A brief Summary on Synchronization
Strategies
in the Muon System
y
Not a comprehensive talk,
talk just basic informatio
Overview
¾
¾
¾
¾
¾
The
ep
problem
ob e as we
e see itt
Tools for synchronization
Synchronization in a changing scenario
Theoretical and practical synchronization
Synchronization steps and schedule
A. Lai – 26/7/2007 – CERN, Commissioning Meeting
1
Synchronization: the problem as we see it
1. Synchronizing the Muon System is:
a)
b)
For a given event, align all the hits within one ODE board and
give
i them
h
the
h SAME Id (relative
( l i alignment)
li
)
Give the hits the RIGHT Id (absolute alignment).
2 O
2.
Owing to the
h nature off the
h Muon Detector, synchronizing
h
is
also:
a)
Measuring the time of the hits inside the clock period, building
the time spectra and centering them inside the 25 ns window, in
order to maximize detection efficiency
In our jargon:
1 is called coarse timing,
timing,
2 is called fine timing.
timing.
• Both tasks 1 and 2 are heavily complicated by the complexity of the Muon
System connectivity
• An automatic calculation of latencies and delays can be useful but does
not spare us from measuring and checking the effective delays on detector
A. Lai – 26/7/2007 – CERN, Commissioning Meeting
2
System Tools for
Synchronization
Fine time tuning (DIALOG, on FEE)
1.6 ns x 31 steps (per single channel)
Coarse time tuning (SYNC, ODE)
Per board (all 192 channels)
• TTCrx settings
Per SYNC (all 8 channels)
• BX offset
• BX preload (at BCR)
• L0 buffer latency.
Per channel:
• Pipeline steps to L0 trigger
• Pipeline
Pi li steps at SYNC inputs
i
Pulses to FEE inputs
Synchronous
Sy
c o ous to a g
given
e
programmable BC#
A. Lai – 26/7/2007 – CERN, Commissioning Meeting
3
SYNC internal dataflow
BCR, L0R
To L0 triigger
Data from FEE
preset/get
t/ t
BC counter
via ECS
BX tag
Start W_Add
L0 buffer
DPRAM
256
Programmable
L0 buffer latency
(R/W pointers)
Start R_Add
L0yes
BXId(3:0)
L0 derandomizer
DPRAM
128
(BCR)
To ODE
controller
t ll
A. Lai – 26/7/2007 – CERN, Commissioning Meeting
4
Synchronization: different strategies for different scenarios
1. On the ODE we have the possibility to measure directly the local time
difference with respect to the start of the orbit.
This would directly
yg
give the RIGHT BX correction (absolute
(
alignment),
g
),
but requires a “good” event rate
Æ not the case in the (long) startup period
Æ method to be used as a crosscross-check later.
2. Preparing the pilot run:
Calo interaction trigger and multi
multi-bunch events.
Æ No urgent need now for this setup
Æ It will be prepared in time for the beam
Present strategy:
Use of pulsing system + SYNC features, both at ECS and DAQ
level
A general comment:
We experimented on our own skin how the “real world of connections - O(105)” is
different wrt our models. We are now checking all the errors in connectivity
(inversions/lacking or defective connections). We still do not know how much the
cable length database differs wrt reality (I can say it surely differs a lot !!).
A. Lai – 26/7/2007 – CERN, Commissioning Meeting
5
Internal vs global timing
Typical scheme of Muon signal flow
LVDS cables (can have
different lengths)
The same FE channel
can belong to
different logical
channels.
Inter-channel skew
i k
is
kept b
below
l
2 ns
IB
Generation of logical
channels (final step)
Chambers
First logic
combinations on
DIALOG
ODE
DAQ only sees
logical channels
Internal equalization can be done:
1. At DAQ level masking FE channels one by one
2. At ECS level by means of the pulsing system
A. Lai – 26/7/2007 – CERN, Commissioning Meeting
6
Synchronization steps foreseen in the present scenario
0. Use the cable length information to calculate an a-priori set of timing
parameters: fine time delays on DIALOG and coarse time delays on SYNC.
1.
Use the information on the ECS cable length and the pulsing system to align
each logical channel (LCH) internally (channel time equalization), as the LCH
are made of different combinations of channels.
channels Once the INTERNAL LCH
components are aligned:
2.
Possible pulse runs, acquiring data on DAQ and analyzing them offline (time
information generated by SYNC).
3.
Optional: track emulation, cosmics ??? (triggering on M3 only, bi-gaps in
AND).
AND)
4.
Same as 2, but with real particles and external trigger (when the beam is on),
to take into account also chamber responses. At this level the Calo trigger
should enter the game.
A dedicated Brunel package is being developed to perform a quick and
efficient data analysis oriented to the detailed calculation of DIALOG and
SYNC timing parameters.
A. Lai – 26/7/2007 – CERN, Commissioning Meeting
7
System Time Alignment SW: previous idea (pilot run oriented)
.xml
Encoded
maps
Condition
DB
BRUNEL
algorithm
Raw data
(multi-bunch
events)
PCfarm
events
Ntuples (.root)
containing
i i
connectivity
i i
attributes as leaves
This structure
is “circular” if
not iterative
Time histograms
(HW map dependent)
Mask
k settings
(fit dependent)
Cable length information
This link is
optional
Histogram fit
Delay & mask
calculation
Muon
System
Sys
configuration
PVSS
SS
HW CU
datapoints
p
Coarse and
fine delays
+
masks
k
(.txt)
(SYNC and
DIALOG chip
p
list)
datapoint
A. Lai – 26/7/2007 – CERN, Commissioning Meeting
8
New Structure of the System Time Alignment SW
.xml
Encoded
maps
Condition
DB
BRUNEL
algorithm
Raw data
(multi-bunch
events)
PCfarm
events
In this case much of the
work
k iis d
done previously
i l b
by
ECS. In principle, the result
can be obtained in one go. If
more runs are needed they
m
y
are independent
Masking should
not be
necessary
Ntuples (.root)
containing
i i
connectivity
i i
attributes as leaves
Time histograms
(HW map dependent)
Mask
k settings
(fit dependent)
Cable length information
Histogram fit
Delay & mask
calculation
Muon
System
Sys
configuration
PVSS
SS
HW CU
datapoints
p
Coarse and
fine delays
+
masks
k
(.txt)
(SYNC and
DIALOG chip
p
list)
datapoint
A. Lai – 26/7/2007 – CERN, Commissioning Meeting
9
Approximate schedule
August:
Connectivity tests continue
Not later than mid September:
Start internal alignment procedures (ECS) in parallel with
connectivity tests,
tests if not yet concluded
concluded.
Mid October:
Pulse runs and analysis Æ first set of System timing parameters
generated.
We aim at concluding the alignment by means of pulse runs by the
end of the year. Pulse runs are also a complete test of the DAQ
chain.
A. Lai – 26/7/2007 – CERN, Commissioning Meeting 10