Simulation upgrade
The L1-trigger upgrade requires us first to improve and redesign the trigger
simulation and secondly to develop new tools to evaluate the DOCAZ design algorithms.
A new DOCAZ analysis package called
L1DczTest
contains the new tools to test the conceptual design for the upgrade. Once ready
the relevant modules will be incorporated into the full trigger simulation after the
appropriate standardization to the Babar framework.
Disadvantages of the current simulation
The trgDC simulation currently suffers from the following disadvantages:
- trgDC is written in Fortran.
- Configuration from the database is difficult. Currently most of the configuration
comes from flat files and is hardcoded in.
- trgDC uses GHITS.
- trgDC does not have any efficiency corrections and is not sensitive to detector conditions.
- trgDC does not have any background overlay.
Associated problems
Hence, we can not make reliable estimates for the effect background has on the trigger
MC simulation. This is especially relevant for:
- How efficient is the DOCAZ?
- How well do we reject background?
- What is the new trigger rate for higher luminosity?
Improve the trgDC efficiency simulation (V.H.)
The trigger simulation takes GHITs as its input. Therefore, the simulation
does not see efficiency corrections done at the DIGI stage. To simulate the
hardware closely we used the GHITs charge as a tool to make the required
efficiency adjustments.
The GHIT charge distribution was found to be too narrow
(charge/path distribution).
The reason is that GEANT does not simulate ionization in thin absorbers or gases but
uses a pure Landau distribution.
To simulate effect due to the read out electronics, a Gaussian smearing
was applied to the GHIT charge. The smearing was studied comparing the GHIT
charge to the DIGI charge distribution for Bhabha events taken from real data
and from MC simulations.
The smearing applied was about 30% with some adjustments to the high charge tail.
Another interesting thing found was gain differences between the different sub
layers of the detector that did not appear in the simulation
(charge vs layer).
The result of this observation was the use of different scaling factors for the
GHIT charge distribution as a function the sub layer.
At the end, in order to know what threshold charge cut to apply to each of the
GHITs we had to study the hit inefficiency for the different sub layers.
A small note:
While studying the TSF efficiency we could find hints that suggest that
low momentum particles should have had their own LUT calibration.
That was the TSF efficiency study in a nut shell. See the following links for the detailed study:
TSF data/MC comparison and efficiency study (V.H.)
Design proposal for incorporating DCH/TrgDC simulation
Valerie's proposal for the design was posted in
meeting-010426.
The full Simulation generates an event using the event generator and then the
different tracks go through a GEANT (perfect detector) that does not have any
electronics noise, digitization, and so on. The output of GEANT is a list of GHITs.
Current configuration:
the trigger and drift chamber simulation gets the GHITs as input from GEANT.
Once the trigger generates the L1Accept the drift chamber uses this time to correct
the drift time that has been calculated. In addition, the drift chamber makes
DIGIs out of GHITs.
The main problems with this scenario are:
- All the efficiency adjustments done by the drift chamber are not seen by the
trigger.
- The trigger does not have any background overlay.
- And since the L3 trigger uses TSF segments as seeds these effects are propagated into L3.
The new configuration
solves these problems: The idea is to split the drift chamber simulation.
The first part is called the
Geant interface:
- DCH would make temporary DIGIs before trgDC++,
temporary in the sense that they don't take into account the L1Accept yet.
- Fill up the correct charge distribution for all DIGIs.
- Overlay background
- Merge DIGIs on the same wire
- Pass the DIGI list to the trigger
The second part
- Incorporate the L1Accept after the trigger to all DCH DIGIs and apply additional time and
efficiency cuts.
In this way we have a realistic simulation that is closer to the hardware, and
we will overlay the background properly and prevent any duplication of work.
L1Sim-MiniWorkshop I
The L1Sim miniworkshop raised intersting points that have to be
addressed during the simulation design meetings:
Conceptual Problems in the Simulation design
- The DCH-trgDC interface needs to change and take DCH DIGIs as input.
- The new simulation should be able to overlay background as many times as needed.
- The simulation should be able to use different TSF LUTs.
