at the 2.4m has extra features for MDM, most importantly the ability to select guide stars and talk to the guide stage. JSkyCalc is interrupted every 30 sec or so to interface with the telescope, which makes it awkward to enter coordinates by hand, so a list is desirable. A blue box will appear on the sky display at the location of your target. (This document does include some JSkyCalc screenshots later on).

  • If the JSkyCalc Guide Stars window isn't open, open it up, and in the guide star window, push the Read Guide Stars button. Notice that the coordinates listed on the guider cartoon are the same as those you entered. Ensure that the Rotator Angle in the Guide Star Selector window matches the Rotator Angle on the TCS; in nearly all cases, both will be zero.

  • Click on an attractive guide star; note that its characteristics appear, including its nominal guider coordinates in the Star X and Star Y fields. Look for a star that is

  • Once you're happy with the star, push Move Guide Probe and click in the pop-up box to confirm the move.

  • The 2.4m utilizes Maxim DL 6 Camera Control from within the TCS control PC; the 1.3m utilizes Maxim DL 5 Camera Control from a devoted PC. In practice, they are nearly identical.  Go over to the Maxim DL Camera Control window. (There are screenshots of the Maxim program in a later section of this document.) Open the Expose tab, and hit the Start button. The CCD Image window update every second or so, and should show a fairly bright star.

  • To ensure you have the right star, select a couple of other stars and move the probe to them. They should land right where the first one did.

  • If there is no guide star, but you can see that the camera is working and getting light, then your instrument may not be centered the same as the guide stage coordinate system. Hunt for the guide star:
    1. Go back to the main computer terminal and find the Multiple Instrument System window;
    2. Use the dx and dy fields to hop the guide probe by increments of 500 units or so. Motions are sent using the Enter key. Since the telescope is pointing right your target, you should find the star very close to the original field.
    3. Once you've got the right star, note X and Y coordinates of the guide probe, and compare to the Star X and Star Y coordinates reported by JSkyCalc.
    4. For now, when you acquire a new target, enter the appropriate offsets in dx and dy after moving the guide probe, to account for the off-center instrument; or, better, you can use a built in target offset to take care of this automatically.

  • If the guide star looks like a donut: You need to focus the guider optics to agree with the telescope (which you already focused, remember?) Using the program focus.exe, found on the desktop, a GUI allows in and out motion of the guider focus.  If you click on the STOP button, you will have to restart the program by clicking the arrow, as seen in the images below.  Do the best you can; if you run off the end, you'll find that the guider works pretty much ok on donuts.  Click STOP to stop the focusing program when not in use.

    focus_01                             focus_02

  • Back in the Maxim window, hit Stop to stop the exposures, then go to the Guide tab.

  • There are three little radiobuttons near the upper right. Select the top radiobutton, labeled Expose.

  • Hit Start. This fires a single exposure, which appears in a window labeled Autoguider image. It should appear similar to your previous image. It may have been left highly magnified; if so, put the mouse in the image and use the scroll wheel to change the size.

  • In principle, the software finds the brightest star in the picture and enters its X and Y coords in the Guide Star fields in the Camera Control window. Check to see that this happened properly; if it hasn't, click on the desired star in the autoguider image window, and its XY coordinates will appear in the Guide Star X and Y fields.

  • Finally, select the Track radiobutton -- the bottom one -- and hit Start. The Autoguider Image becomes a tiny 64x64 pixel postage stamp, and the star should sit in the middle.

  • Check that the guider is working by depressing directional buttons on the paddle for a couple seconds. The star should drift back to the middle of the box. There's more about the settings below - but if it works, it works. If it doesn't, read on below.

  • If you click on the Graph button in Maxim DL, it will display a running graph of the guide corrections. This is very useful and is typical already openned.  

    1.2 : Two target-acquisition "philosophies

    When you acquire a guide star in the FLI camera, you have two choices for acquiring your target:

    1. You can use the simplest method, which starts guiding with the guide star wherever it happens to be and keeps it there. This changes the nominal position of the guide star on the chip every time.

