{"id":"ff557c14-d399-4641-ace2-b26ce05c57ba","arxiv_id":"2501.18235","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Cascading more amplifier stages improves the effective noise figure at a given output power, and AM-PM noise correlation is generally low, except for peaks near the 1 dB compression point in some devices.","lead":"This paper measures how the amplitude and phase noise of a chain of radio-frequency amplifiers changes as you add stages and drive the chain harder. It finds that for the same output power, a longer chain of amplifiers can actually have a better effective noise figure, and that the amplitude and phase noises are usually uncorrelated, with occasional correlation peaks near compression.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Main risk: at fixed phase-shifter setting, the DUT's AM-PM phase shift with input power rotates the I/Q measurement axes, so AM-to-PM coherence peaks near 1 dB compression and the NF comparison may be partially an artifact of quadrature crosstalk; paper does not show per-power re-calibration.","rationale":"Reader focused on cascade additivity; that assumption is standard and likely safe for small noise sidebands, so I did not adopt it as the strongest concern. The more fragile link is the IQ demodulator's quadrature accuracy under large-signal conditions. The paper's own AM-PM derivative data show the phase changes rapidly near the compression points where the coherence peaks appear, so even a small uncalibrated phase error can generate exactly the observed signature. Since the effective NF is derived from the same I/Q channels, the main Figure 9 comparison is also exposed. This is a testable measurement concern, not a fundamental impossibility. If the per-power re-calibration preserves the results, the paper's conclusions are strengthened; a conditional acceptance requiring this check is proportionate.","tokens_in":6519,"tokens_out":18141,"duration_ms":176857,"concrete_test":"Re-measure the one- and two-stage amplifiers of Figures 11-12 at every carrier input power after first tuning the phase shifter at that same power to re-null the quadrature (e.g., minimize Q-branch DC or maximize I-branch AM sensitivity), and compute the coherence and effective NF. Complement this with an injection test: add a calibrated PM tone at each power and measure its leakage into the I channel; if leakage exceeds a few percent near 1 dB compression, the reported AM-PM coherence peaks are at least partly instrumental. If the coherence peaks and the NF-vs-output-power ordering persist after per-power re-calibration, the central claims are supported; if they shift or vanish, the conclusion must be restricted.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central NF-improvement claim and the coherence-peak claim both depend on the IQ receiver actually separating AM and PM. That separation requires a 90° phase relation between the DUT and LO paths at each carrier power. The paper states only that the phase shifter 'must be adjusted to obtain proper settings' and cites [17]; it does not state that the quadrature is re-optimized at every input power. Figure 13 shows that the AM-PM derivative (and hence the DUT insertion phase versus power) peaks near 1 dB compression, which is exactly where Figures 11-12 show coherence peaks. If the phase shifter is left at a small-signal optimum, the true AM/PM axes rotate as power increases, so AM noise leaks into the PM channel and PM noise into the AM channel. The leakage is largest where dφ/dP is largest, producing artificial coherence peaks that coincide with the AM-PM derivative peaks. The same crosstalk biases the measured white-noise levels used to compute effective NF (Figures 8-9); because chains with different numbers of stages have different gain/compression at the same output power, the bias need not cancel and can create or exaggerate the reported NF improvement. The manuscript gives no quadrature-error budget for the nonlinear regime, and the reported 0.15 dB uncertainty from [17] may apply only to the optimized linear setting.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports an experimental study of AM and PM noise in cascaded amplifier chains built from one to four identical stages, measured at 2 GHz from small-signal to strong compression. Using an I-Q receiver with simultaneous AM and PM detection, the authors determine flicker and white noise contributions, compute an 'effective noise figure' from the white AM and PM noise spectra, and measure the AM-to-PM coherence function. The main reported findings are (i) that effective noise figure degrades with input power but, for a fixed output power, improves as the number of stages increases, and (ii) that AM-PM coherence is generally low but shows peaks near the 1 dB compression point in a second amplifier family, coinciding with peaking in the measured AM-PM derivative. The paper includes measured spectra, coherence plots, and effective NF curves, with comparisons across one to four stages.","tokens_in":6805,"tokens_out":3326,"duration_ms":33179,"significance":"If valid, the