{"id":"e1f16225-063f-4a6a-bc89-96519398eeb8","arxiv_id":"1908.01357","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Exact BER expressions for downlink NOMA with imperfect SIC over Nakagami-m fading are derived for N=2 and N=3, with closed forms for Rayleigh fading, and are applied to optimal power allocation.","lead":"This paper derives exact bit error rate formulas for two- and three-user downlink NOMA systems with imperfect successive interference cancellation over Nakagami-m fading, and it uses those formulas to find optimal power allocations. It matters because prior exact BER results were limited to Rayleigh fading or two users, so NOMA designers lacked an exact tool for a general fading model.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Rayleigh closed-form BER expressions (39), (85)-(87) are not valid probabilities: they exceed 1 or become negative at finite SNR, contradicting the paper's central claim of exact closed-form Rayleigh BER.","rationale":"The reader identified the i.i.d. ordering assumption as the weakest point, and that is indeed a real scope limitation. However, a more decisive defect exists in the Rayleigh closed forms, which are a central advertised contribution. Equations (39), (85), (86), and (87) produce BER values outside the unit interval at zero or asymptotically large SNR, and they are internally inconsistent with the conditional BER expressions (35) and (44) from which they are supposed to be averaged. Because the paper explicitly claims these closed forms are exact and uses them in the numerical/simulation section, this is a load-bearing mathematical error, not a matter of interpretation or consensus. The reader's independent check apparently did not cover all four closed forms, or would not have missed this. The general m expressions may be salvageable, but the manuscript as written cannot be accepted; the closed-form Rayleigh results and the corresponding figures/tables must be corrected or removed.","tokens_in":25248,"tokens_out":28922,"duration_ms":254943,"concrete_test":"Evaluate (39) with all γ_{2,c}=Γ and let Γ→∞; the result is Σv_c=2, which is not a valid probability. Similarly, set Γ=0 in (85) and (86) to get 1, and in (87) take Γ→∞ to get -7/8. For a direct numerical check, average the conditional BER (35) over the Rayleigh order-statistic PDF for the stronger of two i.i.d. channels, f(x)=2(e^{-x}-e^{-2x}) for x=α^2, and compare the resulting BER curve with the values produced by (39) at the same SNRs; the two will not agree.","verdict_should_be":"REJECT","load_bearing_attack":"The paper advertises closed-form exact BER for m=1 as a core contribution, but the printed Rayleigh expressions are internally inconsistent. In (39), the weights sum to v=[2,1,-1,-1,1], i.e., Σv_c=2, and each bracket tends to 2 as its argument grows; hence with all γ_{2,c}=Γ and Γ→∞, the expression tends to (1/2)·2·2=2, which is impossible for a BER. At Γ=0, (39) gives 1, while the conditional BER in (35) at zero SNR is 1/2. Similarly, (85) gives 1 at zero SNR, (86) gives 1 at zero SNR, and (87) gives -7/8 as Γ→∞ because its bracket tends to -1/2 and its weights sum to 7. These are not mere scaling conventions: an average of Q(√γ) over any fading distribution must lie in [0,1/2] for Gray-coded QPSK bits at zero SNR, and must stay within [0,1] everywhere. The simulation claims in Sec. VI that cite (38),(39),(85)-(87) cannot be correct as written. This invalidates Contribution 2 and the Rayleigh-closed-form part of the central claim, independent of the i.i.d. channel limitation identified by the reader.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes the bit error rate (BER) of downlink power-domain NOMA with imperfect successive interference cancellation over Nakagami-m flat fading channels, for two-user and three-user scenarios. Exact conditional BER expressions are derived by averaging over the QPSK symbol combinations and the SIC error events, and the average BER is obtained by averaging over ordered Nakagami-m statistics. For the special case m=1 (Rayleigh fading), closed-form average BER expressions are claimed for all users. The derived BER expressions are then used to formulate two optimal power allocation problems, one minimizing the average BER and one achieving equal BER among users. Numerical results and Monte Carlo simulations are presented for various m and power allocations, with claims of perfect agreement between analysis and