    2. Or, you can acquire by offset, by leaving the nominal position of the guide star constant. Operationally, you get the guide star close enough to the nominal position, activate the Track command in Maxim, and let the guider walk the star into position.
    The choice depends on what's efficient for your instrument.

    1.2.1 : Simplest procedure for changing targets

    If you have an instrument where you can quickly see your target and center it -- e.g. CCDS or Modspec -- this may be the most efficient strategy. If you have a long-reading camera of some kind (e.g. direct imaging or OSMOS) you may want to use the guide-probe method detailed below.

    1. Don't forget to Stop the guider in the Maxim window before moving away from a target.

    2. Get your new target into JSkyCalc, select a guide star, and move the guide probe.

    3. Apply any constant dx and dy offsets as needed.

    4. If you have rotated the instrument, go into Maxim DL's Settings box. In the lower left, where it says Manual Calibration, reset the Angle to:

          angle = -1 * rotator angle for the 2.4m.

      angle = -90 - [rotator setting] for the 1.3m.  Note however that the 1.3m rotator is not currently usable so this formula should be considered purely academic.

    5. At the new target, center up, then take an image by selecting the Expose radiobutton (in the Guide tab!), and hitting Start.

    6. Assuming the previous step found the guide star OK, simply select the Track radiobutton, and hit Start. You should be guiding.

    1.2.2 -- Using a guide star to acquire a target.

    If your instrument reads out slowly, it takes a while to center up your target. A faster procedure is to set the guider where you know a guide star will be when you center on your target, and then center on the guide star using the guide camera, which reads out quickly. Basically, you follow steps (1) through (4) above, but then you move the telescope to put the guide star in its nominal position on the guider, and start guiding. Step-by-step instructions for this method are given here.

    1.3 -- A Note on Fine Motions ...

    Suppose you need to make tiny offsets, for example to center a target exactly in the slit. In principle, making small steps with the guide probe should do this -- move the probe a bit, and the telescope will follow eventually (there are about 12 motor steps per arcsec). However, as of 2014 June there appears to be just enough stiction -- or something! -- in the guide probe mechanism to make this a bit problematic. I am now recommending changing the XY pixel coordinates slightly in Maxim DL instead. To do this:

    At rotator angle zero, increasing X moves the telescope East, i.e., "pushes" stars to the west. At the 2.4 m, a full step is 0.223 arcsec; the up and down arrows allow adjustment to +- 0.1 pixel = 22 milliarcsec, which is fine enough control for almost anything.

    2 : Illustrated Walk-Through

    The guide cameras on both telescopes were built by Finger Lakes Instruments, and are operated using Maxim DL software, which provides autoguiding capability.

    2.1 : Camera Startup Procedure.

    1. The Maxim DL software runs under Windows. If you're logged out of the machine, log in as OBSERVER (the password is posted on the monitor frame).  On the 2.4m, simply double-click the Maxim DL icon.
    2. Start Maxim DL by clicking on its icon.
    3. In Maxim DL, click on the camera control icon highlighted in the picture just below. [It's supposed to resemble a camera with a cable hanging off it.] The Camera Control Window (pictured farther below) appears; you'll spend almost all your time here.
    4. Select the Setup tab in the Camera Control window (as shown above).
    5. Press Connect to connect with the camera.
    6. Turn the Coolers controller On.
    7. You can check the temperature setpoint under Camera 1 (which should show FLI-New); the cooler button opens a dialog box. A good setpoint is negative-35 C, which the camera can maintain even when it's fairly hot out. In the winter you might be able to go somewhat cooler, but -35 is cold enough for our purposes.
    8. Go to the Guide tab.
    9. Pop the Options menu (right side, halfway down), and select Camera Settings, which pops this window:
    10. In the dialog box, be sure that the third button down on the right reads Close Shutter, as shown. [The button toggles between Open Shutter and Close Shutter. When it reads Close Shutter, it's waiting for a command to close it, and hence keeps the shutter open. We want to keep the shutter open to avoid wearing it out by running it with every guide cycle.]
    11. Click on OK to commit to open shutter; the window closes.
    12. The Options menu under Guide also lets you set the Track Box size; 64 x 64 is usually a good choice.
    13. Still in the Guide tab, select a 2-second exposure, and set aggressiveness to 4 or 5 in both axes. You can set this higher if you find it too sluggish, but beware of overshoot; in bad seeing, you might want to dial it lower to avoid pointlessly chasing a jumping image.
    14. Go to the Expose Tab (not the expose radio button in the Guide tab!), which looks like this:
    15. Set 1 second exposures for Expose, and select continuous exposures.
    16. Once again, in the Options menu, bring up Camera Settings, toggle the shutter button so that it reads Close Shutter and click OK.
    17. In the Options menu, be sure that No Calibration is checked.
    The camera is all set up. If you go into the Expose tab and hit the Start button, it should fire off repeated exposures and display them.

