MVD installation, fall 2004

1
MVD installation, fall 2004
Pads only
2
Work done with IhnJea Choi
MVD after successful installation
Beam pipe is pushed to the center with a plastic
block (South side)
3
Initially, the South Glinks would not lock.
Traced to bad 10MHz clock on backplane
North crate backplane
South crate backplane
Caused by one bad DCIM module - removed
Now everything reads out as before
4
We also rebuilt the test system in the 1008 side
room, with 3 MCMs
5
Safety improvements
20
Getting up to the lift table meant going op 3
20 steps behind the MVD rack. No floor in that
area
1) Made 3 removable 10 steps
MVD rack
2) Made new piece of false flooring
6

• Problem AMU cell numbers, and event counters, go
  out-of-step
• during multi-event buffering.
• This is a problem in the ALM (address list
  manager) FPGA code.
• Can we reproduce the problem? Yes
• put 5 fast triggers in the GTM file, or
• Insert Imrans fast random triggers into the
  Phenix trigger system

Not quite equivalent Problems show up after
5-10k events, and accumulate. Modify ALM
firmware no more errors using the GTM
triggers, but still errors using the random
trigger some progress... I wrote Perl scripts
to quickly look at the relevant data words
7
Ongoing work
IhnJea will not be back until January 2005, but
he has some time to work on this this fall.
He can prepare a few modified versions of the ALM
FPGA code for testing. I can run the system from
here (with a phone call to John Haggerty to plug
in the random trigger), and try out the new
codes. Unfortunately there is no deterministic
way to find the race conditions that are the
cause of these problems. Keep on plugging away