Distortion Is Not Noise: On the Limits of the Kappa Model for Monostatic ISAC

This paper argues that the aggregate $\kappa$ distortion model is overly pessimistic for monostatic ISAC sensing because the transmitter can monitor its own waveform, and it derives new PA-aware and PN-aware Cramér–Rao bounds to demonstrate that this approach reveals an irreducible velocity-error floor while significantly overestimating sensing degradation compared to practical scenarios.

The Gist

Here is an explanation of the paper using simple language and everyday analogies.

The Big Idea: "Distortion isn't always a mystery"

Imagine you are trying to measure the distance to a mountain by shouting a specific word and listening for the echo.

• The Problem: Your voice box (the Power Amplifier) is a bit broken. When you shout, your voice gets a little distorted, sounding slightly "crunchy" or "fuzzy."
• The Old Way (The $\kappa$ Model): Most engineers used to think, "Oh no! The distortion is just random noise. We don't know what the original sound looked like anymore, so we have to guess." They treated the distortion like static on a radio. This made them think their measurements would be terrible.
• The New Discovery: In a Monostatic system (where the person shouting and the person listening are in the same room), the listener knows exactly what the distorted sound was because they heard it come out of the speaker. They don't need to guess! They can subtract the distortion perfectly.

The paper's main point: The old "noise" model is too pessimistic. It makes the system look worse than it actually is. When you know your own distortion, your sensing ability is much better.

---

The Two Main Characters: The "Crunchy Speaker" and the "Wobbly Clock"

The paper looks at two specific hardware problems that happen in 6G and advanced radar systems:

1. The "Crunchy Speaker" (Power Amplifier Nonlinearity)

• What it is: When you turn up the volume on a speaker too high, it starts to distort the sound.
• The Old Mistake: Engineers thought this distortion ruined the radar's ability to see things.
• The Reality: Because the radar knows exactly how the speaker distorted the sound, it can fix it.
  • Analogy: Imagine you are drawing a picture with a wobbly hand. If you are the one drawing, you know exactly how your hand wobbled. You can still trace the outline perfectly. But if someone else tried to guess what you drew based on the wobbly lines, they would fail.
  • Result: The "crunchy speaker" barely hurts the radar (less than 1 dB of loss), but it hurts the communication (sending data to a phone) a lot, because the phone doesn't know how the speaker distorted the signal.

2. The "Wobbly Clock" (Phase Noise)

• What it is: The internal clock that keeps the signal steady isn't perfect; it jitters slightly. This makes the signal's timing drift.
• The Reality: This is a big deal for radar.
  • Analogy: Imagine trying to time a sprinter running a race. If your stopwatch starts and stops randomly (jitters), you can't tell how fast the runner is going, even if the runner is perfect.
  • Result: This creates a "Velocity Floor." No matter how much power you use or how clear the signal is, you hit a hard limit on how accurately you can measure speed. You can't measure speed better than the jitter of your clock allows. The old models missed this limit entirely.

---

The "Magic Separation" (The Best Part)

The most exciting finding in the paper is that these two problems don't mix. They are orthogonal (independent).

• The Analogy: Imagine you are tuning a car.
  • The Engine (Power Amplifier) controls how fast you can drive (Communication Speed).
  • The Suspension (Oscillator/Clock) controls how smooth the ride is and how well you can see the road (Radar Accuracy).
• The Discovery: You can tune the engine without messing up the suspension, and vice versa.
  • If you want better communication, you just need a better Power Amplifier (make the speaker less crunchy).
  • If you want better radar, you just need a better Clock (make the timekeeper less jittery).
  • You don't have to compromise one to get the other.

Why This Matters for the Future

1. Stop Wasting Money: The old models told engineers, "You need super-expensive, perfect hardware to get good radar." This paper says, "Actually, you can use cheaper, slightly distorted hardware for the radar part because you can fix the distortion in software."
2. Better Design: Engineers can now design systems where the "engine" and the "suspension" are optimized separately. They don't have to fight each other.
3. Real-World Proof: The authors tested this with simulations and showed that even if their software isn't 100% perfect at guessing the distortion (which happens in real life), the system still works great.

Summary in One Sentence

This paper proves that for radar systems that transmit and receive in the same spot, distortion is not a mystery to be feared, but a known variable to be managed, allowing us to build cheaper, more efficient 6G systems where communication and sensing don't have to compromise each other.

Technical Summary

Here is a detailed technical summary of the paper "Distortion Is Not Noise: On the Limits of the Kappa Model for Monostatic ISAC."

1. Problem Statement

Integrated Sensing and Communication (ISAC) systems face significant performance degradation due to radio-frequency (RF) hardware impairments, specifically Power Amplifier (PA) nonlinearity and Oscillator Phase Noise (PN).

• The Status Quo: The dominant analytical approach uses the aggregate $\kappa$ distortion model (Bussgang decomposition). This model treats hardware distortion as additive, unknown noise ($\eta_t$) at the receiver.
• The Flaw: This "distortion-as-noise" assumption is valid for communication (where the receiver does not know the transmitted waveform) but is fundamentally pessimistic and incorrect for monostatic sensing. In a monostatic ISAC system, the transmitter and receiver are co-located; the receiver knows the transmitted symbols ($X$) and the PA characteristics, meaning the distorted waveform ($Z$) is a deterministic, known quantity, not random noise.
• The Gap: Existing literature fails to systematically compare the $\kappa$ model against physics-based models for sensing, leading to overestimated performance degradation and suboptimal system design.

