Evaluation of Simplified Methodology for Obtaining mmWave MIMO Channels from Ray-Tracing Simulations

Aldebaro Klautau; Brenda Vilas Boas; Isabela Trindade

REVIEW 4 major objections 4 minor 10 references

Evaluation of Simplified Methodology for Obtaining mmWave MIMO Channels from Ray-Tracing Simulations

T0 review · 4 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that a simplified method for building mmWave MIMO channels from omnidirectional ray-tracing simulations is accurate only when the transmitter and receiver are far apart.

desk verdict A small, honest conference paper on a practical ray-tracing shortcut; the distance-dependent conclusion is plausible but not yet supported because the error metric is undefined and LOS/NLOS is confounded with distance. read the letter →

arxiv 1908.07126 v1 pith:VJOOKJTV submitted 2019-08-20 eess.SP

Isabela Trindade , Brenda Vilas Boas , Aldebaro Klautau This is my paper · ORCID

classification eess.SP

keywords MIMOmmWaveschannelrepresentationraytracinggeometricmodel60GHzvehicle-to-infrastructure

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tests a common time-saving shortcut: instead of running ray-tracing with antenna arrays at both ends, run one omnidirectional simulation and build the MIMO channel afterwards by attaching a steering vector to each ray path through Eq. (1). It compares that reconstructed channel with a full MIMO ray-tracing simulation at 60 GHz in an urban canyon vehicle-to-infrastructure scenario with four-element linear arrays. The result is distance-dependent: the closest line-of-sight receiver shows the largest error, while a distant non-line-of-sight receiver matches well, and the paper concludes that for long distances the simplified model gives similar results to the full simulation. The practical value is that one omnidirectional ray-tracing dataset can be reused to generate many array configurations without rerunning expensive simulations, as long as the links are long enough.

What carries the argument

The load-bearing object is Eq. (1), the narrowband geometric channel model (also called the virtual or angular channel model). It constructs the $N_{rx} \times N_{tx}$ MIMO channel matrix as $\hat{H}_{mn}=\sqrt{N_{tx}N_{rx}}\sum_{\ell=1}^{L}\alpha_\ell a_r(\phi^A_\ell,\theta^A_\ell)a^H_t(\phi^D_\ell,\theta^D_\ell)$, where the ray complex gains and angles come from an omnidirectional ray-tracing run and the steering vectors supply the array response. The simplified methodology's whole bet is that this post-processing reproduces the full MIMO ray-tracing simulation; the paper's comparison isolates exactly that bet, with the angular separation of rays determining the rank of the reconstructed channel.

What would settle it

Repeat the same V2I comparison at a fixed short distance with varying array size and element spacing, and with a full-wave antenna model on the array: if the simplified channel matches the full MIMO ray-tracing channel there, the claim that distance controls accuracy is wrong; if it does not, the paper's caution is confirmed.

Watch

Extended reading notes

Core claim

The central discovery, on the paper's own terms, is that the narrowband geometric channel model of Eq. (1)—summing complex ray gains times receive and transmit steering vectors—recovers the full ray-tracing MIMO channel matrix well for distant transmitter-receiver pairs in an urban V2I setting, but poorly for a nearby line-of-sight receiver. The paper reports the error across receiver positions and finds that the distant receiver without line of sight has the smallest error, while the closest line-of-sight receiver has the largest. It therefore states that long distances make the simplified results similar to the full simulation, while short links require care. The capacity comparison for the close line-of-sight receiver and the low-error non-line-of-sight receiver shows the approximation remains valid in the low-SNR regime.

Load-bearing premise

The simplified method assumes that the multipath rays collected by an omnidirectional simulation—gains, angles, delays—are sufficient to reconstruct the true MIMO channel with steering vectors, so array-induced effects such as mutual coupling, element patterns, and near-field behavior can be ignored; if those effects matter, the observed distance pattern may reflect model mismatch rather than a general rule.

