## Overview of Electronic Noise in the Readout Chain of CZT Strip Detectors
Electronic noise in the readout chain of double-sided CZT strip detectors fundamentally limits the achievable energy resolution by introducing random fluctuations in the measured charge amplitude for each gamma-ray interaction. The readout chain typically consists of charge-sensitive preamplifiers connected to each anode and cathode strip, followed by shaping amplifiers, analog-to-digital converters, and digital signal processing stages. Because strip detectors require low-noise performance to exploit the excellent charge-transport properties of CZT while maintaining a linear channel count, noise contributions are analyzed in terms of equivalent noise charge (ENC) expressed in electrons rms. The total ENC is obtained by quadratic summation of individual noise components, each with distinct dependence on shaping time, detector capacitance, leakage current, and temperature. In practice, the dominant noise sources shift depending on strip length, bias conditions, and operating temperature, but for typical room-temperature operation with sub-nanoampere leakage and strip capacitances of several picofarads, the interplay of parallel and series noise terms sets the ultimate resolution floor around 0.5–1 keV full width at half maximum at low energies.
## Shot Noise from Detector Leakage Currents
The dominant parallel noise source arises from the shot noise generated by the detector’s leakage current, which includes both bulk leakage through the crystal volume and surface leakage between adjacent strips. This current, typically maintained below 1 nanoampere per strip at operating bias voltages of 50–200 volts through optimized passivation and steering electrodes, produces a white noise spectrum whose power spectral density is 2 q I_leak, where q is the elementary charge and I_leak is the total leakage current per channel. When integrated through the shaping filter, the resulting ENC contribution scales with the square root of the shaping time constant τ as ENC_parallel ∝ √(I_leak τ). In strip detectors, inter-strip leakage is particularly critical because the narrow gaps (50–250 micrometers) experience high lateral electric fields from steering biases; any increase in surface conduction directly elevates this noise term. For high-energy gamma-ray spectroscopy, where longer shaping times are sometimes used to reduce ballistic deficit, this shot noise can become the leading contribution unless leakage is aggressively suppressed through chemical passivation and guard-ring structures.
## Thermal Noise from the Input Field-Effect Transistor
Series thermal noise, often called Johnson noise, originates primarily from the channel resistance of the input field-effect transistor (FET) in the charge-sensitive preamplifier. This noise has a white voltage spectral density of 4 k T / g_m, where k is Boltzmann’s constant, T is absolute temperature, and g_m is the transconductance of the FET. When referred to the input, the ENC contribution is proportional to the total input capacitance C_tot (detector capacitance plus stray and feedback capacitance) multiplied by the square root of the inverse shaping time: ENC_series ∝ C_tot √(1/τ). In CZT strip detectors, long anode or cathode strips result in elevated inter-strip and strip-to-backplane capacitances, often reaching 5–20 picofarads per channel, which amplifies this term. Low-noise JFET or CMOS input stages with high g_m are therefore selected to minimize thermal noise, but cooling the preamplifier (to −20 °C or lower in some laboratory setups) further reduces it by lowering T and increasing carrier mobility. This noise source frequently dominates at short shaping times or in detectors with higher capacitive loading.
## 1/f Noise or Flicker Noise Contributions
Flicker noise, characterized by a 1/f power spectrum, arises from carrier trapping and detrapping at the FET channel–oxide interface or from contact and surface states within the CZT crystal itself. Its voltage noise density increases at low frequencies, making it particularly detrimental when long shaping times are employed to optimize parallel noise. The ENC contribution from 1/f noise scales with C_tot and is largely independent of shaping time within the typical range of 0.5–10 microseconds used for CZT readout. In strip detectors, imperfect electrode–semiconductor interfaces or residual surface damage from polishing and etching can elevate this component. Modern low-1/f-noise FET technologies and careful surface passivation of the CZT reduce this term to a secondary but still measurable level, often contributing 20–40 percent of the total ENC in well-optimized systems. Digital filtering techniques that suppress low-frequency components help mitigate its impact during pulse-height analysis.
