🔭 astrophysics

RNDR noise modeling in first-generation Single electron Sensitive Readout (SiSeRO) devices

This paper presents a noise model for repetitive non-destructive readout (RNDR) in first-generation Single-electron Sensitive Readout (SiSeRO) devices, exploring probabilistic mechanisms like thermal leakage and impact ionization to support the development of ultra-low-noise detectors for future X-ray and UV/visible/near-IR observatories.

Original authors: Tonya L. Peshel, Abigail Y. Pan, Tanmoy Chattopadhyay, Steven W. Allen, Marshall W. Bautz, Michael Cooper, Kevan Donlon, Catherine E. Grant, Sven Herrmann, Jill Juneau, Beverly J. LaMarr, Christopher

Published 2026-07-23

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Tonya L. Peshel, Abigail Y. Pan, Tanmoy Chattopadhyay, Steven W. Allen, Marshall W. Bautz, Michael Cooper, Kevan Donlon, Catherine E. Grant, Sven Herrmann, Jill Juneau, Beverly J. LaMarr, Christopher Leitz, Adam B. Mantz, Eric D. Miller, R. Glenn Morris, Declan O'Neill, Peter Orel, Artem Poliszczuk, Gregory Y. Prigozhin, Haley R. Stueber, Keith Warner

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, silent library where the books are made of light. To read the faintest whispers from the edges of space—like the glow of a distant exoplanet or the birth of a star—astronomers need cameras so sensitive they can count individual photons, the tiny particles that make up light. But here's the catch: every time a camera takes a picture, it makes a little bit of noise, like static on a radio. If that static is too loud, it drowns out the faint signal. For the next generation of space telescopes, scientists need cameras that are not only super fast but also so quiet that they can detect less than one electron of noise. It's like trying to hear a single pin drop in a hurricane, but the camera itself has to be the one making the pin drop. To do this, researchers are building a new kind of camera chip called SiSeRO, which uses a clever trick called "Repetitive Non-Destructive Readout" (RNDR). Think of RNDR as a game of "guess the weight" where, instead of weighing a bag of sand once and hoping you got it right, you weigh it a hundred times, add up the results, and average them out. The more times you weigh it, the closer you get to the true weight, and the less your shaky hands matter.

This paper is about the engineers and scientists who built these super-sensitive SiSeRO chips and asked a very important question: "If we keep weighing that bag of sand a thousand times, will the bag start to get heavier on its own?" In the world of electronics, "getting heavier" means picking up extra, unwanted electrons from nowhere, which looks like noise and ruins the picture. The team developed a computer simulation to act like a crystal ball, predicting what might happen if they pushed these chips to their absolute limits. They looked at two main suspects that could cause this extra noise: "thermal leakage," where heat makes electrons jump out of their seats, and "impact ionization," where speeding electrons crash into atoms and knock loose new ones.

The researchers built a detailed model of the chip's output stage, which is the part that reads the signal. They started by simulating the normal, expected noise from the electronics themselves, which behaves exactly as predicted: the more times you read the signal, the quieter it gets, following a perfect mathematical curve. Then, they added the two "bad guys" to the simulation to see if they would break the pattern. First, they looked at thermal leakage. They calculated how likely it is for electrons to escape due to heat, even at the freezing cold operating temperature of -100°C (173 K). Their simulation showed that at this temperature, the heat is too low to make the electrons jump around enough to cause a problem. It's like trying to boil water in a freezer; the water just stays put.

Next, they investigated impact ionization. This happens when electrons zoom through the chip so fast that they smash into the silicon and create new electron-hole pairs, like a billiard ball hitting a rack of balls and scattering them everywhere. The team simulated this by cranking up the electric field to see if it would cause a cascade of noise. However, when they compared their simulation to the actual data they had collected from real chips running up to 200 cycles, they found a perfect match with the "no extra noise" theory. The real chips didn't show any signs of the billiard balls scattering. This allowed the team to rule out the idea that impact ionization is currently ruining their measurements. In fact, they were able to set a very strict "speed limit" on how much extra noise could possibly be hiding in their data without them noticing.

The bottom line is that for now, the SiSeRO technology is incredibly robust. The team's simulations and real-world tests show that even after hundreds of reads, the chips don't start generating their own noise from heat or electron crashes. They have proven that up to 200 cycles, the "bag of sand" doesn't get heavier on its own. While they can't promise that this will hold true forever if they push to thousands of cycles, their models suggest that any noise that does appear would have to be incredibly tiny. This gives astronomers confidence that these new detectors are ready to help build the next generation of telescopes, capable of seeing the faintest light in the universe without getting confused by their own internal static. The paper concludes that while future work will test even longer cycles, the current generation of SiSeRO devices is already a massive step forward, offering a clear path to ultra-low-noise vision for the cosmos.

Technical Summary: RNDR Noise Modeling in First-Generation SiSeRO Devices

Problem Statement
Future flagship observatories in the X-ray and UV/visible/near-IR regimes require detector technologies capable of sub-electron read noise, ultra-fast readout, and scalability to large focal plane arrays. While conventional Charge-Coupled Devices (CCDs) are insufficient for these requirements, Single-electron Sensitive Readout (SiSeRO) devices offer a promising alternative. SiSeRO technology utilizes Repetitive Non-Destructive Readout (RNDR), where a single charge packet is repeatedly measured to reduce effective read noise proportional to $1/\sqrt{N}$ , where $N$ is the number of cycles.