result that adding stages can improve effective noise figure at a fixed output power is of practical interest for amplifier chain design, and the observed coherence peaks near compression are relevant for nonlinear noise modeling. The paper's strengths are its simultaneous AM/PM measurement capability, the use of a previously published calibration procedure [17] giving a 0.15 dB uncertainty claim, and the lack of any fitted parameters in the reported trends. The experimental data are presented clearly and the main trends are internally consistent. However, the two central claims (effective NF comparison and coherence peaks) rely on the I-Q receiver actually separating AM and PM noise, and the manuscript does not document that the quadrature condition is maintained as the DUT's AM-PM phase shift varies with power. This unresolved calibration issue is the key correctness risk and prevents acceptance in the present form.","major_comments":[{"comment":"The paper does not state that the 90° phase condition between the DUT path and the LO path is re-optimized at every carrier input power. The text only says the phase shifter 'must be adjusted to obtain proper settings' and refers to [17] for the calibration procedure. Since Figure 13 shows that the AM-PM derivative (and hence the DUT insertion phase versus power) peaks near the 1 dB compression point, a fixed phase-shifter setting would produce a quadrature error that grows with power in exactly that region. This would cause AM noise to leak into the PM channel and PM noise into the AM channel, artificially raising the coherence function near compression and biasing the white-noise levels used to compute effective NF. The authors need to either document that the quadrature is re-optimized at each power step or provide a quantitative error budget for quadrature misalignment in the nonlinear regime; without one, the coherence peaks in Figures 11 and 12 and the effective NF comparison in Figures 8 and 9 are not fully trustworthy.","section":"Section 2 (Measurement Setup) and Figures 11–13"},{"comment":"The sentence 'achieving an identical output power implies that the noise power contribution of the last stage is identical for all of the amplifiers, independent of their number of stages' assumes that each stage's added noise depends only on its own output power and is independent of the noise content of its input or the compression state of preceding stages. This assumption is not tested or justified in the paper. While the reported NF improvement is an experimental observation, this argument is used to explain it, and a failure of the assumption could alter the interpretation. The authors should either provide a direct test (for example, comparing the output noise of a single stage driven at the same output power as the final stage of a chain) or explicitly state this as a limitation.","section":"Section 4 (Discussion), argument on last-stage noise contribution"},{"comment":"The coherence peaks are reported for a second amplifier family ('another type of amplifier component') with only the linear gain and noise figure stated (over 26 dB and around 1 dB at 2 GHz). No details of the device technology, bias, or matching conditions are given, even though the paper's own introduction notes that AM and PM noise levels depend on bias and matching conditions. Without this information, the claim that these peaks occur in 'some particular amplifier configurations' is not reproducible, and the reader cannot assess whether the effect is specific to the device or an artifact of the measurement conditions.","section":"Figures 11–13 and Section 3"}],"minor_comments":[{"comment":"The abstract states the effective noise figure 'decreases with the number of stages for a given carrier output power level,' but the results in Section 3 (Figure 9) are phrased as 'significantly improved by adding some amplification stages, except under small-signal conditions for which it remains essentially constant or, more precisely, slightly worse.' Please harmonize the wording so the small-signal caveat is reflected in the abstract or in the conclusion.","section":"Abstract and Section 3"},{"comment":"There is a typo: 'allowing for the simultaneously measuring of AM and PM noise' should read 'allowing the simultaneous measurement of AM and PM noise.'","section":"Section 2"},{"comment":"The sentence 'The desired phase difference can be obtained simply by tuning the phase shifter' is misleading: the phase shifter may introduce loss and its own noise, and the word 'simply' understates the need for the calibration procedure of [17]. Consider rewording.","section":"Section 2"},{"comment":"No error bars are shown on the effective NF data. Since the paper cites a 0.15 dB uncertainty from the calibration procedure, it would be helpful to include error bars or a statement that the uncertainties are smaller than the symbol size.","section":"Figures 8 and 9"},{"comment":"The sentence 'the fact of adding stages does not produce an improvement in AM and PM noise performances' appears to contradict the later statement about improved effective NF. Clarify that AM and PM noise in dBc/Hz worsen with the number of stages, while the effective NF computed from those spectra can nonetheless improve at a fixed output power because the carrier power also increases with gain.","section":"Section 4"},{"comment":"Reference [15] (Adamian and Uhlir) is listed in the bibliography but not cited in the text; reference [19] (Friis) is cited in Section 3 and 4 but appears after [18] in the reference list. Please check the citation order.