Overview of the ATLAS Fast Tracker (FTK) (daughter of the very successful CDF SVT) July 24, 2008 M. Shochet 1

Transcription

1 Overview of the ATLAS Fast Tracker (FTK) (daughter of the very successful CDF SVT) July 24, 2008 M. Shochet 1

2 What is it for? At the LHC design accelerator intensity: New phenomena: 0.05 Hz Total interaction rate: 1 GHz (40 MHz beam crossings) Many possible new phenomena produce heavy b quarks which can only be distinguished from the bulk of the background by reconstructing the individual tracks. The problem! beam pipe few mm We are proposing to significantly enhance the ability of ATLAS to rapidly identify b quarks in the trigger. Currently done in commodity PC s. This is slow and becomes slower as the accelerator intensity and thus track density increase. July 24, 2008 M. Shochet 2

3 ATLAS July 24, 2008 M. Shochet 3

4 Inner Tracking Detectors 3 pixel layers (space point) 8 strip layers (1 coordinate) 11 layers, 14 coordinates Pixels barrel SCT barrel Pixels disks July 24, 2008 M. Shochet 4

5 Getting data into FTK SCT Pixels R O D s on L1 accept FTK dual-output HOLA designed by Tang R O B s R O B s silicon hits silicon tracks Level 2 ask for ROI s July 24, 2008 M. Shochet 5

6 Pixels: 120 Strips: 92 Number of input fibers 12 Number of crates July 24, 2008 M. Shochet 6

7 How does it work? First do pattern recognition, then fit possible candidates. coarse resolution hits (superbins) full resolution hits Prestore patterns (roads) in large content-addressable memory. The Pattern Bank (4-layer) July 24, 2008 M. Shochet 7

8 Massively parallel pattern recognition AM = BINGO PLAYERS PATTERN 4 PATTERN 5 PATTERN 1 PATTERN 2 PATTERN 3 PATTERN N HIT # 1447 track fitter 1 superbin per silicon layer Majority logic allows up to 1 missed layer. July 24, 2008 M. Shochet 8

9 Pixels & SCT Functional layout RODs 50~100 KHz event rate S-links Data Formatter (DF) cluster finding split by layer overlap regions HITS (LVDS links) EVENT # N DATA ORGANIZER Track Fitter PIPELINED DO-board AM ROADS + HITS EVENT # 1 SUPER BINS AM-board ROADS Raw data ROBINs Track Track data data ROBIN ~Offline quality Track parameters July 24, 2008 M. Shochet 9

10 Possible layout for a core crate (after DFs) Track Fitter DO5 DO4 DO3 DO2 DO1 DO0 AM-B1 AM-B0 AM-B2 AM-B3 AM-B4 AM-B5 AM-B6 AM-B7 AM-B8 AM-B9 AM-B10 AM-B11 AM-B12 July 24, 2008 M. Shochet 10

11 Data Formatter Receives raw hits from the detector (RODs) Finds hit clusters pixels silicon strips Store cluster centroids Separates clusters by silicon layer & sends to Data Organizers on 6 LVDS data busses (22 bits each) July 24, 2008 M. Shochet 11

12 Data Organizer Receives hits from Data Formatters. Stores hits at full resolution in a way that is rapidly accessible by pattern number. Sends hits at coarser resolution (superbin) to pattern recognition unit (Associative Memory). Receives patterns from AM, retrieves full resolution hits, and sends road number and hits to the Track Fitter. July 24, 2008 M. Shochet 12

13 Track Fitter Receives road # and associated hits from Data Organizers. Computes all hit combinations Calculate the track parameters curvature, azimuthal angle, polar angle, z 0, impact parameter and the goodness of fit (χ 2 ) using a linear correction to the mean for that sector (excellent precision over a sector). sector: a physical silicon module in each layer pixels: 1" x 2.5 " strips: 2.5" x 5 " July 24, 2008 M. Shochet 13

14 14 measurements, 5 parameters 9 constraints ( χ 2 ) 14 p = a x + b i ij j i j= 1 P i : 5 track parameters & 9 constraints (χ 2 is sum of squares) x j : the hit coordinate in layer j a ij, b i : the stored constants for each sector, calculated in advance from a large sample of training tracks (simulation or data) Cut on goodness of fit; among the combinations in a road, select the track with the best χ 2. Output good tracks to ROBIN. July 24, 2008 M. Shochet 14

