pith. sign in

arxiv: 1906.10598 · v2 · pith:MMWR2THEnew · submitted 2019-06-25 · 🌌 astro-ph.IM · astro-ph.EP

Phase tracking based on GPGPU and applications in Planetary radio Science

Nianchuan Jian , Dmitry Mikushin , Jianguo Yan This is my paper

Pith reviewed 2026-05-25 15:57 UTC · model grok-4.3

classification 🌌 astro-ph.IM astro-ph.EP
keywords phase trackingGPGPUplanetary radio scienceDifferential EvolutionTaylor polynomialDoppler data processingMars ExpressChang'E 4
0
0 comments X

The pith

A GPU-based method fits received radio signals to Taylor polynomials using Differential Evolution to estimate instantaneous phase, frequency, and line-of-sight acceleration for planetary radio science.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents a phase tracking technique that models the radio signal as a Taylor polynomial in phase and amplitude. It uses the Differential Evolution optimizer to find the best coefficients and runs this on GPUs to handle the computation. This produces estimates of phase, frequency, frequency derivative, and total count phase at various scales. These outputs support analysis of planetary occultations and gravitational fields from tracking data of missions like Mars Express and Chang'E 4. The approach achieves real-time processing and specified precision levels on the test datasets.

Core claim

The method replaces traditional phase-locked loop counting with a polynomial fitting approach solved by Differential Evolution on GPUs, yielding instantaneous phase, frequency, line-of-sight acceleration, and integrated phase values from radio tracking signals of Mars Express and Chang'E 4, with integral Doppler precision of 2 mrad/s and 4 mrad/s respectively for 60-second integrations, while enabling real-time operation on specified NVIDIA GPUs.

What carries the argument

Differential Evolution optimization applied to fitting the received signal to a Taylor polynomial model of phase and amplitude, executed on GPGPUs for high computational load.

If this is right

  • The method provides line-of-sight acceleration estimates useful for gravitational field analysis.
  • Real-time processing is achieved on NVIDIA GTX580 and dual K80 GPUs for 400K data blocks with 80,000 objective evaluations in 6.5 seconds.
  • Precision reaches 2 mrad/s for MEX 3-way and 4 mrad/s for Chang'E 4 3-way integral Doppler over 60 seconds.
  • Data supports planetary occultation studies from the estimated parameters at multiple integration scales.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar polynomial fitting on GPUs could extend to other radio science applications where real-time Doppler processing is needed.
  • Convergence of the DE optimizer might vary with signal-to-noise ratios not tested in the MEX and Chang'E datasets.
  • The Taylor expansion assumption limits accuracy for rapidly varying signals beyond the chosen integration interval.

Load-bearing premise

The received radio signal can be accurately represented by a Taylor polynomial in phase and amplitude over the integration interval, allowing the optimizer to recover the coefficients reliably.

What would settle it

Applying the method to a new dataset where the signal deviates from the Taylor model over the interval and checking if the reported precision is not achieved or convergence fails.

Figures

Figures reproduced from arXiv: 1906.10598 by Dmitry Mikushin, Jianguo Yan, Nianchuan Jian.

Figure 1
Figure 1. Figure 1: DE algorithm iteration flowchart 1.3. GPGPU computing The Graphics Processing Unit (GPU) is a specialized proces￾sor originally designed specifically for accelerating video out￾put with complex 3D scenes in games, CADs and visual ef￾fects. In order to optimize the speed of positioning and textur￾ing of complex geometries (which involves interpolation), the GPUs have become rich in floating-point compute th… view at source ↗
Figure 2
Figure 2. Figure 2: Comparison of double-precision computing performance [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Relation between the objective function and the correlation coe [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of convergence rate of two objective functions types [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Data process flow remainder of Taylor polynomial. The second error is random errors in tracking data. The third is system error caused by equipments and medias along the signal path. Remainder of Taylor polynomial is relative to the block length and polyno￾mial order. Random errors come from thermal noise of signal beacon, receiver and medias along signal path. System error can be eliminated by constructio… view at source ↗
Figure 6
Figure 6. Figure 6: Variety of Phase noise σnoise(T) and remainder R3( T 2 ) noise of signal source, transponder on board and receiver at sta￾tion. The second is random interference of solar phase scintil￾lations along signal path. Carrier phase error variance can be expressed as [Sniffin, 2005]: σ 2 noise|1−way = σ 2 S + 1 ρL σ 2 noise|3−way = M2 (BTR − BL) PC/N0|U/L + σ 2 S + 1 ρL (13) Here, • σ 2 S is contribution to carri… view at source ↗
Figure 7
Figure 7. Figure 7: Precision of integration Doppler changes with count interval [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
read the original abstract

