[evlatests] CBE time averaging ---> flip of frequency axis

[Vivek Dhawan]

Dear friend/patron of the VLA:

For data observed between May 27 and June 11, the binary data BDF
files might have the frequency axis reversed relative to the labels.
Each subband is reversed. The overall baseband (1 or 2 GHz) is not
out of order.

The f-axis was inverted only if CBE time averaging was invoked; and
the affected version of the CBE was installed.

CBE integration factor is usually invoked for accumulations longer
than 1sec, upto 1s is done in the correlator memory before the CBE.

We are trying to compile a list of potential affected projects (not
too hard) and figure out whether to fix the SDM labels by editing text,
or by plowing through terabytes of BDF. There are of order 150 projects
using CBE time averaging in the peroid.

Fixes in post-processing and in the archive are being considered.

A complication is that the timeline I have compiled below, with known
version changes, is contradicted by examination of actual data, in the
sense that some data that I expect to be bad is in fact clean.

Another complication for an eventual fix is recirculation on some
subbands often have less CBE integration or none, compared to sub-
bands without recirculation observed in parallel (the final output
integration time of all subbands in a BDF is forced to match.)

Comments on the clarity or content of this message are welcome!

Please feedback your results. Bandpass amplitudes, especially at
baseband edges, are obvious, especially if you have a comparison
dataset from outside the timerange. Phases are not conjugated.
Delays will change sign, and will show discontinuities between
contiguous subbands.

-------------------------------------------------------------------
                 This table is work in progress

* marks a known CBE version update. <<< marks a dataset that fixes
the condition prevailing on that day, either bad or good. A longer
list of projects that either have or have not been affected is
compiled at the end. Drew and I have looked at pipeline output and
unfortunately do not see ANY projects that are clearly bad. This
does not necessarily mean they are fine.

    date       CBE version    Comment

thu jun 11     2015 0611 *   Fix installed after testing.

     jun 10     2015 0603

     jun 09     2015 0603     NEW CM. Phasing broken (2sec) <<<

mon jun 08     2015 0603     Pulsar phasing broken (2sec)  <<<


     jun 04     2015 0603

     jun 03     2015 0603 *   CBE tagged, 'no change in function'

tue jun 02     2015 0526     CM restarted, minor update.


     may 28     2015 0526     ALL phasing good in daytime. <<<

wed may 27     2015 0526 *   CBE, CM, Executor updates.


     may 21     2015 0128     new CBE tested, NOT installed.
                              Phasing broken in daytime tests <<<

     may 20     2015 0128     test MCAF off

     may 19     2015 0128     test MCAF off

mon may 18     2015 0128     test MCAF off


     may 14     2015 0128

     may 13     2015 0128

     may 12     2015 0128     CBE testing.

mon may 11     2015 0128     CBEfail, testMCAF. testing
                                      no change installed.

     may 07     2015 0128     Phasing good: pulsar test.

wed may 06     2015 0128     TestExec for 32 subb.
                              testMCAF off.

-------------------------------------------------------------------

I have the following confirmed contradictions of the timeline above:

15A-239 on may 31, 3s avg., 15:20-17:40 UT - CLEAN!


-------------------------------------------------------------------