Editorial extensions

If this is right

A single omnidirectional ray-tracing run can be reused to produce MIMO channels for many array sizes and orientations, since Eq. (1) does the array work in post-processing.
For long V2I links, capacity estimates from the simplified model track the full MIMO ray-tracing simulation, especially at low signal-to-noise ratio.
For short-range line-of-sight links, simplified channel matrices can mislead: in the paper's setup, the closest receiver gives the largest reconstruction error, so results from that regime should be treated cautiously.
The rank of the reconstructed channel is set by the angular separation of the rays, so the fidelity of the method is tied not just to distance but to how well the clusters are angularly separated.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Distance is likely a proxy: what really controls the error may be angular resolution of clusters, so a close link with widely separated, well-resolved paths could behave like a long link; the paper's distance rule would then be the special case.
The datasets built with this shortcut for machine-learning beam selection inherit its distance bias, so learned models may underperform precisely on short-range links unless trained with full simulations or measured data.
At larger array apertures or higher frequencies, the array-induced effects Eq. (1) omits grow, so the safe-distance threshold should shift; a parameter sweep over frequency, array size, and element spacing would map the boundary.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague—

Short version: this is a legitimate preliminary study of a question that matters if you generate mmWave MIMO datasets from ray tracing: can you simulate with omnidirectional antennas and reconstruct the MIMO response with steering vectors later? They compare that shortcut (Eq. 1) against InSite's full array simulation in a 60 GHz urban-canyon V2I setup. The answer they get is "usually, but the match gets worse at close range." The paper is honest that it's preliminary, and the comparison itself is not circular—the full RT run is an independent benchmark, no parameters are fitted to it.

What's new is narrow: the same simplified methodology already appears in Arikawa and Karasawa and in the authors' own earlier V2I work, but I don't know of a direct check at 60 GHz V2I with a specific distance trend. That makes the paper a useful data point, not a new technique. The writing is clear, the setup is standard, and the authors don't oversell.

Where it gets wobbly: the plotted "Error (%)" in Fig. 1 is never defined. Without knowing whether it's per-element gain, normalized Frobenius, chordal distance, or something else, "similar results" has no quantitative meaning. Second, the distance conclusion rests on too few points, and the two extremes are a close LOS receiver (worst) and an NLOS receiver (best). In a canyon at 60 GHz that is exactly the sort of confound that could make distance a proxy for link condition. They mention the NLOS point in passing, which is good, but the conclusion in Section IV still states a distance rule without separating the two. The Eq. (1) idealizations (no mutual coupling, point-element patterns) are real, but the stress test is right that they're probably not the driver at 4-element ULA scales; they don't naturally scale with distance.

Who is this for? People building V2I or mmWave MIMO training data with RT and wanting to know when the quick-and-dirty method is safe. For that audience the paper is worth reading as a caveat. I would not cite it as a solid result until the metric is defined and a few more receivers or scenarios separate distance from LOS/NLOS. But the question is a good one for a serious referee, especially with a request to add the metric definition and a couple of scatter or box plots that separate LOS/NLOS.

Recommendation: engage with it, but treat it as a preliminary conference result, not as a settled distance law.

Best

Referee Report

4 major / 4 minor

Summary. The paper compares two ways of obtaining mmWave MIMO channel matrices from ray-tracing simulations: a full simulation in which uniform linear arrays are explicitly modeled inside the ray-tracing software, and a simplified methodology in which omnidirectional antennas are simulated and array responses are added in post-processing through the geometric channel model in Eq. (1). The comparison is carried out in a 60 GHz urban-canyon V2I scenario with 4-element ULAs at both link ends. The authors report, in Section III and Fig. 1, that the closest receiver has the largest error while the receiver with the smallest error is in NLOS, and conclude in Section IV that the simplified model gives similar results to the full simulation for long transmitter-receiver distances, with care needed for closer links. The paper explicitly labels the results as preliminary and defers a systematic assessment to future work.

