L. Levinson, sTGC Trigger Demonstrator 1
sTGC Trigger Demonstrator
ATLAS Israel Annual Meeting30 December 2012
Lorne Levinson, Julia Narevicius,Alex Roich, Meir Shoa, Vladimir Smakhtin
30 December, 2012
L. Levinson, sTGC Trigger Demonstrator 3
sTGC trigger demonstratorDemonstrate trigger path only from chamber to Sector Logic input
• Not for studying detector properties and ultimate resolution:ToT is non-linear and saturates, probably not used for final ASD
• use cosmic rays
• BNL VMM1 ASD: 32 ASDs for strips, 12 for pads
• Strips: convert ToT to charge by measuring length of ToT with 1nsec sampling: – Emulates future “TDS” = Trigger Data Serializer ASIC– peak-to-time would be the same
flash ADC would deserialize with 5nsec /bit– Serial output at 6.25Gb/s
• Pads: Simple trigger: dual 3-out-of-4 coincidence logic
• Serial transmission of pad trigger data to TDS (but without BCID assignment)
• Centroid finding, centroid choosing, track vector calc
• Latency measurement30 December, 2012
L. Levinson, sTGC Trigger Demonstrator 4
Not demonstrated
• Because ToTs are saturated, no attempt to get precision centroids: • No Router, instead there is direct 6Gb/s link from TDS to centroid
finder via twinax cables• No readout of VMM
30 December, 2012
L. Levinson, sTGC Trigger Demonstrator 6
Pad over-lapping for 3-out-4 coincidence
• Shown above: pad overlap in a pair of layers• Three rows of 7 pads. Two rows of 8 pads• The second pair of layers in the multiplet is the same.• 37 pads per pair of layers
– Total 4x37 = 148 pads
• Trigger looks in 12ns windows (six 2ns samples): if 3 or more pads are non-zero, a trigger is generated
30 December, 2012
L. Levinson, sTGC Trigger Demonstrator 7
sTGC Trigger Demo-1 FPGA board
30 December, 2012
Power module
Trigger Data Serializer FPGA
16+12 ASD connectors
Pad trigger/ centroidFPGA
10G Quad fibre interface
10G twinax*4 connector
Ethernet readout
L. Levinson, sTGC Trigger Demonstrator 8
Status• PCB boards being tested
– FPGAs correctly receives non-standard LVDS from VMM ASDs
• Firmware done, not tested, but much the same as for Demo-0
• Reassembling pad signal extraction and ASD adapter boards … there are both “left” and “right” VMM boards– Board to bring pad signals to edge of chamber for 2 or 3 rows of pads– Boards to adapt VMM ASDs to the strips on edge of chamber– Boards to adapt VMM ASDS to the board that extracts the pad signals
• We have 46 VMMs in Israel, 44 needed for full test
• Configuration system, HW and SW, being built for up to 48 ASDs
• Getting results for Milestone Meeting will be tight, depends on – no disasters with the PCB boards – understanding of 6.25Gb/sec serial links (deserializers)– We have backup parallel 6Gb/s links if problems with serial links
30 December, 2012
Latency sTGC
NSW Trigger, TDAQ Week Oct 2012 10
min(ns)
max(ns) Notes
TOF from interaction point to SW (10 m) 35 35Propagation delay in chamber 25 25ASD 10 10up- to down-stream package (50cm) 0 3delay for pad trigger & deskew 10 20 strips are pipelined until pad trigger arriveslatency for the last sample of the pulse 64 100 depends on shaping time; ToT is accumulated on the flyToTd: ToT to digital and zero suppression 20 30 uses a 10:1 deserializer to aggregate 1ns samplesOn chamber cabling (up to 4m) to Router 20 20 5 nsec/mToTD-to-Router ser/deserializer 5 10Router 5 10 switch Router to Centroid serializer/deserializer 40 70 GBT does not meet this limit. A different link must be used.fiber to Centroid card in USA15 90m 495 495 5.5nsec/mfind valid centroids and centroid averages 32.5 65 8 layers done in parallel, measured to be 13 clocksdifference of centroid averages 2.5 5calc sub-ROI pointed to at Big Wheel (LUT) 5 5 datasheet BRAM access time =2nsecresynch to BC clock driving output serializer 0 6.25 90° phase chosen to best match pipeline lengthcentroid to Sector Logic serializer 5 10 Total 774 919
L. Levinson
The table shows the latency from the interaction point, through the precision strip trigger logic for the Inner Layer to the Sector Logic in USA-15. We take as a model the Xilinx Virtex6. All times are estimates except that for the centroid finder which has been measured.