## Capacitive Loading and Associated Series Noise Enhancement
The total input capacitance seen by the preamplifier—comprising the strip self-capacitance, inter-strip coupling capacitance, and parasitic wiring capacitance—directly amplifies all voltage noise sources (thermal and 1/f) through the factor C_tot in the ENC formula. In crossed-strip geometries, the orthogonal anode and cathode strips create a distributed capacitive network where each channel couples to multiple neighbors, increasing effective C_tot beyond that of a simple planar detector. For a 250-micrometer-pitch strip over a 10-millimeter length, this capacitance can exceed 10 picofarads, pushing series noise into the dominant regime. Stray capacitances from bond wires or flex-cable interconnects further degrade performance unless minimized through close integration of ASICs directly onto the detector substrate. This capacitance-driven enhancement underscores why strip detectors, despite their channel-count advantage, demand ultra-low-noise front-end electronics tailored to higher capacitive loads than pixelated counterparts.
## Reset Noise and Feedback-Resistor Thermal Noise
In continuous-reset charge-sensitive preamplifiers commonly used for CZT strips, the feedback resistor R_f generates additional parallel thermal noise with current noise density 4 k T / R_f. This term adds directly to the leakage-current shot noise and must be balanced against reset-noise considerations; very large R_f values reduce resistor noise but increase the probability of baseline instability. Pulsed-reset or optical-reset schemes eliminate the continuous resistor noise at the cost of periodic dead time, making them attractive for low-rate applications. In high-count-rate gamma-ray spectrometry with strip detectors, continuous-reset designs with carefully chosen R_f (typically 100 MΩ to 1 GΩ) keep this contribution below 100 electrons rms while maintaining stable operation.
## Quantization Noise and Downstream Digital Processing Contributions
After analog shaping, the analog-to-digital converter (ADC) introduces quantization noise whose rms value is LSB / √12, where LSB is the least-significant-bit voltage corresponding to the system gain. For 12- to 16-bit ADCs sampling at 10–100 MHz with full-scale ranges matched to typical CZT signal amplitudes (tens to hundreds of keV), this term is usually negligible compared with front-end noise but can become relevant at very low energies or when high gain is used to resolve sub-keV features. Digital signal processing stages, including baseline restoration and pile-up rejection filters, may add minor computational noise if fixed-point arithmetic is employed, although floating-point or high-bit-depth processing minimizes this. In multi-channel systems where charge-sharing events require summing adjacent strip signals, the uncorrelated noise from each channel adds in quadrature, effectively increasing the total ENC for shared events by √2 or more.
## Crosstalk and Electromagnetic Interference in Multi-Channel Readout
Capacitive crosstalk between neighboring strips and electromagnetic pickup from external sources represent additional noise pathways that are particularly pronounced in densely packed strip readout chains. Inter-strip capacitive coupling can transfer noise voltage from one channel to another, especially when steering biases create high-impedance nodes. Proper grounding, shielding, and differential signaling in the ASIC-to-FPGA interface are essential to suppress this. In large-area tiled detectors, common-mode noise from power supplies or digital clocks can couple into the sensitive analog front end, requiring careful layout and filtering. Although not intrinsic to a single channel, these system-level effects can elevate the effective noise floor if not controlled to below 50 electrons rms.
## Practical Impact on Energy Resolution and Mitigation Strategies
The quadratic summation of these noise sources ultimately determines the electronic contribution to the total energy resolution, which combines with Fano statistics and charge-trapping variance to set the observed full width at half maximum. At optimal shaping times (typically 1–6 microseconds for CZT strips), total ENC values of 100–300 electrons rms are routinely achieved, corresponding to 0.5–1.5 keV electronic noise at 59.5 keV. Dominant terms shift with operating conditions: low-leakage detectors emphasize series and 1/f noise, while higher-temperature or higher-bias operation elevates shot noise. Mitigation involves selecting high-g_m, low-1/f FETs, cooling the front end, minimizing interconnect capacitance through bump-bonding or direct ASIC attachment, and employing adaptive digital filtering to optimize effective shaping time per event. Advanced ASICs tailored for CZT strips incorporate these considerations, enabling near-Fano-limited resolution in practical gamma-ray spectrometer systems while preserving the scalability advantages of the strip architecture.
CdZnTe Association (CdZnTe.com)
https://www.cdznte.com/blog/what-are-the-dominant-sources-of-electronic-noise-in-the-readout-chain-of-czt-strip-detectors.html
