Page 1 of 35

Transactions on Networks and Communications - Vol. 9, No. 3

Publication Date: June, 25, 2021

DOI: 10.14738/tnc.93.10214.

Neelakanta, P., & Groff, D. D. (2021). Conceiving Inferential Prototypes of MIMO Channel Models via Buckingham’s Similitude

Principle for 30+ GHz through THz Spectrum. Transactions on Networks and Communicaitons, 9(3). 1-35.

Services for Science and Education – United Kingdom

Conceiving Inferential Prototypes of MIMO Channel Models via

Buckingham’s Similitude Principle for 30+ GHz through THz

Spectrum

Perambur Neelakanta

Department of Computer and Electrical Engineering and Computer Science

Florida Atlantic University, Boca Raton, FL 33431, United States

Dolores De Groff

Department of Computer and Electrical Engineering and Computer Science

Florida Atlantic University, Boca Raton, FL 33431, United States

ABSTRACT

Facilitating newer bands of ‘unused’ segments (windows) of RF spectrum falling in

the mm-wave range (above 30+ GHz) and seeking usable stretches across

unallocated THz spectrum, could viably be considered for Multiple Input Multiple

Output (MIMO) communications. This could accommodate the growing needs of

multigigabit 3G/4G applications in outdoor-based backhauls in picocellular

networks and in indoor-specific multimedia networking. However, in contrast with

cellular and Wi-Fi, wireless systems supporting sub-mm wavelength transreceive

communications in the outdoor electromagnetic (EM) ambient could face

“drastically different propagation geometry”; also, in indoor contexts, envisaging

pertinent spatial-multiplexing with directional, MIMO links could pose grossly

diverse propagation geometry across a number of multipaths; as such, channel- models based on stochastic features of diverse MIMO-specific links in the desired

test spectrum of mm-wave/THz band are sparsely known and almost non-existent.

To alleviate this niche, a method is proposed here to infer sub-mm band MIMO

channel-models (termed as “prototypes”) by judiciously sharing “similarity” of

details available already pertinent to traditional “models” of lower-side EM

spectrum, (namely, VLF through micro-/mm-wave). Relevant method proposed

here relies on the “principle of similitude” due to Edgar Buckingham. Exemplar set

of “model-to-(inferential)-prototype” transformations are derived and prescribed

for an exhaustive set of fading channel models as well as, towards estimating path- loss of various channel statistics in the high-end test spectrum.

Keywords: Channel/propagation-models, mm-wave/THz spectrum, Similitude principle,

Wireless communication

INTRODUCTION

With alarming increase in radio-frequency (RF) spectrum utilization and almost fully occupied

status of available electromagnetic (EM) spectral windows, facilitating newer ranges of RF- spectrum for wireless communication is an explicit necessity, so as to accommodate the

growing needs of broad-bandwidth wireless applications. For example, enabling newer ranges

of ‘unused’ segments (windows) of RF spectrum in the mm-wave band (above 30+ GHz) and

Page 2 of 35

2

Transactions on Networks and Communications (TNC) Vol. 9, Issue 3, June-2021

Services for Science and Education – United Kingdom

seeking usable stretches across the unallocated THz spectral range could possibly considered

for Multiple Input Multiple Output (MIMO) communications in accommodating the growing

needs of multigigabit 3G/4G applications enclaving outdoor-based backhauls in picocellular

networks and indoor-specific multimedia networking. However, in contrast with those of

cellular and Wi-Fi systems, usage of smaller, carrier wavelengths (of mm-wave/THz bands) in

outdoor EM ambient could pose “drastically different propagation geometry”; and, in indoor

contexts, pertinent directional mm/THz wave propagations across spatially-multiplexed, line- of-sight (LoS) MIMO links may also face grossly diverse propagation geometry with multiple

number of paths plus nodes of relatively small form-factors. Further, moderately separated

antennas facilitating such directional links, could lead to sparse EM scattering; but, increased

propagation loss in the transreceive paths is inevitable with smaller wavelengths [1] [2].

MIMO techniques, in general, are adopted to combat fading; and, suppose carrier frequencies

falling in the mm/THz wave bands are prescribed for such MIMO systems in order to achieve

multigigabit 3G/4G applications in outdoor-based backhauls (of picocellular networks) and in

indoor-specific multimedia networking. The associated EM propagation artifacts and

interference characteristics of such MIMO-links would pose unique stochastic features and

distinct propagation characteristics thanks to (small) carrier wavelength versus spacing

between the antenna elements and limited EM-scattering involved. Such details could render

the elements of the MIMO-channel matrix not always being independent.

