POSSIBLE WRONG FREQUENCY VALUE IN ARECIBO FITS-HEADERS

Data affected: all previous FITS-files up to 2008-Jan-05
========================================================

A bug affecting the reported center frequency and (extremely rarely)
the sideband selection has been found and corrected.

Note that this bug only affects the values of the header parameters -
the data itself has been taken at the correct frequency and sideband.
The problem here is that there will be a discrepancy between the
center frequency written to the FITS-header and the real center
frequency of the data.

The effect of the bug has been that the reported center frequency
(FITS-header keyword CRVAL1) for some set-ups have been wrong. It is
also possible but very unlikely that the wrong sideband is indicated
(FITS-header keywords UPPERSB and CDELT1).

The problem only occurs for the situation where the WAPPs start taking
data immediately after a Doppler correction. In those cases, the new
updated frequency settings may not yet have propagated to the WAPPs
and the center frequency that was used BEFORE the Doppler correction
will incorrectly be written to CRVAL1 for all data taken in that
scan. Since this is a timing problem, the error is random in nature
and may or may not appear. Also, it may not affect all the WAPPs at
the same time.

This problem affects WAPP FITS-files generated up until 5 January
2008, although it does not affect all files since it only appears for
set-ups where a Doppler correction is made immediately before data is
taken (without any waits, power adjustments or other operation that
takes a bit of time).

A corrected algorithm was installed 5 January 2008. The version number
of the CIMAFITS-files has been updated from '1.08' to '1.09' to
reflect this correction.


DETAILED DESCRIPTION
--------------------

AFFECTED KEYWORDS:

The affected keywords are:

CRVAL1             (center frequency of spectra)
UPPERSB            (indication if spectrum is flipped)
CDELT1             (frequency interval between pixels)

The problem is due to the fact that the frequency set-up is picked up
by the WAPPs from broadcasts occurring once a second, and there is
thus a small delay before they get the updated values. Thus, if the
WAPPs are commanded to take data immediately after a frequency change
which typically would be a Doppler correction, they may or may not
have received the new frequencies. The center frequency is deduced
from the set-up before the data taking starts and is not updated
during the scan. Thus, if an erronous value is picked up, it will be
written to the FITS-header throughout the entire scan. Synthesizer
frequencies will, however, be read after each subscan and will thus
always be correct.

The key word is 'immediately' since any action like a power adjustment
or a wait of any kind will prevent the problem. A typical mode
affected would be 'standard ON's with Doppler correction but without
power adjustments, for example if run in a loop with Doppler
correction on every loop but power adjustment only on the first
loop. In this case the first loop will be good, but all subsequent
ones will probably be affected.

In tests, it appears that the problem do occur most of the time when a
Doppler correction is made immediately prior to an
observation. Although all four WAPPs occasionally were affected it
seems to be more typical that only one or two are affected. Since the
WAPPs are called in sequential order, the problem will appear in the
first WAPPs, i.e. if two WAPPs are bad it will be WAPP-1 and
WAPP-2. Thus, if WAPP-1 is OK, all WAPPs will be OK, and if WAPP-4 is
bad then all WAPPs will be bad.

The error will typically be small since it is the error between the
current and the previous Doppler correction. However, if an observer
has switched between different sources without adjusting the power in
between, the error may be large, since the frequency used then will be
the frequency for the last source.

The bug should rarely affect UPPERSB and CDELT1 since it would only
occur if the Doppler correction changed the frequency in such a way
that the sideband is changed. If the wrong sideband is reported, then
UPPERSB will have the opposite value ('0' instead of '1' or vice
versa) and CDELT1 will have the wrong sign. The absolute value of
CDELT1 is always correct since it is the sign of CDELT1 that carries
the sideband information.



                                        Mikael Lerner

                                        Arecibo, 5 January 2008