131024: dUpdated software on pdevs0 to most recent code.
Switched the working directory on pdevs0 to use the http
server ( it was still using the megs2 repository). Decided to
keep the box names b0spN rather than b0sMgN .. Will
cause a little hassle moving config files between pdevsN and
131007: mock synth's have unlocked error. Dana
measured jumps in 10 mhz reference (at output of 3
synths 10 mhz daisy chain more
info) . Jumps of up to a few cycle. About 6 in a few
days. We saw this a few years ago. The synth with the 10 mhz
from buffers never has jump (just the daisy chained ones). The
synth's re regenerating the 10 Mhz.
replaced pdev-108 (b1sXg1) with pdev-101.
We were having lots of bit flips in the data b1s1g1
polA. polB looked ok.
redid cal.conf for new pdev-101 a/d
replaced pdevs7 spectrometer (pdev-109)
with pdev-127. We were having lots of dump timeouts. Dmesg
on pdev-109 showed some ethernet 0 timeouts.
replaced b5sxg0 box 104 with box 112. Chris had
problems with jumps in b5s1g0
The jumps in the spectra looked like bit flips in
when running adjpower, there were many a/d
overflows (looking at /tmp/pdevsig.log on pdevs1). If the
bit flips were occurring, they must have been occuring
vary early in the pipeline since the a/d overflow bit
check is right after the a/d and mux switch. This bit then
gets carried along with the data sample. Since the
bit and the data sample were probably bad, it really must
have been a problem at the early stages.
Ran diagnostics before box 104 replaced.
rdcomp failed. mainly with b5s1g0. (b1s1g1 also
failed a lot).
11jun11: pdevs6 disc? in raid fails. replaced
17jun11: pdevs2 disc2 in raid fails. replaced
pdev-113, pdevs7, b6sxg0 fan2 (gxb) dropouts).
looks like power supply trouble
pdev-113 replaced with pdev-109
pdev-107, pdevs3,b2sxg0 fan 2 replaced.
b1sxg0 boxes fan failed, fpga
died. This was pdev-109. Replaced with pdev-103.
Swapped I,Q cable in spectrometer box: b6s1g1
(bm 6, high band, polA): dig i0,q0 signal cables swapped
inside spectrometer box. Probably occurred when the
dig levels were changed from 1.5 to .78 volts. Group1 has
not yet been used for datataking. 090216 check all the other
boxes for i/q switches.. None other found.
081205: replaced disc 5 on pdevs1 raid.
The disc i/o had not been keeping up with galfacts
The raid stats showed that disc 5 had more seek
errors, timeout cnts..
Accidentally overwrote pnetctl.html.. Grabbed backup from
12sep08.. and then fill in some of the lost stuff.
B2SxG1 switched out spectrometer box pdev-105 . replaced
with pdev-122. connects to pdevS10 server. digitizer S0 IA was
Used a test program to write as fast as possible to a
psrfits file (input came from memory) (.ps)
080821: svn checkout https://mock.com/svn/pdev/web web on
pdevs1:/home.local/pdev/pdev to get the web files
for the website.
psrv.c was also using the fftlen to compute how many
blks fit in 1 second. Updated so it used the sample rate.
pnetbld now correctly computes how long each block takes in
time domain mode.
or stokesI is stored as unsigned int while fits uses
signed. Changed bpOut.c to clip to maximum signed value for
computation in pnetbld. Shift up by 1 more for
stokesU,V relative to pola,polB. This makes the rms of U,V
the same as polA,B since jeff outputs 2*polA,2*polb, but only
1*U, 1*V. (Note:
13jun08 i shifted 1 less for u,v..17jun08 i corrected it to
shift 1 more). check data