{"id":"22172a60-7a8e-40bd-bb94-4d1134d59b01","arxiv_id":"2607.08068","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":3,"one_line_summary":"Closed-form detection probabilities for multi-band OFDM ISAC signals are derived via characteristic functions of i.n.i.d. exponential variables, and an ADMM-based resource allocator achieves 18 dB detection gain over single-band baselines.","lead":"This paper derives closed-form detection and false-alarm probabilities for multi-band OFDM radar signals used in integrated sensing and communication systems, and proposes an optimization scheme for allocating power and time-frequency resources across bands. It matters because multi-band sensing is a practical approach to overcoming spectrum fragmentation in 6G networks, and the analytical framework enables principled resource allocation rather than heuristic tuning.","discovery_kind":"unclear","skeptic_critique":{"model":"glm-5.2","headline":"The 18 dB gain claim rests on a single-band baseline that randomly selects among bands with ~20 dB path loss variation, making the comparison misleading rather than a true test of frequency diversity or optimization.","rationale":"The reader correctly identifies the analytical core as sound: the GLRT-based derivation, CF factorization, and closed-form expressions (Eqs. 14, 18) are mathematically correct for the i.n.i.d. exponential sum under the stated assumptions. The reader's primary concern about inter-band independence (Eq. 3) is reasonable in principle but overstated for the specific bands used — {2.6, 3.5, 26, 28} GHz have frequency gaps of 0.9–23.4 GHz, far exceeding any realistic target coherence bandwidth, making the Bessel-function correlation in Eq. 2 vanish as the paper argues. The more impactful concern is the single-band baseline. The paper itself reports 26–30 dB variance for single-band detection, which directly evidences that band selection dominates single-band performance. With ~20 dB path loss variation across the four bands, random band selection creates a strawman baseline. The 18 dB headline gain conflates the path-loss advantage of multi-band access with the genuine contributions (diversity + optimization). The ADMM non-convexity concern (Step 1, Eq. 23 solved via SQP without global optimality guarantee) is valid but standard for this type of problem and empirically validated. The verdict remains CONDITIONAL: the analytical contributions are real and novel, but the headline performance claim needs a fair baseline comparison to be credible. If the best-case single-band comparison shows the gain is mostly from path-loss disparity, the paper's framing should be adjusted to emphasize the 6 dB optimization gain and the robustness (variance reduction) benefits, which are the genuine and valuable contributions.","tokens_in":11640,"tokens_out":4800,"duration_ms":243912,"concrete_test":"Recompute the Fig. 4 comparison with a single-band baseline that concentrates all power and time-frequency resources at 2.6 GHz (the lowest-path-loss band) instead of using random band selection. If the optimized multi-band gain over this best-case single-band drops below ~6 dB, the 18 dB headline claim is primarily an artifact of the weak random-band baseline rather than a result of frequency diversity or joint optimization.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The single-band baseline (Section V-B, Appendix B) selects a random band s ∈ {2.6, 3.5, 26, 28} GHz and concentrates all resources there. Since path loss L_b ∝ λ²_b ∝ 1/f²_b (Eq. 1), the SNR difference between 2.6 GHz and 28 GHz is approximately (28/2.6)² ≈ 116, i.e., ~20.6 dB. The single-band performance thus varies enormously depending on which band is randomly selected — and this is confirmed by the paper's own variance results: 30 dB variance for single-band in Fig. 2 and 26 dB in Fig. 4. The mean single-band performance is dragged down by the poor mmWave bands (26, 28 GHz). The claimed 18 dB gain at 90% detection (Fig. 4) therefore likely reflects this path loss disparity — the multi-band system can access the favorable 2.6 GHz band while the single-band baseline is stuck with whatever band it randomly draws — rather than the frequency diversity and resource optimization that are the paper's actual contributions. The 6 dB gain of optimized multi-band over uniform multi-band is the more meaningful number, as it isolates the optimization benefit. The inter-band independence assumption (Eq. 3) that the reader flagged is well-justified for bands separated by GHz-level gaps and is not the primary vulnerability; the baseline comparison is.","agreement_with_reader":"partial"},"referee_report":{"model":"glm-5.2","summary":"This