[daip] question about VLANT

[Bryan Butler]

thanks, eric, this helps.  specific comments/questions below.



Eric Greisen wrote, On 2/29/12 8:45 AM:
> Bryan Butler wrote:
>>
>> eric,
>>
>> i can't find a precise enough description of the algorithm to apply
>> baseline corrections. it's not described either on the web page where
>> you can retrieve them, or in the VLANT help file. i could parse the
>> code, but thought you might know the answer right away and save me
>> some time.
>>
>> let me explain my confusion.
>>
>> the baseline file contains lines like:
>>
>> ;MOVED OBSDATE Put_In_ MC(IAT) ANT PAD Bx By Bz
>> ;
>> JAN06 JAN06 20:30 8 N12 0.000 0.000 0.010
>> JAN03 JAN10 JAN11 19:21 20 N18 -0.0017 0.0020 -0.0020
>>
>>
>> now, for the second line, it's pretty obvious what observations to
>> apply the correction to (although, i have to say that there is still
>> ambiguity because of the unkown *time* of the move on jan03, but let's
>> ignore that for now).
>>
>> but for the first line, there is no identification of a move date, so
>> how can you be clear on how far back from jan06 20:30 to make the
>> corrections?
>>
>> -bryan
>
> The collection of VLA.ant.yyyy files is viewed as a continuous stream of
> information

ahh, that explains how you handle the year boundaries then :).

> and the moved date will appear somewhere probably in the
> previous year - antenna positions are checked after moving but also at
> later dates if they do not move further.

but it doesn't necessarily have to appear in a previous year, does it? 
i mean, a baseline run can find an offset for an antenna that hasn't 
moved (*ever*, in principle), right?  i'm sure i can find specific 
examples in the files, but you probably see what i mean.

and in fact the description on the page states this as well:
"In general, one should use the sum of all available corrections. 
However, antennas may indeed shift slightly over time, even when staying 
on the same pad. The test for an individual data set is either to solve 
for the antenna positions yourself, using those data; or to see what 
happens to your phases before and after these corrections, and judge for 
yourself whether the corrections have improved matters."

i agree that's true - was just wondering if VLANT handled it explicitly 
(but it appears not - and that's probably the right thing to do, because 
you can't know what else *to* do).


> If the putin time is after the
> observation and the antenna number and station match, the correction is
> added to the sum of the corrections. If a moved date appears after the
> observation, then the antenna is removed from the list of those allowed
> to get corrections. At the end, the task looks to see if the list still
> has antennas that could get corrections and says so if there are any.
> There are usually 3 since some antennas only move when they go into the
> barn for maintenance or upgrade.
>
> I found the description on the web pages a bit confusing but it is in
> fact adequate.

i just re-read the web page with your description in mind and, yes, it 
makes more sense now.


> What was not adequate until I got involved were the
> correction files. The ones now in the system have been themselves fixed
> by me and there is at least one range of dates in which it was not
> possible to determine what people had done - VLANT knows that range
> which was short and a long time ago.

i'll plumb the VLANT code.  i think the format of the files is still 
inadequate - there needs to be a time on the MOVED column, because for 
observations on that date you won't know whether to apply them or not.