This paper introduces a phase tracking method for planetary radio science research with computational algorithm implemented fo r NVIDIA GPUs. In contrast to the phase-locked loop (PPL) phase counting method used in traditional Doppler data processing, this method fits the tracking data signal into the shape expressed by the Taylor polynomial with optimal phase and amplitude coefficients. The Differential Evolution (DE) algorithm is employed for polynomial fitting. In order to cope with high computational intensity of the proposed phase tracking method, the graphics processing units (GPUs) are employed. As a result, the method estimates the instantaneous phase, frequency, derivative of frequency (line-of-sight acceleration) and the total count phase of different integration scales. This data can be further used in planetary radio science research to analyze the planetary occultation and gravitational fields. The method has been tested on MEX (Mars Express, ESA) and Chang'E 4 relay satellite (China) tracking data. In a real experiment with 400K data block size and $\sim$80,000 DE solver objective function evaluations we were able to acheive the target convergence threshold in 6.5 seconds and do real-time processing on NVIDIA GTX580 and 2$\times$ NVIDIA K80 GPUs, respectively. The precision of integral Doppler (60s) is 2 mrad/s and 4 mrad/s for MEX(3-way) and Chang'E 4 relay satellite(3-way) respectively.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper presents a GPU-accelerated phase tracking technique for planetary radio science that replaces traditional PLL Doppler counting with Differential Evolution optimization of a Taylor polynomial model in both phase and amplitude. The method extracts instantaneous phase, frequency, frequency derivative (line-of-sight acceleration), and integrated phase at multiple integration scales from 400 k-sample blocks of 3-way tracking data. It reports real-time performance (6.5 s per block on GTX 580; real-time on dual K80) and integral-Doppler precisions of 2 mrad/s (MEX) and 4 mrad/s (Chang’E-4 relay) over 60 s intervals, with the results intended for occultation and gravity-field analysis.

Significance. If the polynomial model is shown to be unbiased and the optimizer reliably global, the approach could supply an alternative high-precision Doppler observable for radio-science experiments that is naturally parallelizable on commodity GPUs. The reported runtimes indicate that the method could support real-time processing pipelines, which is a practical advantage over serial PLL implementations for large data volumes.

major comments (3)
  1. [Abstract and §3] Abstract and §3 (method): the central precision claims (2 mrad/s and 4 mrad/s) rest on the assumption that a low-order Taylor expansion adequately represents the received signal over each 400 k-sample integration window. No residual diagnostics, model-order selection procedure, or test against synthetic signals containing known higher-order dynamics or scintillation are provided; without these, it is impossible to rule out systematic bias in the recovered frequency derivative.
  2. [Abstract and §4] Abstract and §4 (results): the reported precisions are stated without error bars, without direct comparison to a conventional PLL on identical data segments, and without any cross-validation against independent observables (e.g., known gravitational signatures or occultation events). These omissions make the quantitative performance claims unverifiable from the manuscript.
  3. [§3.2] §3.2 (DE implementation): the paper does not specify how the Taylor polynomial degree is chosen, what convergence criteria are used for the DE solver, or how many independent runs are performed to guard against local minima. These hyperparameters directly affect whether the quoted precisions are reproducible.
minor comments (2)
  1. [Abstract] Abstract contains typographical errors (“fo r”, “acheive”, “PPL” instead of “PLL”).
  2. [§3] Notation for the Taylor coefficients and the integration scales is introduced without a clear table or equation numbering, making it difficult to trace which parameters are fitted versus derived.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive report. We address each major comment below and indicate the revisions we will make to improve verifiability and reproducibility.

read point-by-point responses

  1. Referee: [Abstract and §3] Abstract and §3 (method): the central precision claims (2 mrad/s and 4 mrad/s) rest on the assumption that a low-order Taylor expansion adequately represents the received signal over each 400 k-sample integration window. No residual diagnostics, model-order selection procedure, or test against synthetic signals containing known higher-order dynamics or scintillation are provided; without these, it is impossible to rule out systematic bias in the recovered frequency derivative.

    Authors: We agree that residual diagnostics and explicit model-order justification are needed to support the bias claim. In the revised manuscript we will add (i) a short subsection explaining the choice of Taylor degree (typically 3, corresponding to phase, frequency and line-of-sight acceleration) based on the expected signal bandwidth over 400 k-sample windows at the sampling rates of the MEX and Chang’E-4 datasets, (ii) example residual time-series plots for the real tracking segments, and (iii) a brief discussion of the conditions under which higher-order terms or scintillation would become detectable. Full Monte-Carlo tests with synthetic scintillation are beyond the present scope but will be noted as future work. revision: partial

  2. Referee: [Abstract and §4] Abstract and §4 (results): the reported precisions are stated without error bars, without direct comparison to a conventional PLL on identical data segments, and without any cross-validation against independent observables (e.g., known gravitational signatures or occultation events). These omissions make the quantitative performance claims unverifiable from the manuscript.

