RD27 note 20, 14 December 1993
FORWARD CALORIMETRY AND THE LEVEL-1 TRIGGER
N. Ellis
PPE Division, CERN
Introduction
The designs for the level-1 calorimeter trigger that have been performed
so far in RD27 have concentrated on the electron/photon and jet
triggers, although the need for a missing transverse energy trigger has
always been recognised. The forward calorimeters, typically covering
the pseudorapidity range |eta| = 3.0 - 4.5,
are essential for the missing
energy measurement {1}. We must therefore make provision for sending
data from these devices to the level-1 calorimeter trigger processor.
It is worth noting that in ATLAS it is not planned to recompute the
missing transverse energy vector at level-2. However, high-ET objects
that contribute to the missing transverse energy can be validated and
some "corrections" could be made to the missing transverse energy
vector.
Various technologies are under consideration for the forward
calorimeters. In ATLAS, a choice will be made between high-pressure
gas, liquid argon and liquid scintillator calorimeters. The number of
cells in these calorimeters is typically ~2000 with two depth samplings.
The noise per calorimeter cell may be as high as 1-2 GeV energy
equivalent, but it should be kept in mind that sin q < 0.1 so that the
noise expressed in terms of transverse energy should be manageable.
The forward calorimeters may be used to identify forward-going jets in
the physics analysis ("jet tagging"). However, it does not seem
necessary to include this information in the trigger.
Connection of Forward Calorimeters to Level-1 Trigger
As discussed above, the only use of the forward calorimeters in the
level-1 trigger is for the missing transverse energy trigger. There is
therefore no need for the complicated EM cluster logic (and jet logic)
that is used in the central calorimetry. However, the most economical
solution may be to interface the forward calorimetry to the standard
trigger processor inputs, even if much of the trigger logic is not used.
In the bit-parallel design [1], the processor receives zero-suppressed
data, tagged by the bunch-crossing number, over asynchronous serial
links; in the bit-serial design [2], high-speed synchronous serial links
are used.
The data received by the level-1 calorimeter trigger processor is
transverse energy (not total energy). After partial summing, the ET data
are converted using look-up tables to Ex and Ey (components of the
transverse energy vector). Extending the arguments of [2], one should
preserve reasonable granularity in azimuthal angle, phi,
at the input to the
processor. However, one can sum ET over eta
externally without loss of
information for the missing transverse energy calculation.
Two issues that must be considered when summing ET before the input
to the trigger processor are the following:
- Transverse energy rather than total energy has to be summed.
- The electronic and pileup noise must be reasonably small. In the
central calorimeters we normally consider working under
conditions with sigma(ET) < 0.5 GeV before digitisation. After
digitisation a trigger cell threshold can be applied P this has been
shown to be very effective against pileup.
Two possibilities are being considered for the calorimeter readout:
- The FERMI system [3] which digitises the calorimeter signals
before the level-1 pipeline. If this is adopted (for all the
calorimeters), the level-1 trigger data will be derived by digital
summation of the full-granularity data. Note that (from the trigger
point of view at least) the data in FERMI should be transverse
energy. In FERMI, cell thresholds could be applied with the full
granularity, giving optimum performance against noise and pileup.
An added benefit is that calibration constants can be applied
individually for all calorimeter cells (full granularity).
- A system based on analogue pipelines, with an independent trigger
ADC system. Here the trigger signals will be formed by analogue
summation. Following the ADC, a look-up table will be used to
apply calibration constants (output of look-up table is transverse
energy). The same look-up table can be used to apply a cell
threshold with the granularity of the trigger cells.
Interface to the level-1 trigger with FERMI readout for the
calorimeter
If FERMI is used for the calorimeter readout, with cell thresholds
applied at the finest granularity, one could perform digital summation
over the full eta region, giving one input per phi
region (per end plug) to
the trigger processor. For Df = 0.1, this would require 64 processor
inputs, corresponding to one "cluster-finding" module (per end plug)
in the bit-parallel processor design.
The first part of the summation would be done on the FERMI
microsystems, with further summation on the FERMI boards. One
could imagine completing the digital summing tree on forward-
calorimeter trigger interface modules, which would also contain the
standard interface to the trigger processor.
In the case of the bit-parallel
design, this would include zero-suppression, tagging, parallel-to-serial
conversion and electrical-optical translation. This interface would be
connected to the trigger processor via an optical link.
Interface to the level-1 trigger with analogue-pipeline
readout for the calorimeter
If a readout system with analogue pipelines is adopted for the
calorimeter readout, the trigger signals will have to be formed by
analogue summation. Digitisation will then be performed by dedicated
trigger ADCs followed by look-up tables to apply calibration constants
giving transverse energy output and perform trigger cell thresholding.
Here the following considerations apply:
- The inputs to the analogue sum should be (approximately)
proportional to transverse energy.
- The number of calorimeter cells that can be summed before
digitisation may be limited by noise and pileup.
Digitisation could be performed using standard trigger ADC modules,
incorporating the interface to the trigger processor, in which case a
relatively large number of processor inputs may be required.
Alternatively, one could follow the ADC / look-up table system with a
digital summing tree as in the FERMI based system discussed above.
A limitation of the analogue summing scheme described here is that it
may not be possible to adjust the weights in the analogue summation
(or to monitor the calibration of individual calorimeter channels
contributing to the trigger). It would therefore be important to have a
stable calorimeter response.
Conclusions
Forward calorimetry should be included in the level-1 calorimeter
trigger for use in the missing transverse energy trigger. This should be
possible whether FERMI or an analogue pipeline solution is adopted
for the calorimeter readout, although there are some advantages in
using FERMI. In either case, it will be necessary to build forward-
calorimeter interfaces to the level-1 calorimeter trigger processor -
these could make use of standard components (e.g. trigger ADC,
interface to optical data link).
References
[1] RD27 note 8.
[2] RD27 note 6.
[3] FERMI (RD16) status report: CERN/DRDC/93/21.
----
Footnote (1):
If the forward-calorimeter information were ignored, the missing
transverse energy trigger would be dominated by
events with high-pT jets falling outside he
the acceptance of the central calorimeters.