Use of mm/THz wave spectrum hardly prevails in the state-of-the-art communication areas

(including MIMO systems); as such, relevant neoteric paradigms of spectrum utilization consist

of only a limited/sparse and/or mostly insufficient data detailing the underlying EM wave

propagation characteristics. Thus, a lacuna prevails in the comprehension of channel-models at

this spectral range of interest. As such, this study is motivated to address the sparsity of MIMO

channel modeling vis-à-vis fading and path-loss features; and, the scope of this paper is set to

explore the feasibility of conceiving relevant “inferential prototypes” of channel-models in the

said EM spectral range (for possible adoption in MIMO contexts). The proposed method is based

on judiciously sharing the “similarity” of details between already existing (known) “models” of

traditional, lower-side EM spectrum, (namely, VLF through micro-/mm-wave) availed as ex

post data and the corresponding “prototypes” are conceived as ex ante details inferred for the

test-bands (of mm-wave through THz spectrum). Relevant method relies on the so-called

“principle of similitude” due to Edgar Buckingham [3]. Illustrative “model-to-(inferential)-

prototype” transformations are derived and discussed with reference to an exhaustive set of

channel-models on fading and path-loss characteristics pertinent to diverse channel statistics.

Commensurate with the objectives indicated above, this paper is organized as follows: The next

section (Section 2) reviews the general aspects of wireless channel-models and summarizes the

specific aspects of MIMO-channels. Further, focused considerations on such channel

characterizations in the 30+ GHz through THz spectrum are outlined and associated concerns

are identified. In Section 3, the proposed heuristics of inferring “prototypes” of EM propagation

characteristics required at the high-end spectral range (of mm-wave/THz band) using the prior

known (available) “models” at lower bands (like UHF/microwaves) is described. Relevant

systematic procedure in judiciously projecting the associated “similarity” of details in the

“models” (of low-frequency end) into “inferential prototypes” of high-end RF-channels

(representing the test-bands of interest) is described in Section 4 in terms of the heuristics of

Page 3 of 35

3

Neelakanta, P., & Groff, D. D. (2021). Conceiving Inferential Prototypes of MIMO Channel Models via Buckingham’s Similitude Principle for 30+ GHz

through THz Spectrum. Transactions on Networks and Communicaitons, 9(3). 1-35.

URL: http://dx.doi.org/10.14738/tnc.93.10214

the so-called principle of similitude (advocated by Edgar Buckingham [3]). The similitude

considerations involved in the model-to-(inferential) prototype transformations are elaborated

in Section 5 and applied to illustrate fading statistics and path-loss characteristics of channel- models at the test frequency bands of interest, (namely, 30+ GHz through THz EM spectrum).

Section 6 lists the well-known “models” of fading channels with the associated parameters;

hence, the corresponding deduced “inferential prototypes” are presented with descriptive

notes on an exhaustive set of fading statistics. Similarly, “model-to-(inferential)-prototype”

transformations are illustrated to deduce path-loss characteristics the channels in the test

bands of interest. Discussions on the proposed methods are furnished in Section 7 and

pertinent remarks are given in the closure section (Section 8).

WIRELESS- AND MIMO-CHANNEL MODELS: AN OVERVIEW

Modeling wireless communication ambient describing the associated RF propagation and

channel characteristics has been the topic of interest with the advent of mobile communications

and extensive proliferation of cellular networks. Relevant details enclaving transmissions at

traditional HF through UHF/microwave frequencies are addressed exhaustively in the

literature [4-8]. Associated channel-models reflect the propagation characteristics of signals in

pertinent radio environments are essentially needed for evaluating the performances of

wireless communication systems. Typically, commencing from the simple free-space model,

channel characteristics of both outdoor and indoor contexts are indicated for traditional

wireless systems in a gamut of published efforts. To name a few, are the so-called: Lee’s model,

Langley-Rice model, Okumura model, Hata model, Walfish and Bertoni model, Hashemi model,

Rappaport’s SIRCIM model, Saleh and Valenzula model and Seidal and Rappaport model [4-8].

In wireless systems, the EM propagation ambient supporting communication links decides the

integrity of transreceive signals largely specified by the underlying fading and path-loss

characteristics. Fading implies variations (with stochastic attributes) seen in the signal

attenuation versus time, locales of transreceive units and carrier frequency. That is, fading

denotes a random process caused by probabilistic variates on EM propagation due to multipath

and/or shadowing effects from the obstacles. Path-loss is another channel artifact denoting the

EM energy losses encountered in wireless communications that incur on terrestrial paths

through atmospheric environment as well as, in the content-infested indoor ambient.

In all, due to fading and path-loss characteristics, the transmitted wireless signal across its

transceive path would invariably seen morphed at the receiving end. Relevant profile of such

received signals can then be specified in terms of the details on the transmitted signal (known

a priori), if a model for the medium/channel between the transceive entities is also available.

Such characterization of RF signal transit-paths is well-known as a “channel-model”; and, it in

essence, represents the EM propagation profile of the signal and the associated statistics of

signal-strength fluctuations across the channel in question; and, a number of channel-models

have been evolved as listed earlier. The pros and cons of channel considerations in realizing

practical RF communication links are fairly comprehended in the traditional wireless systems

pertinent to classical deployment of VLF through microwave frequencies.

The perspectives of channel-models become even more unique and visibly complex in the

contexts of multiple-input multiple-output (MIMO) technique ushered into wireless

communications in order to achieve a significant increase in data throughput as well as, link

Page 4 of 35

4

Transactions on Networks and Communications (TNC) Vol. 9, Issue 3, June-2021

Services for Science and Education – United Kingdom

reliability (mostly without extra bandwidth or boosting transmission power) [9-11]. MIMO

together with orthogonal frequency division multiplexing (OFDM) has proliferated as a key

technique in the long-term evolution (LTE) of 3G/4G cellular networks. In its conception to

achieve higher throughputs within a given bandwidth via space diversity schemes, the MIMO- links were typically modeled in the early theoretical studies (due to Telatar [12] and Foschini

[13]) as narrowband random MIMO channels. Such efforts refer to evolving a simple channel- model represented by a matrix [H] with M-transmit and N-receive antennas.