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of the journal and the experimental approach is relevant, but the quadrature-calibration question is the central issue that needs to be resolved before publication. The authors should be asked to show, either by a direct measurement or a detailed analysis, that the I-Q receiver maintains AM/PM isolation across the full power sweep, or to quantify how the observed coherence peaks and NF trends would be affected by quadrature error. I would also encourage the editor to ask for more details on the second amplifier family, as the lack of device information limits reproducibility."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the main result—that effective noise figure improves with stage count at fixed output power—is a real, practically useful experimental finding. The AM–PM coherence peaks are more fragile, because the paper never says whether the I-Q quadrature is re-optimized at each carrier power. That's a gap, not necessarily a fatal flaw, but it needs to be addressed.\n\nWhat's genuinely good: the paper gives systematic AM/PM noise spectra and coherence data for one- to four-stage chains, from small signal to compression. The observation that AM–PM correlation is generally low, with occasional peaks near the 1 dB compression point that track the AM–PM derivative, is interesting and extends the authors' earlier single-stage work. The NF improvement at fixed output power is consistent with the physical argument in Section 4: the last stage contributes the same noise, while earlier stages' noise is reduced by the gain of preceding stages operating at lower drive. The measurement method is clearly described and references a prior calibration paper.\n\nWhere the soft spots are: the coherence-peak claim depends on the I-Q receiver actually separating AM and PM. If the phase shifter is set at small-signal optimum and not re-adjusted, the DUT's AM-PM phase shift with power rotates the measurement axes. That can create artificial coherence, exactly where the AM-PM derivative peaks. The paper says the phase shifter 'must be adjusted' but never states it is re-optimized at every input power. The same issue could bias the AM and PM white noise levels used for NF, though the bias might partially cancel if NF is computed from the sum of both contributions. The NF result also rests on a single amplifier family, with no uncertainty intervals on the curves. These are moderate concerns; the NF trend is plausible and the coherence peaks could still be real, but the manuscript needs to document the per-power calibration and show a quadrature-error budget.\n\nWho it's for: RF/microwave engineers designing receiver chains, and anyone doing noise metrology on nonlinear amplifiers. It deserves a serious referee, but the referee should ask for explicit confirmation about the phase-shifter setting across the power sweep and for uncertainty quantification. I'd cite the NF result if it survives review.","headline":"Solid experimental study with a useful NF design rule, but the coherence-peak story needs a clear statement about per-power quadrature calibration before I'd trust it.","tokens_in":7307,"tokens_out":2648,"would_cite":true,"duration_ms":29335,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"For a fixed output power, adding amplifier stages can lower the effective noise figure.","keywords":["AM noise","PM noise","cascaded amplifiers","effective noise figure","flicker noise","white noise","AM-PM correlation","gain compression"],"falsifier":"Measure the effective NF of one-, two-, three-, and four-stage chains while keeping the final stage's output power fixed but changing only the gain placed before it, for example by inserting a calibrated attenuator between stages; if the final stage's noise contribution changes with the number or gain of preceding stages, the cascade argument fails.","tokens_in":6333,"feed_emoji":"📡","tokens_out":9986,"duration_ms":76289,"temperature":0.7,"pith_summary":"This paper reports simultaneous measurements of amplitude-modulation (AM) and phase-modulation (PM) noise in chains of one to four identical 2 GHz amplifier stages, from small-signal operation into strong compression, using an in-phase/quadrature (I-Q) receiver that also measures the AM-to-PM coherence function. The central experimental result is that the effective noise figure, computed from the white-noise parts of both AM and PM spectra, decreases as the number of stages grows when the comparison is made at a fixed carrier output power. At a fixed input power the opposite trend appears: more stages