    2.1.1 : Hidden Guider Parameters

    The guider has a couple of hidden parameters that are sometimes left in less-than-ideal states. They are:

  • Backlash. To check this, pop up the Settings box by pressing the button labeled, uh, Settings. To the left you'll see sub-boxes labeled X Axis and Y Axis, each of which contains a Cal Time and a Backlash. The backlash seems to default to 0.3, which means that if there's a correction, the button is pushed for an extra 0.3 seconds to take up slack in the gears. The 2.4m mount has almost no backlash, so both of these backlash settings should be set to zero at the 2.4m, at least.. If you don't do this, you'll sometimes see an alernating jump in the correction on the guider graph, as it hops back and forth around the equilibrium.
  • Minimum Move If you now go to the Advanced tab of the Guider Settings window you'll find settings for minimum and maximum moves. You want the minimum moves to be set to zero.. The maxima can be left at, say, 2 seconds. That way, if the program derives a badly incorrect center for the guide star, the telescope won't be driven way off.
  • 2.1.2 : Port for Autoguider Output

    If for some reason Maxim DL gets reset to its default values, then you have to tell it which of many methods it should use to control the telescope. You do this on the Settings window (again), in the Autoguider Control box, using Control Via pull-down menu; you want:

    ASCOM Direct for the 2.4m or

    LPT378 for the 1.3m


    2.2 : Manipulating the Image Display

    Above: A sample 512-square frame from the Expose control. The faint vertical stripes are probably due to the fact that we're operating without a shutter, so that the bright star 'paints' those pixels during the (very fast) read cycle.]

    Here are some things you can do to change the display:

    What you're seeing: In the standard orientation of the FLI camera, with the instrument rotator at zero, the guide field appears with North at the top and East to the left. This is rotated by 90 degrees from the guider cartoon in JSkyCalc. At the 2.4 m the scale is 0.223 arcsec per pixel , so that the whole 512-square field subtends 114 arcsec. However, the corners are vignetted -- especially the upper left -- so the useful field is somewhat smaller. [At the 1.3m, the field should be larger, but as of this writing it has not been measured directly.]


    2.3 : Using the Autoguider

    [This is organized a little awkwardly, having been pasted from another document. Read this and the section on Establishing offsets at the start of your run and it should all become pretty clear.]

    In the guide tab there are three important options in the buttons on the right - Expose, Calibrate, and Track.

    Some notes about the user interface and nomenclature:

    If you select the Expose radiobutton under the Guide tab, and hit Start, then:

    In the Options menu (right side of the box) there's a control for Track box size. 64 pixels seems to be a good option. The track box needs to be contained entirely within the image, so the guide star can't be close to the edge of the field.

    The guider has to know how quickly the telescope will move when it hits the buttons, and in which direction. There's an autocalibration procedure for this which is not recommended -- it's not necessary. For completeness, it's described later in greyed-out text.

