Title: Proposed full upgrade of the L1CTT/DFE for Run2b
1Proposed full upgrade of the L1CTT/DFE for Run2b
Meenakshi Narain Boston University
Outline Proposal Reviewers comments and
responses Schedule Cost
2Groups
- Simulation and Algorithm development
- Kansas
- Graham Wilson, Carsten Hensel
- Manchester
- (Liang Han), Terry Wyatt
- Notre Dame
- Mike Hildredth
- Hardware
- Boston University
- Meenakshi Narain, Eric Hazen, Ulrich Heintz,
Shouxiang Wu - FNAL
- Marvin Johnson, Stefan Gruenendahl, Jamieson
Olsen
3Physics Motivations
- In Run IIa and RunIIb, the L1 Track Trigger
Provides - CFT tracks for L1Muon Seeds
- High pT isolated track trigger capability
- CFT tracks CPS clusters ? embryonic electrons
- Found tracks for STT
- As instantaneous luminosities hit 1e31, the
rejection for L1CTT drop drastically. - Accomplish high eff with low fake rates using the
full granularity of the CTT.
4L1CTT architecture
- A multistage system
- Analog Front End (AFE)
- Signals from the tracker
- Mixer
- Sort signals in trigger
- sector wedges
- Digital Front End (DFEA)
- Track Trigger logic
- Octant (CTOC)
- Combine track information
- from several DFE boards
- CTTT
- Construct track trigger terms
- Trigger Manager
- Construct 32 AND/OR terms
- used by the L1 Trigger Framework
- in forming the trigger decision
5L1CTT architecture
- The approved upgrade
- Replace Digital Front End
- daughter boards (DFEA)
- Need to fit 3-5 times
- more equations
- Rebuild these cards using
- larger FPGAs
- Use XC2V6000 Chip with 6M
- system gates
6Lessons from Run2a commissioning
- Current approved plan
- upgrade of the DFEA daughter boards.
- Provide a 5x increase in available logic
resources. - With the experiences over the last year while
commissioning the present Run2a L1CTT system a
few very important issues have surfaced - Lack of extensive testability of the input and
output information from the DFEA - Excessive Firmware download times
- Clock and SCL signal distribution (dependent on
upstream info) - A shortage of spare backplanes (there are none)
7The Proposal
- In order to add testability to the system, one
needs - A new motherboard (daughterboard to a less
extent) design. - Improved diagnostics input and output buffers,
and L3 capability. - To alleviate colossal download times we need
- A new DFE crate controller with a
faster connection to
the D0 online computers. - A new DFE backplane design
- Utilize this chance to put the cables to the
back. - No transition boards.
- Also an independent SCL/clock distribution
- This decouples us from upstream downtime issues
- Detailed specifications of the boards are being
developed follow links through the Run2 trigger
web page
8Installation Commissioning
- The proposed scheme is driven by the desire to
enhance testability and reliability of firmware
downloads. - Eases Installation and commissioning
- Plan to run on a partial crate of the new system
on the platform in parallel with the existing
DFEA using LVDS splitters - Key to successful commissioning of the new system
- Advantages
- Can assemble the whole crate outside of the
collision hall - Extensive testing of the entire chain possible
before putting in collision hall - Enhanced testing capability
- Use Run2a Data derived test vectors to verify
- Take out the old crates (2 of them) and replace
them with new ones - Use the existing Run2a tools used for
commissioning - Low level changes will be transparent at user
level
9Rescope Review
- Review held on April 5th, 2004
- Reviewers
- Darin Acosta (UF), John Anderson, Bill Freeman
(Fermilab) - Charge to the Committee
- Total cost of the new CTT project.
- Is the total cost well estimated and does it
include all necessary cost for materials and
labor? - Contingency
- What is the uncertainty of the cost estimate
(i.e. how much contingency should be assigned to
the total cost of the project?) - Schedule of the new CTT project
- Is the schedule reasonable and attainable?
- Milestones and tracking.
- Are there sensible milestones in the schedule
that will allow management to track the progress
of the project? - Technical Feasability
- Is the project feasible as described and does it
optimize installation and commissioning time?
This is not a detailed technical review more of
an overall sanity check.
10Responses to Review Comments
- DFEA chip decision, merging MB/DB
- Already received gift from Xilinx for all
XC2V6000 - Start with merged scheme (shorter prototype
cycle) and split later on if needed - 48V power distribution system
- Q The risk is that the system may induce EMI
effects to other sensitive D0 electronics - A In the last 2 months done various tests no
affects seen on the calorimeter. Add shielding as
a precaution.
