1. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 1 Synchronization and alignment requirements

2. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 2 Synchro Problem Signals generated by the same BX interaction and coming from different Physical Channels have different delays To reach the maximum efficiency of the muon detector we must: Measure time arrival of signals with a precision of 1-2 ns Have the possibility to delay signals with the same accuracy (Fine time) Synchronize signals with the same BX Id (Coarse time) We can perform this tasks with DIALOG, SYNC and TTCrx devices

3. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 3 Synchro Tools SYNC chip can: histogram incoming signals with a precision of ~ 1,56 ns (16 bins in the 25 ns window) delay signals of an integer number of BX ID, up to 10 clock cycles TTCrx chip can delay signals of an integer number of BX ID, up to 15 clock cycles DIALOG chip can delay signals at steps of 1.67 ns up to 50 ns

4. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 4 Synchro tools The control of these devices is done by ECS via CANbus protocol Synchronization procedure must get access to DIALOG and SYNC by the ECS infrastructure (CANbus protocol – PVSS software) CANbus is managed by the ELMB board (one for each ODE, four for each SB)

5. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 5 Synchro in practice Muon system is composed of about 120000 physical channels For each of these the fine time histogram is required to align the signal distribution at the centre of the reference clock period (Fine time) It translates into two main actions: put SYNC chips in FT histogram mode, wait for acquiring sufficient statistics and then download histo data to analize assign the appropriate delay to signals writing DIALOG delay registers

6. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 6 Synchro in practice We will shift the signal distribution as follow: BX n TIME DISTRIBUTION BX n TIME DISTRIBUTION

7. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 7 Synchro in practice Next step take is to set the correct BX tagging (Coarse time) It translates into two main actions: put SYNC chips in BX histogram mode, wait for acquiring sufficient statistics and then download histo data to analize assign the appropriate delay to signals writing SYNC delay registers

8. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 8 Synchro in practice We will shift the BX tagging ID as follow: Coarse time procedure takes into account the LHCb bunch orbit BX n (wrong) TIME DISTRIBUTION BX n–3 (right) TIME DISTRIBUTION

9. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 9 Time problems BUT… This procedure would take a huge amount of time (too much!!!) The idea is to consider the same delay for close channels and to parallelize the procedure where possible (different I2C chains)

10. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 10 Synchro procedure Synchronization procedure for Fine time is sketched in the flow chart below:

11. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 11 Synchro procedure Synchronization procedure for Coarse time is sketched in the flow chart below:

12. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 12 Time evaluation Here a time evaluation to synchronize all the muon system Thanks to Caterina Deplano

13. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 13 Syncro & ECS All the registers involved in the synchronization tasks are only accessible via ECS (CANbus – I2C) These procedures will be developed inside the ECS - PVSS environment Crucial requirement is to have implemented all the CANbus features for SB and ODE (inside the PVSS framework…)

14. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 14 DIALOG & SYNC reg queries The synchronization procedure have to get access to DIALOG chip only to write delays The main question is: how much the distributions will move in time? (to be investigated…) SYNC access will be more frequent to monitor FT distributions We suppose Coarse time delays will not change in time

15. Enrico Fois - LHCb Muon ECS Meeting - Roma 06-04-2006 15 Histo download time A single histogram download from SYNC to PVSS takes about 350 ms… (I2C clock?...Is it a negligible detail?) …up to 24 SYNC per ODE  less than 9 seconds! Can we parallelize downloads? Data transfer from SYNCs to ODE ELMB is implemented via SDO, thus the client have to wait server response… parallelization is possible for each single CAN branch but not for nodes Question: how many nodes for each branch???