    Authors: We will augment §4 with (i) uncertainty estimates obtained from the ensemble of converged DE solutions on each block, (ii) a side-by-side comparison of the integral-Doppler values against a conventional PLL processor run on the identical MEX and Chang’E-4 segments, and (iii) a short discussion of the limited opportunities for cross-validation with independent gravity or occultation signatures in the chosen datasets. These additions will make the quoted 2 mrad/s and 4 mrad/s figures directly verifiable. revision: yes

  3. Referee: [§3.2] §3.2 (DE implementation): the paper does not specify how the Taylor polynomial degree is chosen, what convergence criteria are used for the DE solver, or how many independent runs are performed to guard against local minima. These hyperparameters directly affect whether the quoted precisions are reproducible.

    Authors: The revised §3.2 will explicitly state: Taylor degree is fixed at 3; the DE convergence criterion is an objective-function threshold of 10^{-6} (or 5000 generations, whichever occurs first); and five independent DE runs with different random seeds are performed per block, retaining the lowest-cost solution. These implementation details were omitted for brevity but are part of the code used to produce the reported results. revision: yes

Circularity Check

0 steps flagged

No circularity: method is empirical fitting on external data

full rationale

The paper describes a numerical procedure that fits a Taylor polynomial model (phase and amplitude coefficients) to radio tracking signals via Differential Evolution, implemented on GPUs, and reports achieved precisions when applied to independent MEX and Chang'E-4 datasets. No derivation chain, prediction, or uniqueness claim reduces by construction to fitted inputs or self-citations; the central results are direct outputs of processing external observations rather than self-referential quantities. The method is self-contained against external benchmarks with no load-bearing self-citation or ansatz smuggling.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The method depends on the modeling choice that a low-order Taylor polynomial suffices for the signal and on the numerical behavior of the DE optimizer; no free parameters are explicitly named in the abstract but the polynomial degree and DE control parameters are implicit choices.

free parameters (2)
  • Taylor polynomial degree
    The order of the polynomial used to model phase and amplitude is a modeling choice that must be selected and affects the fit quality but is not specified.
  • DE solver hyperparameters
    Parameters such as population size, mutation factor, and number of objective function evaluations (~80,000 cited) are chosen to reach convergence but not detailed.
axioms (1)
  • domain assumption The received signal phase and amplitude can be adequately modeled by a Taylor polynomial over the integration interval.
    This modeling assumption underpins the entire fitting procedure described in the abstract.

pith-pipeline@v0.9.0 · 5796 in / 1487 out tokens · 60137 ms · 2026-05-25T15:57:13.847063+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages

  1. [1]

    Asmar , author N

    author S. Asmar , author N. Renzetti , title The deep space network as an instrument for radio science research , journal NASA STI/Recon Technical Report N volume 95 ( year 1993 ) pages 21456

    work page 1993

  2. [2]

    author S. W. Asmar , author K. Aksnes , author R. Ambrosini , author A. Anabtawi , author J. D. Anderson , author J. W. Armstrong , author D. Atkinson , author J.-P. Barriot , author B. Bertotti , author B. G. Bills , et al., title Planetary radio science: Investigations of interiors, surfaces, atmospheres, rings, and environments , journal A White Paper ...

    work page 2009

  3. [3]

    Lorell , author W

    author J. Lorell , author W. Sjogren , title Lunar gravity: Preliminary estimates from lunar orbiter , journal Science volume 159 ( year 1968 ) pages 625--627

    work page 1968

  4. [4]

    Konopliv , author A

    author A. Konopliv , author A. Binder , author L. Hood , author A. Kucinskas , author W. Sjogren , author J. Williams , title Improved gravity field of the moon from lunar prospector , journal Science volume 281 ( year 1998 ) pages 1476--1480

    work page 1998

  5. [5]

    author A. S. Konopliv , author C. F. Yoder , author E. M. Standish , author D.-N. Yuan , author W. L. Sjogren , title A global solution for the mars static and seasonal gravity, mars orientation, phobos and deimos masses, and mars ephemeris , journal Icarus volume 182 ( year 2006 ) pages 23--50

    work page 2006

  6. [6]

    author F. G. Lemoine , author D. E. Smith , author D. D. Rowlands , author M. Zuber , author G. Neumann , author D. Chinn , author D. Pavlis , title An improved solution of the gravity field of mars (gmm-2b) from mars global surveyor , journal Journal of Geophysical Research: Planets volume 106 ( year 2001 ) pages 23359--23376

    work page 2001

  7. [7]

