Date: Thu, 31 Jul 2003 10:57:37 -0700
From: Daniel Sigg <sigg_d@ligo.caltech.edu>
To: Gabriela Gonzalez <gonzalez@lsu.edu>,
     Keith Riles <keithr@tenaya.physics.lsa.umich.edu>,
     John Zweizig <jzweizig@ligo.caltech.edu>,
     Ed Daw <edaw@lsuligo.phys.lsu.edu>,
     Peter Fritschel <fritschel_p@ligo.mit.edu>,
     Brian O'Reilly <irish@ligo-la.caltech.edu>,
     Rauha Rahkola <rrahkola@darkwing.uoregon.edu>,
     Robert Schofield <rmss@conch.uoregon.edu>,
     Peter Shawhan <shawhan_p@ligo.caltech.edu>,
     Stan Whitcomb <stan@ligo.caltech.edu>, Daniel Sigg <sigg_d@ligo.mit.edu>,
     Rana Adhikari <rana@ligo-la.caltech.edu>
Subject: S2 Data Quality: byte swap error

S2 Data Quality: byte swap error:

Here is an email from Rolf describing the recently found byte swap error
that affects H1 IOO PSL, SUS coil readbacks and PEM channels acossiated
with H1.
> http://blue.ligo-wa.caltech.edu/ilog/pub/ilog.cgi?group=detector&date_to_view=07/25/2003&anchor_to_scroll_to=2003:07:25:17:18:39-rick
 

This was happening during S2.
Daniel

>During signal injection testing last week, a problem was noted with data from
>two ADCU (h1adcusus and h1adcupem). This was subsequently tracked down to a
>software problem in these two ADCU. These two ADCU are different than all others
>in LIGO in that they are the only two directly connected to the framebuilder network.
>
>The problem originated when the framebuilder network at Hanford was switched
>over to a newer high speed network back in January. The new net was installed to
>accomodate the high data load. The older network cards had hardware byte
>swapping (needed to send data from a Pentium to the Sun), but the new network
>boards do not have this feature. Therefore, the software was changed in these
>ADCU to do the byte swapping. Unfortunately, it was not noted that while the
>ADCU writes the data as shorts, the framebuilder, using 32 bit DMA to recieve
>the data, actually reads integers. This results in inproper byte swapping.
>
>The net result in the stored data frames is that data samples for these two ADCU
>are correct in value but out of time order. So the time series, from the
>beginning of a frame, instead of being (S)ample 0, S1, S2, S3, S4, ..., got
>stored as S1,S0,S3,S2,S5,S4 .... ie pairs of data samples are swapped in the
>time series.
>
>The software was corrected, reloaded and tested yesterday. Unfortunately, the S2
>data must somehow be corrected or at least be noted that this problem exists in
>that data.
>
>-rolf