[daip] Ref freq. & ref pixel question you may know an answer to

[Eric Greisen]

Michael Bietenholz wrote:
> Hi Eric
> 
> In AIPS uv-data sets, there is a frequency axis, and a nominal
> frequency in the header, which is the frequency at some particular
> pixel along the frequency axis (ie. typically CRVAL3 and CRPIX3).
> (I'm aware of the further, per-IF shifts from the FQ table).  My
> question is this: Do you know why the reference pixel is *ever*
> anything other than integer (or perhaps half-integer of pixel=1.0
> represented the bottom of the channel)?  Since we're setting
> up the reference frequency when the observations are planned,
> why make life complicated and place your channels with some
> wierd offset from the freq-freq?
> 
> I ask because I'm trying to reconcile some small differences between
> data correlated with DiFX and that same data correlated with the
> RadioAstron ASC correlator, and in one version the REFPIX is 0.75
> while in the other its 1.0.
> 
>                   thanks,   michael

Observations almost always end up with an integer pixel (ref freq in 
center of channel) but with channel averaging (NCHAV and CHINC) rather 
strange reference channel values may occur unless one changes the 
reference frequency to avoid them.  Of course, writers of FITS files and 
the like sometimes do not get this right either and view the left or 
right edge of a voxel as the reference point (it is not - it is the center).

Eric Greisen