Pdev maintenance procedures
090907: exchanging/replacing pdev
spectrometer
There are two sets of 7 pdev spectrometers: group 0
and group 1. These can be used for two separate experiments that are
run simultaneously. Each of the 7 bbmixers feeds the same signal
into the group 0 spectrometer box and the group 1 spectrometer box. For
various software reasons (cal mainly) you must run with group 0, group
1 is optional.
If a spectrometer box in group 0 dies (eg.. fan
failure, fpga overheats) you can replace it with the corresponding
group 1 box using the following procedure. Datataking can then continue
(but without the group 1 experiment).
- Login to pdevs1 as root or phil
- edit the files:
- /usr/local/pdev/etc/boxes.conf.templ. You need to swap the ip
addresses for the boxes you are switching.
- Eg.. suppose b1sxg0 failed and you want to switch it with
b1sxg1.
- before:
- b1s0g0 pdev-103
1
0 0
gxa pdevs2
- b1s1g0 pdev-103
1
1 1
gxa pdevs2
- b1s0g1 pdev-108
1
0 0
gxa pdevs9
- b1s1g1 pdev-108
1
1 1
gxa pdevs9
- After:
- b1s0g0 pdev-108
1
0 0
gxa pdevs2
- b1s1g0 pdev-108
1
1 1
gxa pdevs2
- b1s0g1 pdev-103
1
0 0
gxa pdevs9
- b1s1g1 pdev-103
1
1 1
gxa pdevs9
- /usr/local/pdev/etc/cal.conf
- You should switch the entries of b1sxg0 with b1sxg1 so that
the correct a/d offsets are used. if you have time you can just rerun
the a/d cal.conf routine
- HardWare mods:
- Move the serial cables for the attenuator settings from
pdev-103 to pdev-108
- Ethernet cable from pdevBox to switch.
- As long as you are only switching 1 set of boxes you don't
have to mess with the ethernet cables since the switch for group 0 is
connected to the switch for group 1 via a 1 gbit link.
- If you switch more that 1 set, then you should probably also
switch the ethernet cables that plug into each spectrometer box. For
fast datataking more than one spectrometer box would have to share the
single 1 gbit link between group0,group1 switches to get to their
respective spectrometers.
Permanent replacement:
If you want to put a new spectrometer box in
to replace a failed one:
- Offline:
- /share/megs/phil/svn/pdev/pdev/pnetbld/boxes.conf.templ
- Make the same changes you did for the above online file.
- svn commit boxes.conf.templ .. update in
reposistory.. add message when requested..
- Online:
- edit the file /usr/local/pdev/etc/boxes.conf.templ to put the
ip address for the new box in (see above)
- Make sure the bbm serial cables are on the group 0 cpu. (eg
b2sXg0) spectrometer box.
- Remeasure the a/d cal values.
- Warning:You need to run
pnet on the new box as root once before letting cima take over.
- When the routine ssh's into the new box, ssh gives a
warning
messages and asks whether it is ok to connect to this machine... This
mesasge screws up cima so you need to get it out of the way
before running cima.
- !! This has to be
done as root since cima runs pnetctl as
root. Each user will get their own ssh message requesting if it is ok
- As root on pdevs1
- cd /dump/pdev/files
- pnet --conf=pnetAll.conf --sigstat --extclk
- If you get an ssh message just enter yes.
- After it runs a few loops, just ctrl-c
- You can then run cima.
090907: Making sure all boxes select extclk
When the power to a spectrometer box is cycled, the
spectrometer comes back with the internal clock selected. If you then
run cima you may get an error that the clock selection is not the same
for all the boxes in use. You can fix this with:
090907: Verifying the synthesizer levels.
The mock spectrometer uses 3 synthesizers. They are
located in the middle rack of the mock spectrometer. They are
controlled manually. These are the things to check:
- clock: 172.032Mhz 12dbm (top)
- lo 0 : 175Mhz 12dbm
- lo 1 : 325Mhz 12dbm
- rf on light is on (lower right of each synth).
- The rf on light or frequency changes could occur if their power
is cycled. This shouldn't happen as long as the ups's are working
090907: updating the a/d offsets cal.conf
file
The spectrometer uses the file
/usr/local/pdev/etc/cal.conf to remove the a/d offsets (it also does
scaling but we don't use that). If change a spectrometer box, or the
a/d offsets reported while adjusting power is large you should remake
the file.
Remaking the files entails: setting the downstairs
if/lo to the noise source and then running pnet with the
--sigcal=filename command to regenerate the files. This can be done
with a script:
- On a vxWorks window enter: if2noise .. this selects
the if2 noise source.
- slogin to pdevs1 as phil
- cd /dump/pdev/files
- caldoit
- hit return when it prompts you to be sure that if2noise has
been entered. It will then recreate the file
/usr/local/pdev/etc/cal.conf.
- Make a copy of the new file for historical purposes:
- from /dump/pdev/files:
- cp /usr/local/pdev/etc/cal.conf rmsyymmdd.out
<-
page up
home_~phil