In the expanded applications, MIMO channel models have been subsequently framed for

different EM environments; and, comparison between channel-models is done on the basis of

the associated theoretical considerations versus measured data. Relevantly, two kinds of

channel models, namely, correlation-based stochastic models (CBSMs) and geometry-based

stochastic models (GBSMs), are prescribed (commensurate with IEEE 802.11n standards) in

order to evaluate the performances of the wireless communication systems; and, the former

being less complex is typically used in theoretical evaluations of MIMO system performance.

However, the accuracy of CBSM is limited for wireless channels exhibiting non-stationary

phenomenon and spherical wave effects. In contrast, GBSM is considered as being accurate in

portraying the realistic channel properties of massive MIMO channel applications (in spite of

the associated higher computation complexity).

MIMO techniques in general are indicated for combating fading. The underlying channel

characterizations, however are unique and mostly differ from those of non-MIMO strategies,

even at lower carrier frequencies. As such, exclusive MIMO-channel models have been

developed. Typically, the MIMO channel characterization is exemplified via a lamp-post based

outdoor deployment model (for example, as in a mesh backhaul), which duly accounts for

fading due to ground and wall reflections; and, in indoor contexts, the MIMO infrastructure with

space diversity and spatial multiplexing (for example, as done in “streaming high-definition

television from a set-top box to a television set”), relevant (MIMO)-channel model becomes

even more unique with a number of eigen modes framing the form-factor of the channel

involved [1].

The MIMO-specific channel-models have been comprehensively studied at traditional lower

end of EM spectrum [9-11]. Though restrictedly addressed (on ad hoc basis), pertinent channel

considerations and channel-modeling in the millimeter wave regime of RF systems [14]

including the MIMO-links [1] also exist. Yet, such a comprehension is rather sparse and even

absent for the EM spectrum ranging 30+ GHz through THz. However, such details are highly

desirable and imminent in any aggressive suite of implementing the current needs of robust

wireless communication links including MIMO systems. Specifically, MIMO channel-modeling

in the high-end spectral range of 30+ GHz through THz is needed towards profitable use of the

large extents of bandwidth available in the said test bands of interest. Pertinent uses as stated

before refer to realizing multi-Gigabit wireless networks with applications ranging from indoor

multimedia networking to accommodating outdoor backhauls for picocellular networks.

However, using the spectral range of 30+ GHz through THz in such applications, the carrier

wavelengths involved are obviously an order (or orders) of magnitude smaller than those in

the existing cellular and Wi-Fi systems. As such, significantly different propagation geometry

can be foreseen in the underlying EM ambient. For example, omnidirectional transmission is

almost ruled out because of the increased propagation losses encountered at smaller

Page 5 of 35

5

Neelakanta, P., & Groff, D. D. (2021). Conceiving Inferential Prototypes of MIMO Channel Models via Buckingham’s Similitude Principle for 30+ GHz

through THz Spectrum. Transactions on Networks and Communicaitons, 9(3). 1-35.

URL: http://dx.doi.org/10.14738/tnc.93.10214

wavelengths. So, highly directive transreceive MIMO-links can be envisaged with electronically

steerable beams formed by compact antenna arrays. As a result, only a small number of paths

would dominantly prevail (in such directional small-wavelength links), posing a sparse

scattering ambient unlike, the rich scattering environment at lower carrier frequency

operations. Usage of small wavelengths also implies achieving spatial-multiplexing gains even

in LoS transreceive links or more generally, in the ambient of sparse scattering consistent with

the dispersed of antennas having moderate separations.

As stated earlier, knowledge on EM propagation characteristics (or channel models) in the said

mm/THz bands (henceforth addressed as the “test-bands”) is still primitive and mostly

unknown for applications in MIMO-systems or otherwise. Filling this niche will conform to

transformational improvements in using the radio spectrum of test-bands containing ‘unused’

segments (windows) in mm-band (above 30+ GHz) and usable stretches across the unallocated

THz frequencies. Hence, it will add a dimension of realism in conceiving the exploding

broadband communication needs (including novel MIMO applications), so as to support the

associated traffic intensity of triple-play wireless communication services in the near-future

and beyond.

Hence proposed in this paper is a feasibility consideration on a very practical problem: How

can channel/propagation models be inferentially deduced for the spectral range of test-bands

from the existing (available) knowledge on the lower strata of EM spectrum? And the solution

sought thereof, is to find one-to-one correspondence between the channel characteristics

(dubbed as “models”) known ex post at low-frequencies and corresponding ‘scaled-up’ channel

features (called “prototypes”) attributed ex ante, to higher (test) frequencies of interest.