11Responses to Review Comments
- Crate Controller
- Question design choice, suggest to use commercial
solutions - A We had explored commercial off the shelf
choices. - These solutions will force use of either a VME or
PCI interface, adding complexity to the DFE as
well. - Radiated EMI from the various busses on a
commercial board presents a risk to cal
electronics - Considered a hybrid solution for the CC using
commercial VME module like L3 single board
computer and a PMC for SCL. - Rejected
- as simple hardware transforms into a large
software project (including a new driver). - Software to implement SCL interrupts is a good
bit of effort. - 16 MHz VME clock would require noise shielding.
- Our design is based on the CMS/OSU design a
commercial solution - Send raw Ethernet frames over a gigabit link
- Reuire PC/Linux software effort (drivers
available) - Use OSUs VirtexII Pro rocket I/O transceiver
(dedicated EE at OSU) - Also add bi-directional parallel port interface
to board.
12Responses to Review Comments
- DSAT DFEM/DFEA tester
- Question investment of resources in this project,
suggest this be merged into DFEM itself. And
include SCL on the tester. - A
- We need a tester somewhere.
- In the two weeks after the review we invested in
a tester schematic. We are holding sending it
for layout. - However, we do prefer a standalone tester as
something is needed to test/ power-up DFEA/DFEM
prior to the setup at FNAL is put together. - We feel this is not a huge resource investment,
while merging it with DFEM/DFEA will put
unnecessary load on the one engineer doing the
DFEA work and may delay the DFEA (which we want
to avoid at any cost).
13Responses to Review Comments
- Test Setup
- Q A lot of value put it in the in-situ tests of
the upgraded L1CTT system during this years
shutdown, need to develop contingency plans to
test outside of the collision hall in case this
is not met - A
- We do have a plan for test-setup outside the
collision hall - For the Fall 2004 shutdown all we really need to
achieve is installation of a new crate with power
supply and backplane, the LVDS cable extensions
and the splitters. - Crate controller and DFEA prototypes can be added
(and replaced) during shorter accesses later. - However, we still know from experience with the
current system that there is no replacement for
testing in situ, especially with regard to
timing.
14COST SCHEDULE
15Schedule
- Aggressive but manageable schedule
- Needed to start the project at yesterdays
timescale - Timing of anticipated Fall Shutdown and CTT
slice test a critical ingredient for ensuring
success
16Milestones
Backplane and Crate Controller
17Milestones
DFEA/DFEM/ FPGAs
Production
18Fall 2004 Slice Test
- Fall 2004 shutdown
- Aug 23rd and lasts 13 weeks.
- During this time plan to run a partial crate of
the new system on the platform in parallel with
the existing DFEA using LVDS splitters. - Plan to assemble the partial test crate prior to
the shutdown. - Extensive testing of the entire chain possible
before putting in collision hall
19Readiness for Slice Test
20Readiness for Slice Test
21Readiness for Slice Test
- Optical DCL/LVDS Splitters
22Readiness for Slice Test
23L1 CTT Cost Change
WBS Description Original Cost Revised Cost MS FNAL Labor
1.2.3 Level 1 Tracking 870,182.47 1,131,251.00 114,228.53 146,840.00
1.2.3.1 Prototype L1 Central Track Trigger Algorithm Coded And Simulated With FPGA Simulation Tools 0.00 0.00 0.00 0.00
1.2.3.2 Develop Target CTT Algorithm 85,272.00 85,272.00 0.00 0.00
1.2.3.3 Target L1 Central Track Trigger Algorithm Firmware Coded And Simulated With FPGA Simulation Tools 0.00 0.00 0.00 0.00
1.2.3.4 Develop Test Procedures 1,550.40 1,550.40 0.00 0.00
1.2.3.5 DFEA Preproduction I 106,347.60 106,347.60 0.00 0.00
1.2.3.6 DFEA Preproduction II 51,620.12 49,433.00 (2,187.12) 0.00
1.2.3.7 DFEA Production 625,392.35 537,370.00 (67,742.35) (20,280.00)
1.2.3.8 DFEA Backplane (BP) N/A 33,617.00 15,137.00 18,480.00
1.2.3.9 DFEA Crate Controller (CC) N/A 96,656.00 7,096.00 89,560.00
1.2.3.10 optical DCL (host) N/A 41,480.00 0.00 41,480.00
1.2.3.13 LVDS Splitter N/A 20,438.00 2,838.00 17,600.00
1.2.3.11 DFE Motherboard (DFEM) N/A 140,449.00 140,449.00 0.00
1.2.3.16 DSAT - DFEM/DFEA Standalone Tester N/A 18,638.00 18,638.00 0.00
All costs are shown in FY02 Dollars 114,228.53 146,840.00
24L1 CTT Cost Change
- Total Cost change to DOE MIE (escalated,
burdened, etc) - Original Revised Change
- Cost Cont. Cost Cont. Cost Cont
- 361k 155k 717k 186k 356 31k
- Total difference in cost new contingency
estimate is 387k - This cost will come from the management reserve.
- Lowers the contingency on the total project cost
from 33 to 30. - Have not touched money earmarked for AFE II
upgrade -