Although first-generation SiSeRO prototypes have demonstrated exceptional performance (achieving ~0.5 $e^-$ noise in under 57 cycles) with no deviation from the theoretical $1/\sqrt{N}$ trend up to 200 cycles, there is a concern that extending RNDR cycles further to meet future mission requirements may eventually encounter a noise floor. Specifically, stochastic physical mechanisms—such as thermal leakage and impact ionization—could introduce excess dark current-like noise that scales linearly with read time, thereby limiting the ultimate noise performance.

Methodology
To investigate these potential limits, the authors developed a Monte Carlo-based noise model tailored to the first-generation SiSeRO architecture (CCID-93 prototypes developed by MIT Lincoln Laboratory). The model simulates the output stage behavior over extended RNDR cycles (up to 1000 cycles) and incorporates the following components:

Baseline Noise Modeling: The model generates digitized waveforms matching the 10 ns resolution of the Archon controller. It simulates the standard read noise using:
- Thermal Noise: Modeled as a random scatter from a Poisson distribution, scaled by the square root of temperature ( $\sqrt{T}$ ).
- 1/f (Flicker) Noise: Modeled using a Poissonian sequence of pulses in the time domain with fixed amplitude and random sign, following established prescriptions.
- Note: The model excludes thermal dark current from the image region, focusing solely on the output stage.
Excess Noise Mechanisms: The authors specifically modeled two probabilistic mechanisms that could introduce dark current-like noise:
- Trap-Assisted Thermal Leakage: Using the Shockley-Read-Hall (SRH) model, the authors calculated electron emission rates from crystal defects in the depleted internal gate region. This depends on trap density, temperature, and trap energy levels (fixed at ~0.6 eV).
- Impact Ionization: The model calculates the probability of electron-hole pair generation near the MOSFET drain implant due to high electric fields. This utilizes the Van Overstraeten-De Man model, incorporating a correction factor ( $K$ ) to account for the non-linear electric field near the drain.
Statistical Analysis: The model injects these probabilistic events into the RNDR waveform. The authors compared the simulated cumulative noise against the theoretical $1/\sqrt{N}$ trend. Furthermore, they applied $\chi^2$ -minimization and F-tests to laboratory data (100–200 cycles) to place statistical upper limits on any potential excess charge signal.

Key Results

Baseline Behavior: The inclusion of thermal and 1/f noise in the model did not cause any divergence from the $1/\sqrt{N}$ trend, consistent with the behavior of an ideal RNDR device in the absence of time-dependent excess noise.
Thermal Leakage: The model predicts that at the operating temperature of 173 K (-100°C), the contribution of trap-assisted thermal leakage is negligible. Significant excess noise from this mechanism is only predicted at higher operating temperatures.
Impact Ionization: The probability of impact ionization increases with electric field strength but decreases with temperature (likely due to increased lattice scattering). At 173 K, the model indicates that the electric fields present in the current devices are insufficient to cause measurable excess noise.
Experimental Constraints: Analysis of existing laboratory data (up to 200 cycles) shows no statistically significant deviation from the $1/\sqrt{N}$ $1/ N$ trend. The authors placed statistical upper bounds on the amplitude of any potential excess charge:
- At 173 K, the 2 $\sigma$ upper bound for excess charge is $3.08 \times 10^{-4}$ ADU/cycle ( $9.61 \ e^-$ /pixel/s).
- The minimum detectable signal is constrained to $7.23 \times 10^{-5}$ ADU/cycle ( $2.26 \ e^-$ /pixel/s).
- The data indicates that any excess charge generated is too small to measurably affect detector performance under current test conditions.

Significance and Claims
The paper claims that the maturation of SiSeRO technology addresses critical technology gaps for future X-ray and UV/visible/near-IR observatories, such as the Habitable Worlds Observatory (HWO). The primary significance of this work lies in:

Validation of Scalability: The results confirm that first-generation SiSeRO prototypes, even without specific optimization for RNDR, exhibit no evidence of excess noise mechanisms up to 200 cycles. This supports the feasibility of extending RNDR cycles to achieve the sub-electron noise requirements of future flagship missions.
Predictive Modeling: The developed Monte Carlo model provides a generic framework for simulating RNDR noise behavior. It allows researchers to predict the limits of detector performance based on physical parameters like temperature, trap density, and electric field strength.
Design Guidance: By establishing statistical constraints on dark current-like noise, the work informs the design iterations for future second-generation SiSeRO devices. It suggests that current operating conditions are safe from impact ionization and thermal leakage, but future testing at significantly longer cycle runs will be necessary to fully characterize the noise floor as the technology matures.

The authors conclude that while no deviation from the ideal noise trend has been observed yet, the probabilistic modeling of thermal leakage and impact ionization prepares the community for potential noise floor limitations as RNDR cycles are pushed further to meet the stringent requirements of next-generation space and ground-based telescopes.

More like this