Produzione del chip standard cell AMchip03 Per SVT a CDF • Breve descrizione dei tests del prototipo e risultati 1. il chip funziona (up to now) 2. ma yield e’ basso 3. 37% perfetti, ~55-60% perfetti+ “quasi” perfetti • Descrizione dell’ upgrade AM++ e sua flessibilita’ • Descrizione diverse possibilita’ di produzione e loro costi Servono almeno 70 kE aggiuntivi AMchip03 status . • 126 dies produced by IMEC, 116 of which came packaged. Arrived on September 30th The SCAM chip • Simulation test vectors translated for the Test stand. • Test of all the chips in the Test stand Output to Logic Analyzer Inputs from Pattern Generator AMChip socket Test Vectors Three types of tests: • Random Tests: 1) Generate Random Patterns 2) Generate Random Configurations 3) Send random Hits + good hits Roads to next AMchip • Memory Tests • Feature Tests 1) 2) 3) 4) FSM Jtag Extest I/O Opcode Jtag Pattern Flux Pattern Bank Roads From AM SCAMChip internal Structure AMchip03 Test: Results • We passed all the 116 chips through an intensive test at 100 nS. • 8 AM chips tested @30nS to be mounted on a first LAMB++ Yield (flawless) ≈ 37% Yield (flawless + defective Patterns < 2) ≈ 53.5% Yield (flawless + defective Patterns < 5) ≈ 59.5% Yield Expected 68%, based on IMEC’s predictions. Yield puo’ fluttuare moltoproduzione calcolata su 35% BLOCCO 0 Pattern Layer Chip 0x0 0x100 0x800 1A3 2 23 BLOCCO 1 Pattern Layer 819 1 Chip 34 0x900 0x1000 0x1100 BLOCCO 2 Pattern Layer 106B 4 1000 0 1190 2 BLOCCO 3 Pattern Layer Chip 8 76 0x1800 2 0x1900 0x200 240 1 78 0xA00 AFC ? 18 0x1200 122A 12E0 3 2 25 98 0x1A00 0x300 378 4 109 0xB00 B0C 0 1 0x1300 133A 1367 0 2 9 33 0x1B00 0x400 4CA 1 17 0xC00 C39 ? 87 0x1400 1421 ? 111 0x1C00 0x4FF 4 0xCFF 0x14FF 4 Chip 19C6 1924 5 2 103 6 196C 5-4-1-0 62 0x1CFF 8 3 1 solo pattern wrong (preliminary tests) 1 < Pattern wrong < 5 Pattern wrong>4 o other errors = 19/116 = 7/116 = 46/116 16.4 % 6.0% 39.7% Good = 43/116 37.1% Test Stand – Vme Crate A complete system with an AMS/Pulsar, a merger and an AM++ with FPGAs has been tested with success @33MHz. Currently Testing with a Lamb++ equipped with 8 AMchips Next step: adding a Lamb++ with 16 “defective”AMchips. At the moment we can load 5000 patterns in AMchips and send Random Test Hits + good Hits. # Lambs (16 Amchips)/ wedge SVTupg/ tot SVTnow chips 2 Lamb: 160 kpatt/wedge 8 Lamb: 640 kpatt/wedge Gain of the 2004 SVT upgrade SVT Phyiscs Trigger rates: examples @3x1032 5 20 Current SVT 13 KHz keuro 384 + spare ~500 1536 + spare ~2000 53 2003 93 2004 Upgraded SVT 23 KHz Maximum L1 rate for 5% L2 dead time @3x1032 Trigger Z b-b Hadronic B decay L1 Trigger rate 26 KHz 177 KHz Installazione 2005 Eventuale completamento AM++ (9U VME) LAMB GLUE MAX 2005: 640 Kpatt/wedge AM VME INTERFACE INDI LAMB CONNECTORs PIPELINE REGISTERs Input Control Clock Distrib. HIT/ROAD CONNECTOR TOP GLUE To AMS Year chip boards devel. Total 2003? 120 kE 10 kE (test b.) 5 kE 135 kE 2004 - 10 kE (protot.) 