MCCD Fast guider use
M. F. Waterson 1997 Aug 06
The MCCD fast guider was built so that it could operate on either a
bright field/dark spot or a dark field/bright spot guiding object and was
extensively lab tested by Buzz Graves and Herb Ryerson in both modes. However
it is important to be mindful of the fact that they were using an artificial
image that was very well defined in their tests; real life is seldom so
Inner and outer loops
The fast guider sensor itself does not distinguish the image on the sensor,
rather, when the loop is closed simply tries to drive the voltage difference
between the opposite edges of each axis to zero by moving the fast mirror
proportionate to the difference voltage multiplied by the gain pot setting.
(Thus, if the gain is set to zero, the mirror will not be moved). At short
intervals, the software executes a monitoring loop that samples the mirror
drive voltage in each axis, and if it is close to full scale, drives the
Coude mirror in a direction that is supposed to reduce the required drive,
centering the image. The light level is also checked, and if the large
error or low light persists for a (settable) time, the loop is opened automatically.
One might ask how large the error had to be to start the Coude mirror drive;
basically, if the X or Y meter is pegged and the gain is turned up even
to 0.1 there will be a large enough drive signal to cause the coude system
to try correct it.
Because almost every potential target will have an intensity gradient across
the guide sensor, a feature called "Limb Darkening Correction" is built
into the system. Despite the fancy title, this is nothing more than a command
that measures the instantaneous error voltage at the preamp outputs, and
injects an opposite DC level to (supposedly) cancel it out. If this is
done on a quiet sun region near a spot on which you will attempt to guide,
the (quiet sun) error will basically be dominated by the limb gradient
and thus will allow you to effectively flatten the field at the spot. This
useful feature can be used to help many other problems as well, but one
needs to remember that all it is doing is measuring the error voltage (with
no offset active), then setting a DC DAC voltage into the summing junction
of the loop. It does not check to see if the correction was effective.
Bright or dark guide spot
The ability to guide on a "dark" or "bright" spot is provided by enabling
the system to change the sign of the loop feedback by a software flag;
this effectively reverses the sense of the mirror correction movement (rather
then swapping the sensor "+" and "-", for example). This means that the
sense of the error signal as seen on the meters will be "backwards", but
that is seldom particularly confusing in operation. What can be confusing
is the definition of "Bright" vs "dark"- it may seem as though the "invert
loop gain" flag (image stabilizer screen #2 on the MCCD) is backwards to
normal logic - just use the opposite if the obvious choice doesn't work!
Tricks, treats, traps
Finally, closing the loop depends on having a reasonable error voltage
when the command is given; this system has many degrees of freedom and
it is easy to find a non-linear combination (!). There are a few tricks
I've found that make this more or less easier to achieve.
First, the system will guide on all the photons in the guide sensor, despite
how ugly the image is. if you have vignetting, ghosting, etc., you need
to think about how that will sum into the (square) guide sensor area; sometimes
you can use the asymmetries to advantage. Also, if you have a marked intensity
difference in one axis relative to the other (such as a ghost at the top
of the field) you may have different optical gain in that axis; the gain
pots should always be adjusted independently (see below)
Second, when to try to close the loop, you should have only a small error
voltage for best results. Since the scale of the meters is quite expanded
around the zero point, this is tricky, especially if you do have a substantial
image gradient, however by careful movement of the beam-splitter ("Buzz
mirror") you can roughly locate the electrical center of the sensor by
observing the meter crossovers. While it is helpful to have the meters
actually reading on-scale rather than pinned, this is not always possible
to achieve in a reasonable time if you are trying to guide on a difficult
object. In this case, true limb darkening is not likely to be relevant
anyway, so I use the limb darkening function to help me out: I put the
image as near the crossover as I can find and activate the limb darkening
choice. (NOTE - it is a Good Idea to manually remove the limb darkening
correction by setting the line to N if you need to make large adjustments
in the image setup or position as the old offset can cause the guider tuning
to be more difficult). The limb darkening function will attempt to remove
the large static offset you have (from not being in the right place) with
a much better precision than you can achieve with the Buzz mirror. If you
have both meters on-scale (hopefully near zero!) when the function completes,
you are all set and can go ahead and close the loops; it should be possible
to adjust the gains as described below. If either meter is still off scale,
remove the darkening by setting the line to N and try again. . . but maybe
you will have to change some optical parameters (Zoom, e.g.) to get things
in a more reasonable configuration.
Third, when you do close the loop, do so with the gain pots set to ZERO.ZERO
so that the loop is effectively still open. This will let you slowly increase
the loop gain while watching the spot and/or the meters; as you turn up
the gain the spot will either stabilize or jump off to the side. If the
meters peg and it jumps to a limit position in both axes, OPEN THE LOOP
IMMEDIATELY! If you leave it closed, the Coude mirror will begin driving
to try to center the image, but it will be heading in the wrong direction,
and you'll have to manually drive it back before proceeding (which is a
Really Good Reason to write down the position of your desired observation
position referred to the coude encoders before you start this whole process!)
Be careful you don't accidentally correct a Coude movement with the Buzz
mirror--that only moves the image on the guider! If you have been reasonably
careful about the center position and limb darkening, you probably have
the polarity wrong, and should try reversing that choice first, going through
the whole business again from the start.
If, when you turn the pots up a little (no higher than 0.3), the meter
moves towards the center (or at least does not peg!) you are on the right
track, and should try to adjust the loop gain further. Slowly, turn up
the gain of each channel a little (no more than 1 turn at a time) iteratively,
watching the behavior of the meter. You should see the meter stabilize
near the center zero point, but you will soon reach a point beyond which
there is no visible improvement; reduce the gain somewhat (to prevent oscillations)
and go around to the optics bench to listen for a high-pitched, raspy whine.
That noise is the mirror mechanically oscillating, and is a Bad Thing--reduce
the gain, in both channels, until it goes away! It will blur the image
badly, but you can't see it as it is faster than the video frame rate.
I have found that things will work better than you visually realize using
the minimum gain possible while still keeping the image quiet on the monitor.
Gain is like that--while some is good, more is not necessarily better.
If it seems as if the spot is not guiding in the center of the detector,
but you can't remove the offset, you can try to apply the limb correction
with the loop closed. This is useful of you have a very diffuse spot, or
are so near the limb that you can't find a good quiet-sun region to use.
The loop may lose lock when the initial offset is removed for the measurement
if it was a large offset, but I have generally had good luck when I tried
it; because the loop is closed, the error voltage measured will be more
stable that if it was open, so the compensation will come closer to nulling.
Finally, the Coude Mirror drive feature is there to allow the system to
take out large drifts in the image position, and usually works well. However,
if you are guiding at extreme zoom levels you may have to adjust the length
of time that the mirror is driven to make its movement match the image
scale of the guider (too long will drive it out of the guider field, too
short and the loop will open because the error persists too long. The value
is set on screen #2, at the bottom, in hundreds of a second; think about
how fast the image moves on the guide video when you use the paddle to
estimate a setting (< .1 sec is probably too short, 5 sec is probably
too long). You'd like the mirror to move the image back towards the "center"
by no more than 1/2 the sensor width at a time.