Significance. If the central claim holds, the simplified methodology would be practically valuable: it allows MIMO channel matrices to be generated from a single omnidirectional ray-tracing run for arbitrary array sizes and orientations, reducing simulation time. The paper addresses a real methodological question for mmWave MIMO studies, and the comparison setup is appropriate in that the full RT simulation is an independent benchmark and no parameters are fitted to it. The strengths of the paper are the clarity of the underlying idea, the explicit channel model in Eq. (1), and the honest statement of the preliminary nature of the results. However, as presented, the evidence is insufficient to establish the distance-dependent conclusion because the error metric is undefined, the distance trend is confounded with LOS/NLOS link condition, and the study covers only one scenario and one array size.

major comments (4)

[Section III, Fig. 1] The quantity plotted as 'Error (%)' in Fig. 1 is never defined. Without an operational definition of the error between the full RT channel matrix and the simplified-model channel matrix, the claim that 'the simplified model gives similar results' has no precise meaning. Different error measures, such as per-element gain error, normalized Frobenius-norm error, or capacity loss, can rank receiver positions differently and yield different distance thresholds. The paper must state the exact formula used for the reported percentages.
[Section III, Fig. 1] The distance-based conclusion is confounded with link condition. The closest receiver, RX 6, is in LOS and has the largest error, while the receiver with the lowest error, RX 10, is in NLOS. In a 60 GHz urban-canyon scenario, LOS/NLOS status strongly affects path loss, angular spread, and multipath richness, so the observed error pattern could be a link-condition effect rather than a distance effect. To support the conclusion that distance is the governing factor, the authors should compare LOS receivers at multiple distances or otherwise control for LOS/NLOS status.
[Section III, Section IV] The generality of the conclusion is not supported by the presented evidence: only one scenario, one array geometry (4-element ULA), and one frequency are considered, and the results come from a single set of receiver locations without error bars or statistical tests. The paper itself acknowledges in Section IV that a 'systematic assessment' is future work; that admission is accurate but means the central claim in the conclusions goes beyond what Sections II and III establish. At minimum, the conclusions should be rephrased as a hypothesis to be tested in the planned systematic study, or the reported results should be accompanied by a metric that quantifies uncertainty.
[Section II-B, Eq. (1)] Eq. (1) reconstructs the MIMO channel by applying ideal steering vectors to rays obtained from omnidirectional antennas, implicitly neglecting mutual coupling, element patterns, and near-field effects. For the specific 4-element ULA at 60 GHz these effects are plausibly small, but the paper does not state this assumption or provide any check. Since the full simulation would include such array-induced effects, the authors should either explicitly list the idealization as a known limitation or provide a concrete test of its impact, for example by comparing against a full RT simulation with realistic element patterns at the shortest simulated distance.

minor comments (4)

[Fig. 1] The caption 'Bar distribution of the errors' is vague: it does not say what each bar represents, how the receiver distances are grouped, or why the abscissa is labeled 'Distance(m)' while the figure is described as a bar distribution. The caption should describe the construction of the bars.
[Fig. 2] The capacity plot lacks axis labels and a definition of the SNR used in the comparison. The text states the approximation is 'still valid for low SNR regime,' but without specifying how capacity is computed and at which SNR values, this claim cannot be checked.
[Section III] The text states that 'it could be inferred that it is less difficult to estimate the channel from the receiver far away' based on a few receiver points; this inference is presented more strongly than the data support. A more measured wording would avoid overstating the observational evidence.
[Section II-A] The manuscript alternates between 'full RT MIMO simulation' and 'all-MIMO simulation' (in the introductory paragraph of Section II) for the same method; one consistent term should be used throughout.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: Eq. (1) is an independent post-processing model compared against a full RT benchmark, with no fitted parameters.

full rationale

The paper's central comparison is between two independently computed channel matrices: the full MIMO ray-tracing simulation (Section II-A) and a simplified geometric-model reconstruction (Section II-B, Eq. (1)) that uses omnidirectional ray data as input. No parameter of the simplified model is fitted to the full-simulation output, and the error and capacity comparisons in Section III are genuine external benchmarks rather than restatements of the model's inputs. The only self-citation is the reuse of the V2I scenario from reference [4], which is co-authored by one of the present authors, but this scenario is an input setting, not a load-bearing derivation of the result. The conclusion that accuracy improves at larger distances is a reported empirical observation, not a consequence of the model's definition. The main weaknesses are methodological (the error metric in Fig. 1 is undefined, LOS/NLOS is confounded with distance, and the results are explicitly preliminary), but those are correctness or support concerns, not circularity. Accordingly, the circularity score is low.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The central claim rests on the validity of the geometric channel model, the accuracy of ray-tracing outputs, and the assumption that array effects can be postponed to post-processing. No new physical entities are introduced.