simulation.","tokens_in":25469,"tokens_out":15337,"duration_ms":140440,"significance":"If correct, the paper would provide the first exact BER analysis of downlink NOMA with imperfect SIC over Nakagami-m fading, including closed forms for Rayleigh fading, and would enable BER-aware optimal power allocation. The conditional-error-probability derivation is a substantial combinatorial exercise, and the general-m averaging framework using ordered Nakagami-m statistics is a useful contribution. However, the claimed Rayleigh closed forms contain serious algebraic errors that make them invalid as probabilities, so the central contribution is not currently established. The paper also silently assumes i.i.d. channel statistics, which limits the scope of the claims.","major_comments":[{"comment":"The Rayleigh closed-form expressions (39), (85), (86), and (87) are not valid BER probabilities. For (39), if all γ_{2,c}=Γ and Γ→0, each bracket tends to 1, so the expression tends to (1/2)∑_{c=1}^5 v_c = 1; but the conditional BER in (35) at zero SNR is (1/2)∑ v_c Q(0) = 1/2. Similarly, (85) tends to 1 as Γ→0, whereas the conditional expression (44) gives 1/2; (86) also tends to 1 as Γ→0; and (87) tends to (1/4)·7·(-1/2) = -7/8 as Γ→∞ because ∑ v_c = 7 and the bracket tends to -1/2. A BER must lie in [0,1] and, for Gray-coded QPSK at zero SNR, must equal 1/2. Thus these closed forms are internally inconsistent with the conditional expressions from which they are supposedly derived, and the simulation matches claimed in Sec. VI for (39),(85)-(87) cannot be correct as stated.","section":"Sec. III.C, Eq. (39); Sec. IV.D, Eqs. (85)-(87)"},{"comment":"The ordered channel PDF in (93) is derived under the assumption that the N channel gains are i.i.d. Nakagami-m random variables, which is why it factors as f(α)[F(α)]^{n-1}[1-F(α)]^{N-n}. The system model in Sec. II does not state this i.i.d. assumption, and in a downlink NOMA setting users generally have different average powers (different Ω_n). For non-identically distributed Nakagami-m gains, the order-statistics PDF does not have this simple product form, so the average BER expressions (36)-(37), (82)-(84), and (101) do not apply to the general downlink NOMA scenario. The paper should either explicitly restrict all claims and simulations to the i.i.d. case or extend the analysis to non-identical fading parameters.","section":"Appendix I, Eq. (93); Sec. VI"},{"comment":"The simulation validation is not reproducible from the information given: no number of Monte Carlo trials, no description of how the ordered channels are generated when users have different average SNRs, and no code or pseudocode for the SIC detector are provided. Moreover, the claimed 'perfect match' between the analytical curves (39),(85)-(87) and simulation cannot be true given the invalid probability limits identified above. The figures and the optimization tables in Sec. V must be regenerated with corrected closed forms, and the simulation methodology should be described in enough detail to allow independent verification.","section":"Sec. VI, Figs. 6-12"}],"minor_comments":[{"comment":"The caption says 'N = 2' but the text and the figure refer to the three-user scenario; this should be 'N = 3'.","section":"Fig. 7 caption"},{"comment":"The subscript in γ_{1,2} is written as α_{2n}; it should be α_1 (the channel gain of the first user).","section":"Eq. (11)"},{"comment":"The function Φ(a,z) is called the lower incomplete gamma function in (92) but is described as the upper incomplete Gamma function in the surrounding text; the terminology should be made consistent.","section":"Appendix I, Eq. (92)"},{"comment":"The infinite series representation of [Φ(m,·)]^μ uses coefficients S_i that depend on μ, but this dependence is not indicated in (99), where S_i appears for each k without a subscript denoting the exponent n+k-1; the notation should be clarified.","section":"Appendix I, Eqs. (94)-(95)"},{"comment":"The phrase 'perfectly match' should be replaced with a quantitative statement of agreement (e.g., maximum deviation or confidence bounds), since no simulation parameters or error bars are given.","section":"Sec. VI"}],"recommendation":"major_revision","confidential_remarks":"The central claimed contribution includes closed-form Rayleigh BER expressions that are demonstrably invalid as probabilities (they exceed 1 or become negative at certain SNRs). This is a load-bearing error affecting Contribution 2, the simulation