    It's quickest to simply set the guide parameters by hand. In the Guider Settings window, at the lower left, under Manual Calibration, simply set:

    For the 2.4m:

    X Speed = -7.8, Y Speed = -7.3, Angle (deg) = 0 for the 2.4m

    X Speed = -8, Y Speed = +8, Angle (deg) = -90 for the 1.3m

    These numbers are for 2.4m, and similar values are obtained at the 1.3m, provided that the guide rates are set at standard values. These are:

    If you rotate the instrument (very seldom done on the 1.3m), change the Angle parameter to adjust for the rotation. The formula is

    angle = -1 * rotator angle.

    For example, with the rotator at +45, the guider wants an angle of -45 degrees.

    NOTE: the calibration procedure described in the greyed-out text below is deprecated at the 2.4m -- it is not working correctly as of 2014 January, apparently because a hot pixel is causing it to get confused. It is unnecessary in any case.

    The 1.3m camera may not have this problem; Jules Halpern's notes discuss some considerations for calibrating the 1.3m guider, in the event it's needed.

    Now you can move on to the Calibrate button. This automatically establishes the scale, orientation, and parity of the camera image, and the sensitivity of the guide buttons, so that the program knows what buttons to push to bring the star back when it drifts. Note that you probably only need to do this at the start of your run; you may even be able to skip this if you know nothing changed since the last observer. Note especially that you don't have to recalibrate the guider at every new target; doing so is a waste of time.

    You're finally ready to guide!

    2.4 : Establishing Offsets at the Start of the Run.

    Here's a procedure to get you going.

    1. Turn to one of the main Linux computers (not the guider PC), and bring up JSkyCalc24mGS or JSkyCalc13mGS (depending on your telescope). There's a manual for JSkyCalc24mGS here.; if the link doesn't work, look at the mountaintop web page.
    2. Point at a bright star, or some other unmistakable target, and find it with your science instrument. Focus the telescope on your science instrument.
    3. Load the exact coordinates of the bright star (or target) into JSkyCalc24mGS. The entry boxes are in the main window, with the Dartmouth-green bar; to make this easier, hit "Stop Update" and then "Resume Update" when you're done. Don't forget the equinox. Hint: Change the "sleep for" field to 30 seconds (instead of the default 15), and you'll be a lot happier. [It's highly recommended that you load the coordinates from an MDM-style pointing list, using the Object Lists facility of JSkyCalc.]


    4. Click on the Guide Stars button at the bottom of the main JSkyCalc window. This brings up the guide star selector, which looks like this:


    5. Verify that the coordinates in the upper right corner of the guide cartoon are appropriate, and that the Rotator angle is set correctly.
    6. Select a guide star by clicking on it -- a little blue box appears around it, and its particulars are reported. Guide stars in the range 10-14 work well. [Sometimes the thread that updates the guider cartoon stops. If this happens, you need to press "Read Guide Stars" to force an update.]
    7. Press Move guide probe to move the guide star to the star's expected location in the focal plane; it won't move until you confirm it.
    8. Once the probe is in position it'll report its position in the xmis window on the main computer; they show in green when the probe is at its destination. [The numbers don't agree with the earlier figure - this is a later screenshot.]
    9. The little red circle in the guider cartoon is supposed to indicate the guide probe position, but it usually doesn't work. Ignore it.

    10. In Maxim DL, select the Expose tab and hit the Start button. You should see the guide star (see below if you don't).
    11. Focus the guider using the program focus.exe, found on the desktop.  Click STOP when done focusing to quit the program from running.  Click the white arrow to restart the program and focus further as needed.  If the guider focus reaches the end of its travel without getting into focus, you're not dead - Maxim DL guides amazingly well on out-of-focus images.
    focus_01                                    focus_02
    1. The guide star won't necessarily be exactly centered (that's ok). If your guide star has little friends bright enough to be in UCAC3, you may see them -- the pattern will be rotated 90 degrees from the JSkyCalc diagram.
    2. Stop exposing, then go to the Guide tab and select the Expose radiobutton.
    3. Hit Start -- a single image will be taken and the coordinates of the brightest star in the field -- hopefully your guide star -- will appear in the Guide Star X and Y fields. WRITE THESE DOWN.
    4. Grab the mouse and draw a nice box around the star -- this will be useful later.

