Dear Patrick,

Thanks for your message. For now this ARTS run concerns a sensitivity study to 
see the impact of the sidelobes, compared to an ideal Gaussian response. I 
already suspected that this will not be useful for day-to-day simulations and 
probably requires some combination of smoothing/averaging/approximation and 
probably a transition to 1d, possibly even just the first Stokes parameter as 
you suggested. However, it would be fun to run it in full configuration once, 
even it takes ages. 

I'll work with Oliver Lemke to get a simple custom response running, and then 
will use your comments to slowly add dimensions/Stokes vector elements and see 
where it breaks. I'll make sure to get back to you if and when that happens (or 
even if it doesn't to share the results). 

Best wishes,
Frank

--
Frank Werner
Mail Stop 183-701, Jet Propulsion Laboratory
4800 Oak Grove Drive, Pasadena, California 91109, United States
Phone: +1 818 354-1918

On 10/18/22, 10:56 AM, "Patrick Eriksson" <[email protected]> wrote:

    Dear Frank,

    First a warning. 2D antenna patterns in ARTS are implemented in a crude 
    way and result in slow calculations. In addition, hardly used and 
    accordingly not very well tested.

    Not sure where you need help. I would suggest to start without 
    polarization. You seem to have what is needed. The max response of the 
    antenna pattern, the boresight, should be placed at zenith=0 and 
    azimuth=0. If comparing to lat/lon, zenith=0 should be placed at the 
    equator, so a change in e.g. 0.001 deg in zenith and azimuth give equal 
    changes at some distance. I hope the description of antenna_response is 
    clear enough to understand how to fill the variable/file.

    Even if your antenna data are normalized, I still recommend to set 
    sensor_norm=1. Otherwise, discretization inside ARTS could have an effect.

    Set antenna_dlos to [[0]].

    The tricky part is to set mblock_dlos_grid. This variable is a bit 
    special. It contains a set of relative deviations from the boresight, as 
    pairs of zenith and azimuth deviations. mblock_dlos_grid descrives the 
    "pattern" of pencil beam calculations done, to estimate the radiance 
    field seen by the antenna.
    If you look at sensor_responseAntenna, there are two options. I 
    recommend the "interp_response" option. Here you basically define a 
    dza_grid and daa_grid, and put all combinations in mblock_dlos_grid in 
    the order described.

    When you get this to work and results look OK, then add polarisation. 
    You could be the first one to use 2D antennas with polarisation. So 
    maybe also do this with 1D antenna to verify.

    I hope this is of help. Don't hesitate to ask if something still is 
    unclear, or you get stuck.

    Bye,

    Patrick



    On 2022-10-17 18:42, Werner, Frank (329D) wrote:
    > Dear ARTS team,
    > 
    > After playing around with a 1d Gaussian antenna response (using the 
    > workspace variable ‘AntennaConstantGaussian1D’), I now want to set a 
    > custom antenna pattern from actual lab measurements.
    > 
    > I have vectors for the zenith and azimuth angles, as well as a grid of 
    > normalized antenna responses for each Stokes vector element. The 
    > normalization follows the ARTS behavior, where the integral of the 
    > responses over all angle increments is 1.
    > 
    > Can you guys help me out with correctly using that info in ARTS? I 
    > briefly looked at creating a workspace variable named ‘antenna_response’ 
    > with ‘GriddedField4’, but somehow the format confuses me. Also, I am 
    > pretty sure that this is not enough and there probably are more steps I 
    > need to take for ARTS to accept that variable as the antenna pattern.
    > 
    > Thanks for your help and best wishes,
    > 
    > Frank
    > 
    > --
    > Frank Werner
    > Mail Stop 183-701, Jet Propulsion Laboratory
    > 4800 Oak Grove Drive, Pasadena, California 91109, United States
    > Phone: +1 818 354-1918
    > 

Reply via email to