paper derives closed-form detection and false alarm probabilities for multi-band OFDM ISAC signals via the generalized likelihood ratio test (GLRT) and the characteristic function (CF) method for sums of independent non-identically distributed (i.n.i.d.) exponential random variables. Building on these expressions, the authors formulate a joint power and time-frequency resource allocation problem to maximize detection probability and solve it using an ADMM-based algorithm. Simulations validate the theoretical derivations and demonstrate detection gains for multi-band signals over single-band baselines.","tokens_in":11826,"tokens_out":1072,"duration_ms":132959,"significance":"The derivation of closed-form detection performance for multi-band OFDM ISAC is a solid contribution. The mathematical treatment of the GLRT statistic as a sum of i.n.i.d. exponential variables and the subsequent use of partial fraction expansion to obtain tractable closed-form expressions (Eqs. 11-18) is rigorous and provides a useful theoretical foundation. The deflection-coefficient optimality proof for the weights (Appendix A) is clean. The proposed ADMM-based resource allocation scheme is a practical approach to exploiting the non-linear characteristics of the derived detection probability.","major_comments":[{"comment":"Section V-C, Fig. 4: The claim of an '18 dB detection gain over traditional single-band baselines' is misleading due to the construction of the single-band baseline. As detailed in Appendix B, the single-band baseline selects a random band $s$ from the set $B$ and concentrates all resources there. Because path loss $L_b$ scales with $1/f_b^2$ (Eq. 1), the SNR varies drastically across the chosen bands (2.6, 3.5, 26, 28 GHz). A random draw that lands on 26 or 28 GHz will suffer enormous path loss compared to 2.6 GHz. The paper's own variance results (26-30 dB for single-band) confirm that the single-band mean is heavily dragged down by these high-frequency draws. Consequently, the 18 dB gain largely reflects the multi-band system's ability to access the favorable sub-6 GHz bands while the baseline cannot, rather than isolating the benefits of frequency diversity or the proposed resource优化","section":null},{"comment":"Section V-C, Fig. 4: The comparison between optimized multi-band and uniform multi-band (6 dB gain) is the more meaningful metric for evaluating the proposed ADMM optimization scheme, as it holds the available frequency bands constant. The authors should reframe the primary contribution and abstract to emphasize this 6 dB optimization gain, rather than the 18 dB gain which is heavily conflated with path loss disparities.","section":null},{"comment":"Section IV-B, Eq. (20)-(23): The ADMM update for $E_b$ in Step 1 (Eq. 23) is noted to be a non-convex problem solved via sequential quadratic programming (SQP). Because the overall problem (P2) is non-convex, the standard convergence guarantees of ADMM do not apply. The manuscript should briefly discuss the theoretical or empirical convergence behavior of alternating between SQP for the non-convex subproblem and CVX for the convex subproblems, as this impacts the reliability of the optimization results.","section":null}],"minor_comments":[{"comment":"Section II-B, Eq. (2): The correlation coefficient $ho_{i,j}$ involves Bessel functions $J_0$ and $J_2$. It would help the reader to briefly state the physical intuition behind the transition from this continuous correlation model to the binary intra-band/independent inter-band assumptions.","section":null},{"comment":"Section IV-A, Eq. (19): The notation $S_b = M_b N_b$ is introduced, but it is not explicitly defined that $M_b$ corresponds to $M^b_{sym}$ and $N_b$ to $N^b_c$ from Section II. Please clarify.","section":null},{"comment":"Section V: The simulation parameters mention 'average target RCS $[-20, 0]$ dBsm'. It is unclear if this means a uniform distribution across this range or a specific mean value within it. Please specify the distribution used for the Swerling II model.","section":null},{"comment":"Figures 2 and 4: The axis labels and legends are somewhat sparse. For instance, Fig. 2 lacks explicit axis titles in the provided text. Ensure all figures have clearly labeled axes, units, and legends for standalone readability.","section":null}],"recommendation":"minor_revision","confidential_remarks":"The stress-test note correctly identifies the baseline comparison as the primary vulnerability. The 18 dB gain is a red herring because the single-band baseline is penalized by random path loss draws. The authors need to address this framing issue, but the core mathematical derivations (Eqs. 11-18) and the 6 dB optimization gain over uniform multi-band are sound and represent a valid contribution. Minor revision is appropriate to fix the framing of the gains and clarify the ADMM convergence."