compress earlier, and their AM/PM noise performance degrades faster with drive. The paper also finds that AM and PM noise are largely uncorrelated, but in particular amplifier types the coherence shows peaks near the 1 dB compression point, coinciding with peaking in the measured AM-to-PM derivative. If the cascade result holds, receiver and radar front ends can trade extra stages for lower effective noise at a required output power.","feed_headline":"More amplifier stages can cut noise figure at same output power","feed_subtitle":"Measured AM/PM noise in 1 to 4-stage 2 GHz chains shows effective NF improves with more stages at fixed output power.","key_machinery":"The argument is carried by the I-Q receiver measurement setup, which simultaneously demodulates AM and PM noise and computes the coherence function, the normalized cross-power spectral density between the two noise channels, ranging from zero for uncorrelated to one for fully correlated. The cascade analysis rests on a simple noise-accounting rule: each stage amplifies the AM and PM modulation noise arriving from earlier stages and adds its own noise, so when the output power is held fixed the last stage's added noise is identical and the earlier stages' contributions are weighted down by the gain that follows. For the correlation result, the key object is the measured AM-to-PM derivative, whose peaking near the 1 dB compression point tracks the peaking of the coherence function.","core_discovery":"The authors claim that, for a cascade of identical amplifier stages, the effective noise figure at a given carrier output power is improved by increasing the number of stages. Their reasoning is that at equal output power the final stage adds the same noise regardless of how many stages precede it, while the noise added by earlier stages is reduced by the gain of the stages after them and is generated at lower drive levels. Measured effective noise figures from AM and PM white noise in one- to four-stage amplifiers support this: at low input power the cascade follows the standard noise-figure behavior, and under compression the effective NF rises for all chains, but the absolute degradation is larger for the chains with fewer stages. A separate finding is that AM-to-PM coherence is generally low, yet some amplifier configurations show clear correlation peaks close to the 1 dB gain compression point, in both flicker and white noise regions, and the same peaking appears in the derivative of the AM-to-PM characteristic, linking the underlying noise-conversion mechanisms to the device's large-signal AM-PM response.","pith_inferences":["A testable design extension: if the cascade result generalizes to other transistor technologies, system designers could deliberately lower the drive level of early stages and add gain later to reduce effective NF at a fixed output power, accepting extra stages and DC consumption.","The coincidence of coherence peaks with AM-to-PM derivative peaks suggests that a static AM-PM measurement could serve as a screening test for operating points where AM and PM noise become correlated, before full noise measurements are made.","The load-bearing assumption that a stage's added noise depends only on its own output power is unlikely to hold in strong compression, where noise conversion becomes input-dependent; the reported NF improvement may vanish or reverse in that regime.","Because cyclostationary noise theory allows AM-PM correlation to grow under strong periodic drive, the observed near-compression peaks are consistent with that picture, but the device-dependence of the peaks points to a role for specific bias and matching conditions."],"forward_implications":["For a fixed carrier output power, a cascade with more identical stages can exhibit a lower effective NF than a shorter chain, even though the same chain has worse AM/PM noise figures when compared at the same input power.","Adding stages moves the onset of gain compression to lower input power levels, so input-referred AM/PM noise performance degrades faster with drive as the stage count grows.","Under small-signal conditions the effective NF is nearly independent of the number of stages, and slightly worse with more stages, consistent with the standard cascade noise-figure formula.","AM and PM noise in the tested amplifiers are mostly independent, so an amplifier's AM noise performance cannot be inferred from its PM noise performance; both must be measured.","In some devices, operating near the 1 dB compression point can create significant AM-to-PM correlation at specific offset frequencies, meaning noise at those offsets should not be treated as two independent channels."],"supporting_citations":[{"why":"Defines the effective NF obtained from AM and PM white noise and shows that AM and PM noise contributions differ in the nonlinear regime; used to compute the paper's effective NF values.","marker":"[12]"},{"why":"Shows that carrier-power dependence of AM and PM noise in a stage can be linked to derivatives of the AM-AM and AM-PM curves; used to explain the coherence peaks via the measured