REQUIREMENTS FOR DETECTORS SUPPLYING TRIGGER INPUTS
- Jusko and O. Villalobos Baillie
University of Birmingham
November 2nd 2005
1.Requirements
The Central Trigger Processor has several requirements for detectors supplying trigger inputs. These have been presented, discussed and approved on previous occasions. The purpose of this document is to collect all the requirements together, so as to ensure that detectors comply with them, in order to avoid problems during commissioning.
•TOGGLING SIGNAL.The requirement for trigger detectors to be able to generate a special continuous pattern on request in order for the CTP to perform synchronization of inputs has been set out in several documents. The description given in the CTP Design Report is repeated here (Appendix 1) for completeness.Note that the hardware described must be controllable by software. (See next point.)
•SOFTWARE.In order to manage the synchronization operation easily, it is necessary for it to be under automatic control. A description is given of how to do this. The basic requirement is to be able to switch between normal mode and test mode. The required commands are described in Appendix 2. In addition, the proposal has consequences for the ALICE ECS system, which must not issue incompatible commands (e.g. start run) during synchronization. These are discussed in Appendix 3.
•TRIGGER INPUT TIME WINDOW. The requirement that trigger detectors should ensure that their inputs arrive at the CTP within 16 BCs (400 ns) of the decision time for their respective trigger level (approved at the October 2004 TB) is re-iterated. Specifically, this means that, with the current design latencies for the three trigger levels, the inputs should arrive in the following time-windows:
400-800 ns (L0);
5.7 – 6.1 µs (L1);
87.2 – 87.6 µs (L2).
Fine tuning is done by the CTP provided the inputs are delivered to the CTP trigger inputs inside these windows. However, it is still not excluded that these latencies could change slightly, if the trigger input latencies for crucial detectors turn out to be longer than planned. For this reason, it is recommended that detectors keep the capability to change the latency of their inputs in discrete steps, so as always to be able to stay within the window. Shifts of more than 100 ns (L0) or 200 ns (L1, L2) are not expected.
APPENDIX 1. Automatic Synchronization of Trigger inputs.
(From: P. Jovanovic, CTP Preliminary Design Report – June 2004.)
3.6.3Synchronization of trigger inputs
There are 60 trigger inputs to the CTP (24 L0, 24L1, 12L2), generated at nearly as many locations and all connected via individual cables. It is the system requirement that their synchronization with the CTP’s BC clock shall be automated and their phase monitored.
Also, in order to minimize the overall L0 trigger latency, the CTP’s BC clock shall “track” the last-arriving among the L0 trigger inputs.
In order to meet the above requirements:
- there shall be a programmable delay line on the BUSY board (Data Delay Devices model PDU15F-1A4, 32 delay steps, 1ns per step);
- there shall be an ADC on the L0, the L1 and the L2 boards (Linear Technology model LTC1096LCS8, 8-bit serial ADC), with full access to all the corresponding trigger inputs.
The same hardware has been successfully used on the LTU board for a nearly identical application. The measurement of the signal phase in respect to the BC clock is explained in [20], section 3.14.12. The same method shall be used on the CTP boards; during the phase measurement, the trigger sub-detectors are required to generate a pattern of alternating ones and zeros (Figure 3.6.3).
APPENDIX 2. DIM server control of synchronization mode.
Control of the synchronization mode should be performed via a DIM server, to be provided by each detector. The DIM server which is used for DCS operations would, for example, be suitable. It should respond to the following command services:
START_TOGGLEstarts toggling signal N (N specifies the input form the detector in the case where a detector provides more than one input.
STOP_TOGGLEstops toggling signal N.
These two commands would be issued by the CTP software during the synchronization operation. An example of DIM server code, written in C, can be found in
APPENDIX 3.ECS communication
Co-ordination between the CTP and ECS is achieved as follows. The CTP_proxy has two states:
RUNNINGThe CTP performs its normal tasks
DORMANTTo be used when the CTP is synchronizing. In this state, the CTP_proxy is able to respond to the I_SYNCHRONISE command by changing its state to the I_SYNCHRONISING state. When the CTP_proxy is in this state it is allowed to issue commands so as to manage the synchronization procedure with the input detectors.
The synchronization procedure is not itself available directly from ECS, as it is executed only rarely.