},"author_rebuttal":null,"desk_editor":{"model":"glm-5.2","letter":"The real contribution here is the closed-form detection and false alarm probabilities for multi-band OFDM ISAC (Eqs. 14, 18). The characteristic function method for sums of i.n.i.d. exponentials is correctly applied, the partial fraction expansion is clean, and the deflection-coefficient optimality proof for the weights (Appendix A) is tight. This is a genuine analytical result that I haven't seen in the prior multi-band ISAC literature, which has focused on CRB analysis and signal processing rather than detection probability expressions. The GLRT framework and the inter-band independence assumption are well-motivated — for bands separated by GHz-level gaps, the coherence bandwidth argument is physically sound and not the weak point. The reader flagged independence as load-bearing, but I think that concern is overstated; the approximation is standard and defensible for non-contiguous CA bands. The Monte Carlo validation in Fig. 1 matches the closed forms almost perfectly, which is good evidence the math is right. The ADMM-based resource allocation is reasonable but has a real gap: Step 1 (Eq. 23) is non-convex, solved via SQP with no global optimality guarantee, and convergence is shown only empirically for three noise levels. That's a soft spot but not fatal — the algorithm clearly works in practice, and the 6 dB gain of optimized over uniform multi-band (Fig. 4) is the meaningful number that isolates the optimization benefit. The 18 dB gain over single-band is where I part ways with the paper. The stress-test note is correct: the single-band baseline randomly selects among {2.6, 3.5, 26, 28} GHz, and since path loss goes as 1/f², the SNR spread between 2.6 and 28 GHz is roughly 20 dB. The paper's own variance numbers (30 dB for single-band in Fig. 2, 26 dB in Fig. 4) confirm this. So the 18 dB gain largely reflects the multi-band system's ability to access the favorable sub-6 GHz band while the single-band baseline is stuck with whatever it randomly draws — not frequency diversity or optimization per se. The paper should have used a best-single-band baseline (all resources on 2.6 GHz) to make a fair comparison. No code or data is shipped, which limits reproducibility of the optimization results specifically. The closed-form expressions themselves are parameter-free and independently verifiable, which is the stronger part of the work. This paper is for ISAC researchers who need analytical detection performance expressions for multi-band systems. The analytical core deserves a serious referee; the optimization and simulation claims need tightening before publication.","headline":"Closed-form multi-band OFDM detection expressions are a genuine new result; the 18 dB gain claim is inflated by a misleading single-band baseline.","tokens_in":12702,"tokens_out":624,"would_cite":true,"duration_ms":88866,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"glm-5.2","headline":"18 dB detection gain from multi-band radar-communication signals","keywords":[],"falsifier":"If the inter-band scattering correlation is non-negligible for realistic targets or frequency configurations, the characteristic function factorization in Eq. 11 fails, the closed-form detection probability in Eq. 18 no longer holds, and the optimization objective becomes invalid.","tokens_in":11865,"feed_emoji":"📡","tokens_out":887,"duration_ms":159972,"temperature":0.7,"pith_summary":"This paper argues that splitting a radar-sensing signal across multiple non-contiguous frequency bands, rather than concentrating it in one band, yields large detection gains for weak targets. The central mechanism is frequency diversity: when the frequency gap between bands exceeds the target's coherence bandwidth, the radar cross-section fluctuations at each band become statistically independent. Under this independence assumption, the paper derives exact closed-form expressions for the detection and false-alarm probabilities of multi-band OFDM signals by factoring the characteristic function of the sum of independent-but-not-identically-distributed exponential random variables. It then uses these expressions as an optimization objective, jointly allocating transmit power and time-frequency resources across bands via an ADMM-based algorithm to maximize detection probability. The paper reports an 18 dB detection gain over single-band baselines at 90% detection probability.","feed_headline":"18 