time, tunnelling, stacking

[Stupendous Man <richmond@a188-l009.rit.edu>]

TIME:

  Mike G. asks:

> Both Michael and I are working on reatime control programs. Michael's will
> run under Linux, mine will run under Windows.  Just as a check, my program
> says the Julian date right now (December 16, 2001, 23:11 EST) is 2452260.
> How does that compare with other programs?

  Good question.  We ought to make checks like this frequently.
No item is too trivial to check. 

  My program, when run at 01:19:00 UT Dec 16 2001, took an image
which was called

           2259013E.fit

The first four characters, 2259, are indeed the last four digits of
the Julian Date (2,252,259.5556).

  The FITS header of this image contains keywords:

           UT      = '01:19:00'
           DATE-OBS= '2001-12-16'

And these, too, are correct.  So I'm pretty sure that my control program
is getting the dates and times right.


TUNNELLING:

  Thanks to Dirk and Chris A. for hints on tunnelling via SSH.
I have SSH installed on all the computers here, so I gave the
system a test:
      
         computer "remote"     is connected to Mark IV
         computer "local"      is in my office

            step 1: log into "remote" and start "tait" control program
                       it listens to a socket, say, number 999.
                       In production, this would be running all the time.

            step 2: on "local", use ssh to connect to the remote system,
                       and associate port 1024 on local to port 999 on remote

                            ssh tass.rit.edu -L 1024:tass.rit.edu:999

            step 3: now, on "local", I can run the "taitcl" program,
                       telling it to send commands to local port 1024

                            taitcl -p 1024 exp dark 60

  The command "exp dark 60" is sent to local port 1024, but that turns
into the remote computer's port 999 via the magic of SSH tunneling.
So, commands run on the local machine can end up at the remote machine.
Very convenient under some circumstances.

  Of course, if I want to LOOK at any images as they are collected,
I need to be logged into the remote machine.  But if I just want to,
say, take series of darks or flats, tunneling might be a good idea.

  
STACKING:

  Tom Droege writes:

> I suppose that one could get the same accuracy result averaging the stars 
> after detection, but then this results in fewer detections.  Any comments?

  It depends on what you want to detect, and how well you want to detect it.
Suppose that I take a series of 60-second exposures.  If I look at each
one individually, using a technique which finds "10-sigma" peaks, then
I'll end up with a (relatively) small list of bright stars.  Each of the
faintest stars will have individual measurements accurate to 
around 10 percent.

  I could stack, say, 8 exposures together, and then again look for 
"10-sigma" peaks.  Now I'll find more sources, but again, each individual
measurement will be good to around 10 percent.

  On the other hand, if I average together measurements from 8 consecutive
individual images, I have only a few sources, but each measurement is 
more precise -- down to around 3.5 percent.

  So the questions are: 

       1. are you interested in sources which vary so much
          that even 10-percent accuracy is sufficient to measure them?  
          Or do you require 3-percent, or 1-percent, or 0.1-percent, 
          in each measurement?

       2. How many sources do you want?

  If your goal is to find the maximum number of sources, then stacking
is the only way to go -- but the precision of measurements of the
faintest sources will always be pretty poor.  If your goal is to make
very precise measurements, then stacking probably won't help much.

                                           Michael Richmond