AM-PM derivative.","marker":"[13]"},{"why":"Supplies the definition of the coherence function used to quantify AM-to-PM correlation as a function of offset frequency.","marker":"[16]"},{"why":"Provides the setting and calibration procedure that lets the I-Q receiver measure AM and PM noise simultaneously with 0.15 dB uncertainty.","marker":"[17]"},{"why":"Gives the standard cascade noise-figure formula used as the small-signal baseline for interpreting effective NF versus stage count.","marker":"[19]"},{"why":"Provides the conversion-matrix and harmonic-balance noise analysis that models additive and converted noise as small-index AM/PM modulation of the carrier.","marker":"[20]"},{"why":"Supports the treatment of noise as cyclostationary under strong periodic excitation, the basis for expecting possible AM-PM correlation growth.","marker":"[21]"},{"why":"States that additive white noise gives zero AM-PM correlation, the baseline against which the measured low coherence is interpreted.","marker":"[22]"}],"fun_headline_variants":["Cascading amplifiers lowers noise figure at constant output","More stages, same output: better effective noise figure","AM-PM noise coherence peaks near compression in cascades","Effective NF improves with more amplifier stages at fixed power","Adding amplifier stages reduces noise figure at matched output"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusion that more stages improve effective noise figure at a fixed output power assumes that each identical stage adds a noise power depending only on its own output power, so the final stage contributes the same noise whether it is preceded by one stage or by three.","fun_headline_variants_meta":{"raw":{"variants":["Cascading amplifiers lowers noise figure at constant output","More stages, same output: better effective noise figure","AM-PM noise coherence peaks near compression in cascades","Effective NF improves with more amplifier stages at fixed power","Adding amplifier stages reduces noise figure at matched output"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000195,"raw_usage":{"total_tokens":1335,"prompt_tokens":902,"completion_tokens":433,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":518,"completion_tokens_details":{"reasoning_tokens":357}},"tokens_in":518,"tokens_out":433,"duration_ms":4299,"temperature":1.0,"reasoning_tokens":357,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T00:13:26.459213+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the effective NF of one-, two-, three-, and four-stage chains while keeping the final stage's output power fixed but changing only the gain placed before it, for example by inserting a calibrated attenuator between stages; if the final stage's noise contribution changes with the number or gain of preceding stages, the cascade argument fails.","supporting_citations":[{"cited_title":"Phase noise and AM noise measurements in the frequency domain","cited_arxiv_id":null,"evidence_quote":"Defines the effective NF obtained from AM and PM white noise and shows that AM and PM noise contributions differ in the nonlinear regime; used to compute the paper's effective NF values."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows that carrier-power dependence of AM and PM noise in a stage can be linked to derivatives of the AM-AM and AM-PM curves; used to explain the coherence peaks via the measured AM-PM derivative."},{"cited_title":"Noise figure vs","cited_arxiv_id":null,"evidence_quote":"Supplies the definition of the coherence function used to quantify AM-to-PM correlation as a function of offset frequency."},{"cited_title":"100-GHz cooled amplifier residual pm and am noise measurements, noise figure, and jitter calculations","cited_arxiv_id":null,"evidence_quote":"Provides the setting and calibration procedure that lets the I-Q receiver measure AM and PM noise simultaneously with 0.15 dB uncertainty."},{"cited_title":"The Room Temperature Multi-Channel Heterodyne Receiver Section of the PHAROS2 Phased Array Feed","cited_arxiv_id":null,"evidence_quote":"Gives the standard cascade noise-figure formula used as the small-signal baseline for interpreting effective NF versus stage count."},{"cited_title":"Accurately Modeling of Zero Biased Schottky-Diodes at Millimeter- Wave Frequencies","cited_arxiv_id":null,"evidence_quote":"Provides the conversion-matrix and harmonic-balance noise analysis that models additive and converted noise as small-index AM/PM modulation of the carrier."},{"cited_title":"Noise as a diagnostic tool for quality and reliability of electronic devices","cited_arxiv_id":null,"evidence_quote":"Supports the treatment of noise as cyclostationary under strong periodic excitation, the basis for expecting possible AM-PM correlation growth."},{"cited_title":"Investigations of AM, PM noise, and noise figure in an SiGe-HBT amplifier operating in linear and non- linear regime","cited_arxiv_id":null,"evidence_quote":"States that additive white noise gives zero AM-PM correlation, the baseline against which the measured low coherence is interpreted."}],"review_version":1}