Skip to content

Perfect Agreement on Strip clustering Acts vs Athena outcomes

Carlo Varni requested to merge cvarni/athena:SkipClusterizationOnGroupSize into main

The Acts and Athena strip clusterization algorithms were producing different results, with the Acts algorithm producing more strip clusters than the Athena counterpart. In particular, the extra clusters were all included in collection with hashIds that were not available in the Athena strip cluster collections.

The effect is very minimal, on 2.8 k events (tt-bar PU200):

INFO Strip Clusters
INFO    Number of InDet Clusters: 643518519
INFO    Number of xAOD Clusters : 643523155
INFO    Number of mismatches    : 4636

Turns out this is because the Athena strip clusterization algorithm was applying a cut on the getGroupSize of the RDO collections (the maxFiredStrips property is set to 384)

// If more than a certain number of RDOs set module to bad
unsigned int nFiredStrips{0};
for (const SCT_RDORawData* rdo : *rd) {
   nFiredStrips += rdo->getGroupSize();
}
if (nFiredStrips > m_maxFiredStrips.value()) {
   [...]
   continue;
}

This cut was not applied in the Acts algorithm, this MR adds it skipping the clusterization should this condition occur. As a result, there is now perfect matching between the outputs of the two algorithms.

Edited by Carlo Varni

Merge request reports