STT session at L2 workshop
Friday, 24 April, 13:30 to 15:30
Agenda:
- STT baseline design vs STT with SLIC/FIC --- performance comparison [Ulrich Heintz] (15 min.)
- Cost and engineering impact of L2 options on STT [Hal Evans] (15 min.)
- status report on track fitting algorithm implementation, STT baseline processor card [John Hobbs] (15 min.)
- status report on queueing studies [Terry Heuring]
- impact of possible L1CFT changes (Glink to Cypress) on STT [Marvin Johnson] (5 min.)
- define and agree on criteria for decision on STT options (discussion, decision?) (20 min.)
- Requirements for STTpp, and performance comparison:
STT baseline design vs STT with FIC/SLIC/Alpha [Ulrich Heintz]
Ulrich showed the
list of requirements for the STTpp, an updated version of that shown in the
meeting of April 7;
for comparison baseline vs SLIC/..…, see
see here
for timing comparison, and
here
for general comparison -- pros and cons.
It appears that possible engineering amd maintenance benefits of the
FIC/SLIC option are outweighed by less performance.
Note that some of the numbers quoted in the timing comparison need to be
checked, in particular the processing times for the track fitting;
more reliable numbers will be available once John Hobbs has finished his<
algorithm and timing studies.
- Cost and engineering impact of L2 options on STT [Hal Evans]
This is a comparison of various options for realizing the TFC with standard
L2 boards, only with regards to cost and engineering needs, without
consideration of feasibility or performance. In all options it is assumed
that the baseline STC (with slight modifications depending on the option)
is used. Among the options considered, only the the ones with the classic
or the standalone SLIC has an advantage in the form of less engineering,
while all the other options have no engineering advantage and higher cost.
Transparencies of Hal's presentation are available
here
- Track fitting algorithms, TFC [John Hobbs]
John Hobbs is continuing his studies in search of the optimal track fitting
algorithm, as well as various implementation options (DSP's, Altera FPGA's,
other PLD's)
- queueing studies [Terry Heuring]
Terry Heuring is doing STT studies using the RESQ package; no results yet.
- Impact of possible L1CFT changes (Glink to Cypress) on STT [Marvin Johnson]
Marvin is against slowing down read-out -- advocates keeping G-link for SVX read-out and L1CFT.
- Main decisions affecting the STT taken at the L2 workshop:
- discard March 17 FIC/SLIC/Alpha option
- go ahead with design studies for STC baseline option
- leave track fitting (processor) card (TFC) open until after algorithm
and timing studies are done
- keep G-link for SVX read-out
- keep G-link for L1CFT
- abolish 20 bit G-link -- keep only 16 bit
