[daip] DFTPL question/remark
Eric Greisen
egreisen at nrao.edu
Wed Mar 11 15:19:38 EDT 2009
Hanno Spreeuw wrote:
> Hello,
>
> The sampling time of the visibilities in my dataset is 5s.
> When I set bpar(2)=5, I get 2879 flux density values.
> When I set bpar(2)=1 or 4.99, I get 3842 flux density values.
>
> Perhaps this cannot be fixed, but I just wanted to mention it.
The issue is the perennial issue of time binning (esp given that time
recording is not perfectly accurate). If an interval starts with a
sample at T1 does it include all samples <= T1+BPARM(2) or all samples
< T1+BPARM(2)? The code was the former so I changed it to the latter.
I put a remark in the help file to provide some guidance. Since times
are not perfectly accurate (days in single-precision floating point)
even this change will probably get you 2 samples averaging together on
occasion. The use of 4.9 or 1 is recommended. To get 2 together most
of the time (i.e. where possible), a BPARM(2) value like 6 rather than
10.0 would be recommended.
Eric Greisen
More information about the Daip
mailing list