- TSF segment fine phi must be transmitted for all layers and not only for the axial superlayers
(as currently done).
- The option to use PDTM or DOCAzM
- Write out the input and output information for different modules
to the format suitable for the test stand, and for the purpose of hardwar-MC comparisons.
- The configuration must be consistent for all parts of the L1 trigger simulation (DCT, EMT, GLT).
- A preliminary flow chart for the new L1/DCT simulation can be found here:
Top level design
of L1/DCT
L1Sim-MiniWorkshop II
On 13 March 2002, a second L1Sim MiniWorkshop was held at SLAC.
There were two informal presentations and some useful discussions
prelimanry design proposal summarized below.
- Eric Torrence presented a high-level view of the upgrade, and
presented a proposal for the flow of information between the
various simulation modules.
Further details are here, while the high level diagram can
be found here.
- Valerie Halyo presented a detailed design proposal of the new
trigger simulation,
raising the different problems and solving the subtle
issues of the design.
Some of the questions are mentioned below.
Details of the work done in collaboration with Gerald G. and Anders W.B.
are given in the
Trigger Upgrade Design Proposal .
Minutes from the disscusions is given here
The following questions were raised at this workshop:
- What configuration information is needed for each module in the simulation?
- What is the best mechanism for controlling this configuration?
- How should the code properly switch between simulating the old and new hardware in the
production simulation executable?
- What sort of extra functionality and output is needed for the teststand and
other expert use?
- How much of this extra functionality should be included in the production code?
- Exactly what information is needed at each step of the simulation?
It was agreed that the answers to these questions would be decided at regular
meetings devoted to the L1Sim upgrade.
An initial list of the packages which need modification can be found here.
Background overlay
We are using G4 to generate our MC.
Our plan is to first:
- Generate an event that decays to neutrinos using G4.
A file named nosignal.dec was added to ProdDecayFiles/
- Skim flat Bhabha events from data
Once the neutrino MC is generated the skimmed Bhabha events will be treated
as background.
To find out the time offset for the background overlay
we are going to compare the Tdc time or T0 time
between the MC generated as explained above and the Bhabha events treated as signal.
After the global tuning every sub detector system will have to adjust
their time to the background overlay correspondingly.
EMC
The EMC simulation is using Wave Forms (WF) for background overlay.
In order to do the background overlay the plan is to record complete WFs
for cyclic triggers if data flow band width will permit.
The problem is that the full WF calibration will be a solution
for RUN 1 but at about March this year the features extraction changed.
This change rendered the background overlay problem more complicated and challenging.
EMT
Progress on
EMT Simulation
DCH
M. Van Hoek discussions on the changes to DchSim
can be found in
DchSim upgrade I and
DchSim upgrade II.
Current status:
M. Van Hoek succeed to split the DCHsim package and made a list of Digis before trgDC
and A. Borgland changed trgDC to feed trgDC with the wire layer and drift time from the
temp Digi information instead of GHIT. We can therefore now run the trigger simulation using Digis.
DCT
Achieved goals are:
- Generating the right Digi charge distribution
to help make the efficiency adjustments
- Finally DATA/MC comparison between DCH/TSF hit efficiency is reasonable
a trg summary report is given in
trigger TSF efficiency report
- DCH raw chrage simulation report
- The conclusion are :
- Temp DIGI are made by DchSim which mix bkg and merge DIGIs and Xtalk
-
- The raw charge should be proparely simulated:
- Use a scaling factor to simulate the right voltage
and global gas gain (condition DB)
- Include all $7104$ wire by wire gain adjustment based
on the initial calibration done using Bhabhas
- Include an additional smearing to simulate noise
- The last conclusion from trg report is that the charge based efficiency
adjustment should be done by DchSim and not by trgDC anymore
since the results were equilavent therefore, the DCH which keep track
on all drift chamber conditions and configurations naturally
adjust the raw charge proparely and hence the TSF/DCH hit efficiency
- Next goal Merging the background overlay
Once these steps work we can use temporarily the modified old version
of trgDC until the new version is ready.
Hope to have news soon ....
HOWTO
Useful stuff, instructions etc can be found here.
| 