INFERRING “PROTOTYPES” FROM “MODELS”: PROPOSED METHOD AND APPROACH

As indicated above, the present study objectively proposes an approach to develop

“prototypes” of EM propagation and channel characterizations of RF signal transmissions in the

test frequency bands covering the new frontiers of (almost) unused 30+ GHz spectral windows

and mostly unexplored terahertz span (above 275 GHz sans spectrum allocations). As briefed

above, relevant concept is illustrated in Figure 1.

The proposed approach towards developing “prototypes” of EM propagation and related

channel characterizations of RF signal transmissions in the spectral bands of (almost) unused

30+ GHz frequencies indicated above (and illustrated in Figure 1) is summarized as follows:

▪ A set of channel details are presumably known a priori at lower frequency bands

(like UHF/microwaves). The set of such ex post features are designated as the

“model”

▪ The said “model” details are listed and judiciously projected to frame a “inferential

prototype” at the desired high-end frequency (test-band); and, the ex ante set of

features so conceived denote an inferred “prototype” that shares the “similarity” of

all the pertinent details in the “model”

▪ The projection of shared similarity indicated above is done via a systematic

procedure based on the so-called principle of similitude advocated by Edgar

Buckingham [3] in terms of dimensional analyses perspectives detailed in the next

section.

Page 6 of 35

6

Transactions on Networks and Communications (TNC) Vol. 9, Issue 3, June-2021

Services for Science and Education – United Kingdom

Figure 1. Specifying a “model” function denoting a set {a, b, ...} of known data on channel

attributes at lower-end frequency spectrum (fm) towards formulating a corresponding -

function (for use to infer the “prototype” function as indicated in Step 2). The set {ao, bo...,}

denotes reference entities that render the variable set {a, b, ...,} dimensionless; and, {(m1, m2,

...)} are pertinent exponents prescribed to frame the -relations consistent with Buckingham’s

similitude principle.

The steps in conceiving “inferential prototypes” of RF channel characteristics in a desired test- band using the “model” data available at lower frequencies are as follows:

o Step 1: Construction of the “model”: This is illustrated in Figure 1. The channel

parameters {a, b, ...} known a priori at lower frequencies, fm (say, HF through

microwaves) are first identified and specified as the “model” features denoted

by a functional relation: Func(fm: a, b,...). This model function is then modified

and rewritten, following the so-called as Buckingham’s similitude principle

described in the next section. Relevantly, a Buckingham π-function of the

model is specified as: g[(fm/fo): (a/ao)m1, (b/bo)m2...], where fo is a reference

frequency prescribed to normalize, fm; and, {ao, bo, ...} etc. are normalizing

parameters indicated respectively, for the elements of the set {a, b,...}; and,

the set {m1, m2, ...} depicts appropriate exponents indicated for the aforesaid

-relation. This “model” prescription as above depicts ex post details of the

channel in question.

103

106

109 Hz

“MODEL

Model function: Func (f

m

: a, b, ...)

Buckingham π-function of the

model:

g{(f

m

/f

o

): (a/a

o

)

m1

, (b/b

o

)

m2...}

Ex post data of the

“model”

Similitude principle

INPUT

(Known ex post regime)

Available channel-characteristics

HF..., VHF ..., UHF ..., Microwave ... ... mm-Wave ...

. THz

Page 8 of 35

8

Transactions on Networks and Communications (TNC) Vol. 9, Issue 3, June-2021

Services for Science and Education – United Kingdom

(p1, p2, ...)} are appropriate exponents imposed on -relations consistent with Buckingham’s

similitude principle (as will be explained in the pursuant section).

BUCKINGHAM’S SIMILITUDE PRINCIPLE

The similitude principle has been deployed extensively in various disciplines, to name a few:

Model-based testing and evaluations of structures in civil, aeronautical/aerospace, naval

architecture [15], reliability predictions [16], radar cross-section (RCS) studies [17] etc. The

underlying considerations of similitude concept can be outlined as follows:

▪ Evolving a “prototype” refers to elucidating a set of new descriptive details formulated

a posteriori and prescribed on an entity (for example, a structure or a phenomenon).

Such inferred (ex ante) details correspond to an existing and available set of ex post data

(referred to as the “model”). Such data is presumably known a priori on the entity in

question. The underlying one-to-one transformation is inferred and interpreted via

similitude considerations based on the theory of modeling consistent with the norms of

dimensional analysis

▪ The underlying theoretics and mathematical description of theory of dimensions

(evolved in terms of dimensions and units) is aptly applied to portray the similitude

aspects of the prior knowledge on the said “model” (depicting the physical entity or

phenomenology) versus the “prototype” posterior information being sought, subject to

the following norms:

o The mathematical description (of transforming a “model” into a “prototype”)

should comply with the laws of nature and specified in a dimensionally

homogeneous form. That is, regardless of the choice of dimensional units (in which

the associated physical variables are measured), the governing equation must be

valid

o All relevant governing equations must be dimensionally homogeneous. That is , an

arbitrary equation of the form, [func(x1, x2, ..., xn) = 0] can be alternatively

expressed as, [g(1, 2,... m) = 0], where the -terms denote dimensionless

products of n physical variables of the set {x1, x2, ..., xn}; and, m = (n – r) with r

depicting the number of fundamental dimensions that are involved in the physical

variables.