comparisons, and the power-allocation results. The error appears fixable by re-deriving the Rayleigh averages from the conditional expressions, but the authors must also address the unstated i.i.d. assumption and regenerate the numerical results. I therefore recommend major revision rather than rejection, provided the corrected analysis remains within the paper's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline: the paper's central claim—exact BER for N=3 downlink NOMA over Nakagami-m with imperfect SIC—does not survive contact with the numbers. The conditional BER expressions for the second and third users in the three-user case are not normalized: at zero SNR, where every Q(·) equals 1/2, the U2 expression (67) evaluates to 1 and the U3 expression (81) to 7/8. A valid BER for Gray-coded QPSK must be 1/2 at zero SNR. The Rayleigh closed-forms (39), (85)–(87) inherit these errors; (85) gives 1 at zero SNR and (87) goes negative at high SNR. These are not typos in the averaging step—the underlying case counting has the wrong weights.\n\nWhat's genuinely good: the systematic enumeration is clearly laid out, the literature review is fair, and the N=2 first-user expression (38) survives an independent check. The idea of using exact BER as the objective for power allocation is a sensible one.\n\nThe soft spots beyond the normalization: no code or data are shipped, and the claimed perfect match with Monte Carlo is only shown in figures. The ordered-statistics averaging assumes i.i.d. Nakagami links, so users with different average powers are outside the model—a real limitation the authors do not state. The optimization tables also contain at least one constraint violation (Table III, N=2, Eb/N0=0 sums to 1.024).\n\nProportionally: this is a load-bearing flaw, not a cosmetic one. The paper as written cannot be used as an exact BER tool. The method is salvageable; someone who redoes the case counting for N=3 could likely produce correct expressions, and the N=2 U1 part may be correct.\n\nWho is this for? NOMA performance analysts who want exact expressions. As is, they should not trust the formulas without re-derivation. I would not cite it in its current form.\n\nRecommendation: Send it to peer review anyway. The topic is of interest, the errors are concrete and fixable, and a good referee can identify what needs to change. But it would need major revision before acceptance.","headline":"The exact BER expressions for the three-user case and the Rayleigh closed-forms violate probability bounds (BER > 1/2 at zero SNR, negative at high SNR), so the central claim fails as written despite a systematic derivation.","tokens_in":25991,"tokens_out":7703,"would_cite":false,"duration_ms":65894,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["94A12","62G30"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper derives exact bit-error-rate expressions for two- and three-user downlink NOMA over Nakagami-m fading with imperfect successive interference cancellation.","keywords":["NOMA","bit error rate","successive interference cancellation","Nakagami-m fading","optimal power allocation","fairness","QPSK","order statistics"],"falsifier":"Run a two- or three-user Monte Carlo BER test with strongly asymmetric average channel powers, such as $\\Omega_1=1$, $\\Omega_2=5$, and $\\Omega_3=25$, at a fixed SNR and power split, and compare with (36)--(37) or (82)--(87): a clear mismatch would show that the i.i.d.-power ordering assumption, not the conditional Q-function derivation, is the limit of the claim.","tokens_in":25059,"feed_emoji":"📶","tokens_out":9785,"duration_ms":94132,"temperature":0.7,"pith_summary":"This paper tries to put the bit error rate (BER) of power-domain downlink NOMA on an exact footing. It derives per-user BER expressions for two- and three-user systems that use Gray-coded QPSK over Nakagami-$m$ flat fading, and it does so under imperfect successive interference cancellation, where an earlier user's decision error leaves residual interference. The conditional error probability of each user is shown to be a short weighted sum of Gaussian Q functions, and averaging over the ordered fading statistics yields the BER formulas. The paper then uses those formulas to find optimal power coefficients for two goals: equal BER across users and minimum average BER. If the derivation is correct, system designers can predict each user's exact reliability and choose power splits for constrained-power applications