2. Methodology

The authors propose a physics-based analytical framework that explicitly models the known nature of the distorted waveform in monostatic sensing.

• System Model: A monostatic OFDM-ISAC system where the transmitter applies a memoryless PA (modeled via the Rapp model) and introduces Phase Noise (modeled as a Wiener process with Common Phase Error, CPE).
• Key Assumption: The sensing receiver has access to the distorted waveform $Z[k,m]$ (via digital pre-distortion feedback or direct knowledge of the PA input/output), whereas the communication receiver does not.
• Analytical Tools:
  • Cramér–Rao Bound (CRB): Derived using the Slepian–Bangs formula for PA analysis and the Augmented Fisher Information Matrix (FIM) with Schur complements for Phase Noise analysis.
  • Comparison: The derived "Physics-Based CRB" is compared against the standard "Bussgang/$\kappa$ Model CRB" to quantify the overestimation gap.
  • Validation: Results are validated via Monte Carlo (MC) simulations using grid-search estimation.

3. Key Contributions

A. PA-Aware Sensing CRB (Theorem 1 & 2)

• Derivation: The authors derive a CRB for delay and Doppler estimation that accounts for PA nonlinearity while treating the distorted signal as known.
• Finding 1 (Low Degradation): The actual sensing degradation due to PA nonlinearity is SNR-independent and small (< 1.1 dB for typical Input Back-Off). This is because PA spectral regrowth adds high-frequency content that partially compensates for signal compression.
• Finding 2 ($\kappa$ Model Failure): The $\kappa$ model overestimates sensing degradation. The error ratio grows linearly with SNR, creating a "false sensing floor" that does not exist in reality. At high SNR, the $\kappa$ model can overestimate degradation by factors exceeding 100x.

B. PN-Aware Doppler CRB (Theorem 3)

• Derivation: Using an augmented FIM approach, the authors analyze the impact of Phase Noise.
• Finding 1 (Delay): Delay estimation is unaffected by Phase Noise under symmetric spectra (coupling term vanishes).
• Finding 2 (Velocity Floor): Doppler (velocity) estimation suffers from an irreducible error floor at high SNR. This floor scales with the square root of the oscillator linewidth ($\sqrt{\beta}$).
• Significance: This velocity floor is invisible to the $\kappa$ model, which incorrectly predicts that increasing SNR will indefinitely improve velocity estimation.

C. Orthogonal Design Space (Corollary 1)

• Separability: The joint PA+PN CRB reveals that the design space for ISAC is orthogonal:
  • PA Linearity (IBO) primarily controls Communication Rate and has negligible impact on sensing (except for minor spectral moment changes).
  • Oscillator Quality ($\beta$) primarily controls Sensing Velocity Accuracy and has negligible impact on communication.
• Implication: System designers can optimize PA and Oscillator specifications independently for their respective tasks.

4. Key Results

• Overestimation Gap: Simulations show that at 30 dB SNR, the $\kappa$ model overestimates sensing degradation by 12 dB (for IBO=3dB) compared to the physics-based truth. The gap widens as SNR increases.
• Velocity Floor: For a typical oscillator linewidth ($\beta=100$ Hz), the velocity estimation error saturates at ~1.4 m/s at high SNR, regardless of how much transmit power is added. This exceeds the 0.5 m/s requirement for automotive applications, highlighting a critical hardware bottleneck.
• DPD Robustness: The assumption that the distorted waveform is "known" remains robust even with practical Digital Pre-Distortion (DPD) template errors. An NMSE of -25 dB (typical) results in less than 1 dB overhead in range CRB.
• Pareto Frontier: The true sensing-communication trade-off curve (Pareto frontier) is significantly better than predicted by the $\kappa$ model. The gap widens at higher SNRs, proving that the $\kappa$ model leads to overly conservative system designs.

5. Significance and Impact

• Paradigm Shift: The paper fundamentally challenges the "distortion-as-noise" dogma in ISAC, proving that for monostatic sensing, distortion is a deterministic signal that can be perfectly accounted for.
• Design Optimization: It provides a clear roadmap for hardware selection:
  • Invest in high-quality oscillators (low $\beta$) for sensing performance.
  • Invest in linear PAs (high IBO) for communication performance.
  • Avoid over-designing PAs for sensing, as the returns diminish rapidly due to the irrelevance of PA nonlinearity to sensing accuracy (once the waveform is known).
• Resource Efficiency: By correcting the pessimistic $\kappa$ model, the paper suggests that ISAC systems can achieve target sensing accuracies with lower transmit power or less aggressive hardware linearization, leading to more energy-efficient 6G systems.

In conclusion, the paper demonstrates that the aggregate $\kappa$ model is inadequate for monostatic ISAC analysis. A physics-based approach reveals that hardware impairments affect sensing and communication asymmetrically, enabling an orthogonal design strategy that maximizes joint performance.