CAMAC Progress Report

                      Week Ending: July 27, 2001
                               Week: 26

Tom Folkers:
     1. Other then several sessions of consulting with Tom Sargent 
        about the control system, I spent little time this week on 
        CAMAC related issues.
     2. Spent most of my time working on the impending move to the 
        first floor offices.

Tom Sargent:
     1. We timed execution of the position servo loop in the Linux 
        host at 23 msec.  This executes once every 10ms, so the duty 
        cycle is very low. 
     2. I eliminated the use of the Traquair server/loader software, 
        after it gets the DSP software downloaded and running.  Their 
        server/loader was using async. I/O which has the side effect 
        of spawning hundreds of short-lived tasks per second. This was 
        not impacting overhead, but it did make the servo process more 
        vulnerable to disruption when Linux gets busy doing other 
        time-shared tasks. So now, the servo program on Linux calls 
        the Linux DSP driver directly.  So far I've not seen the servo 
        program delayed by Linux running other tasks by more than 
        about 250ms.
     3. I looked at altering the scheduling algorithm in Linux to a 
        more real-time-friendly method.  I have not been able to get 
        any benefit from this yet. I'm still looking at it.  Jeff 
        Fookson is working on the same problem at the moment for a 
        different Linux application, so there may be some improvement 
        shortly.
     4. Fixed a bug in the DSP program that had it pausing if the 
        communications FIFO with Linux filled up. It now keeps 
        running. 
     5. Added some safety checks to the servo program: The telescope 
        is now stopped if the commanded positions stop getting updated 
        for 60 seconds if the telescope is commanded to its position 
        limits, or if the servo program is blocked on I/O to the DSP 
        for more than 2 seconds.  The servo also now monitors the time 
        differential between the Linux host and the DSP and resets the 
        time on the DSP if it gets more than 10ms off.
     6. Altered the Linux servo and DSP software so the servo program 
        can be stopped leaving the DSP program running.  Previously 
        the DSP was always started and stopped with the servo task.  
        The servo program can now be restarted at any arbitrary time. 
        It recognizes that the DSP is running and up to date and skips 
        initializing it. 

Pete Hubbard:
     1. Continue Mux wiring from board to connectors.
     2. Add line termination resistors to Mux board.
     3. Bill provided the signal grouping out of the VCO 25 pin 
        connectors that feed DRUMBEAT. Connecting to Mux board will 
        have two groups of twenty inputs. This places the nineteen 
        channel Bolometer on the "A" inputs and the rest on "B" 
        inputs.

Dave Ashby:
     1. No Report.