15 Readout Buffer (ROBIN) Stores tracks for access by the level-2 trigger PCs. July 24, 2008 M. Shochet 15

16 Track Fitter details GigaFitter a simplified version built for the CDF SVT 2D reconstruction (transverse to the beamline) 6 detector layers 3 track parameters (curvature, azimuthal angle, impact parameter) 3 constraints χ 2 July 24, 2008 M. Shochet 16

17 GigaFitter scheme INPUT FiFo Lay0-Ram Lay1- Ram Lay2- Ram Lay3-Ram... Lay10-Ram Comb - FiFo DSP: Fit Tracks Choose best χ 2 track Constants RAM DO roads & hits input FIFO RAMs according to detector layer Combinations (one hit/layer) calculated sequentially & stored in Comb-FIFO Each combination & the constants are sent to DSP for fitting & selection July 24, 2008 M. Shochet 17

18 C1 Hit C2 18 C3 18 DSP algorithm for SVT 39 ACC ACC ACC 39 FIFO C4 onstants ACC 39 DSP48E RST EV CTRL READY RST Comb-FIFO data serialized: 1 hit and its constants sent to parallel DSP slices each computing a track parameter or constraint. An additional DSP computes total χ 2 from individual constraints Total of 7 DSP slices in parallel each working at 200 MHz July 24, 2008 M. Shochet 18

19 DSP Slice One DSP slice computes a track parameter in 14 clock cycles, plus 4 for the first one. July 24, 2008 M. Shochet 19

20 Xilinx XC5VSX95T V5 slices (4 LUT + 2 FF) 1% used for each fitter 1520 kbit of distributed RAM 1% used for each fitter 640 DSPs 2.5% used for each fitter in the FTK version 5 parameters, 9 constraints, 1 to calculate overall χ 2 40 fitters/chip (remember: many combinations per road) 1 Mbyte of block RAM plenty for the SVT prototype not even close for the FTK!! July 24, 2008 M. Shochet 20

21 Missing hit problem To obtain high reconstruction efficiency, we must allow one physical detector layer to miss a hit. The constants in the parameter and constraint equations are different when there is a missing hit. Store 12 sets of constants (all hits, a miss in one of 11 layers). A lot of memory that has to be accessed very quickly. Can one look ahead for the constant set that will be needed next? Alternative is to estimate the hit location in the missing layer. How long does it take? July 24, 2008 M. Shochet 21

22 Size of the constant memory 210 words/constant set ( ) If we need 2-byte precision 420 bytes/constant set One constant set/sector. Currently estimate 100k sectors in an FTK crate. 42 Mbytes of fast memory 0.5 Gbyte if solve missing hit problem with more memory We have heard that with the latest FPGAs, there is very fast access to external computer-like memory. Is that true? July 24, 2008 M. Shochet 22

23 How many fitters are needed? The number of cycles from the time the data from a road is in the input FIFO until the track parameters are in the output FIFO is approximately: NfitCycl = Ncomb Nhits (all 14 calculations done in parallel) A road packet takes Nhits + 1 Data Organizer clock cycles to be sent to the Track Fitter. Thus if we are to have 0 deadtime from track fitting, we need the number of parallel fitters (there are 40 in a chip), Nfitters, satisfying: Nhits + 1 NfitCycl 1 = DO freq fitter freq Nfitters This translates into a maximum average Ncomb of Nfitters ClockRatio (Nhits + 1)/Nhits where ClockRatio is the ratio of the fitter to DO clock speeds. We will have to satisfy this: road width, # of FPGAs. July 24, 2008 M. Shochet 23

24 Hit Warrior function One can easily get many tracks (ghosts) from a single real particle due to presence of extra random silicon hits. Compare a new track with those already found. If new one has 8 or more hits in common with a stored track, keep only the best χ 2 track. If space permits, add this function to the Track Fitter. July 24, 2008 M. Shochet 24

25 Spy Buffer Data flows very quickly through this system. By the time any PC using its output detects a problem, the data would already be long gone from the FTK. That makes diagnosing a problem that is occurring internally in the FTK extremely difficult. We found it very useful in the SVT to have deep buffers at the input and output of every board in the system. Then, when a problem is detected, these spy buffers can be frozen in the entire system and read out. The buffer is deep enough so the event with the error is still inside it. This allows diagnosing and fixing subtle problems. July 24, 2008 M. Shochet 25