CAMAC Progress Report

                     Week Ending: August 23, 2002
                               Week: 82


Tom Folkers:

Tom Sargent:
     1. I looked at the velocity estimation being produced by servo 
        and tried a few other algorithms to compute velocity.  The 
        current method fits a parabola to the last 8 points and 
        evaluates the derivative of the parabola at the current time.  
        I tried a simple digital filter, linear fits to the last N 
        points, simple dx/dt computations spanning several intervals, 
        etc.  The parabola method seems to be the best. It has the 
        smallest phase lag and doesn't "oversmooth" the data.
     2. I fixed a few things on the "to do" list:  The program now 
        handles moving to a new position which is very close by, 
        better.  If the distance is less than 20 arcsec, it doesn't 
        slew at all.   I added an error check to the sun avoidance 
        slew to pause if the reading of the az & el path lists get too 
        far out of synchronization.  I altered  the tracking algorithm 
        to invoke the slew algorithm if the position error gets over 1 
        arcminute.
     3. I looked at the "sluggish" tracking problem again. When a slew 
        is finished, the tracking algorithm sometimes takes 10 or 20 
        seconds to close the last few arcseconds of error.  This 
        appears to be due to the accumulated error term in the servo 
        loop growing very slowly under some conditions.  When the 
        telescope is tracking well, the accumulated error term reaches 
        a nonzero equilibrium value.  The problem is getting to that 
        equilibrium value quickly.  I finally realized that this was a 
        function of the target velocity and did a series of 
        observations to determine the relationship.  It turns out that 
        it is very linear and very well behaved.  I changed the 
        program to estimate the equilibrium value of the accumulated 
        error term when the slew ends. This seems to have fixed the 
        sluggishness problem. 

Pete Hubbard:
     1. Wiring changes for Digital Back End chassis.
     2. Update drawing changes DBE.
     3. Move three Sun 20 work stations and PC,s to lab.
     4. Helping Dave F. with AOS interface requirements to replace 
        Camac hardware with PC I/O.

David Forbes:
     1. I concentrated on the driver code, producing a usable API for 
        Bill to work with. The AOS driver now has ioctl functions to 
        control the CCD baseline DACs and to enable or disable each 
        channel.
     2. I also developed the CCD baseline correction algorithm and 
        tested it out. The algorithm tracks the CCD dark level changes 
        well and converges quickly with no overshoot. I found that the 
        low-pass filters in the DAC amplifiers were way too slow to 
        allow the correction program to work well. I will be changing 
        these filters from 1 second to .01 second response time. This 
        involves replacing 3 capacitors on each of 3 A/D cards.
     3. I spent the weekend learning the Green Bank filterbank design 
        and coming up with a suitable replacement. I'll be working 
        with Harry on this project as the AOS project permits. 
     4. I will bring Bill up to speed on the AOS application software 
        this week.