dB detection gain from multi-band radar-communication signals","feed_subtitle":"Splitting sensing across non-contiguous bands and jointly optimizing resources yields exact detection formulas and large gains over single-b","key_machinery":"The characteristic function (CF) method applied to a sum of independent-but-not-identically-distributed (i.n.i.d.) exponential random variables. Each band contributes a weighted exponential detection statistic whose rate parameter depends on the per-band SNR. The CF of the joint statistic factorizes into a product of per-band CFs (Eq. 11, Eq. 16), which is then inverted via partial fraction expansion to yield closed-form PDFs and integration to CDFs. The ADMM algorithm splits the non-convex resource allocation problem into three subproblems (auxiliary variable, time-frequency resource, power) solved iteratively with SQP and CVX.","core_discovery":"The detection and false-alarm probabilities of a multi-band OFDM ISAC system with a Swerling II target can be expressed in exact closed form as sums of weighted exponentials (Eq. 14 and Eq. 18), provided the inter-band scattering coefficients are statistically independent. This independence holds when the frequency gap between non-contiguous bands far exceeds the target's coherence bandwidth. The optimal combining weight for each band is the per-band average SNR, which maximizes the deflection coefficient. Using these closed forms as an optimization objective, a joint power and time-frequency resource allocation scheme solved by ADMM achieves a reported 18 dB detection gain over single-band,","pith_inferences":["If inter-band correlation is non-negligible (e.g., for targets with large projected lengths or small bistatic angles), the CF factorization breaks down and the closed-form expressions would require modification to account for correlated exponential components, potentially reducing the diversity gain.","The framework could be extended to multi-target or MIMO scenarios, but the independence assumption and the GLRT grid search would need re-examination, as multi-target interference and spatial correlation would introduce additional coupling between bands."],"forward_implications":["Multi-band ISAC systems using carrier aggregation can achieve substantially higher weak-target detection sensitivity than single-band systems with equivalent total resources, supporting the case for fragmented spectrum as an asset rather than a liability.","The closed-form expressions enable rapid resource optimization without Monte Carlo simulation, making real-time adaptive sensing frame design tractable.","The 18 dB gain claim, if validated, suggests that dynamic resource allocation across heterogeneous bands (e.g., sub-6 GHz and mmWave) can compensate for frequency-selective fading that would otherwise degrade single-band detection."],"fun_headline_variants":["Exact detection formulas for multi-band radar-communication signals","Multi-band OFDM sensing yields 18 dB gain via closed-form optimization","Closed-form multi-band ISAC detection enables 18 dB gain over single-band","ADMM-based multi-band resource allocation achieves 18 dB detection gain"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The inter-band statistical independence assumption — that radar cross-section fluctuations at different non-contiguous bands are uncorrelated — is the load-bearing premise. The entire characteristic function factorization and resulting closed-form expressions depend on it. The paper justifies this via a coherence bandwidth argument, but its validity depends on target geometry, bistatic angle, and specific frequency gaps.","fun_headline_variants_meta":{"raw":{"variants":["Exact detection formulas for multi-band radar-communication signals","Multi-band OFDM sensing yields 18 dB gain via closed-form optimization","Closed-form multi-band ISAC detection enables 18 dB gain over single-band","ADMM-based multi-band resource allocation achieves 18 dB detection gain"]},"model":"glm-5.2","effort":"low","cost_usd":0.0,"raw_usage":{"total_tokens":598,"prompt_tokens":538,"completion_tokens":60,"prompt_tokens_details":null},"tokens_in":538,"tokens_out":60,"duration_ms":22269,"temperature":1.0,"reasoning_tokens":null,"cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-10T00:36:39.135086+00:00","model_set":{"reader":"glm-5.2"},"falsifier":"If the inter-band scattering correlation is non-negligible for realistic targets or frequency configurations, the characteristic function factorization in Eq. 11 fails, the closed-form detection probability in Eq. 18 no longer holds, and the optimization objective becomes invalid.","supporting_citations":[],"review_version":1}