30 kE 40 kE 2005 53 kE +40 kE 100 kE (produc.) +70 kE (produzione) +40 kE tests di qualifica +25 kE package sottile 153 k +40 kE Upgrade 2004: 600 chips Upgrade 2005: 2000 chips buoni Ipotesi 35/50% yield – fondi 95 + 60 keuros = 155 ke Due strategie di produzione: MPW e Pilot Run MPW: buono per piccole produzioni usando maschere del prototipo Si paga solo il silicio ed i 5/6 del wafer si buttano! Con 128 wafers 1700/2430 chips - costo=240 keuro Pilot Run: nuove maschere 210 k$ 1 wf 250 chips contro i 38 MPW PILOT RUN #wafers 12 25 #good chips costo keuro 1050/1500 215 2188/3125 225 Mancanti keuro 60 70 costo altrettanti chip 26.4 k$ 55 k$ Conclusioni •Il chip funziona (up to now) •Yield ~ 37% •Forse si possono usare anche i quasi perfetti: 55-59% •Yield puo’ fluttuare moltoproduzione calcolata su 35% •Pilot run: migliore garanzia per avere 2000 chips •Si richiedono 70 keuro aggiuntivi per il pilot run Backup slides Upgrade 2004: 600 chips Upgrade 2005: 2000 chips buoni Ipotesi 35/50% yield – fondi 95 + 60 keuros = 155 ke MPW Wafers #good chips costo keuro Mancanti keuro costo /chip 50 75 100 128 665/950 998/1425 1330/1900 1700 130 185 190 240 0 30 35 85 184/128.5 E 177/124 E 138/96 E 138 E PILOT RUN 12 25 #good chips costo keuro 1050/1500 215 2188/3125 225 Mancanti keuro 60 70 costo /chip 198/139–32/22 E 101/71 –21/15 E Tsukuba Chicago Tempi di processamento: come agisce l’upgrade ? SVT exec time ~ proporzionale # candidati da fittare raw data from SVX front end NUOVA AM piu’ grande Hit Finders Sequencer Associative Memory COT tracks fromXTRP roads x 12 phi sectors Road Warrior 12 fibers hits Track Fitter Merger hits to Level 2 Hit Buffer Ricetta per velocizzare il tempo di esecuzione di SVT: 1. pattern piu’ sottili (AM grande) meno fits. 2. Road Warrior per rimuovere i ghosts TEMPI DI REALIZZAZIONE • Nuova AM-board: inizio estate 2004 (Pisa) durante estate 2004: test con FPGA (Pisa) • Progetto prototipo AM-chip: luglio 2004 (Ferrara-Pisa) consegna chip ~2 mesi – disponibile ad ottobre. • Nuova LAMB: montare nuovo AM-chip a ottobre 2004 (Pisa) • test del chip + scheda: ottobre – dicembre 2004 (Pisa-Ferrara) • produzione: inizio 2005 (Pisa-Ferrara) • installazione: estate 2005 (Pisa-Ferrara) • Altri DAQ/Trigger upgrade: previsti nel 2006 Road Warrior: . . . (~60 k$ Fermilab) messa in opera entro fine 2003 F. Spinella (in funzione) RW+TF: F>50ms = 21% Pulsar (Pulser And Recorder) Design Custom Mezzanine AUX card Mezzanine slots Pulsar Works up to 100MHz Top view Bottom view I/O Mezzanine cards for: S-LINK (CERN/LHC) Hotlink TAXI to be specified Three ALTERA APEX 20K400 FPGAs Self-testable Modular, lego-style open design Replacing > 10 CDF board types all CDF, many ATLAS connectors/standards AMchip L. Zanello Trigger/DAQ Upgrades for Run IIb • Need to Maintain or increase bandwidth – Luminosity Rate (=s*L) – Luminosity inter/x-ing Complexity (and fakes) Event Size, Exec. time – All sphysics to tape • L1 Bandwidth (output to L2) – XFT Purity strigger(L1, L2) – SVT, L2 L2 Exec. t L1 Bandwidth • L2 Bandwidth (output to L3) – COT TDC Readout rate – Level 3 Processing L3 Exec. t – Event Builder Readout rate • L3 Bandwidth (output to tape) – CSL Tape rate (not under IIb project)