Skip to content

[ALICE3] TRK: rInnRow column for staggered ML layers in the config file - #15992

Open
lorber98 wants to merge 4 commits into
AliceO2Group:devfrom
lorber98:trk-ml-rinnrow
Open

lorber98 wants to merge 4 commits into
AliceO2Group:devfrom
lorber98:trk-ml-rinnrow

Conversation

@lorber98

@lorber98 lorber98 commented Oct 21, 2026 •

Copy link
Copy Markdown

The TRK configuration file (TRKBase.configFile, layouts kSegmented and kSimplifiedRealistic) accepts two new optional columns for the ML layers:

  • 8th column, rInnRow: in the staggered layers (3 and 4) it chooses which staves sit at rInn:

    • 0: even staves at rInn (default, current behaviour);
    • 1: odd staves at rInn, i.e. stave 0 in the shifted row.

    In the flipped layer 3 the rInn row is the outer one and the shift points inwards, so there rInnRow = 1 puts stave 0 in the inner row; in layer 4 in the outer row.

  • 9th column, phiOffset (degrees, counterclockwise as phi, default 0): rotates all the staves of a ML layer about the beam axis. They keep their tilt with respect to their own radial directions.

The columns are positional: to give rInnRow the mode column (7th) must be written too, to give phiOffset also rInnRow.

  • Files without the new columns and the built-in layouts give the same geometry as before.
  • An rInnRow other than 0 or 1 is fatal. Non-default values are reported in the log, e.g. Layer 3 (staggered): even staves in the inner row (rInnRow = 1) (or as having no effect on a layer that is not staggered).

Why

To reproduce in O2, with the native TRK builder, the layout of the current CAD model of the middle layers:

  • in the CAD, the L6 stave at phi = 0 is in the inner row; without rInnRow the native layer 3 always has stave 0 in the rInn row (the outer one, since the layer is flipped);
  • in the CAD, the staves of L3, L4 and L5 are rotated about the beam with respect to the O2 placement (by -1.01°, -0.26° and -0.07°); the radius and the tilt could be set in the file, but not the azimuth.

Configuration file with the CAD layout

Parameters derived from the silicon of the CAD staves (latest model). Columns must be separated by TABs. Usage:
--configKeyValues "TRKBase.layoutMLOT=kSegmented;TRKBase.configFile=<file>"

/ ALICE 3 TRK, TRKBase.layoutMLOT=kSegmented: ML layout of the CAD model of the middle layers (21/10/2026)
/ rInn = silicon face nearest to the beam at the stave centre; mode 1: thick is x/X0 (chip 0.095 cm)
/ --- ML LAYERS (0..4) ---
/ rInn	thick	tilt	nStaves	nMods	stagOffset	mode	rInnRow	phiOffset
7.2610	0.01	6.974	10	11	0.0000	1	0	-1.0125
10.5799	0.01	5.486	14	11	0.0000	1	0	-0.2561
13.7799	0.01	4.990	18	11	0.0000	1	0	-0.0731
20.1800	0.01	0.000	26	11	0.3600	1	1	0.0000
29.0800	0.01	0.000	38	11	0.4000	1	0	0.0000

With this file every stave of the five ML layers has the same phi, tilt and row as the CAD stave.

The TRK configuration file (TRKBase.configFile) accepts an
optional 8th column, rInnRow, which chooses the staves of a staggered layer
(layers 3 and 4) that sit at rInn: 0 = even staves (default, previous
behaviour), 1 = odd staves, i.e. stave 0 in the shifted row. Existing files and the
default layouts give the same geometry as before. A value other than 0 or 1 is
fatal.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@lorber98
lorber98 marked this pull request as ready for review October 21, 2026 07:31
@lorber98
lorber98 marked this pull request as draft October 21, 2026 15:31
The TRK configuration file (TRKBase.configFile) accepts an optional 9th column, phiOffset
(degrees, counterclockwise as phi), which rotates all the staves of a ML layer about the beam
axis, keeping their tilt. Existing files and the default layouts give the same geometry as before.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@lorber98
lorber98 marked this pull request as ready for review October 21, 2026 21:00
@njacazio
njacazio enabled auto-merge (squash) October 22, 2026 02:03

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

2 participants