1. Laboratoire de l’Accélérateur Linéaire (IN2P3-CNRS) Orsay, France Olivier Callot 4 July 2002 The L0 Calorimeter Trigger Basic principles Overall organization Front-End cards Validation cards L0 trigger review, 4 July 2002

2. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 2  Identify hot spots A shower has a 'small' size Detect a high energy in a small surface Use a square of 2 x 2 cells area  8 x 8 cm 2 in the central region of ECAL (may loose a few % of the energy)  more than 50 x 50 cm 2 in the outer region of HCAL  Select the particles with the highest E T Due to its high mass, a B particle decays into high P T particles 'High P T ' is a few GeV For the Level 0 decision, we need only the particle with the highest P T.  Maybe also the second highest in HCAL, see later Basic principles

3. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 3 One can then select locally the highest candidate Process further only these candidates  Reduced complexity and cabling  Only ~200 for ECAL and ~50 for HCAL starting from 6000 and 1500 cells.  Validate the candidates Electron, photon,  0 : Electromagnetic nature using the PreShower, charge using the SPD Same granularity, projective  Look only at the cells with the same number in the other detectors Hadron Would like to add the energy lost in ECAL, in front of the candidate  Complex connectivity, expensive Useful only if the ECAL contribution is important  If small, it can be ignored without too much harm Look only at ECAL candidates !  Manageable number of connections

4. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 4  Select the highest validated candidate One wants simple decisions at this early level Using the second highest hadron was shown to improve marginally the efficiency in some cases. The studied implementation allows to produce this second highest. No need for a second highest electron or photon. The number of links, and consequently the cost is by far smaller if one reduces locally the number of candidates…  Processing entirely synchronous No dependence on occupancy and on history Easier to understand and to debug Pipeline processing at all stages.

5. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 5

6. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 6 Hardware implementation  Inputs About 6000 ECAL cells (same number for PreShower and SPD) Variable cell size, but identical structure for SPD/Prs/ECAL HCAL different, see later Front-end electronics located on top of the detector Order of 100 rads / year  Use SEU immune components, I.e. ACTEL anti-fuse PGA 32 channels per card for ECAL/HCAL, 64 for Prs/SPD We want to minimise cabling complexity Integrate the first selection in the front-end card. Quantity to manipulate: E T, converted from the 12 bits ADC. 8 bits are OK, with full scale around 5 GeV. Use a dedicated backplane for as many connections as possible Use LVDS levels, multiplexed signals, as soon as there are several bits See Daniel’s talk for details.

7. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 7

8. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 8

9. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 9

10. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 10

11. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 11 First selection  Build the 2x2 sums Work inside a 32 channels (8x4) front-end card To obtain the 32 2x2 sums, one needs to get the 8 + 1 + 4 neighbours Via the backplane (9) or dedicated point-to-point cables(4)

12. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 12  Select the local maximum in the card Simple comparison of the summed E T. Iterative, 16  8  4  2  1 comparisons in 2+1 cycles Currently implemented in 3 ACTEL PGA's

13. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 13  In the Ecal/Hcal front-end card More details in Christophe’s talk

14. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 14 ECAL Validation For each candidate, one needs to access the SPD+PreShower information, i.e. 2 times 4 bits for the SPD+Prs cells corresponding to the ECAL candidate. The address is sent from the ECAL to the PreShower FE card  One PreShower card handles 64 channels, exactly 2 ECAL cards. The 2x4 bits are extracted synchronously at each BX and sent to the Validation Card See Remi’s talk

15. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 15  Usage of the PreShower/SPD information SPD signs the charged particles. The PreShower detects early showers, i.e. electromagnetic particles Electrons give signal in SPD and PreShower. Photons give a signal in PreShower only. To remove pile-up signal, one look only at SPD cells matching a PreShower signal. One can also kill ‘dirty’ candidates, for example if the four PreShower cells are fired. The decision will be in a Look-Up Table, 8 bits of address. Each output will be 3 identical bits, followed by triple voting for SEU protection.

16. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 16  From ECAL, produce 4 candidates Electrons and photons are just validated FE-candidates ‘local  0 ’ are detected by a high total energy on the card. ‘global  0 ’ are detected by summing the FE-candidates of two consecutive cards. No SPD/PreShower validation foreseen for the  0, but this will be integrated in the card, in case…  This is just a few more output bits of the previous LUT, plus the validation of a register.  Only the highest Et candidate is interesting We select the highest of the 8 on the Validation card Very similar to the selection on the FE-card.  Output 4 candidates Each has 8 bits Et and 8 bits address, plus BX-ID.

17. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 17 HCAL Validation  Purpose: Add the ECAL energy to HCAL candidate Frequently, a hadronic shower starts in ECAL. Ideal case: Add the ECAL cells in front of the HCAL candidate. But this implies a lot of connections, at 40 MHz. But this addition is important only if the ECAL energy is important If important, it has a large chance to be a local maximum in the ECAL card. This is now a manageable problem About 200 ECAL local maximum, to send to 50 HCAL candidates But one could also sent the HCAL candidates to the ECAL ones,  Less connections  Some duplication  But can use the ECAL Validation Card. One ECAL card matches only one HCAL card. One validation card receives at most four HCAL cards One HCAL card goes at most (30 of 50) to two Validation cards

18. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 18

19. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 19  Principle of operation Compare the addresses of the ECAL and HCAL candidates First select which HCAL card to use for this ECAL card  This is a configuration parameter, in a multiplexer. Compare the 5 bits addresses  10 bits LUT, as this varies a lot from Validation card to Validation card  Use 3 identical output bits, and a Triple Voting technique to be SEU immune. There may be several matches An HCAL candidate may be in front of more than one ECAL card ! One should select the highest ECAL contribution matching The ECAL Et is added to the original HCAL Et No other changes  One has no information of which ECAL energy was used. Each (modified) HCAL candidate is sent to the Selection Crate  5 bits address, 8 bits Et + BX-ID  Same format as ECAL candidates…

20. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 20

21. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 21 Overall Validation Card (prelim.)

22. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 22 SPD Multiplicity  Count the total multiplicity in the SPD New idea, under study. Reject dirty events, to cut tails in computing time distribution Should help L1, and reconstruction Biases to be understood.  Re-use of the connections In the PreShower card, count the number of SPD bits fired Answer in the range 0-64 = 7 bits Send this number to the ‘validation’ slot, using the same backplane links as in the ECAL crate The processing is simple: Add the 8 numbers and send the 10 bits to the Selection Crate. This can be added on the (existing) SPD control card which will anyway exist on this slot.

23. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 23 Selection Crate  Receive several similar inputs 28 electrons, 28 photons, 28 local  0, 28 global  0. Build the final 14 bits address, select the highest, send it to the L0 Decision Unit  Minor difference for the SPD multiplicity Should add the inputs on 13 bits, instead of selecting the highest  HCAL processing more delicate Removal of pseudo-clones Same HCAL sent to two ECAL Validation  Keep only the one with the highest energy. Address is the same. We want also the second highest, and the total Et Details in Umberto’s presentation.

24. Olivier Callot 4 July 2002L0 review - L0 Calorimerter overview 24 Summary  Relatively complex object 350 FE-cards, 28 (+16) Validation cards, Selection crate. Many links, but a lot on a dedicated backplane About 500 cables inside and between the 12 racks on the Calorimeter Platform. Fully synchronous.  More technical information PGA on the FE and Validation CardChristophe Beigbeder Backplane and linksDaniel Charlet PreShower and SPDRemi Cornat Optical links and Selection CrateUmberto Marconi