free parameters (1)

L (number of most prominent rays per TX/RX pair) = not specified
Eq. (1) sums over L rays; L is chosen by the simulator or authors but never reported, and it directly controls the accuracy of the simplified channel reconstruction.

assumptions (3)

domain assumption Narrowband geometric channel model (Eq. 1) with L prominent rays is a valid representation of the true MIMO channel for arbitrary array configurations.
The simplified method's accuracy is judged against full RT via this model; if the model is incomplete, the distance trend reflects model error rather than methodology. Invoked in Section II-B Eq. (1).
domain assumption Ray-tracing software (Remcom InSite) outputs accurate ray parameters (gains, angles, delays) for omnidirectional antennas.
The paper relies on RT accuracy citing [6],[7] but does not validate against measurements. Invoked in Section II.
ad hoc to paper Array-induced effects such as mutual coupling, element patterns, and near-field effects are negligible or identical between full and simplified simulations.
The simplified methodology ignores these effects; the paper does not discuss their influence on error. Invoked in Section II-B Eq. (1).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Evaluation of Simplified Methodology for Obtaining mmWave MIMO Channels from Ray-Tracing Simulations." pith.science (2026). https://pith.science/paper/VJOOKJTV

@misc{pith2026190807126,
  author       = {Pith},
  title        = {Pith review of: Evaluation of Simplified Methodology for Obtaining mmWave MIMO Channels from Ray-Tracing Simulations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VJOOKJTV}},
  note         = {Machine review of arXiv:1908.07126}
}

read the original abstract

The use of higher frequencies and MIMO is important in many 5G use cases. However, the available channel models for millimeter waves (mmWaves) currently demand investigation and the number of measurements is still limited. Using simulators is a current practice in mmWave MIMO research and ray tracing is considered one of the most accurate techniques. Due to the relatively long time of ray tracing simulations, it is common practice to adopt a simplified simulation methodology in which omnidirectional antennas are simulated and, later, the results are used together with a geometrical model to consider that antenna arrays were used. This allows flexibility and decreases the overall time spent with simulations. This paper investigates the corresponding assumptions and how accurate are the results of the simplified methodology when compared to effectively using antenna arrays in the ray tracing simulation. The preliminary results indicate that the distance between transmitter and receiver needs to be sufficiently large.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references · 10 canonical work pages

[1]

R. W. Heath, N. Gonz\'alez-Prelcic, S. Rangan, W. Roh, and A. M. Sayeed, ``An Overview of Signal Processing Techniques for Millimeter Wave MIMO Systems ,'' vol. 10, no. 3, pp. 436--453, Apr. 2016

work page 2016
[2]

A. M. Sayeed, ``Deconstructing multiantenna fading channels,'' IEEE Transactions on Signal Processing, vol. 50, no. 10, pp. 2563--2579, 2002

work page 2002
[3]

Tse and P

D. Tse and P. Viswanath, Fundamentals of Wireless Communication. 1em plus 0.5em minus 0.4em Cambridge University Press, 2005

work page 2005
[4]

Klautau, P

A. Klautau, P. Batista, N. Gonzalez-Prelcic, Y. Wang, and R. Heath, ``5 G MIMO D ata for M achine L earning: A pplication to B eam- S election using D eep L earning,'' in 2018 Information Theory and Applications Workshop ( ITA ) , Feb. 2018. [Online]. Available: http://ita.ucsd.edu/workshop/18/files/paper/paper_3313.pdf

work page 2018
[5]

V. Va, J. Choi, T. Shimizu, G. Bansal, and R. W. Heath, ``Inverse Multipath Fingerprinting for Millimeter Wave V2I Beam Alignment ,'' 2017, IEEE Early access

work page 2017
[6]