    Now you're all set up.

    2.5 : Acquiring targets by offset.

    This is especially appropriate if your instrument reads slowly, so that it's awkward to center your target by simply looking at it. The basic idea is to put the guide probe where you know the guide star will be, and then center using the guide camera:

    1. Be sure JSkyCalc has your target's coordinates in the RA and dec window. [This is easiest if you read them from a list.]
    2. Set the telescope on your object's coordinates.
    3. If the JSkyCalc guide star selector tool is not open, open it using the Guide Stars button at the bottom of the JSkyCalc window.
    4. If you have rotated the instrument, be sure that the exact rotation angle is entered in the box in the guide star selector window.
    5. Look at the JSkyCalc guide star selector cartoon. Select a suitable guide star, and move the guide probe into position as described earlier.
    6. In Maxim DL, select the Expose tab, and press the Start button. The image should start updating.
    7. Find the guide star in the image. You may have to paddle around a little. If the rotator is at zero, the image has north at the top and east to the left. [If you're trying to match the pattern in the guide star cartoon, note the directions indicated in the cartoon.]
    8. Using the hand paddle, move the telescope to put the guide star in the box you marked earlier.
    9. If you're concerned that you may not have the right guide star, you can select another guide star and move the guide probe to it. It should land right in the box.
    10. Once you're confident of your centering, stop taking exposures with the Maxim-DL Stop button, select the Guide tab, and the Track radiobutton; then press Start to start guiding.
    11. If there is fine adjustment to be done -- e.g. for getting a star into a spectrograph slit -- then it can be done using the dx: and dy: entry boxes in xmis. Note that the OSMOS has an elaborate centering sequence built around this.

      2.6 : Building in a target offset.

      As noted earlier, if the center of your instrument's field is offset from the nominal center of the guide field, you'll have to adjust for this, because JSkyCalc's nominal coordinates won't be quite right.. One way to do this is to offset the guide probe away from the JSkyCalc coordinates every time you set on a new object, by manually entering dx and dy into xmis. This is tedious and error prone.

      It's easier in the long run to build this into the calculation, using the provision in JSkyCalc. Suppose you have already determined the dx and dy values for your instrument. Install them as follows:

      1. Compute these values:
        Target X offset = dy / 12.8
        Target Y offset = dx / 12.8

        That's not a misprint - the X offset depends on dy, and the Y offset depends on dx.

      2. In the Guide Star Selector, click the Other radiobutton (to indicate you're not using either OSMOS slit).
      3. Enter the number in the fields of the same name in the guide star selector.

      Now that you have non-zero numbers in those fields, when you read guide stars, you'll two sets of coordinates in the selector -- the yellow numbers are for the target, and the green the nominal center. A yellow box marks the target's off-center location in the diagram.

      2.7 : Miscellaneous oddities etc.

      Guide star coords are not with respect to full frame. I've seen Maxim get into a state where it reads a subframe while guiding, and then the guide star coordinates are not with respect to the full frame. In principle it's faster to read a subframe, but in practice the read is fast enough that it makes no difference. To cure this, In the guide tab, in the Settings pop-up window, lower right side, in the Exposure Settings sub-box, hit the Reset button, and then the OK button so that it 'takes'. It may not look as if it has, if you've zoomed the guider image display to see only the guide box; just de-zoom with the mouse wheel to see the whole field.

      Appendix : Jules Halpern's notes on guiding the 1.3m.

      In 2013 December, Jules Halpern had a run at the 1.3m in which he paid careful attention to the guider. Here are his notes, which should be very useful.