    Sniffin , title DSMS Telecommunications Link Design Handbook , type Technical Report , JPL/NASA Technical Report, year 2005

    author R. Sniffin , title DSMS Telecommunications Link Design Handbook , type Technical Report , JPL/NASA Technical Report, year 2005

    work page 2005

  8. [8]

    a tzold , author F. Neubauer , author L. Carone , author A. Hagermann , author C. Stanzel , author B. H \

    author M. P \"a tzold , author F. Neubauer , author L. Carone , author A. Hagermann , author C. Stanzel , author B. H \"a usler , author S. Remus , author J. Selle , author D. Hagl , author D. Hinson , et al., title Mars: Mars express orbiter radio science , in: booktitle Mars Express: The Scientific Payload , volume volume 1240 , pp. pages 141--163

    work page

  9. [9]

    author D. E. Highsmith , title The effect of USO stability on one-way Doppler navigation of the Mars Reconnaissance Orbiter , type Technical Report , year 2002

    work page 2002

  10. [10]

    Aung , author W

    author M. Aung , author W. Hurd , author C. Buu , author J. Berner , author S. Stephens , author J. Gevargiz , title The block v receiver fast acquisition algorithm for the galileo s-band mission ( year 1994 )

    work page 1994

  11. [11]

    L faldli , author R

    author A. L faldli , author R. Birkeland , title Implementation of a software defined radio prototype ground station for cubesats , in: booktitle Proceedings of the ESA Small Satellites Systems and Services Symposium, Valletta, Malta , volume volume 30

    work page

  12. [12]

    author R. Storn , title Differrential evolution-a simple and efficient adaptive scheme for global optimization over continuous spaces , journal Technical report, International Computer Science Institute volume 11 ( year 1995 )

    work page 1995

  13. [13]

    author T. D. Moyer , title Formulation for observed and computed values of Deep Space Network data types for navigation , volume volume 3 , publisher John Wiley & Sons , year 2005

    work page 2005

  14. [14]

    Price , author R

    author K. Price , author R. M. Storn , author J. A. Lampinen , title Differential evolution: a practical approach to global optimization , publisher Springer , year 2006

    work page 2006

  15. [15]

    author K. J. Price , author P. J. Hanson , author B. J. Whittle , title Stimulation by carbachol of mucus gel thickness in rat stomach involves nitric oxide , journal European journal of pharmacology volume 263 ( year 1994 ) pages 199--202

    work page 1994

  16. [16]

    Storn , author K

    author R. Storn , author K. Price , title Minimizing the real functions of the icec'96 contest by differential evolution , in: booktitle Proceedings of IEEE International Conference on Evolutionary Computation , organization IEEE , pp. pages 842--844

    work page

  17. [17]

    author K. V. Price , title Differential evolution vs. the functions of the 2/sup nd/iceo , in: booktitle Proceedings of 1997 IEEE International Conference on Evolutionary Computation (ICEC'97) , organization IEEE , pp. pages 153--157

    work page 1997

  18. [18]

    Nvidia , title Programming guide , year 2010

    author C. Nvidia , title Programming guide , year 2010

    work page 2010

  19. [19]

    Rupp , title CPU , GPU and MIC hardware characteristics over time , year 2013

    author K. Rupp , title CPU , GPU and MIC hardware characteristics over time , year 2013

    work page 2013

  20. [20]

    Hoberock , author N

    author J. Hoberock , author N. Bell , title Thrust: A parallel template library , year 2010

    work page 2010

  21. [21]

    author J. H. Yuen , title Deep space telecommunications systems engineering , publisher Springer Science & Business Media , year 2013

    work page 2013

  22. [22]

    a tzold , author M. H \

    author M. P \"a tzold , author M. H \"a usler , title Mars Express Orbiter Radio Science MaRS , publisher IGM, Bibliothek , year 2000

    work page 2000

  23. [23]

    Jian , author J.-G

    author N.-C. Jian , author J.-G. Yan , author J.-S. Ping , author J.-P. Barriot , author J. A. P. Rodriguez , title A new method to determine the gravity field of small bodies from line-of-sight acceleration data , journal Research in Astronomy and Astrophysics volume 19 ( year 2019 ) pages 048

    work page 2019

  24. [24]

    author T. D. Moyer , title Mathematical formulation of the double-precision orbit determination program (dpodp). , journal Mathematical formulation of the Double-Precision Orbit Determination Program (DPODP)., by Moyer, TD. NASA, Jet Propulsion Lab., California Inst. Technology, Pasadena, California. Techn. Report 32-1527, 10+ 160 pp. volume 1 ( year 1971 )

    work page 1971

  25. [25]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

    work page