SVT: an online silicon vertex tracker for the CDF upgrade Page: 4 of 6
This article is part of the collection entitled: Office of Scientific & Technical Information Technical Reports and was provided to Digital Library by the UNT Libraries Government Documents Department.
The following text was automatically extracted from the image on this page using optical character recognition software:
HI INDER MERGER A -
x 12CUT 2
2 MERGER LEVELS (4 brds)
U ~ ~t* I I
Fig. 1. The SVT architecture (left) and processing time (right)
of 250 pm transverse size). An average of three low-resolution tracks (roads) per event are
output from the AM to the HB, which reassociates each road with a list of hits. The road+hits
information is subsequently sent to the Track Fitters (TFs) which perform a full resolution (-
15 pm) fit using a fast linearized algorithm. Each processor in the TF farm gets one road and
reconstructs one or more tracks within that road. The TFs run in parallel on different roads
and the number of TF processors is large enough that each TF will only work on one road per
event for the majority of the events. The final output of the TFs is a set of track parameters
(Pt, 0 and impact parameter), to be used by the L2 trigger processors for the final decision.
Fig.l(right) shows the overall timing of SVT operation on the average event, starting from Li
accept. Readout, hit finding and AM input overlap in time, and when data are sent to the TFs
(- 7 s) a new event can be fed into the HF-AM chain, so that the SVT works basically as a
2-stage pipeline. Under the assumption that all the boards work on a 30 MHz clock, simulations
show that the average event processing time for SVT is about 13 ps.
2.1 Associative Memory
In the AM, patterns corresponding to real tracks are stored as a combination of superstrip
addresses in the various planes of the detector. The AM recognizes roads while receiving the
stream of hits coordinates by successively comparing them with all stored patterns . Roads
which have fired in the current event are flagged, and output as a stream of road addresses at
the end of the input stage (Fig.2). The AM is subdivided into banks, each AM bank working on
one of the 12 30 0 sectors of the SVXII (wedge). Since the size of the AM needed to store all
possible roads increases more than exponentially with the number of channels, the algorithm is
only applied with limited resolution: the size of the superstrips, and thus the size of each AM
bank, is dictated by a compromise: reducing the resolution increases the number of fake roads,
while reducing the number of patterns to store. It also increases the probability of multiple hits
Here’s what’s next.
This article can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Article.
Bardi, A.; Belforte, S. & Berryhill, J. SVT: an online silicon vertex tracker for the CDF upgrade, article, July 1, 1997; Batavia, Illinois. (digital.library.unt.edu/ark:/67531/metadc690535/m1/4/: accessed November 14, 2018), University of North Texas Libraries, Digital Library, digital.library.unt.edu; crediting UNT Libraries Government Documents Department.