Title: Front End Driver (FED) Crates and Racks Tracker PSU racks
1Front End Driver (FED) Crates and RacksTracker
PSU racks
- Matthew Pearson
- RAL
- m.pearson_at_rl.ac.uk
Electronics Week November 2002
2Considerations
- How many FEDs per crate?
- How shall we arrange the FEDs?
- Dont mix FEDs from different partitions
- In crates?
- In racks?
- Separate into /-?
- Do we need to reserve space in case we need to
use more - FEDs than we expect? (if data rate per FRL is
gt200Mbytes/s) - Can we keep the S-LINK64 short enough?
- (Has to be less than 15m)
3How many FEDs per partition?
- 440 FEDs in total
- 4 Partitions TEC, TEC-, TIBTID, TOB
- Partition Layer
FEDs - TEC (1-3,4-6,7-8,9)
(40,32,16,8) 96 - TEC- (1-3,4-6,7-8,9)
(40,32,16,8) 96 - TIBTID 2(1,2,3,4)2(1,2,3) 2(12,16,6,8)
2(5,5,5) 114 - TOB 2(1,2,3,4,5,6)
2(11,12,7,8,14,15) 134 - Whats the best way to arrange these?
- Note Data rate per FED is highest for TIB1 and
TEC7,8,9
4How many FEDs per crate?
- VME64 crates have 21 9U slots (see reference 1)
- 17-19 FEDs per crate ---- Depending on
other -
modules used.
- Other modules
- Spare slots (eg. Tester board)
- 1 or 2 less slots in each crate
- FMM (Fast Merging Module)
- Or something similar, which may
- be a transition card on the back-plane.
- 1 less slot in each crate (?)
- Crate or rack controllers?
- 1 less slot in each crate
5Crate/Rack Controllers
- Remote PC based crate controllers are favoured.
- Need VME interface module in each crate
- Interface module has to be 6U
A 6U module needs 6U-9U adapter. (only takes up
one slot)
VME Backplane
6Number of crates
Assume a given crate contains FEDs from only 1
partition. If 17 or 18 FEDs per crate
Partition Crates TEC
6 TEC- 6 TIBTID
7 TOB 8
For 19 FEDs per crate
Partition Crates TEC
6 TEC- 6 TIBTID
6 TOB 8
7How many racks?
- Dont mix partitions in a rack.
- Partition Racks
- TEC 2
- TEC- 2
- TIBTID 3
- TOB 3
- TOTAL 10 racks
- true for both 17 or 18
- FEDs per crate.
- for 19 FEDs per crate we
- save one rack.
- also, dedicated DAQ racks for each partition (64
FRLs per DAQ rack).
3 crates per rack
8Suggested FED rack layout
1.0m 1.5m
A B C
- Need PC racks for
- FEDs and DAQ
- FED racks are grouped
- to assist in cabling
- (Francois)
- Each DAQ is close to
- FEDs from same
- partition
-
1 DAQ DAQ
2
TIBTID TOB 3
TIBTID TOB 4
TIBTID TOB 5
DAQ DAQ 6
FED PCs DAQ 7
DAQ DAQ 8
TEC- TEC 9
TEC- TEC 10
TEC- TEC 11
12 13 . .
S-LINK length minimised (lt10m)
9TIBTID Crates
Assuming 17 FEDs per crate (The number of FEDs in
each crate is shown in brackets)
FRLs
FRLs
TIB2 (16)
TID2 (5)
TIB1 (12)
TID2- (5)
TIB2- (16)
TIB3 (12)
TIB1- (12)
B5
B1
B2
B4
B3
10TOB Crates
FRLs
FRLs
TOB2 (12)
C2
C3
C4
C5
C1
11TEC/- Crates
FRLs
C7/B7
C9/B9
C10/B10
C8/B8
12S-LINK64 cable length
- The S-LINK64 standard will be used for the
FED-FRL connection. - If using S-LINK PCI source card on VME transition
card, - the link runs at 66MHz.
- This has been shown to be OK for cable
length of 15m. -
- Therefore
- the S-LINK cable lengths between FEDs and
- FRLs of lt10m are OK.
13Some other S-LINK points
- We would like to run FED at 40 or 80MHz (320 or
640Mbytes/s). - For convenience
- To run as fast as possible
- (to cope with unexpected high data rates)
- If FED runs at 80MHz, we will get back-pressure
signals from the S-LINK source card. - An alternative is to drive the S-LINK directly
from the FED FPGA. - We could then run at any speed up to 100MHz (the
S-LINK standard limit) but would prefer to run
at 80MHz - This would place tighter restrictions on the
S-LINK cable length. How much tighter?????
14Tracker PSU racks
Current base-line plan is to place TK PSUs in
the Experimental Hall (UXC55).
- 20 racks for SST
- 2 racks for Pixel
- 4 racks for 48V DC PSUs
(estimates from Horst)
Allocate space
- 29 racks on Upper Balcony in central position
- 24 racks on Middle Balcony in central position
Still things to consider (accessibility,
maintenance, cost, radiation and magnetic field
issues ...)
15Summary
- A scheme has been presented detailing the FED
- layout within the crates and racks, and the
placement - of the racks in USC55.
- Note
- If the scheme is approved, and once the cables
- are cut by Francois, nothing can be changed.
16References
- Technical specification for subracks for LHC
experiments - http//atlas.web.cern.ch/Atlas/GROUPS/FRONT
END/rackscrates.html - 2. The S-LINK 64 bit extension specification
S-LINK64 - http//his.web.cern.ch/HIS/s-link/spec/
- 3. CMS DAQ Technical Design Report
- 4. Rack layout in USC55
- http//cmsdoc.cern.ch/wsmith/USC55_racks_v3
6.html