T. S. Rappaport, R. W. Heath, R. C. Daniels, and J. N. Murdock, Millimeter Wave Wireless Communications. 1em plus 0.5em minus 0.4em Prentice Hall, 2014

work page 2014
[7]

Stabler and R

O. Stabler and R. Hoppe, `` MIMO channel capacity computed with 3D ray tracing model,'' in 3rd European Conf. Antennas Propagation , Mar. 2009, pp. 2271--2275

work page 2009
[8]

Arikawa and Y

S. Arikawa and Y. Karasawa, ``A Simplified MIMO Channel Characteristics Evaluation Scheme Based on Ray Tracing and Its Application to Indoor Radio Systems ,'' vol. 13, pp. 1737--1740, 2014

work page 2014

Show all 10 references

[9]

U bertragungskanal; MIMO , abstract = Das enorme Wachstum des mobilen Datenaufkommens wird zu substantiellen Ver \

11em plus .33em minus .07em 4000 4000 100 4000 4000 500 `\.=1000 = #1 \@IEEEnotcompsoconly \@IEEEcompsoconly #1 * [1] 0pt [0pt][0pt] #1 * \| ** #1 \@IEEEauthorblockNstyle \@IEEEcompsocnotconfonly \@IEEEcompsocconfonly \@IEEEauthorblockAstyle \@IEEEcompsocnotconfonly \@IEEEcomp...
[10]

11em plus .33em minus .07em @technote 4000 4000 100 4000 4000 500 `\.=1000 = #1 #1 #1 0pt [0pt][0pt] #1 * \| ** #1 \@IEEEauthorblockNstyle \@IEEEauthorblockAstyle \@IEEEauthordefaulttextstyle \@IEEEauthorblockconfadjspace -0.25em \@IEEEauthorblockNtopspace 0.0ex \@IEEEauthorbl...

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

R. W. Heath, N. Gonz\'alez-Prelcic, S. Rangan, W. Roh, and A. M. Sayeed, ``An Overview of Signal Processing Techniques for Millimeter Wave MIMO Systems ,'' vol. 10, no. 3, pp. 436--453, Apr. 2016

work page 2016

[2] [2]

A. M. Sayeed, ``Deconstructing multiantenna fading channels,'' IEEE Transactions on Signal Processing, vol. 50, no. 10, pp. 2563--2579, 2002

work page 2002

[3] [3]

Tse and P

D. Tse and P. Viswanath, Fundamentals of Wireless Communication. 1em plus 0.5em minus 0.4em Cambridge University Press, 2005

work page 2005

[4] [4]

Klautau, P

A. Klautau, P. Batista, N. Gonzalez-Prelcic, Y. Wang, and R. Heath, ``5 G MIMO D ata for M achine L earning: A pplication to B eam- S election using D eep L earning,'' in 2018 Information Theory and Applications Workshop ( ITA ) , Feb. 2018. [Online]. Available: http://ita.ucsd.edu/workshop/18/files/paper/paper_3313.pdf

work page 2018

[5] [5]

V. Va, J. Choi, T. Shimizu, G. Bansal, and R. W. Heath, ``Inverse Multipath Fingerprinting for Millimeter Wave V2I Beam Alignment ,'' 2017, IEEE Early access

work page 2017

[6] [6]

T. S. Rappaport, R. W. Heath, R. C. Daniels, and J. N. Murdock, Millimeter Wave Wireless Communications. 1em plus 0.5em minus 0.4em Prentice Hall, 2014

work page 2014

[7] [7]

Stabler and R

O. Stabler and R. Hoppe, `` MIMO channel capacity computed with 3D ray tracing model,'' in 3rd European Conf. Antennas Propagation , Mar. 2009, pp. 2271--2275

work page 2009

[8] [8]

Arikawa and Y

S. Arikawa and Y. Karasawa, ``A Simplified MIMO Channel Characteristics Evaluation Scheme Based on Ray Tracing and Its Application to Indoor Radio Systems ,'' vol. 13, pp. 1737--1740, 2014

work page 2014