Explicitly, the second consideration above namely, [func(x1, x2, ..., xn) = 0]  [g(1,

2,... m) = 0] implies the following: (i) The -form of physical occurrence can be

deduced via proper use of dimensions of the n physical quantities, {xi}i = 1, 2, ..., n by

resorting to dimensional analysis. (ii) In physical systems that differ only in

magnitudes of the units (used to measure the n quantities, namely {xi}i = 1, 2, ..., n ),

the “model” entity and its reduced or enhanced scale-version (“prototype”) will

have identical functional, g(.); and, (iii) similitude requirements for modeling

result from forcing the -terms, {i}i = 1, 2, ..., m to be equal in the “model” as well as

in the “prototype”. (This is a necessary condition for the full functional

relationships to be equal).

In short, a “model” denotes a representation of the physical system with details (known ex post)

that can be adopted to infer the behavior of a corresponding “prototype” in some respect. That

is, the principle of similitude (due to Buckingham [3]) enables predicting the performance of a

Page 9 of 35

9

Neelakanta, P., & Groff, D. D. (2021). Conceiving Inferential Prototypes of MIMO Channel Models via Buckingham’s Similitude Principle for 30+ GHz

through THz Spectrum. Transactions on Networks and Communicaitons, 9(3). 1-35.

URL: http://dx.doi.org/10.14738/tnc.93.10214

new design (called “a prototype”) based on mimicking the data from an existing, similar design

(designated as “the model”).

In such representations, the first consideration is to show that the two systems can be described

in terms of dimensionless numbers as per Buckingham’s heuristics. This can be outlined as

follows: Given any element (m) of the -set of the model and the corresponding element (p)

of the prototype, ideally the similarity principle demands, p/m = 1. But, in practice, at least

some first or second-order deviations may be encountered rendering exact model-to-prototype

extrapolation, somewhat “distorted” [18]. Conceiving an approximated version via piecewise

linearization in similarity relations is, however, feasible as will be indicated later in this paper.

In general, the similitude problems based on Buckingham’s heuristics can be classified as

follows: (i) Geometric similitude, which denotes the “model” in question representing an “exact”

scale-copy of the “prototype” consistent with physical geometry involved; (ii) kinematic

similitude concerned with flow considerations, where the “model” flow is set proportional to

the corresponding flow in the prototype, preserving the associated vector nature of the flowing

entity; and, (iii) dynamic similitude, which refers to forces associated in the “model” being

proportional to the forces seen in the “prototype”. (This is a necessary condition, but not

sufficient for dynamic similitude) [11] [19-23].

Relevant to the present study of applying Buckingham’s similitude principle in the contexts of

channel modeling, first the underlying similitude considerations pertinent to EM theory should

be identified. That is, the associated set of postulations and constitutive relations consistent

with the linearity aspects of Maxwell’s equations [24] should be specified in dimensionless

forms by invoking Buckingham’s -theorem. Then, the EM propagation/channel-modeling is

rendered viable for the similitude approach. Existing studies on applying such similitude

principle in electrodynamics and EM contexts are as follows: (i) Geometrical inferences

(relative to wavelength) such as in antenna designs [25], radar cross-section (RCS) evaluations

[17]; (ii) channel characterizations [26] (studied in a limited extent) and (iii) frequency-scaling

of rain attenuation (such as in ITU-R Rec.P.618-7, 2001 and similar models [27] [28]).

Basically, the kinematic/dynamic similitude is applied in EM contexts considering (i) the flow

of electric current signified by the time-varying states or dynamics of electric charges, (ii) the

rate of change of electric/magnetic flux and (iii) the Poynting vector flow. EM similitude

technique has also been advocated in studying laser scattering [23]. Typically, electrodynamic

similitude is applied to model the nondispersive targets as indicated in [22].

Applying -theorem in EM contexts

In all, the heuristics of adopting Buckingham’s -theorem in the contexts of electromagnetism

(EM) and hence, applying the principle to elucidate the “prototypes” of channel-models vis-à- vis EM propagation characteristics in wireless communications involve the following basics of

EM theory outlined to deduce the associated -relations.

As well known, the EM waves possess a characteristic wavelength length ( in m), which is

related to their frequency (f in Hz) by the relation,  (m) = [v in m/s)/f in Hz] where, v denotes

the velocity of propagation of EM wave in the medium of interest and f is the frequency of the

Page 10 of 35

10

Transactions on Networks and Communications (TNC) Vol. 9, Issue 3, June-2021

Services for Science and Education – United Kingdom

EM wave. For free-space propagation, v ≡ c (= 3  108 m/s) as decided by the associated

constants of the free-space, namely, the absolute permittivity o = (1/36)  10−9 farad/meter

and absolute permeability, o = (4)  10−7 henry/meter; and, c =1/(oo)

1/2 m/s, (= 3  108 m/s

depicting the speed of the light). And, v = c/(rr)

1/2 m/s [24] expressed in terms of

(dimensionless) relative permittivity and permeability, (r, r) of the medium; hence, for free- space, (r =1; r = 1).

