On-board Calibration System

Updated 9 January 2026

On-board calibration systems are integrated hardware–software solutions that autonomously calibrate sensor arrays using controlled stimuli and real-time data acquisition.
They employ methods like LED amplitude scanning, linear response modeling, and per-cell corrections to maintain sub-percent accuracy over wide dynamic ranges.
These systems are essential in high-energy physics and space missions, mitigating issues like pedestal shifts and cross-talk to ensure scalable and reliable calibration.

An on-board calibration system is an integrated hardware–software solution that enables autonomous, in-situ calibration of electronic, optical, or sensoric subsystems without external intervention. These systems inject well-controlled calibration stimuli (electrical charge, light pulses, RF tones, photon or particle beams, etc.) and execute precise measurement protocols to monitor, correct, and verify gain, linearity, timing, and cross-talk in high-density readout chains, often in challenging environments such as space missions or high-energy physics experiments. Modern on-board calibration architectures employ programmable drivers, embedded reference sources, signal distribution networks, and real-time DAQ pipelines to achieve sub-percent accuracy across wide dynamic ranges.

1. Architectural Principles and System Components

A typical on-board calibration system, such as the QMB6+HBU0 complex for SiPM calorimetry, consists of:

Dedicated calibration hardware boards: e.g., the QMB6 PCB hosting quasi-resonant UV-LED drivers with dual control voltages (V1, V2) for amplitude tuning, placed directly above a sensor array (HBU0) (Cvach et al., 2011).
Optical/electrical stimulus delivery network: Notched optical fibers—precisely milled and polished—to distribute light to sensor tiles with uniformity σ/μ ≈ 15%, alignment fixtures to guarantee reproducible coupling, and routing for scalable multiplexing.
Sensor and front-end ASIC chain: Silicon photomultipliers (SiPMs), multi-channel ASICs (SPIROC2), with high-gain/low-gain preamplification, ring-structured analogue memory (16 cells/channel), and 12-bit ADC conversion, supporting per-cell pedestal interpolation and gain tracking.
Control and synchronization interfaces: PC↔CANbus↔calibration board chain, enabling programmable pulse amplitude/frequency, automated scan protocols, and data acquisition sequencing with sub-millisecond resolution.

Textual block-diagram (as in the QMB6+HBU0 system):

LabVIEW PC ↔ CANbus ↔ QMB6
   ↓
Quasi-Resonant LED drivers (6x)
   ↓
Milled notched optical fiber (12 notches/row)
   ↓
Scintillator tile + SiPM + SPIROC2 ASIC (36ch) on HBU0
   ↓
Analogue memory + ADC → DAQ/PC

2. Calibration Methodologies and Key Algorithms

2.1 Amplitude and Response Scanning

LED amplitude scans are performed by stepping V1 (with fixed V2), probing delivered light in pixel/MIP equivalents, covering 0–4095 ADC bins at 1-bin steps (0.08% resolution) (Cvach et al., 2011), synchronized with hardware triggers (beam-off/internal, beam-on/external).
Response modeling: In the non-saturated regime, a linear fit L(V_LED) = a·V_LED + b is constructed for each channel, calibrating both stimulus and sensor-specific gain.
ADC-to-MIP conversion: After subtracting per-channel pedestals P₀ and applying channel-wise gain G, calibrated MIP equivalents are computed as N_MIP = (A_SiPM − P₀)/G.

2.2 Memory Cell Effects and Corrections

Pedestal profiles: Analogue memory cells show non-uniform pedestal shifts (first 3 cells elevated by ~50 bins, others by ~20 bins); high-gain mode accentuates pedestal bias, necessitating per-cell correction to avoid 1–2 MIP errors at low signal (Cvach et al., 2011).
Cross-talk and non-uniformity: Optical notch variability and SiPM gain spread contribute to uniformity metrics (σ/μ ≈ 15% for fiber, ≈10% for SiPMs); software compensation via individualized calibration constants corrects inter-channel illumination bias.

3. Performance Metrics and Operational Results

Calibration efficacy is quantified through dynamic range, nonlinearity, and channel uniformity metrics:

Metric	Range/Spread	QMB6+HBU0 Example (Cvach et al., 2011)
Dynamic range	0–250 MIPs	~2500 fired pixels, linear to ~30 MIP
Initial linearity	δ < 1% (up to ~30 MIPs)
Saturation nonlinearity	δ ≈ 5% (at 200–250 MIPs)
Optical-fiber uniformity (σ/μ)	~15%
SiPM gain uniformity (σ/μ)	~10% (across 12 notches)
ASIC gain ratio (HG/LG)	Mean ≈ 17.5, σ/μ ≈ 2%
Pedestal shift at 80% saturation	~30 bins (1.2% dyn. range)

Accuracy in the linear region supports tracking sensor gain drift to sub-1%. Saturation region corrections and cell-wise pedestal fitting are essential for maintaining calibration precision in high-rate regimes.

4. Limitations and Mitigation Strategies

Pedestal shifts and first-cell offsets impose systematic errors in low-signal channels, notably under high illumination. Corrections require pre-measured zero-LED pedestal profiles or exclusion of saturated regions in high-gain operation.
Inter-channel crosstalk due to overlapping fiber notches leads to non-uniform stimulation of detectors; individualized calibration constants and mapping of notch efficiencies are employed for software correction.
Memory cell structure and saturation artifacts must be characterized for each ASIC, with on-board per-cell calibration tables in firmware recommended for future large-scale deployment.

5. Design Implications for Large-Scale Systems

On-board calibration architectures like QMB6 support high-density, scalable calorimetry or sensor arrays by:

Providing direct, per-channel stimulus calibration that operates across the full dynamic range required for physics or mission objectives.
Enabling real-time correction for environmental drift, cross-talk, and instrumentation aging, essential in operational contexts such as particle physics or spaceflight.
Facilitating future extensions—integration of adjustable pulse-width drivers, notched fiber uniformization, and firmware-level per-cell calibration tables improve practical deployment and scalability to thousands of channels.

6. Comparative Cases and Broader Context

Additional systems (DAMPE BGO-Cal (Zhang et al., 2015); EUDET HCAL (Cvach et al., 2011)) employ similar paradigms—a digitally programmable charge injection circuit or amplitude-tunable LED drivers combined with notched fiber optic light distribution—demonstrating the widespread adoption of autonomous, programmable calibration concepts in contemporary high-energy and space instrument front-ends. Practical results consistently demonstrate sub-percent gain accuracy, robust environmental tolerance, and effective mitigation of biasing effects, informing design recommendations across domains.

In summary, an on-board calibration system combines programmable stimulus generation, distributed delivery networks, real-time acquisition/processing chains, and channel-wise calibration algorithms to establish metrologically robust, scalable, and autonomous calibration for dense sensor arrays. In advanced implementations, per-cell corrections, individualized fitting, and firmware-level adaptation ensure sub-percent accuracy and operational resilience (Cvach et al., 2011, Zhang et al., 2015, Cvach et al., 2011).