such as satellite links.","feed_headline":"Exact BER formulas for downlink NOMA in Nakagami fading","feed_subtitle":"Two- and three-user downlink NOMA gets exact per-user error rates, with closed forms for Rayleigh fading.","key_machinery":"The argument is carried by the superimposed QPSK constellation and by the ordered Nakagami-$m$ order-statistic density. Each transmitted superposition is one of 16 points for $N=2$ and 64 points for $N=3$, and every error event reduces to the probability that a Gaussian noise component crosses one of the amplitude levels $A_{u_1u_2u_3}=u_1\\sqrt{\\beta_1}+u_2\\sqrt{\\beta_2}+u_3\\sqrt{\\beta_3}$, $u_i\\in\\{0,1,-1,2\\}$, so the conditional error rate is a Q function $Q(\\sqrt{\\gamma})$. The averaging step uses the ordered PDF in (93), which contains powers of the lower incomplete gamma function; the infinite-series expansion (94)--(95) and binomial expansion (96) turn it into a gamma-like density, and Craig's integral representation of the Q function together with the identity $\\int_0^\\infty x^t e^{-bx}\\,dx = t!/b^{t+1}$ produces the final average-BER expressions.","core_discovery":"The central claim is that the exact BER of each user in a downlink power-domain NOMA system with QPSK and imperfect SIC is a signed sum of Gaussian Q functions whose arguments are functions of the power coefficients. For $N=2$, the second user's conditional BER is a five-term sum with coefficient vector $[2,1,-1,-1,1]$; for $N=3$, the second user needs ten terms and the third user eighteen terms. The conditional sums are then averaged over the ordered Nakagami-$m$ channel statistics, giving the expressions (36), (37), and (82)--(84), and for the Rayleigh special case $m=1$ the closed forms (38), (39), and (85)--(87). Throughout, \"imperfect SIC\" means the receiver explicitly accounts for the possibility that a previously decoded user's symbol is wrong, so the resulting terms reflect both correct and incorrect SIC outcomes. The paper reports that these analytical results match Monte Carlo simulation closely, and that the same expressions allow power allocation either to equalize BER among users or to minimize the average BER.","pith_inferences":["The same conditioning recipe suggests a general structural pattern: for any $N$-user system the per-user conditional BER should be a signed sum of Q functions over constellation amplitudes, with coefficients fixed by the combinatorics of SIC error events; the two- and three-user formulas are then instances of one larger family.","Because the average-BER formulas use an ordered density built from identical link distributions, a natural extension is to non-identically distributed Nakagami links, where the users have different average powers; that would require an order-statistic density for non-i.i.d. channels and would likely change the power-allocation tables.","The near-monopoly of power assigned to the weakest user under fairness suggests that in practice a NOMA system might combine the exact BER expressions with adaptive modulation to avoid extreme power splits.","The closed-form Rayleigh results could be inverted to give, for a target BER, the required power split as a function of SNR; such an explicit design rule is not written out in the paper but follows directly from (38), (39), (85)--(87)."],"forward_implications":["At $m=1$ (Rayleigh fading), the BER formulas become closed forms, so per-user error rates and power-optimization costs can be evaluated without numerical integration.","The fairness-optimal power allocation gives almost all the power to the weakest user at high signal-to-noise ratio (over 98% for $N=2$ at 30 dB), so equalizing BER is expensive and requires accurate knowledge of operating SNR.","The exact results differ from the union-bound approximation by up to about 3 dB in the low-SNR regime, so designers using the bound can underestimate the power needed to hit a target BER.","Imperfect SIC has a large effect on the farthest user in the three-user case; assuming perfect SIC overstates reliability most where SIC errors are common."],"supporting_citations":[{"why":"Supplies the union-bound BER for downlink NOMA over Nakagami-$m$ fading that the exact expressions are compared against and improve upon.","marker":"[21]"},{"why":"Prior closed-form symbol error rate for NOMA with imperfect SIC, limited to two users and Rayleigh; the paper extends to exact BER in Nakagami-$m$.","marker":"[23]"},{"why":"Prior two-user downlink and uplink NOMA BER in i.i.d. Rayleigh fading; the paper generalizes to ordered Nakagami-$m$ and three users.","marker":"[27]"},{"why":"Provides the order-statistics formula used to write the ordered channel-gain PDF for the $n$-th user.","marker":"[30]"},{"why":"Supplies the infinite-series representation of the powered lower incomplete gamma function that makes the averaging integrals tractable.","marker":"[32]"},{"why":"Gives the integral identity $\\int_0^\\infty x^t e^{-bx}\\,dx=t!