[daip] SETJY

[Eric Greisen]

Andreas Brunthaler wrote:
> Dear Eric,
> 
> I hope you are all back at work now in Socorro. I'm having some problems 
> with the velocity calculation in SETJY. I compared the results I get 
> when using the 31DEC2011 and the 31DEC2013 versions and get different 
> results for the same input parameters.
> 
> I'm using OPTYPE '' and the following information:
> 
> SYSVEL       18
> RESTFREQ   6.668E+09   519200
> VELTYP     'LSR'
> VELDEF     'RADIO'
> APARM       501,0
> 
> Now, If I use the 31DEC2011 version, I get in the SU table:
> 
> 
> COL. NO.            2                18                19
>       ROW      SOURCE            LSRVEL            RESTFREQ
>    NUMBER                        M/SEC             HZ
>         2      G000.31-0.20      1.978371D+05      6.668519D+09
> 
> 
> In the 31DEC2013 version, I get:
> 
> 
> COL. NO.            2                18                19
>       ROW      SOURCE            LSRVEL            RESTFREQ
>    NUMBER                        M/SEC             HZ
>         2      G000.31-0.20      179825.49691      6668519456.0
> 
> 
> If I use OPTYPE = 'VCAL' in the 31DEC2013 version, I get yet another value:
> 
> 
> COL. NO.            2                18                19
>       ROW      SOURCE            LSRVEL            RESTFREQ
>    NUMBER                        M/SEC             HZ
>         2      G000.31-0.20      233796.30089      6668519456.0
> 
> Apart from the different values I also get a different format. Now, if I 
> run CVEL on this, I get shifts of < 1 channel for the 31DEC2011 version 
> (as expected). However, for the 31DEC2013 version, I get shifts of 50 
> channels (using OPTYPE='') and 100 channels (using OPTYPE='VCAL'), which 
> corresponds to the velocity offsets above.
> 
> Any idea where this difference comes from?
> 
> Andreas
> 
Actually I think I know what version of aips you are using $ARCH='LINUX'
which is the 32-bit version of Linux.  I wonder if your machine could 
use the 'LNX64' version which is also available in Bonn.  You might ask 
Walter Alef or whoever knows about such things there.  The LNX64 version 
works correctly and is the modern and slightly faster way to go.  I will 
try to see what is going on with the 32 bit version.

Eric