Transcript MuCapWfd

WFD analysis for runs 10/11 –
first steps
Rob Carey
MuCap Meeting
July 24, 2008
Pulse parameters
1. Gondola number (1-17)
2. Tube number (1-4)
3. Start of pulse
4. End of pulse
5. Block Time
6. Position of maximum on island
7. Leading Edge Time
8. Weighted time (top 7)
9. Maximum value
10. Integrated pulse height (top 7)
11. Width
Data Integrity
WFD features/problems
(mostly run 10)
•
•
•
•
Hole in the trigger
Misformatting disease
Start-time ambiguity
Bit dropping
The hole in the trigger –
(continuation block logic)
4 missing
Samples
Including
unseen
trigger
Original
trigger
pulse
From
next pulse:
NOT trigger
24 samples
WFD is triggered
anyway
No clear sign of events
lost
Look at muSC data delta-Ts
Should we worry about the hole in
the electron data?
spike
Discriminator effect?
Afterpulses?
bump
hole
ns
Time between WFD blocks – one channel
No, there are much bigger things
going on just after each pulse
spike
hole
ns
Misformatting disease
• Symptom: Ask for DBC=4 (for example), get
data that looks a lot like DBC=2 (for example)
with corrupt timewords
• Observation: Trouble starts or stops only on
begin run (with CR/CSR manipulation) and
specifically by setting block count (DBC), #
presamples, personality
“Golden run list” would include none of these
Will Ag suffice?
• Traced to corruption of CR/CSR space caused
by …
When is a pullup resistor not a
pullup resistor?
• Answer – when it appears in the firmware
and disappears in the instantiated circuit.
Turns into
Q: In what state is the
line (i.e. geographical
address) when the
level is latched?
A: Who knows?
Resolution – force the compiler’s hand by
inserting an explicit resistor, eliminate “listening
on the line” (or reduce # of WFDs in crate)
Run 10 solution (Tim,Volodya)
• Reduce number of WFD boards per crate
• Flag outbursts of misformatting disease in
WFD analyzer (basis of mu WFD code)
turned off (sad emoticon)
After April 11, disease mentioned only
once in e-log, corresponding to adjustment
of WFD thresholds. Devastates data from
MuSC WFD (run 41032).
The good news about
misformatting disease
• It’s very severe – destroys time words,
spoils coincidences, should be easy to
spot in mta data (next time in mu)
• Lasts whole runs – it’s there or it isn’t
• It’s very severe – data cannot be
recovered. Especially bad for mu+
analysis
Start Time Ambiguity
Ch 1
Ch 2
Assignment of time 0
depends on when start
signal arrives
Ch 1
Ch 2
a) Run 10 – WFD clocks
(1,2,3,4) are distinct and
not in phase
1
Inner Scintillator
3
Outer Scintillator
2 b) Run 11 – WFD clocks (1,2)
and (3,4) are separately in
phase but (1,3) and (2,4)
4
are not in phase
Average
delta-t
between
channels
1 and 3
over a
segment
Now current code in
mta assumes that
midpoint of 2 bumps on
previous page does not
change over the run …..
… but it does 
calculate midpoints on
the fly
Still to do
• Scan mu output for evidence of
misformatting disease (run 10)
• Profile channels for bit dropping (run 10)
• Look for evidence of hole in trigger in
muSC (run 10) : none yet observed
• Time-word goes backward (seen on rare
occasions in MuLan 2006/7)
• Eliminate (flag) fills with few pulses (can’t
resolve ambiguities)
Longitudinal studies
• Time offsets
 WFD-TDC offsets (muon and electron)
 channel/channel offsets within 1 gondola,
used in forming coincidences, helpful in
resolving Start-time ambiguity
Channel assignments – OK
WFD Golden run list – particularly for run 10
+ list of problem channels
Physics requirements
• Time pickoff should be stable vs. time since kick
: no problems anticipated, but since no laser
calibration, check for consistency between
detectors
• In the same spirit, amplitude should be stable vs.
time since kick
Challenge – background pulses don’t look like
decay electrons
Especially for mu+ want homogeneous response
ADC
Cnts
ns
Pedestal vs Time since Kick (all pulses)
ADC
Cnts
ns
Amplitude vs time since kick (all pulses)
ADC
Cnts
ns
Pedestal vs time since kick (4-folds)
ADC
Cnts
ns
Amplitude vs. Time since Kick (4-folds)
ADC
Cnts
ns
Definitely more particle background here (4 –folds)
ADC
Cnts
ns
Pedestal is flat for these 4-folds
For Mu+ Analysis, want uniform
coverage
• Want uniform (and high) detector response
over the solid angle
• Check efficiency of WFD 4-folds over the
Gondolas using ePC tracks as reference
Efficiency gondola 4
mm
mm
Efficiency of scintillator gondola 3
WFD/PMT/scintillator response is
• High (96-97 percent)
• Uniform over the length of the detector
 Little effect from attenuation
Max sample vs phi from run 10 (no pedestal subtraction)
Average of four MaxVals – pedestal vs phi
 Response is pretty flat over phi (apart from missing sector) : run 11
Hole in the ePC
Why you need to be careful with
time offsets
• Timing offsets for run 11 wander. Coincidences
(like these in 62800) sometimes disappear.
Not from
Production
mta !
ns
ns
ns
Gondola 9 : dt layer 1 vs. dt layers 2
Gondola 4 is often bad
Conclusions
• Mu reconstruction looks OK, as far as I can tell only MuSC might want a PFA upgrade
• Longitudinal studies needed to assess extent of
run 10 problems - start with mta output, revise
mu, no big problems anticipated for run 11
• Need to arrive at WFD golden run list
• If time correlations and tight coincidences are
important, need 2-pass mta code (NBD) to
establish timing offsets on a run-by-run basis