Leonard MacEachern

MALTA2 Boundary-Loss Surrogate for Asynchronous MAPS Readout

CERN, event-driven MAPS readout, and an analytic model connecting charge sharing, pixel grouping, address encoding, and tracking loss.

What this research is about

MALTA2 uses an event-driven pixel readout for recording particle tracks. To reduce the amount of data leaving the sensor, nearby pixel addresses that arrive almost simultaneously can be combined with a bitwise OR. That saves bandwidth, but it can also create a decoded address that no longer points to the original hit; if the shift is too large, tracking may reject a real particle hit.

This work condenses that chain of effects into a compact model linking charge sharing and timing, pixel-group boundaries, address encoding, and tracking loss. It makes it possible to compare group shapes and coincidence windows without rerunning a full detector simulation for every design choice. At the calibrated MALTA2 operating point, the two pixel directions behave almost symmetrically, making an 8×8 arrangement the preferred shape for a 64-pixel group.

CERN collider tracking environment and a MALTA2-style event-driven sensor array
MALTA2 pixel matrix containing an inner dashed eight-by-eight group, dimension arrows aligned to that group, and asynchronous OR-merged readout logic
Five-stage surrogate with arrows linking charge-sharing and time-walk kernels, directional coincidence probabilities, boundary densities, encoding loss factors, and tracking loss
Φtrk(Nx,Ny) = Lx px/Nx + Ly py/Ny Physics + geometry + encoding → algebraic design rule
Normalized seed-pixel and adjacent-pixel physics kernels
Geometry regime map showing the square-preferred MALTA2 operating point
Constant-area aspect-loss curve showing the eight-by-eight optimum for a 64-pixel group

Calibrated MALTA2 insight

  • Directional merge physics is nearly isotropic (ρ ≈ 1).
  • Effective encoding loss is nearly symmetric (Lxeff ≈ Lyeff ≈ 1).
  • For 64 pixels, the fixed-area optimum is 8×8.
  • The regime map supports inverse design of geometry and coincidence window τ.

Growing CERN relationship: detector expertise, electronics, modelling, and future measurements meet at a shared calibration interface. For MALTA3 in 65 nm TPSCo, the kernels must be retuned to the process-specific timing and analog response.

01 · Physical process

From charge sharing to a merged address

The model begins with the seed- and adjacent-pixel response kernels. These determine how likely two pixel addresses are to arrive inside the same coincidence window and therefore be merged by the asynchronous readout.

02 · Architectural consequence

Boundaries turn local coincidences into tracking loss

A merged address can cross a group boundary or acquire bits from two neighbouring hits. Boundary density, directional coincidence probability, and encoding loss therefore connect the pixel geometry to a track-level observable.

03 · Design use

Compare group geometry without a full detector run

Once calibrated, the surrogate supports rapid comparison of group shapes and timing windows. It does not replace detailed detector simulation; it provides a compact inverse-design rule for deciding which configurations merit that expense.