                                                         February 9, 2014
       Supplemental notes on autoguiding at the 1.3m, with help from the
       Maxim DL User Manual; some of this would apply to the 2.4m as well.
      This writeup assumes that you are already familiar with John Thorstensen's
      manual on Autoguiding at MDM, therefore, with the basic operation of the
      Maxim DL guiding system.
      If you or the autoguider (figuratively) depress a guide button on the
      hand paddle. the telescope will move a certain number of arcseconds
      per second (the guide rate).  This is also what happens during the
      Calibrate function of the guider.  If you use the Calibrate function
      at the 1.3m, as described below, you should do it with a star on the
      celestial equator, and enter 0 in the "Scope Dec" field at the top center
      of the Guide window.  This is necessary because the 1.3m TCS does not
      adjust the HA guide rate for declination; the EW guide motions are slower
      across the sky at high absolute values of the declination.  By the same
      token, the guider doesn't know at what declination you are pointed unless
      you tell it.  So you have to enter manually the declination in the
      "Scope Dec" field of the Guide window each time you move to a new target.
      The guider will use the declination to adjust how long it presses the buttons.
      Moral: at the 1.3m, YOU have to tell the autoguider what the declination is.
      (At the 2.4m, ignore the "Scope Dec" feature of guider because the TCS takes
      care of it. The 2.4m guider always sees the same speed in pixel units.)
      Here. we refer to the telescope guide "rate" in arcseconds per second, while
      a guide "speed" is in pixels per second, the pixel of the autoguider CCD.
      Automatic calibration is optional
      In order to get an accurate calibration, the telescope should move at
      least 20 pixels in each of X and Y.  The length of the move will be
      proportional to the product of the "Cal time", which you enter in the
      Settings pop-up of the Guide window, and the paddle gain, which you set
      in the Command line of the TCS Control Panel (running on mcgraw). 
      The "Cal time" is how long the guider will press the guide buttons to
      determine the guide speeds.  The default Cal times are 5 seconds in both
      X and Y, and the recommended paddle gain is 0.02.  To see or change
      the paddle gain, type in the Command line of the TCS:
           tx paddle gain       (to see what it is)
           tx paddle gain=0.02  (to set it)
      Moral: The guide speeds are only valid in conjunction with a particular
      paddle gain.
      Check the paddle gain at the beginning of each night, as the staff
      may have increased it to slew rate if they were working on the
      telescope during the day.  Also, any time you restart the TCS, the
      paddle gain may reset to a default value.  The last time I restarted
      the TCS it came up with paddle gain=0.01, which is not a bad value to
      use.  But you would have to calibrate the guider for that gain if that
      is what you are going to use.
      For paddle gain=0.02 and Cal time of 5 seconds, the calibration should
      give the following speeds AT THE CELESTIAL EQUATOR.  The values are only
      repeatable to +/-0.5 and are rounded off here to that level of precision:
           X Speed  -8.5
           Y Speed  +6.5
             Angle  -90
      Remember that these speeds are in pixels per second, so it implies that the
      telescope moved about 30-40 pixels during the calibration, which is good
      Sometimes the calibration will fail because of backlash in declination;
      the telescope may not move in declination (X) even if the guide signal
      is activated for 5 seconds.  If this occurs, a little error message
      will pop up to the effect that the star moved less than 5 pixels in X,
      and you will see that the speeds were not updated. If this persists
      after several tries, you could try increasing the Cal time to 10 seconds
      to get it to move.  It's okay to do this as long as it doesn't cause the
      star to move completely off the field.  However, you may never take out
      the backlash (more on this below), so you may not get a reliable result.
      As a last resort, just manually enter the above values, and they should
      work well enough, providing that you also set the paddle gain to 0.02.

      Manual calibration is preferred

      Now that you understand how the fancy auto-calibration works, and
      why it can fail at the 1.3m, you may realize that the guide speeds
      can be determined during the day by using the physical hand paddle,
      the clock, and the coordinates on the TCS GUI. In fact, the values
      that I tabulated above were determined in this way, and are consistent
      with what the guider measured by itself on the sky. You could just
      adopt these parameters (including paddle gain=0.02), and stop reading
      here. I only really wanted the calibrate function of the guider
      to determine the signs of the Speeds and the Angle, which I would
      otherwise have to guess. To calculate the calibration, I just had
      to know the pixel scale of the guider CCD, which is 0.67", to convert
      the speeds that I measured into pixels per second. I have only
      determined the pixel scale to within about 10%. It should be possible
      to be more accurate, but I haven't done that yet.

