{"id":"89d1a8da-cd32-420a-98e5-891b46a560b5","arxiv_id":"1908.02304","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A recursive least squares step-tracker, using a parabolic beacon model with pre-fitted curvature coefficients, tracked Brasilsat B3 for 24 hours on a real antenna system.","lead":"This paper shows a real 3-meter antenna tracking an inclined-orbit satellite by fitting a parabola to beacon signal strength readings and moving to the estimated peak. It replaces the usual batch least squares with an online recursive version and reports a 24-hour tracking run with a mean beacon level of 2.89 dB.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No independent orbit reference: the 24-hour beacon trace alone cannot prove the antenna followed Brasilsat B3, so the 'satisfactory tracking' conclusion rests on an unquantified proxy.","rationale":"The reader's verdict was CONDITIONAL, and I agree with that verdict but for a different primary reason. The reader focused on the assumed parabolic beacon model with fixed Kx/Ky. That assumption is not the most load-bearing issue: within a local, symmetric displacement pattern, a curvature mismatch mainly rescales the estimated offset from the pattern center and, during tracking, that offset is small, so the induced pointing bias is modest. The more serious gap is empirical: the paper's evidence that the antenna actually followed Brasilsat B3 consists of a beacon level time series and a figure-8 trajectory, but neither is anchored to the satellite's true position. The beacon level is an uncalibrated proxy; the figure-8 shape is the expected qualitative form but has no quantitative comparison. The unspecified tracking-cycle period, pattern size, forgetting factor, and RLS reset behavior compound the reproducibility problem. A repeat run with TLE-derived reference angles would settle whether the algorithm truly tracks the satellite within a small fraction of the beamwidth. This is a standard and inexpensive check for this kind of engineering-validation paper, so the verdict remains CONDITIONAL rather than ACCEPT or REJECT.","tokens_in":4729,"tokens_out":12991,"duration_ms":146113,"concrete_test":"Repeat the 24-hour run while logging GPS timestamps alongside azimuth, elevation, and beacon level. For the same UTC date, obtain the Brasilsat B3 TLE and propagate with SGP4 to the Guama STP coordinates to produce independent azimuth and elevation predictions. After removing a constant mount misalignment (a single azimuth/elevation bias fitted over the first hour), compute the RMS and 95th-percentile tracking error. If the RMS error exceeds about 10% of the 3-m antenna's HPBW (approximately 0.18 degrees at 4 GHz) or if the -24 dB beacon minima coincide with prediction mismatches larger than a beamwidth, the 'satisfactory tracking' claim fails; also report the beacon level at best known pointing to convert the 2.89 dB mean into pointing loss.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim in the Conclusions is that the LS/RLS system 'is able to track the Brasilsat B3 satellite satisfactorily', supported by Fig. 10 (beacon level, mean 2.89 dB, std 1.15 dB, minima near -24 dB) and the figure-8 trajectory in Fig. 6. The load-bearing weakness is that no independent ground-truth position (e.g., TLE/SGP4 propagation) is provided for the test interval, and the beacon-level axis is not calibrated: the paper never states the maximum beacon level at perfect pointing, the receiver noise floor, or the mapping from pointing error to beacon loss. Consequently, a relatively high mean beacon level could coexist with substantial pointing errors if the 3-m antenna beam is broad (HPBW approximately 1.75 degrees at C-band), and the -24 dB minima indicate episodes of large error whose duration and frequency are not quantified. The phrase 'satisfactorily' is thus not tied to any performance metric. The Kx/Ky assumption flagged by the reader is real but secondary: if the pattern coordinates are local offsets from the pattern center and the pattern is symmetric, an error in Kx or Ky scales the estimated offset but does not bias it when the offset is near zero; the missing reference comparison is the decisive gap.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper describes the implementation of an automatic satellite tracking system for a 3-meter parabolic antenna. The tracking algorithm models the beacon signal level as a 2D parabola in azimuth and elevation, estimates the parabola peak using least squares (or a recursive least squares variant) from measurements taken along a rectangular displacement pattern, and then moves the antenna to the estimated peak. The system was tested for about 24 hours while tracking the inclined-orbit satellite Brasilsat B3. Reported results include a figure-8 azimuth/elevation trajectory, a mean beacon level of 2.89 dB with standard deviation 1.15 dB, and beacon dips near -24 dB. The authors conclude that the least squares technique tracks the satellite satisfactorily and that the RLS adaptation removes the need for matrix inverses and improves adaptability through a forgetting factor.","tokens_in":5008,"tokens_out":3076,"duration_ms":34572,"significance":"If the reported results are valid, the paper demonstrates a practical, real-time implementation of a