[mmaimcal] Text of AI on antenna quardupod design from Beasley

[Jeff Mangum]

Hi Mark,

Understood.  I fear, though, that we are slowly sliding into a situation 
where we *can* correct for many of the problems two antenna designs 
present, but the effort involved prevents many of these corrections from 
being made, independent of the good intentions of those involved.  I 
think that this was Stephane's main point.  Taking the pointing case as 
an example, most telescope operations spend a great deal of time during 
their operational lifetime tweaking pointing performance.  For many of 
these telescopes, these efforts consume a significant fraction of the 
operational resources, when integrated over the lifetime of the 
observatory.  The fraction of the total resources devoted to pointing 
issues was about 20% for the 12m (cryogenic/receiver 
development/maintenance/diagnosis/repair was the big "winner" for the 
12m).  Most of this time is personnel resources.

So, echoing what I believe Stephane was suggesting in his 2003 comments, 
two antenna designs adds significantly to the operations cost of the 
observatory.  Note too that this is a recurring cost, so the integral 
over the lifetime of the observatory makes this a big number.

I still worry about polarimetry, though, as being a possible 
"showstopper" for the two antenna problem.  As Steve is our expert, I 
hope he chimes-in on this issue.

Cheers,

Jeff


Mark Holdaway wrote:

> Jeff Mangum wrote:
>
>> Hi Mark,
>>
>> Comment below...
>>
>> Mark Holdaway wrote:
>>
>>> On Beam issues:
>>>
>>> I'm not sure that the different quadripod positions will be a big 
>>> deal for
>>> polarization.  Or more exactly stated, I don't know how to do that 
>>> calculation
>>> to get the polarized beams.
>>>
>>> The total intensity beams are simple to calculate, and the feed legs 
>>> will make
>>> sidelobes of order 1-3%.   However, we never fixed the incorrect 
>>> specification
>>> that we know the primary beam to 6% accuracy, so it won't be too
>>> convincing a problem.  Doing the calculations will at least tell us 
>>> what level this
>>> is, and using real feed leg data (the two times I did this in the 
>>> past, I just made
>>> stuff up and passed  it by people like Tim Cornwell or Darrel Emerson
>>> or JingQuan, who said "Yup, a bit more like *this* and it should be 
>>> sort of
>>> realistic").
>>
>>
>>
>> But doesn't this issue fit into what Stephane called "common mode 
>> errors"?  If all antennas have the same feed leg induced beam error 
>> pattern, then the "signature" of this error term in a polarimetry 
>> measurement would allow for its calibration/cancellation.  Won't 
>> dissimilar designs make this cancellation process much more 
>> difficult, or perhaps not doable at all?
>>
>
> If we have one design, we've got one primary beam to deal with, and one
> polarization beam to deal with.  That is even overstating it, as each 
> antenna's
> VP will be a bit different, so we'll have an average beam with small 
> increments.
> To perform wide field imaging or wide field polarization imaging, we 
> will need to
> take this beam into consideration  --   thats what mosaicing is.
>
> Now, if we have two antenna designs, we actually have three primary 
> beams and
> three polarization beams.  We'll have one average beam, and the 
> increments are
> from that average are larger, but clustered into three camps:  Vert x 
> Vert, EIE x EIE, and
> EIE x VERT.
>
> Low to moderate fidelity imaging can be done using the average beam.
> Higher fidelity imaging will require that we use the three different beam
> models.
>
> In some things, "common mode" errors really just disappear -- if you 
> suddenly
> added 1 mm to the cables to each antenna, the phase error is zero, and 
> you would
> never know about the 1mm excess.    Just beause we have the same 
> primary beam among
> all antennas doesn't mean we don't have to treat that primary beam -- 
> and if we can
> treat the effects of one primary beam, we can treat the effects of 
> three primary beams.
> The cost?   Manpower to characterize those three primary beams, 
> implement them in
> software, and make judgements about when we actually need to go to 
> that trouble....
> and when we DO go to that trouble (??? 10% of the time ???)  we will 
> need to do
> 3 FFT's for every FFT we used to have to do --- if these FFT's 
> dominate the
> computing budget, and IF we need to deal with the 3 beams X% of the time,
> our computer budget needs to increase by  a factor of 1 + 2*X/100  ---
> 10% ==> 1.20  --- a modest factor.
>
>>>
>>> And a final security blanket statement: in principle, we will be 
>>> able to deal with
>>> either the differing total intensity or the polarization beams in 
>>> software;  this is an
>>> effort we knew we would have to address at some point -- these 
>>> algorithms may have
>>> been priority 3 in the SSR's list, having two antenna contracts 
>>> probably brings
>>> that priority up to 2.
>>
>>
>>
>> Can the same be said of either pointing calibration or terms which 
>> are due to gravitational deflection?  The commonality of errors due 
>> to pointing, pathlength, and gravitational deflection seems to be a 
>> major advantage to being able to do all of these things which haven't 
>> really been done before.
>
>
>
>
> OK -- if ALL antennas are mispointed the same way and we know it, we 
> just say
> "the antennas were actually pointing HERE", change the header in the 
> data, and pretend
> that is what we intended to do and proceed with mosaicing (this is 
> similar to OTF mapping
> when the wind blows you off-- as long as you know where you are 
> pointing).  Generally,
> it will be more complicated than that, and you will have different 
> pointing errors for each
> antenna.
>
> BUT, lets take the analog of ALL antennas mispointed the same way for 
> each type of antenna.
> SO -- Vertex ants are mispointed by (x1, y1),   and EIE are all 
> mispointed by (x2, y2).
> SO, just as we have three different primary beams and need to do 3 
> different FFTs.
> we treat all Vert x Vert baselines as pointing to (x0+x1, y0+y1) -- 
> then we treat all
> EIE x EIE baselines as pointing to (x0+x2, y0+y2)  ---   and EIE x 
> Vert would be a bit more
> complicated ---- the Primary Beam then depends on the two 
> mispointings, but all
> EIE x Vert baselines will have the same effective primary beam and 
> effective pointing center.
> And we are just back to using three different beams, pointing centers, 
> and FFTs.
>
> SO:  its a pain in the BUTT -- but it isn't the end of the world.
> I think the maintenance and operational issues are a much bigger 
> argument in the end.
>
>   -Mark
>
>>
>> Cheers,
>>
>> Jeff
>>
>
> _______________________________________________
> mmaimcal mailing list
> mmaimcal at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/mmaimcal