An example of elucidating dimensionless parameters, functions, equations etc. in the contexts

of electrodynamics as required in similitude applications is as follows: Considering the well- known Coulomb’s law of force (|F|) between given two charges of q1 and q2 units, separated by

a distance d units in a dielectric medium characterized by a (dimensioned) constant, k, the said

law is specified by the equation, {func(x1, x2, ..., xn) = 0}  func[|F|, k, (q1, q2), d] = 0; explicitly,

this Coulomb’s law can be written as: {[|F| − k q1q2/d2]} = 0. In terms of the associated units

term-by-term, Coulomb’s law can be rewritten as: func{[K1 force unit] − [K2 (capacitance

unit/length unit)−1  K3 (charge unit)2]/[K4(length unit)2] = 0}, where, the func{.} is expressed

in terms of corresponding fundamental measures that describe the associated quantities; and

K1, K2 and K3 are relevant proportionality constants.

The above relation holds for any system of units, (CGS, SI or FPS) for force, capacitance, charge

and length; and as such, it is dimensionally homogeneous. Now, in terms of the notions of

Buckingham [3], the implicit functional relation of Coulomb’s law, namely, func[|F|, k, (q1, q2),

d] = 0 can also be represented as a product of powers in a form as follows: |F| = CD × [k

(q1,

q2)

d

] newton, where CD is a dimensionless parameter, (which by itself could be a function of

dimensionless groupings of some pertinent physical entities; or, it can be a simple constant,

otherwise); and, the exponents  = + 1,  = − 2 and  = + 2 are explicitly pertinent to the

Coulomb’s law under discussion.

Next, the functional relation, F = CD[k

(q1, q2)

d

] can be cast in the -form, {g(   m) = 0}

using similitude principle. Relevantly, consistent with Buckingham’s -theorem, the

dimensionally homogeneous equation, {func[|F|, k, (q1, q2), d] = 0}n = 4 is reduced to an

equivalent equation involving a set of dimensionless products, namely, {g(1, 2, ..., m) = 0}m =

2 ≡ {g(kq2/Fd2, q2F/kd2) = 0}. In other words, the given equation, {func(x1, x2, ..., xn) = 0} 

{func[F, k, (q1, q2), d] = 0}n = 4 implies the same Coulomb’s law specified in a dimensionless

format as: {g(   m) = 0}  {g(kq2/Fd2, q2F/kd2) = 0}m = 2 .

Likewise, all the postulations and constitutive relations of EM theory based on the linearity

aspects of Maxwell’s equations [24] can be specified in dimensionless forms by invoking

Buckingham’s -theorem via relevant algorithmic suites as indicated above. However,

distortions may not be uncommon in such similitude comparisons of EM contexts due to

dissimilarities that may prevail in the overall geometry, boundary/initial conditions and/or

(nonlinear) material (EM medium) implications. In such cases, still a piecewise linearization

can be considered [18] as will be illustrated later.

Similitude considerations on EM propagation channel models

Following the similitude notions narrated above relevant to the contexts of electromagnetics,

relevant -relational details can be developed towards deducing EM propagation profiles and

Page 14 of 35

14

Transactions on Networks and Communications (TNC) Vol. 9, Issue 3, June-2021

Services for Science and Education – United Kingdom

Figure 3. A hypothetical set of -functions, {g[.]} specified across several stretches of frequency

bands, (Band-1, Band-2, ..., Band-j ...); and, each g[.] is approximated by an asymptotic

(piecewise linear) representation. Each frequency band is specified with a value of the

normalized frequency set: {(fm/fo)m = 1, 2, 3,...}.

The variable set {fm; a, b, ...} can further be expressed in a normalized form and rendered

dimensionless with a corresponding set of reference values, {fo; ao, bo, ...}. Each of the resulting

dimensionless variable is also specified with a dimensionless exponent. In all, FUNC(fm; a, b, ...)

 g[fm/fo; (a/ao)

, (b/bo)

, ...]. The function, g[.] denotes the Buckingham’s −function

designated with the exponent set {  ...}.

Now, suppose a hypothetical set of -functions {g[.]} is presumably known ex post relevant to a

spectral range of frequencies. The function g[.] is then approximated by an asymptotic

(piecewise linear) representation and attributed to represent each segmented stretch of

frequency bands across the EM spectral range as shown in Figure 3. (Note: The bands shown

need not however, be continuously and successively adjacent. There could be a window of

unknown features between any two bands wherein, g[.] is unspecified).

Considering a set of successive frequency bands as illustrated in Figure 2, the EM details

prescribed for each band via the function g[.]. That is, g[.] denotes the “models”, namely, model- 1, model-2, model-3 etc. respectively for the bands, Band-1, Band-2, Band-3 etc. Now, for each

frequency segment of the set {(fm/fo)m = 1, 2, 3,...} depicting {(Band)m = 1, 2, 3,...}, the associated

“model” prescription can be transformed into a corresponding “prototype” of the next,

(successive) higher frequency band. This is done via a successive iteration method using

Buckingham’s similitude principle as described in the pseudocode presented in Table 2.

....

Frequency bands: (fm

/f

o

)m = 1, 2, 3,...

Band-2 ...

f

2

Band-j

f

j

Band-1

f

1

g[.]