model-based step-track algorithm on a real antenna control system, with an RLS formulation that avoids matrix inversions and can track a moving inclined-orbit satellite. The paper is transparent about the system architecture and provides logged data over a 24-hour period, which is a useful engineering contribution. However, the central claim of 'satisfactory tracking' rests entirely on an uncalibrated beacon level proxy and lacks any independent ground-truth comparison, so the significance is currently limited to a proof-of-concept demonstration rather than a validated tracking performance result.","major_comments":[{"comment":"The central claim that the tracking was 'satisfactory' is supported only by the beacon-time series in Fig. 10, yet that series is uncalibrated: the paper does not state the beacon level corresponding to perfect pointing, the receiver noise floor, or the mapping from pointing error to beacon attenuation. Without an independent orbital reference (e.g., a TLE/SGP4 propagation of Brasilsat B3 over the test interval) or a comparison with an alternative tracking baseline, the reported mean of 2.89 dB cannot be distinguished from a substantial steady-state pointing error on a broad 3-meter C-band antenna beam. The dips to about -24 dB also need to be quantified in duration and frequency, since sustained deep fades would contradict the 'satisfactory' conclusion.","section":"Section VI, Fig. 10"},{"comment":"The estimation procedure treats the curvature coefficients Kx and Ky as known and constant over the 24-hour run, but the paper gives only a rough range for Kx and no uncertainty or sensitivity analysis for either coefficient. If the true beam pattern deviates from a parabola, or if Kx and Ky vary with elevation or temperature, the estimated peak position p, q will be biased in a way that the paper does not quantify. The authors should at least examine the least-squares residuals to validate the parabolic model, and report how sensitive the final pointing angles are to the assumed Kx and Ky values.","section":"Section II, Eq. (1) and paragraph after Eq. (10)"},{"comment":"The RLS implementation is a key contribution, but crucial timing and parameter choices are missing: no values are reported for the forgetting factor λ, the tracking-cycle period, the number of samples per displacement pattern, or the actual sampling rate in the logged run. Since the paper notes that λ=0.98 gives a memory horizon of about 50 samples and that measurements are taken every 20 ms, the reader cannot assess whether the forgetting factor is consistent with the duration of the measurement pattern or with the timescale of the satellite motion. This omission prevents independent verification of the claimed advantage of RLS over batch least squares.","section":"Section III, Eq. (11)-(16) and Section IV"}],"minor_comments":[{"comment":"In the sentence following Eq. (10), 'the values if Kx' should read 'the values of Kx'.","section":"Section II"},{"comment":"The phrase 'in out setup' is a typo for 'in our setup', and the sentence 'these measurements are taken every 20 ms' needs to be reconciled with the earlier statement that the antenna stops for a predetermined time at each vertex.","section":"Section IV"},{"comment":"The text refers to 'Figure 8 and 9' and 'Figures 8 and 9' inconsistently; use the plural consistently. Also, the azimuth and elevation axes in Figures 8 and 9 should state their units (degrees) in the axis labels.","section":"Section VI"},{"comment":"Units for x and y (azimuth and elevation) are never specified; if they are in degrees, the magnitudes of Kx and Ky should be interpreted in those units, and the reader should be told explicitly.","section":"Section II, Eq. (1)"},{"comment":"The beacon receiver is described as linearized to 0-10 V, but the plotted beacon axis is in dB; the conversion from voltage to dB should be stated so that the reported mean and standard deviation are reproducible.","section":"Section VI, Fig. 10"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is an honest, well-scoped engineering demonstration of Laine's model-based step-track on a real 3-meter antenna, with an RLS wrapper and a rectangular scan pattern. The math is standard, the hardware description is concrete, and the 24-hour beacon trace shows the loop stayed up. The problem is that the paper's main conclusion—'satisfactorily' tracking—is not tied to any independent measure of pointing accuracy.\n\nWhat is actually new: the RLS formulation avoids matrix inversion and adds a forgetting factor, and the rectangle pattern adapts to asymmetric beam shapes. Both are incremental but sensible extensions. The paper also deserves credit for being upfront that Kx and Ky must be pre-estimated, and for noting the elevation dependence of Kx. The derivations in Sections II and III are correct, and the figure-8 trajectory in Fig. 6 is a plausible signature of an inclined-orbit satellite.