CAMAC Progress Report

                      Week Ending: June 01, 2001
                               Week: 18


Tom Folkers:
     1. On vacation all week.

Tom Sargent:
     1. Compared  two algorithms for dynamically determining 
        corrections for interpreting the encoder outputs. One was the 
        Ellipse Least Squares fit and the other, a simpler step wise 
        method patterned after the calculations in the original AXPUG 
        thesis.  Both were run on the T.I processor using 40-bit 
        arithmetic for the ellipse fit and 32-bit for the AXPUG 
        algorithm.  The ellipse fit algorithm is known to be unstable 
        in the presence of noise or if the ellipse runs through the 
        origin, but neither of these things is true in our 
        application.  I tested both methods on the T.I. processor with 
        varying amounts of noise.  As long as the sums are formed 
        carefully to avoid round off, the Ellipse method is faster and 
        more accurate. It will do a fit to 1200 points in 2ms.  Noise 
        levels of 1 part in 1000 degrade it enough to make one 
        consider the simpler method, but our data is expected to be 
        much better than that. So, I have essentially copied Johann 
        Schraml's method for analyzing the encoder outputs. 

     2. I began writing tracker.c - the code that will run in the T.I. 
        DSP processor.  It's about half done. I have defined a bi-
        directional communication protocol to use between the Linux 
        host and the T.I. chip.  I should be able to start testing it 
        within a day or so. It will probably be a week before it has 
        the minimum required list of capabilities to actually use at 
        the telescope.

     3. The DSP software can run with real encoder data or in a test 
        mode that internally generates artificial encoder data. In the 
        test, mode which is written, but not tested yet,  interrupts 
        come from the RTC instead of the A/D unit, but the data flow 
        is virtually the same. At each tick, the time is used to 
        compute R.A. and dec. for a simulated object which it then 
        converts to Alt-Az and then to equivalent encoder outputs. So 
        it should provide realistic tracking data.  I can later fairly 
        easily add slewing and raster scanning to it.

     4. The Traquair A/D demo program has been compiled and is ready 
        to test with the output from the A/d units.

Pete Hubbard:
     1. Machine front panel for fan guards.
     2. Make junction box for A/D sample clock wiring.
     3. Finished A/D chassis wiring.

Dave Ashby:
     1. No Report.