Page 16 of 35

16

Transactions on Networks and Communications (TNC) Vol. 9, Issue 3, June-2021

Services for Science and Education – United Kingdom

 g[(f1/fo); (a1/ao)

1, (b1/bo)

1;(c1/co)

1, ...]

≡ g[(f2/fo); (a2/ao)

2,(b2/bo)

2;(c2/co)

2, ...]

Next

Deduce

→ [k2×P1(f2/fo)]C

2

 This is a scaled version of [k1×G1(f1/fo)]C

1 with (k1/k2) as a

scaling constant and {C1 and C2} being the exponents that decide

a power-law relation

 Prototype-2 of Band-2

Next

→ Use Prototype-2 as a pseudo- “model” for prototyping Band-3

→ Consider: Band-2 − −format of the g[.] function

 P2: g[(f2/fo); (a2/ao)

2, (b2/bo)

2;(c2/co)

2, ...]

→ Rename P2(.): Prototype - 2 as: Pseudo-model-2

→ Consider: Band-3 - -format of the g[.] function

 P3: g[(f3/fo); (a3/ao)

3, (b3/bo)

3;(c3/co)

3, ...]

→ Designate P3(.) as: Prototype-3

Next

Equate

→ Band-2 - -format of the g[.] function: P2(.) and Band-3 - -format of the g[.]

function: P3(.)

→ g[(f2/fo); (a2/ao)

2, (b2/bo)

2;(c2/co)

2, ...]

≡ g[(f3/fo); (a3/ao)

3, (b3/bo)

3;(c3/co)

3, ...]

Deduce

→ [k3×P3(f3/fo)]C

3

→ This is a scaled version of [k2×P2(f2/fo)]C2 with (k2/k3) being a scaling

constant; and, {C2 and C3} are the exponents that decide the associated

power-law relation

 Prototype-3 of Band-3

Next

→ Use Prototype-3 as a pseudo-model for prototyping Band-4

 (and so on ...)

Continue

→ Iteration until the entire spectrum of interest is exhausted

List

→ “Prototypes” evolved across the entire spectrum of interest

%% Comment: The above procedure allows elucidating the features in a given

band, say, Band- M, by knowing the features in the previous band, Band (M −

1) via scaling and similitude principle.

→ That is, the features of Band (M – 1) depict the known ex post

profile or the “model”; hence, the unknown profile of the Band- M inferred ex ante depicts the desired “prototype”

End

------------------------------------------------------------------------------------------------------------------------

Ascertaining “prototypes” of channel models in the desired spectral range (of mm-wave

through THz band stretching into suboptical regions) following the procedure described above,

Page 20 of 35

20

Transactions on Networks and Communications (TNC) Vol. 9, Issue 3, June-2021

Services for Science and Education – United Kingdom

suggested here is to invoke the thematic of Buckingham’s similitude principle to deduce the

“prototypes” of path-loss characteristics at the desired test bands vis-à-vis corresponding

known details of the “models” in the available stretches of lower-end frequency bands.

Hence, commensurate with the objective as above, the model-to-prototype transformation as

needed, a compatible set of path-loss functions (PL1 through PL9) can be formatted in

dimensionless forms as listed in Table A.1 wherein, the pertinent details include both LoS and

non-LoS transmissions across single- and/or multi-path signal transits, each path possibly

posing unique angle-of-arrival (AoA).

Constructing a “prototype” of a path-loss-model

The underlying concept of similitude-based scaling of path-loss up into higher frequency

spectra can be specified via “model”-to-prototype transformation. Using the relevant path-loss

ratio, PLR (= PR/PT) functions indicated in dimensionless forms in Table A.I, the underlying

similitude consideration can be illustrated with the following illustrative example (of Figure 4):

Figure 4 A typical, relative path-loss ratio profile expressed as RF power attenuation (per km)

versus frequency (from 5 GHz through 1000 GHz with fo: Normalization (reference frequency) =

900 MHz [52]. (A: Actual measured data as in [52] and B: Inferred data as per linearized

similitude principle; Band-1 and Band-2 are designated spectra on either side of the corner

frequency, fc/fc = 575 and similitude-based inference is made in Band-2).

Suppose as an example, the data available in [52] on attenuation (per km) versus frequency

range up to 1000 GHz (normalized with a reference frequency, fo = 900 MHz) is plotted as

illustrated in Figure 4 (pursuing band-by-band segmentation strategy indicated in Figure 3). In

Figure 4, the profiles of path-loss versus decades of frequency changes are shown for the bands,

(Band-1 and -2). There are two asymptotes marked (in dotted-lines) with a point of intersection

at a corner frequency, fc/fo = 575. Now for analysis, the left-side of fc/fo taken as Band-1

representing the ex post details of the “model”: FUNC(Band-1)  [(Path-loss)Band-1]; and, the

0

0.2

0.6

1

0 200 400 575 800 1000 1200

g1

= m1

 (f1

/fo

)



g2

= c2

+ m2

 (f2

/fo

)



(f/fo

): Normalized frequency (Dimensionless

)