\n\nWhere it falls short is validation. There is no TLE/SGP4 reference orbit for the test interval, so the figure-8 shape is asserted, not verified. The beacon level is used both as the tracking input and as the success metric, with no calibration: the paper never states the max beacon at perfect pointing, the receiver noise floor, or the mapping from pointing error to beacon loss. A 3-meter C-band antenna has a beamwidth on the order of 1.7 degrees, so a mean beacon of 2.89 dB could coexist with substantial pointing error. The -24 dB minima in Fig. 10 show that large errors occurred, but their duration and frequency are not quantified. Nor is there a baseline comparison with conventional least squares or another tracking method, and key parameters like the forgetting factor and step sizes are not reported. The Kx/Ky uncertainty is a real but secondary issue: if the pattern is symmetric, a scale error in Kx/Ky does not bias the estimated peak near boresight, though it would bias larger offsets.\n\nThe central claim is likely true in the weak sense that the system kept the antenna pointed well enough to hold the beacon, but the paper does not establish the stronger claim of satisfactory tracking against a standard. For a short conference paper, this is acceptable as an implementation note; for a journal, it needs a reference orbit comparison and a calibrated beacon-to-pointing-error curve.\n\nWho is this for? Engineers building small ground stations who want a worked example of step-track with RLS. It is not a methods paper and not a definitive validation. I would send it to a knowledgeable referee if the venue is a specialized antenna or satellite comms venue, with the expectation of a revision that adds ground truth or tones down the claim. It deserves a serious referee, but I would not cite it as evidence that the technique outperforms anything.","headline":"A credible but under-measured engineering demo: the math is right and the antenna really moved, but 'satisfactory tracking' is never tied to an independent orbit reference.","tokens_in":5517,"tokens_out":1762,"would_cite":false,"duration_ms":22655,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Least-squares beacon tracking keeps a real antenna locked on an inclined-orbit satellite for 24 hours.","keywords":["satellite tracking","least squares","recursive least squares","antenna control system","beacon signal","inclined orbit","step-track","parabolic model"],"falsifier":"Measure the antenna's actual beam pattern by sweeping a fine grid of azimuth and elevation offsets around the peak with the satellite stationary, fit the assumed parabola to those measurements, and check whether the residuals are comparable to receiver noise; systematic, non-quadratic residuals (or curvature values that drift with elevation or weather) would mean the peak estimate is biased and part of the reported tracking gain is an artifact of the model.","tokens_in":4550,"feed_emoji":"📡","tokens_out":8592,"duration_ms":84883,"temperature":0.7,"pith_summary":"This paper reports a working implementation of least-squares satellite tracking in a real 3-meter antenna control system. The system tracked the inclined-orbit Brasilsat B3 satellite for about 24 hours while logging azimuth, elevation, and beacon level. The central demonstration is that the 2D parabolic beacon model used in earlier step-track work, solved with recursive least squares, holds the antenna on the satellite in hardware; the run's mean beacon level was 2.89 dB with a standard deviation of 1.15 dB. The paper also argues that the recursive filter improves on batch least squares because it avoids matrix inversion and its forgetting factor lets the model adapt as the satellite moves.","feed_headline":"3-meter dish tracks inclined-orbit satellite for 24 hours","feed_subtitle":"Least-squares beacon tracking keeps the dish on Brasilsat B3 with a mean signal of 2.89 dB.","key_machinery":"The central mechanism is the 2D parabolic beacon model $L(x,y)=K_x(x-p)^2+K_y(y-q)^2+L_{pq}$, where $(p,q)$ is the antenna pointing direction that maximizes the beacon and $L_{pq}$ is the peak level. Expanding this model converts the peak-finding problem into the linear regression $Y=X\\beta$, solved by batch least squares or, in the paper's adaptation, by recursive least squares; RLS avoids computing $(X^TX)^{-1}$ and uses a forgetting factor $\\lambda$ to weight recent samples more heavily. Around this estimator the tracking loop executes a rectangular displacement pattern, samples the beacon every 20 ms, updates the estimated peak, moves to it, and then waits for the next tracking cycle.","core_discovery":"On the paper's own terms, the central claim is that a least-squares step-track algorithm can track an inclined-orbit geostationary satellite in real hardware, not just in simulation. The algorithm models the beacon level near the peak as a 2D parabola $L(x,y)=K_x(x-p)^2+K_y(y-q)^2+L_{pq}$, expands it into the linear regression $Y=X\\beta$, and recovers the satellite direction $(p,q)$ from the least-squares estimate of $\\beta$. Over a 24-hour test on Brasilsat B3 the antenna produced a figure-eight trajectory typical of an inclined orbit, with azimuth varying more than $30$ degrees and elevation less than $2$ degrees, and the beacon level was restored at each tracking cycle; the mean beacon level was $2.89$ dB with standard deviation $1.15$ dB. The paper takes this as confirmation that the model-based step-track procedure works in a real control system, and reports that the recursive least-squares adaptation performed well, removing the need for matrix inverses and adding adaptability through the