/b^{t+1}$ used to evaluate the average over the gamma-like ordered density.","marker":"[33]"},{"why":"Provides Craig's alternative representation of the Gaussian Q function used to turn conditional BER terms into integrals over the fading PDF.","marker":"[34]"}],"fun_headline_variants":["Exact BER for NOMA with imperfect SIC","Closed-form BER for NOMA over Nakagami fading","NOMA BER formulas enable optimal power split","Exact per-user BER for two- and three-user NOMA","Imperfect SIC BER analysis for NOMA downlink"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that all user links experience the same Nakagami-$m$ fading law with the same average power, so a single order-statistic formula can describe the ordering of weak to strong users; if the users sit at different distances with different average received powers, the averaging formulas stop applying.","fun_headline_variants_meta":{"raw":{"variants":["Exact BER for NOMA with imperfect SIC","Closed-form BER for NOMA over Nakagami fading","NOMA BER formulas enable optimal power split","Exact per-user BER for two- and three-user NOMA","Imperfect SIC BER analysis for NOMA downlink"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000222,"raw_usage":{"total_tokens":1457,"prompt_tokens":954,"completion_tokens":503,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":570,"completion_tokens_details":{"reasoning_tokens":424}},"tokens_in":570,"tokens_out":503,"duration_ms":4841,"temperature":1.0,"reasoning_tokens":424,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:16:10.469658+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a two- or three-user Monte Carlo BER test with strongly asymmetric average channel powers, such as $\\Omega_1=1$, $\\Omega_2=5$, and $\\Omega_3=25$, at a fixed SNR and power split, and compare with (36)--(37) or (82)--(87): a clear mismatch would show that the i.i.d.-power ordering assumption, not the conditional Q-function derivation, is the limit of the claim.","supporting_citations":[{"cited_title":"Error probability analysis of non-orthogonal multiple access over nakagami- m fading channels,","cited_arxiv_id":null,"evidence_quote":"Supplies the union-bound BER for downlink NOMA over Nakagami-$m$ fading that the exact expressions are compared against and improve upon."},{"cited_title":"Closed-form symbol error rate expressions for non-orthogonal multiple access systems,","cited_arxiv_id":null,"evidence_quote":"Prior closed-form symbol error rate for NOMA with imperfect SIC, limited to two users and Rayleigh; the paper extends to exact BER in Nakagami-$m$."},{"cited_title":"BER performances of downlink and uplink NOMA in the presence of SIC errors over fading channels,","cited_arxiv_id":null,"evidence_quote":"Prior two-user downlink and uplink NOMA BER in i.i.d. Rayleigh fading; the paper generalizes to ordered Nakagami-$m$ and three users."},{"cited_title":"On the error performance bound of ordered statistics decoding of linear block codes,","cited_arxiv_id":null,"evidence_quote":"Provides the order-statistics formula used to write the ordered channel-gain PDF for the $n$-th user."},{"cited_title":"Power allocation for robust distributed best-linear- unbiased estimation against sensing noise variance uncertainty,","cited_arxiv_id":null,"evidence_quote":"Supplies the infinite-series representation of the powered lower incomplete gamma function that makes the averaging integrals tractable."},{"cited_title":"Gradshteyn and I","cited_arxiv_id":null,"evidence_quote":"Gives the integral identity $\\int_0^\\infty x^t e^{-bx}\\,dx=t!/b^{t+1}$ used to evaluate the average over the gamma-like ordered density."},{"cited_title":"A new, simple and exact result for calculating the probability of error for two-dimensional signal constellations,","cited_arxiv_id":null,"evidence_quote":"Provides Craig's alternative representation of the Gaussian Q function used to turn conditional BER terms into integrals over the fading PDF."}],"review_version":1}