[daip] Re: DELZN questions

[Leonia Kogan]

George Moellenbrock wrote:

>
> Hi Leonia-
>
> A couple of quick questions about DELZN:
>
> 1. What are the units of the "SQRT of variance of the residuals"
> that is reported at the end of execution (for PRTLEV=2)?  I assume
> it is in millimeters, but the value reported seems rather larger than
> the rms implied by the data in the plots.  In fact, it looks like
> a plausible estimate of the rms from the plots is something like
> the sqrt of the value reported.  (Has the sqrt of the variance
> actually been taken?)


Yes you are right I forgot to take the SQRT.  I'll fix it

>
> 2. Why does DELZN insist on writing a txt file at more than just the 
> single reference timestamp used in the solution?  My CL table has a 
> large number of entries, and DELZN fails complaining about more than 
> 10000 rows. I need both to update my CL table and create the txt file 
> because I have to use the solution in the original dataset, and 
> transfer it to another dataset.  But I am forced to turn off the CL 
> update to get a txt file. It seems to me that a single set of 
> coefficients per antenna---i.e., those reported during execution--- is 
> all that is really required in the txt file.  (This makes the txt file 
> more useful in record-keeping and
> external analysis, too.)

The text file is created for using it by CLCOR ('ATMO')  The file is at 
the format required by the 'ATMO'
You can use the outfile to solve the 10000 rows problem repetin CLCOR 
several time with different time rag interval

>
> 3. A couple of suggestions:
>
>   a. Why not do the plots in time units, since this will make comparison
>      with plots of the input MBDs easier?

The plots are in time units (days).

>
>   b. Why not have an option (e.g., APARM(1)=3) to create all 3 types of
>      plots in a single run?

Let me discuss tha within the AIPS group. I persannelly consider it does 
not need

>
>   c. Why not have an option to plot the residuals corresponding to
>      each type of current plot?

Let me discuss tha within the AIPS group. I persannelly consider it does 
not need
At the beginnig I thought that the plots themselve are not reqired, 
bcause the could confuse the people.
And it happened I spent a time to explain to the people what do the 
plots show.
The residual can add misunderstanding

>
>   d. Why not have APARM(2)=0 force no atm solution (clock only)?  (This
>      would be useful for diagnostic purposes when trying to decide
>      what orders to use for atm and clock.)

There is an option to solve atm only . My opinion is that the option to 
solve the clock only is a wrong addition.  But I'll
ask other people opinion

>
>   e. It would be good to have a REFANT parameter that would be used
>      to force the solution to use only input MBDs that share the 
> specified
>      reference antenna.  I have found that just a few MBDs with a
>      different ref ant can yield DELZN solutions differing wildly from
>      what you get with a consistent refant.  It took me awhile to 
> discover
>      that I had a few MBDs with differing refant lurking in my input
>      SN table, and things improved tremendously when I avoided them.
>      In any case, DELZN should complain in this case that it is using
>      solutions with differing refants.

I spent so much time to find the solution for the rare case of several 
refant. although the uniq refant is the most jeneral case. I'll look if 
it s simple change of the codes

>
> I am really finding DELZN an interesting task.  With data carefully 
> observed to provide good leverage to separate the clock and the atm, I
> think I am getting very good zenith delay solutions.
>
> Thanks,
> George