Path-loss

ratio

Band-1 Band-2

fc

/fo

0.2

600 800 1000 1200

1.0

Path-loss

ratio

Band-2

0.6

A B

Page 22 of 35

22

Transactions on Networks and Communications (TNC) Vol. 9, Issue 3, June-2021

Services for Science and Education – United Kingdom

Table 3. Relative path-loss ratio: Measured empirical data [52] versus values inferred via

similitude principle

Path-loss ratio

f2/fo

Actual

data [52]

Inferred

values

600 0.31 0.29

650 0.36 0.29

700 0.42 0.31

750 0.49 0.34

800 0.55 0.37

850 0.63 0.42

900 0.70 0.48

950 0.78 0.60

1000 0.87 0.64

1050 0.96 0.74

1100 1.05 0.90

Exemplars of lower frequency band “models” of path-loss characteristics

Pursuing the procedure indicated with an example above, the art of similitude adopted towards

framing the model-to-prototype transformations of path-loss characteristics, the following

examples of lower frequency band “models” are indicated to obtain their corresponding

prototypes.

A. Lee’s model [5]

This path-loss model suggests an asymptotic variation in path-loss across decades of frequency

changes as illustrated in Figure 5. For example, suppose, the attenuation versus frequency is

specified and known in the range up to 1000 MHz (denoted as Band-1).

Figure 5. Lee’s model [5]: An example depicting the change in signal attenuation (expressed in

terms of the slope dB/decade) over a range of: 40 MHz through 4000 MHz observed in the path- loss profile. The data is specified within the radius of coverage based on radio horizon distance

and base-station antenna height under non-obstructive point-to-point RF transits.

The signal attenuation (in dB/decade) shown represents a “model”: FUNC(f1-band; 1) 

[(PR/PT)f1-band]

1. Relevantly, the EM power-loss is regarded mainly due to free-space, far-field

EM wave proliferation; as such, it follows the (1/R)-law (as limited by antenna-beaming

considerations). (However, in the event of EM absorption versus frequency due to path- 0

40

80

120

10 1000 100000

Frequency in Hz

Signal

attenuation

in dB/decade

Band-1

Band-2

Page 24 of 35

24

Transactions on Networks and Communications (TNC) Vol. 9, Issue 3, June-2021

Services for Science and Education – United Kingdom

m

m

o m o o

1m "Model" at f 2 1/2 1/2

o m o o

2m "Model" at f T m o o

(2 / ) [0.5 ( / ) ( / ) {1

[a ]

[( / ) / (2 f / f ) ( / )] } 1]

and

[a ] [(d / d ) / (d / )]

             +

=           −

= 

(13)

where do and dT denote the Rayleigh-criterion specified on terrain irregularity (shape) across

Ro and RT respectively; and,  depicts the normalized conductivity responsible for the EM

absorption in the medium.

Now, the scaling of equation (13) can be done via similitude principle to get the “prototype” at

a desired elevated frequency fp >> fm as follows: For the “prototype” being elucidated at fp:

1 2 3

4 A 4 T

R p o o p p o p T p R p o p

( ) ( )

1p p o p 2p p o p

[P (R / ) / P (R / )] [(R / ) / (R / )] [(h / ) / (h / )] [( / )]

[{exp( a (R / ) / (R / )} {exp( a (R / ) / (R / )} ]

− − − 

 

  =        

−    −  

(14)

where the subscripted index ‘p’, as stated before denotes the “prototype” considerations

applicable at fp with the prototype exponent set being:     The normalized set of

parametric data in equations (13) and (14) are:

{PR(R/λm or p), Po(Ro/λm or p), (hT/λm or p), (hR/λm or p), (λo/λm or p), a1 m or p, a2 m or p}

(15)

These entities and the exponents indicated enable formulating the required similitude of

“model” -to- “prototype” function transformation concerning the path-attenuation under

consideration.

With reference to a typical Okumura path-loss model presented in Figure 6, there are three

observable aspects on path-loss characteristics. Relevant path-loss slope varies over the

decades of frequency, the path distance (R) and EM propagation across more than one type of

environment. In essence, the variation of path-loss versus decades of frequency is indicated by

varying path-loss slope for a given antenna height ratio and transceive distance. Thus, the

results in Figure 6 depict segmented asymptotic variations across some (Band-1)-to-(Band-2).

Hence, corresponding pair of, “model”: FUNC(f1-band; 1)  [(PR/PT)f1-band]

1 and a “prototype”:

FUNC(f2-band; 2)  [(PR/PT)f2-band]

2 can be inferred via similitude considerations.

C. Medium-specific absorption losses – Crane model/JPL model on rain-attenuation [53-55]

Precipitation (rain, snow and fog specific) losses are highly dependent on seasonal variations,

rain structure and rain-rate statistics [43], and conform to outdoor EM absorption resulting

from dielectric losses. The extent of such losses per unit distance will depend on the density

and dispersive pervasion of absorbing (lossy) contents being present; and, such lossy dielectric- based EM absorption is also approximately proportional to the square of the frequency of the

propagating wave. Further, the vertical structure as well as, the rain cell-size would play

significant role in making of a relevant global attenuation model. Typically, the so-called Crane

attenuation model is an exemplary depiction, which duly accounts for the parameters on the

path-length (R), rain-rate, rain-height, rain- volume etc. over typical sets of terrains.