Transcript cost-estimate-july-2..

Cost Estimate
(and configuration)
Drew Baden
Univ. Of Maryland
July 2001
1
HB, HO, and HE Geometry
• Layer 0 in HB and HE eliminated in readout
• HE h=28 and 29 combined into 1 TPG
• HE h>21 has 10° segments, therefore 2 TPG/channel
2
HF Geometry
• For this estimate, use h/f scheme:
 36 f-towers, 12 h-towers, 2z, 2F/B
 should be ok to 10% for cost estimate but will clean it
up ASAP
3
Constraints
• Magic numbers:
 3 channels per fiber
- 1 channel = 1 TPG for HB, HE<21
- 1 channel = 2 TPG for HE>21
- HF is complicated (more later)
TPG:
- 8 TPG per SLB
- 3 (or fewer) SLB per HTR
DCC input bandwidth
- 140 bytes/HTR
- 100kHz L1A
- <200MB/s total
 limits crates to 15 or fewer HTR
4
Proposal
• HTR
 24 channels, or 8 fibers, per HTR
- limits HTR to 3 or fewer SLB
 SLB are the “daSilva” TPG transmitter boards to Level 1
 boards are large - can maybe accommodate 4 but would be at expense
of front-panel space
 8 TPG pair per SLB, therefore 24 TPG/HTR. Going to 4 would make 32
TPG/HTR, but 32 channels, 3/fiber…doesn’t fit!
 We are SURE that 24 channel HTR cards will assume a minimal of risk
for going to the HTR prototype
• VME
 15 or fewer HTR per VME crate at full 24-channel occupancy
- for 120-140bytes/HTR @ L1A, bandwidth is 180-210 MB/sec
 this is for 10 time buckets + TPG + 1 byte/remaining channels + 15%
occupancy
- my feeling: we will be ok here
 e.g. 10 time buckets is probably overkill?
5
HB/HE  TDAQ Mapping
• HB 1-h-10
 2 crates, 15 HTR/crate, 1 channel = 1 TPG, 3 SLB/HTR
• HB 11-h-14
 1 crate, 12 HTR, 1 channel = 1 TPG, 3 SLB/HTR
• HB h=15
 1 crate, 9 HTR, 3 channels per TPG sum, 9 SLB
• HB+HE h=16
 1 crate, 12 HTR, 4 channels per TPG sum, 12 SLB
- 3/4 SLB per HTR, not sure if it’s really doable (ask WSmith!)
• HE 17-h-20
 1 crate, 15 HTR, 2 channels/TPG for h=17 and 1:1 for 18-20
6
HB/HE  TDAQ Mapping
(cont)
• HE h=21 and 22
 since 1 channel = 2 TPG, can only put 4 fibers into HTR
 use the unused fibers for HB PINS
- 4 fibers x 3 PINS = 12 PINS/card
 need 3 HTR cards for each annulus
- 3 x 4 fibers x 3 fibers/channel = 36 x 10° = 360°
- 36 PINS total per HTR x 2 HTR = 72 PINS
 takes care of all HB PINS
• HE h=23 and 24
 same as above, only here we put all 72 channels of HE PINS
• 1 crate for HE 21-h-24
 12 cards, still 3 SLB/card = 36 SLB
7
HB/HE  TDAQ Mapping
(cont)
• HE h=25 and 26
 2 longitudinal channels in sum, still 10°
 therefore half number of SLB as previous crate
 same number of HTR cards (12)
• HE h=27
 5 longitudinal channels in sum, still 10°
 5 fibers x 3 channels/fiber = 15 channels, or 30° per HTR
 12 HTR covers this annulus
- still 72 TPG/annulus, 8 channels/SLB, or 9 SLB total
8
HB/HE  TDAQ Mapping
(cont)
• HE h=28 and outer “radiation” annulus (29)
 add together to form 1 TPG annulus
 same 12 HTR as previous annulus now covers 0<f<p
 each HTR must receiver:
- 3 fibers for 2 channels of h=28 and “29” (Df=20°)
- 9 HTR makes 180° coverage
- 180 means 36 TPG channels, or 4.5 SLB
 will this work? WSmith says maybe….TBD
 must use 2 VME crates to cover these annuli
9
HO  TDAQ Mapping
• No SLB!
• 15 HTR/crate:
 24 channels/card, 120° of Df per HTR, 3 HTR/annulus
• Total of 3 crates for each HO, 6 total
10
HF  TDAQ Mapping
• Need 12 HF towers to make 1 TPG:
 3h, 2f, 2z
 12 channels means 4 fibers
- therefore 2 TPG per 1 HTR, or .5 SLB/HTR
• 1 HTR covers 3 units h, Df 40°
 would need 9 HTR to cover 3 units of h, all 360°
 4 sets of 9 would cover all 12 units of h
- 36 HTR total, can stuff into 3 VME crates of 12 HTR/crate
- total of 6 VME crates for entire HF
- NOTE: will go up when get better understanding of mapping
11
Unit Costs
• HTR
 $2600/HTR latest estimate
- includes a $1200 Xilinx XCV815EM for all 24 channels
 decreases card connectivity, has 1.12Mbit block RAM!
• Can perhaps go to 405EM, 560kbits, but no understood yet
• would save about $600/card
- Optical-to-Electrical finally sorted out (we hope)
 StratusLightwave SQR-45-9-1 - quad 1.6Gb,850nm,receiver
 “quotes” (courtesy of Rob’s persistence!) $130, or $260/card
• DCC/HRC
 stick with the $5k/card estimate
• VME
 $5.5k/crate, $3k/rack, 3 crates/rack
• only 10% spares on everything (except VME)
12
Cost Estimate
• Cards:
 HTR: 469 cards, $1,235k
 DCC: 38 cards, $ 190k
 HRC: 38 cards, $ 190k
• VME:
 Crates: 34, $187k
 Racks: 12, $ 36k
• TPG Cables:
 Total $36k
• Grand Total:
 $1.930M
13