      Aggressiveness is not a virtue

      An exposure time of 2 seconds for guiding works pretty well.
      The recommended value of "Aggressiveness" is 4-5. Aggressiveness is the
      fraction of the calculated offset that it actually corrects, with 10
      being the full move. So 4-5 is a conservative amount that prevents
      overshoot. Overshoot causes an oscillation at a small multiple of the
      period of the guide exposure, with an amplitude of +/- a couple of pixels.
      (In normal guiding, with good seeing and no wind, the typical errors
      can be 0.1-0.2 pixels.) If the guider appears to be oscillating,
      first check that the paddle gain is correct for the guide speeds that
      you are using. As a second resort, reduce the Aggressiveness.
      Make these changes to avoid chasing after momentarily jumping images
      when the seeing is bad or when wind buffets the telescope.

      "Nodding" is not bad guiding

      An oscillation caused by the guider is NOT the same as the intermittent
      "nodding" behavior of the 1.3m, which is a feature of the TCS alone.
      It's easy to tell the difference. A guider-induced oscillation will stop
      as soon as you stop guiding. On the other hand, if the TCS is nodding,
      it is a more pernicious effect that occurs whether or not you are guiding.
      The TCS nodding can manifest itself as repeated jumps of 10"-60" that
      quickly return to the original position. Or it can be a jump followed
      by a return in several discrete steps. Or an oscillation of smaller
      amplitude that is more-or-less continuous. These motions occur only in
      declination. The guider is not capable of compensating for the nodding.
      The user can't prevent such nodding either, except perhaps by wishing,
      but more often by moving to another part of the sky.

      Backlash is unpredictable

      Sometimes backlash in declination will prevent the guider from correcting
      for many cycles, up to a few minutes. The telescope will drift away by
      a couple of pixels in declination before eventually coming back. That's
      when your lack of aggressiveness becomes a handicap, because it takes
      a longer time to take up the backlash. There is a Backlash parameter
      in the Guider Settings next to the Speeds. It units are seconds, and it
      is supposed to account for the time delay due to backlash in exactly the
      way that you might imagine. It presses the button for that much extra time
      that it takes to start the telescope moving. Backlash is usually a problem
      in declination but not in RA. However, you should probably set both backlash
      times to zero unless you are exactly sure about the behavior of the telescope,
      which I'll bet you aren't. The 1.3m backlash is intermittent, and sensitive
      to the mechanical adjustment of the telescope as well.

      Conflicting coordinate conventions are confusing

      The guider image is oriented such that North is up, and East is to the left.
      The X coordinate increases to the West (right), and Y increases to the South
      (down) as you pan the cursor across the image. This is for the default
      telescope rotator position of 0 degrees, and applies at both telescopes.
      The Guide Star X,Y coordinates and the scrolling X Err and Y Err
      in the Guider Information window are consistent with this convention.
      The sign of the error is the direction that the star moved. However,
      the points plotted in the Tracking Error Graph have X and Y switched
      from the image coordinates, and one of their signs is reversed, too.
      X Err is graphed as -(Y Error) and Y Err is graphed as X Error. So you
      have to look at the X Error graph to see errors in declination. If you're
      confused now, join the club. These conflicting coordinate conventions
      appear to be an unfortunate design feature.

      This raises the question of which coordinate system applies to the
      Speeds and the Aggressiveness, which are the properties you can evaluate
      and modify. The X Speed is for declination because that's the one that
      usually fails in the self-calibration. And the guide rates determined
      "by hand" for RA and declination are different; their magnitudes are
      consistent with X Speed being for declination. So X Aggressiveness
      is probably also for declination. In summary, the graph's coordinates
      are correct and the image's are not, if that helps you think about this.