forgetting factor.","pith_inferences":["The paper fixes $K_x$ and $K_y$ from calibration, but the same RLS equations could estimate them online; that would remove the site-specific calibration step and would be a direct test of whether curvature drift matters for tracking accuracy.","Because azimuth and elevation varied at very different rates, the rectangular instead of square search pattern is only a first adaptation; a natural experiment is to adjust the rectangle's aspect ratio and orientation from the estimated $K_x/K_y$ and measure the effect on convergence and beacon hold.","The single 24-hour run on Brasilsat B3 leaves open how the tracker behaves under rain fade, multipath, or a satellite with a different inclination; replicating the test on another inclined-orbit satellite would separate algorithm behavior from satellite-specific geometry.","The beacon variation during the run includes weather and diurnal effects, so the 2.89 dB mean is not purely a tracking-quality measure; comparing the tracked beacon with a fixed-pointing baseline over the same period would quantify the net gain of the tracker."],"forward_implications":["A 3-meter antenna with a beacon receiver can automatically track an inclined-orbit satellite for a full day, making step-track with the parabolic model a practical option where continuous tracking hardware is not available.","Sampling the beacon every 20 ms during the search pattern is feasible because RLS stores only a filter state; a batch least-squares implementation would have to keep and invert a large sample matrix.","The forgetting factor gives the tracker a finite memory horizon $\\tau=1/(1-\\lambda)$, so the estimate can follow both satellite motion and slow changes in the received beam during a track.","The logged 24-hour mean beacon level of 2.89 dB with 1.15 dB standard deviation is a concrete benchmark for comparing future step-track algorithms on similar antennas."],"supporting_citations":[{"why":"supplies the 2D parabolic beacon model and the step-track displacement pattern that the paper implements and extends with RLS.","marker":"[7]"},{"why":"defines the Brasilsat B3 satellite and its inclined orbit, the target that determines the tracking geometry.","marker":"[4]"},{"why":"provides the digital tracking receiver that converts the beacon signal to the voltage measured and logged by the control system.","marker":"[5]"},{"why":"gives the reference orbit trajectory used to interpret the observed figure-eight motion.","marker":"[8]"}],"fun_headline_variants":["Least-squares satellite tracking works on real antenna","24-hour test: least-squares tracking keeps dish on target","Beacon-seeking least squares locks onto inclined orbit","Real-time least-squares step-track achieves 2.89 dB mean"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The tracking assumes the beacon strength around the correct aim point is a 2D parabola whose curvature coefficients are known in advance and stay fixed for the whole 24-hour run; if the actual beam departs from that shape, the estimated satellite position is biased.","fun_headline_variants_meta":{"raw":{"variants":["Least-squares satellite tracking works on real antenna","24-hour test: least-squares tracking keeps dish on target","Beacon-seeking least squares locks onto inclined orbit","Real-time least-squares step-track achieves 2.89 dB mean"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000722,"raw_usage":{"total_tokens":3177,"prompt_tokens":824,"completion_tokens":2353,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":440,"completion_tokens_details":{"reasoning_tokens":2286}},"tokens_in":440,"tokens_out":2353,"duration_ms":17458,"temperature":1.0,"reasoning_tokens":2286,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:47:23.821442+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the antenna's actual beam pattern by sweeping a fine grid of azimuth and elevation offsets around the peak with the satellite stationary, fit the assumed parabola to those measurements, and check whether the residuals are comparable to receiver noise; systematic, non-quadratic residuals (or curvature values that drift with elevation or weather) would mean the peak estimate is biased and part of the reported tracking gain is an artifact of the model.","supporting_citations":[{"cited_title":"Tracking systems for satellite communications","cited_arxiv_id":null,"evidence_quote":"supplies the 2D parabolic beacon model and the step-track displacement pattern that the paper implements and extends with RLS."},{"cited_title":"Variable Resolution, 10-Bit to 16-Bit R/D Converter with Reference Oscillator , 2008","cited_arxiv_id":null,"evidence_quote":"defines the Brasilsat B3 satellite and its inclined orbit, the target that determines the tracking geometry."},{"cited_title":"Caracter\\' i sticas do Sat\\' e lite Brasilsat B3 em \\' O rbita Inclinada, para Projeto T\\' e cnico de Redes de Comunica c \\ o es Digitais em Banda C","cited_arxiv_id":null,"evidence_quote":"provides the digital tracking receiver that converts the beacon signal to the voltage measured and logged by the control system."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"gives the reference orbit trajectory used to interpret the observed figure-eight motion."}],"review_version":1}